**About the Editors**

#### **Georgios Tsantopoulos**

Georgios Tsantopoulos is a Professor at the Department of Forestry and Management of the Environment and Natural Resources, Democritus University of Thrace, Orestiada, Greece. He achieved his Ph.D. at the Department of Forestry and Natural Environment of the Aristotle University of Thessaloniki in 2000. His research interests are forest extension, environmental communication, and public relations. He has written more than 160 papers in international journals, chapters in books, proceedings of international and national conferences, etc. He has also participated in 23 research programs financed by the European Union and National Funds.

#### **Evangelia Karasmanaki**

Evangelia Karasmanaki is a Ph.D. Candidate in Environmental Communication and Policy at the Department of Forestry and Management of the Environment and Natural Resources, School of Agricultural and Forestry Sciences, Democritus University of Thrace, Orestiada, Greece. She has published papers in peer-reviewed journals and conference proceedings. Moreover, she has reviewed papers in various international scientific journals. Some of her research interests are environmental education, environmental communication, environmental policy, and environmental awareness.

## *Editorial* **Energy Transition and Climate Change in Decision-Making Processes**

**Georgios Tsantopoulos and Evangelia Karasmanaki \***

Department of Forestry and Management of the Environment and Natural Resources, Democritus University of Thrace, Pantazidou 193, 68200 Orestiada, Greece; tsantopo@fmenr.duth.gr **\*** Correspondence: evagkara2@fmenr.duth.gr

Humans have been using fossil fuels for centuries, and the development of fossil fuel technology reshaped society in lasting ways. From an economic perspective, countries with significant quantities of fossil fuel deposits are regarded as privileged. They have been able not only to avoid expensive fuel imports but also to develop cost-effective electricity sectors, which, in turn, brought economic development to rural areas lacking other avenues for economic development [1].

The problem, however, is that the combustion of fossil fuels in the energy, residential, and transport sectors is a major source of carbon dioxide emissions which trigger climate change, the most dangerous environmental problem that threatens the survival of all living beings on the planet [2,3]. As concerns about the environmental impact of fossil fuels are growing, the idea of producing clean, inexhaustible, and sustainable energy from alternative energy sources such as wind and sun is gaining attention around the world. Renewable energies generate comparably lower emissions, and even when estimating the emission rates of a renewable facility at all stages (construction, installation, operation, decommission), the emissions of renewables in comparison to fossil fuels are still notably lower. Moreover, the deployment of renewables can contribute to the diversification of energy supply, establish locally-produced power, help countries decrease their dependence on expensive fuel imports, create new job positions, etc.

The risk of climate change and the potential of renewable energy to mitigate emissions are reflected in the policy agenda. Over the past years, strong policies aiming at lowering the emissions of the fossil fuel-fired electricity system have been established. Indicatively, tight regulations are forcing businesses across sectors to reduce their environmental impact while many countries provide incentives to encourage investments in renewable energy sources [4].

That being said, the technology of renewable energy sources is not as developed as that of fossil fuels, with the latter having the momentum of two centuries of development. This means that it is much easier and more affordable to establish new fossil fuel projects rather than renewable ones [5]. Beside this obstacle, the deployment of renewable energy requires public support and acceptance because the public can influence actions aimed at realizing this transition. To avoid conflicts and bolster the efforts to deploy renewables, public attitudes towards climate change, energy transition, and energy must be examined [6,7].

The aim of this Special Issue is to publish research and review papers that will offer insights new into various aspects of the new energy landscape. Such insights may help policymakers reach decisions that will facilitate the shift to a low-carbon economy. The Special Issue includes papers focusing on various topics, including the effectiveness of energy policies, the technological performance of renewable systems, informatics tools and software, as well as public attitudes towards energy topics.

The Special Issue includes the following reviewed works:


**Citation:** Tsantopoulos, G.; Karasmanaki, E. Energy Transition and Climate Change in Decision-Making Processes. *Sustainability* **2021**, *13*, 13404. https://doi.org/10.3390/ su132313404

Received: 30 November 2021 Accepted: 2 December 2021 Published: 3 December 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).


Finally, we are grateful to many people for helping us complete this Special Issue successfully. It would be no exaggeration to say that nothing would have been possible without their contribution. First, we would like to thank all authors who responded to our invitation and submitted their works to our Special Issue. We would like to thank Julie Suo, our Special Issue's Managing Editor, for her continuous support, attentiveness, and kindness. Her role in the successful completion of this Special Issue has been critical. The support and conscientiousness of the Editorial Board of *Sustainability* must also be acknowledged. We would like to thank the academic editors responsible for each submission as well as the reviewers who have generously dedicated a part of their valuable time to reviewing papers for our Special Issue. The success of the journal relies on their meticulousness and competence. Having served as Guest Editors of *Sustainability*, we are certain that *Sustainability* will continue to publish high-quality research and review

papers that provide state-of-the-art knowledge about topics related to the environment, energy, and decision-making. We would also like to express the hope that this Special Issue will make a notable contribution to energy transition and will be used by policymakers in decision-making processes.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Christina Diakaki 1,2,\* and Evangelos Grigoroudis <sup>1</sup>**


**Abstract:** Improving energy efficiency in buildings is a major priority and challenge worldwide. The employed measures vary in nature, and the decision analyst, who is typically the architect, the engineer, or the building expert that has undertaken the task to suggest energy efficient solutions, faces a complex decision problem comprising numerous decision variables and multiple, usually competitive objectives. The solution of such multi-objective problems typically involves some sort of objectives aggregation, which reflects the preferences of the involved final decision maker that is the building's user, occupant, and/or owner. The preferences elicitation, however, is a difficult task, and this paper aims to provide an interactive framework that will allow their consideration in a relatively easy manner. More specifically, a mathematical programming approach is proposed herein, which allows the elicitation and incorporation of the decision maker's preferences in the decision model via the assessment of his/her utility function with the assistance of the multicriteria decision aid method UTASTAR. To study the feasibility and efficiency of the proposed approach, the case of a simple building is examined as an application example. The study results suggest that the proposed approach is capable of helping the decision analyst to suggest energy measures that satisfy, as much as possible, the decision maker's preferences, without having to precisely prescribe them beforehand.

**Keywords:** buildings; energy efficiency; energy efficiency improvement; multi-objective optimization; preference disaggregation; preference elicitation; value system; utility function

#### **1. Introduction**

Despite the long-lasting research and development in the particular field, the problem of improving energy efficiency in buildings still remains under investigation, according to recent reviews [1,2], due to its inherent complexity. The complexity of the problem stems from the involvement of several, typically competitive objectives (e.g., cost versus energy consumption) and the availability of numerous alternative measures (e.g., addition of insulation, change of color, use of cool coatings and renewables, etc.) [3], based on which, a final choice has to be made.

In practice, the specific measures to be adopted are typically suggested by the architect, the engineer, or the building expert, who undertakes the task to study the problem, thus playing the role of the decision analyst (DA). However, for any suggestions to be accepted by the final decision maker (DM), who may be the building's user, occupant, and/or owner, they have to satisfy his/her specific requirements and preferences. This further increases the complexity of the problem, and calls for solution approaches that allow the realistic comparative evaluation of all the available alternatives [4]. Such an approach has been proposed by Diakaki et al. [5], who developed a relevant multi-objective decision model based on the principles of mathematical programming.

The aforementioned model considers as objectives to minimize the primary energy consumption of a building and the released CO<sup>2</sup> emissions during operation, as well as the initial investment cost. The particular formulation lends itself for solution via mathematical

**Citation:** Diakaki, C.; Grigoroudis, E. Improving Energy Efficiency in Buildings Using an Interactive Mathematical Programming Approach. *Sustainability* **2021**, *13*, 4436. https://doi.org/10.3390/ su13084436

Academic Editor: Georgios Tsantopoulos

Received: 13 March 2021 Accepted: 14 April 2021 Published: 15 April 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

optimization techniques [5], as well as evolutionary methods like genetic algorithms [6], should the problem complexity become such that a solution via analytic techniques is no longer feasible. Despite the reduced precision compared to the simulation models typically employed for the evaluation of alternative measures [2], the mathematical programmingbased approach has been proved to allow for the realistic comparative evaluation of all the available, alternative measures [7], it has thus been adopted by several researchers in the field (see, e.g., [8–13]).

Irrespective of the particular technique that one may employ for the solution of a multi-objective mathematical programming problem, to reach a single, final solution, which will be satisfactory, thus acceptable by the corresponding DM, weights need to be assigned to the different objectives [2,14]. These weights reflect the relevant importance of each considered objective to the DM, and/or the trade-off that exists among them, due to their competitive nature. The determination of such weights is a difficult task, as it is very unlikely for a DM to be able to explicitly state his/her preferences and satisfaction levels for each and every considered objective. Thus, rather than trying to determine the criteria weights [14], the implicit elicitation and learning of the preferences and value system of the DM, and their incorporation and use in the decision making process, seems more convenient. The development of such an approach for the multi-objective decision problem of improving energy efficiency in buildings is the purpose of the work presented herein.

Specifically, it is the aim of this paper to present an approach, whereby the DA will manage to reach a single, final solution of maximum utility to the DM, as an outcome of an interactive process of individual inter-alternative preference modelling. To this end, the main principles and rationale of a two-phase, iterative procedure proposed by Siskos and Despotis [15] for similar decision problem settings have been adopted. The procedure starts with identifying an initial compromise solution for the energy efficiency improvement problem established in Diakaki et al. [5] (first phase), and then runs iteratively (second phase) as many times as necessary to extract the DM's aspiration levels for each objective, and estimate a respective utility function, which is used in order to reach a single, final solution, which is as close as possible to the DM's actual preferences and value system. Throughout the iterative procedure, interaction is offered at two levels: (i) interactive modification of the DM's satisfaction levels on the different pursued objectives; and (ii) interactive assessment of the DM's utility function via the development and use of the UTASTAR multicriteria decision aid model [16]. UTASTAR is a preference disaggregation approach, which aims at inferring the value or utility function(s) of a DM, given his/her expressed preferences over a reference set of alternatives.

Through the aforementioned interactive procedure, the proposed approach allows the DA to (a) develop the DM's overall utility function for the considered problem; (b) solve the problem by optimizing the developed utility function, rather than aggregating the individual objective functions of the considered problem via potentially arbitrary weights, like in the original multi-objective problem formulation in [5]; and (c) reach a single, final solution of maximum utility to the DM.

To study the feasibility and efficiency of the proposed approach, the case of a simple building is examined as an application example. The study results suggest that the proposed approach is capable of helping the DA to select and suggest energy measures that satisfy, as much as possible, the DM's spectrum of desires, without having to precisely prescribe them beforehand.

The rest of the paper is structured in three more sections. Section 2 introduces the proposed approach, while Section 3 presents the application example. Section 4 discusses the results and findings, and Section 5, finally, summarizes the conclusions of the study and highlights future research directions.

#### **2. Materials and Methods**

#### *2.1. Overview*

The approach proposed herein builds upon the mixed-integer, non-linear, multiobjective optimization problem developed by Diakaki et al. [5], which may be generally defined as follows:

$$\begin{array}{l}\min[\mathcal{g}\_1(\mathbf{x}), \mathcal{g}\_2(\mathbf{x}), \dots, \mathcal{g}\_n(\mathbf{x})] \\\text{subject to } \mathbf{x} \in X\_\prime \end{array} \tag{1}$$

where **x** = (*x*1, *x*2, . . . , *xm*) is the vector of *m* binary or continuous decision variables reflecting alternative choices (e.g., doors and windows types that can be used in the building, structure of multi-layer components such as walls, ceilings, and floors, materials to be used for their construction, and systems that can be used for heating, cooling and hot water supply), *X* ⊆ *R <sup>m</sup>* is the feasible region or decision space of the problem under study, which is implicitly dictated by a set of constraints concerning the decision variables and their intermediary relations; and *g*1(**x**), *g*2(**x**), . . . *gn*(**x**) are the values of *n* considered objectives. In the problem defined in [5], *n* = 3, as the considered objectives are the total annual primary energy consumption (MJ/year), the CO<sup>2</sup> emissions (kg CO2/year) released to the environment by the operation of the heating, cooling, and/or hot water supply systems, and the investment cost for the construction or retrofit of the building envelope and the acquisition and installation of the aforementioned systems, respectively.

The decision model (1) is used herein in the following two-phase procedure:


$$\begin{array}{c} \max \, u[\mathbf{g}(\mathbf{x})] \\ \text{subject to } \mathbf{x} \in X, \end{array} \tag{2}$$

where, **g**(**x**) = (*g*1(**x**), *g*2(**x**), . . . , *gn*(**x**)) is the vector of the values of the objectives of the initial Problem (1). The decision Problem (2) is solved in the third step of the process, the solution is presented to the DM, and the iterations restart until a solution is reached that will be sufficiently satisfactory for the DM, so that he/she will not wish to further improve it.

Figure 1 presents the flowchart of the aforementioned procedure, while the following subsection provides the details of its different phases.

**Figure 1.** The flowchart of the interactive multi-objective mathematical programming approach.

#### *2.2. The Interactive Mathematical Programming Approach*

#### 2.2.1. Phase One

As mentioned earlier, within the first phase of the proposed interactive mathematical programming approach, the individual objectives of decision Problem (1) are minimized and maximized to establish the initial lower and upper bounds of the objectives. More specifically, the lower bound *l<sup>i</sup>* , which is the ideal solution for each objective *i*, with *i* = 1, 2, . . . , *n*, is calculated as follows:

$$l\_i = \min[g\_i(\mathbf{x})] \\ \text{subject to } \mathbf{x} \in X\_{\text{\textquotedblleft}} \\ \tag{3}$$

while for the upper bound *h<sup>i</sup>* , which is the anti-ideal solution, the following problem is solved:

$$h\_i = \max[g\_i(\mathbf{x})]
\\
subject\ to \mathbf{x} \in X.\tag{4}$$

Then, an initial compromise solution is obtained via the solution of the following problem:

$$\begin{array}{c} \min z \\ \text{subject to } \mathbf{x} \in X \\ z > m\_i(g\_i(\mathbf{x}) - l\_i), i = 1, 2, \dots, n \\ z > 0 \end{array} \tag{5}$$

where

$$m\_{\mathbf{i}} = d\_{\mathbf{i}} / \sum\_{\mathbf{i}} d\_{\mathbf{i}\prime} \mathbf{i} = 1, 2, \dots, n\_{\prime} \tag{6}$$

and

$$d\_{\mathbf{i}} = (h\_{\mathbf{i}} - l\_{\mathbf{i}}) / h\_{\mathbf{i}\prime} \text{ } i = 1, 2, \dots, n. \tag{7}$$

The solution of Problem (5) is the closest one to the ideal values of the objectives calculated via (3) in the sense of the weighted Tchebycheff norm.

#### 2.2.2. Phase Two

The second phase of the proposed interactive mathematical programming approach is the iterative one, so let *q* be the number of iteration. Let also *X <sup>q</sup>* be the feasible region, *h q i* the upper bound of objective *i*, and **g** *q* the vector of the optimal values of the objectives reached in iteration *q*.

When entering for the first time in phase two, for the upper bounds and the objectives values, the following hold:


In addition, *X* <sup>0</sup> = *X* holds.

Given the above initial values, as well as the lower bounds *l<sup>i</sup>* , i.e., the ideal solutions of the objectives, the three steps described below are successively executed.

#### Step 1

At the first step of phase two, interaction takes place in order to learn the trade-offs among the objectives for the DM. More specifically, the DM is asked to express his/her satisfaction with respect to the values of the objectives that have been reached so far, i.e., for the values in **g** *q*−1 .

If the DM does not find any objective value satisfactory, the multi-objective decision problem has no satisfactory solution. In such case, the problem should be reviewed and revised, and the procedure should restart from phase one. However, if some values in **g** *q*−1 are satisfactory, the DM is asked to suggest the objectives, which he/she insists to further decrease, and the whole set of objectives *G* is split in the following two complementary sets:


Given the split of *G* in the two subsets, the DM is asked again to suggest, if there are any objectives in *GD*, which could be increased to make room for the desired further decrease of the objectives in *GD*. If the response to this question is no, there is no room for further improvement, the procedure stops, and the solution reached so far is the best compromise solution to the examined problem. If, however, the response of the DM is yes, the upper bounds of the objectives are updated as follows:

$$h\_i^q = \begin{cases} \ g\_i^{q-1} \text{ if } g\_i \in \mathcal{G}\_D \\\ h\_i^{q-1} \text{ if } g\_i \in \overline{\mathcal{G}}\_D \end{cases} \tag{8}$$

For each *g<sup>j</sup>* ∈ *GD*, the following problem is solved:

$$\begin{array}{l}\min g\_{j}(\mathbf{x})\\\text{subject to } \mathbf{x} \in X\\g\_{i}(\mathbf{x}) \le h\_{i}^{q} \; i = 1, 2, \dots, n \text{ and } i \ne j,\end{array} \tag{9}$$

and the feasible region is finally reduced as shown below:

$$X^{q} = X^{q-1} \cap \left\{ \mathbf{x} \in \mathbb{R}^{m} / g\_{i}(\mathbf{x}) \le h\_{i}^{q}, i = 1, 2, \dots, n \right\}. \tag{10}$$

#### Step 2

The second step of phase two is also a learning process aiming at the DM's preferences elicitation. To this end, for an arbitrary chosen integer *s*, *s* + 1 reference alternative profiles *ak* , with *k* = 0, 1, . . . ,*s*, are generated. Each profile comprises a coordinate *aik* for each objective *i*, which is calculated as follows:

$$a\_{ik} = l\_i + (k/s) \left( h\_i^q - l\_i \right) \tag{11}$$

Apparently, any other number of alternative profiles, as well as profiles generation procedure, can be adopted, as long as the generated profiles are representative of the trade-off among the objectives, and do not dominate each other. As their purpose is not to be offered to the DM as possible problem solutions, the generated profiles do not need to be efficient or feasible. They are just presented to the DM, who is asked to rank order them. The ranked set of alternative profiles is then used in the UTASTAR method [16] to assess the DM's utility function *u*[**g**(**x**)], as described in Appendix A.

#### Step 3

The utility function assessed in Step 2 is maximized in this last step over the feasible region. In other words, the Problem (2) is modified as follows, to take into account the reduction of the decision space according to (10):

$$\max \mu[\mathbf{g}(\mathbf{x})]$$
 
$$\text{subject to } \mathbf{x} \in X^{q}. \tag{12}$$

and then solved.

The solution of Problem (12) marks the end of the current iteration, and the procedure restarts from step 1 with the new solution, feasible region, and upper objective bounds.

#### **3. Application Example**

#### *3.1. Overview of the Decision Problem*

To assess the feasibility and efficiency of the proposed approach in suggesting measures that satisfy the competitive objectives of the energy efficiency improvement problem in a way that is compatible with the preferences and value system of the DM, the case of a simple building is studied. The building, taken from the study of Diakaki et al. [5], assumes an envelope, which comprises a floor and ceiling area of 100 m<sup>2</sup> , 2 walls of area 24 m<sup>2</sup> , 2 walls of area 30 m<sup>2</sup> , and a door and window area both of 6 m<sup>2</sup> .

The decisions regarding the considered building concern appropriate choices for:



#### **Table 1.** Alternative door types.

#### **Table 2.** Alternative window types.


#### **Table 3.** Alternative wall structures.


#### **Table 4.** Alternative ceiling structures.


#### **Table 5.** Alternative floor structures.


#### **Table 6.** Alternative insulation materials.



#### **Table 7.** Alternative heating-only systems.

**Table 8.** Alternative heating–cooling systems.


**Table 9.** Alternative heating–hot water supply systems.


#### **Table 10.** Alternative hot water supply-only systems.




The values of the thermal and solar transmittance, and the thermal conductivity of construction materials and components in Tables 1–6 have been taken from the ASHRAE database [17], while the cost values in all the aforementioned tables were obtained through a short, unofficial market survey that took place for the needs of the study described in [5].

The application of the multi-objective decision modelling approach to the particular decision problem leads to a mathematical model of the form (1), which includes 18 continuous and 57 binary variables. The model, which is summarized in Appendix B, aims at determining measures that minimize the following three objectives:


These objectives are competitive, since, typically, the cost-efficient solutions are less environmentally friendly and vice versa. Thus, the search for a globally optimal solution is infeasible, and the DA has to search for a feasible solution, which will comply as much as possible with the DM's preferences and value system. To assist the DA in this search, the multi-phase iterative procedure described in Section 2 is applied.

## *3.2. Application of the Interactive Mathematical Programming Approach* 3.2.1. Phase One

In the first phase of the proposed approach, the individual objectives of the examined decision problem are minimized and maximized, according to (3) and (4), respectively, in order to establish the ideal and anti-ideal solutions of the problem. In addition, an initial compromise solution is identified via the solution of Problem (5).

Table 12 summarizes the outcomes of this phase. The outcomes clearly demonstrate that the choices made depend on the pursued objective(s). For example, when the objective is solely to minimize the primary energy consumption, the most energy efficient choices are made in contrast to the choices made when aiming solely at the reduction of the initial investment cost. In this latter case, the cheapest choices are made, which are the worst from the energy efficiency perspective. These two objectives are clearly competitive to each other, but also to the emissions objective. The release of CO<sup>2</sup> emissions does not depend solely on the generation efficiency of the heating, cooling, and hot water supply systems, but also on the utilized fuel. Thus, some energy efficient choices are no longer efficient when emissions come into the picture.

Table 13 summarizes and highlights the basic information about the problem at hand, which has been generated by the proposed approach in phase one. More specifically, the table comprises the ideal and anti-ideal objective values, the initial upper bound for each objective, the initial compromise solution, as well as the rate of closeness of the objectives to their ideal values, being calculated as follows:

$$\text{Rate of closseness to the ideal solution} = 100 \frac{\text{g}\_i^q - l\_i}{h\_i - l\_i} \tag{13}$$

with *q* the number of iteration; for phase one, *q* = 0 holds. Apparently, the lower the value of the rate, the better.

Table 13 makes clear that the initial compromise solution comprises choices that lead the objectives of primary energy consumption and release of CO<sup>2</sup> emissions very close to their ideal solutions (rates of closeness are 0.85% and 1.59%, respectively). The initial investment cost, on the other hand, is not similarly close to its ideal value (rate of closeness is 38.80%), and this may cause dissatisfaction to the DM. For this reason, the second phase of the proposed approach is activated, to examine the satisfaction level of the DM and refine, if necessary, the problem solution.


**Table 12.** Summary of phase one outcomes.

EHC: electrical system that will be used for both heating and cooling (see Table 8); EHW: electrical system that will be used for both heating and hot water supply (see Table 9); EC: electrical system that will be used only for cooling (see Table 8); NEH: electrical system that will be used only for heating (see Table 7); NEW: non-electrical system that will be used only for hot water supply (see Table 10).



3.2.2. Phase Two-Iteration 1-Step 1

Entering in phase two, the basic information of Table 13 is presented to the DM. Assuming that he/she is satisfied by the performance on objectives 1 and 2, but asks for an improvement on objective 3, i.e., a further cost reduction, even at the expense of the other two objectives, the following sets are formed:


and the upper bound of objective 3 is updated as follows:

$$h\_3^1 = h\_3^0 = 15540.\tag{14}$$

Being members of set *GR*, the upper bounds of the other two objectives remain equal to their initial values, i.e.:

$$\begin{aligned} h\_1^1 = h\_1^0 = 722123, \\ h\_2^1 = h\_2^0 = 74559. \end{aligned} \tag{15}$$

Then, the Problem (9) is solved for the third objective, which is the only member of set *GR*:

$$\begin{array}{l} \min g\_3(\mathbf{x}) \\ \text{subject to } \mathbf{x} \in X \\ g\_1(\mathbf{x}) \le h\_1^1 \\ g\_2(\mathbf{x}) \le h\_2^1 \end{array} \tag{16}$$

and the feasible region of the decision problem is reduced as follows:

$$X^1 = X^0 \cap \left\{ \mathbf{x} \in R^{75} / \mathcal{g}\_i(\mathbf{x}) \le h\_{i\prime}^1 \, i = 1, 2, 3 \right\} \,\tag{17}$$

with *X* <sup>0</sup> being the decision space *X* of the initial problem.

3.2.3. Phase Two-Iteration 1-Step 2

On the basis of information from step 1 and assuming *s* = 9, 10 alternative profiles *a<sup>k</sup>* , *k* = 0, 1, . . . , 9, are generated, according to Equation (11), and presented to the DM in order to rank order them. Table 14 presents the profiles of these alternatives for each objective, along with their assumed ranking *r, r* = 1, 2, . . . , 10.


**Table 14.** Reference set of alternatives and DM's ranking.

The information of Table 14 is then used in UTASTAR, leading to the marginal utility functions graphically displayed in Figure 2, which define the global utility of the DM via the following additive function:

$$u[\mathbf{g}(\mathbf{x})] = u\_1(g\_1(\mathbf{x})) + u\_2(g\_2(\mathbf{x})) + u\_3(g\_3(\mathbf{x})),\tag{18}$$

or the equivalent:

$$u[\mathbf{g}(\mathbf{x})] = 0.300u'\_1(\mathcal{g}\_1(\mathbf{x})) + 0.525u'\_2(\mathcal{g}\_2(\mathbf{x})) + 0.175u'\_3(\mathcal{g}\_3(\mathbf{x})),\tag{19}$$

where *u* 0 *i* , *i* = 1, 2, 3, are the normalized, in the range [0, 1], values of the marginal utilities *ui* , graphically displayed in Figure 3.

**Figure 2.** Marginal utility functions of: (**a**) Primary energy consumption; (**b**) release of CO<sup>2</sup> emissions; (**c**) initial investment cost.

**Figure 3.** Normalized marginal utility functions of: (**a**) Primary energy consumption; (**b**) release of CO<sup>2</sup> emissions; (**c**) initial investment cost.

3.2.4. Phase Two-Iteration 1-Step 3

In this last step of phase two, the utility Function (18) or (19) is maximized over the decreased feasible solution space *X* 1 *,* defined in (17). More specifically, the following problem:

$$\max \iota[\mathbf{g}(\mathbf{x})] = \iota\_1(g\_1(\mathbf{x})) + \iota\_2(g\_2(\mathbf{x})) + \iota\_3(g\_3(\mathbf{x})) \tag{20}$$

or its equivalent:

$$\max \, u[\mathbf{g}(\mathbf{x})] = 0.300u'\_1(\mathbf{g}\_1(\mathbf{x})) + 0.525u'\_2(\mathbf{g}\_2(\mathbf{x})) + 0.175u'\_3(\mathbf{g}\_3(\mathbf{x})) \tag{21}$$
 
$$\text{subject to } \mathbf{x} \in X^1$$

is solved.

The solution of any of the aforementioned problems generates the new compromise solution, displayed in Table 15, the current iteration is terminated, and a new iteration starts from step 1.


**Table 15.** Basic information generated in iteration *q* = 1 of phase two.

3.2.5. Phase Two-Iteration 2-Step 1

The second iteration of phase two starts with the results of Table 15 being presented to the DM. Apparently, the cost objective has been reduced as desired, coming closer to its ideal value; a rate of closeness 17.53% has been achieved, which is also reduced compared to its previous value (38.80%). This improvement, however, has come at the expense of the other two objectives, the values of which, as well as their corresponding rates of closeness, present an increase.

If the consequences of the obtained solution are not satisfactory, the interaction with the DM should continue, like in the previous iteration, until reaching a satisfactory solution. Otherwise, the procedure stops here and the final choices made through this multi-phase procedure (see Table 16) are presented to the DM.

**Table 16.** Initial and final compromise solutions.



EHC: electrical system that will be used for both heating and cooling (see Table 8); NEW: non-electrical system that will be used only for hot water supply (see Table 10).

#### **4. Discussion**

The previous two sections presented an interactive mathematical programming approach to the problem of improving energy efficiency in buildings, and demonstrated its use via an example case study. The problem is difficult to solve as it involves multiple, competitive objectives, and a large number of decision variables, given the large number of available, alternative measures, which can be adopted in this respect. In addition, the solution of the problem requires the DM to express his/her preferences to the considered objectives, a fact that further increases the problem's complexity.

The approach proposed herein exploits the mathematical programming model proposed by Diakaki et al. [5] and the UTASTAR value elicitation method proposed by Siskos and Yannacopoulos [16] under an interactive decision framework, which has been developed following the rationale and principles of the decision-oriented method for multiobjective linear programming problems proposed by Siskos and Despotis [15]. The proposed framework assists the decision making procedure so that decisions are made, which comply with the value system of the DM, without the need to prescribe it beforehand.

The proposed approach can be also adopted in other decision settings within, but also beyond, the field of energy and environment. A similar approach, for example, lies on the basis of ADELAIS, an interactive computer program developed to support the search for a satisfactory solution in multi-objective linear programming problems, which has been used as a tool for the selection of stock portfolios [18]. In contrast, however, to both the initial conception in [15] and the ADELAIS program, the decision framework developed herein concerns a mixed-integer nonlinear mathematical programming problem, which aims at minimizing rather than maximizing the considered multiple objectives. This means that the overall framework can be adopted to any possible decision settings, should adequate care be taken to consider any potential particularities; e.g., in a case where objectives with a positive preference direction (e.g., comfort) should also be considered, to incorporate them, preserving at the same time the required cohesiveness of all considered objectives, their preference direction should be reversed by changing the sign of their corresponding functions. In addition, the mathematical programming formulation is quite flexible, allowing the incorporation of additional DM's objectives and preferences.

#### **5. Conclusions**

The study presented herein demonstrated the feasibility as well as the strengths of applying an interactive mathematical modelling approach to the problem of energy efficiency improvement. The application of such a systematic approach allows for the simultaneous consideration of all available combinations of alternative actions, the consideration of any logical, physical, technical, or other constraints that may apply, and the incorporation of the preferences and value system of the DM without having to explicitly prescribe them beforehand. In addition, the application of the proposed approach ensures that a single, final solution will be reached, which will be satisfactory, and thus acceptable by the corresponding DM.

The proposed approach addresses the problem of improving energy efficiency in buildings in a systematic way. Thus, it can provide the basis for the development of a corresponding decision support system (DSS), which could assist the respective DAs in their difficult task of identifying, among the large volume of available measures, those that will satisfy the needs, requirements, and preferences of the DMs. According to Li et al. [19], there is still plenty of room for the enhancement of the existing relevant toolkits and the development of new ones, and the proposed approach provides the ground in this direction.

**Author Contributions:** Conceptualization, C.D. and E.G.; investigation, C.D. and E.G.; writing original draft, C.D. and E.G.; writing—review and editing, C.D. and E.G. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Appendix A. The UTASTAR Method**

The UTASTAR method proposed by Siskos and Yannacopoulos [16] is a variation of the UTA method, which aims at inferring a set of additive value functions from a given ranking on a reference set *A<sup>R</sup>* of alternative actions *a* ∈ *AR*.

UTASTAR assumes an unweighted additive value function of the form:

$$\mu(\mathbf{g}) = \sum\_{i=1}^{n} \mu\_i(g\_i) \tag{A1}$$

under the normalization constraints:

$$\begin{cases} \sum\_{i=1}^{n} \mu\_i(\mathbf{g}\_i^\*) = 1 \\\ u\_i(\mathbf{g}\_{i\*}) = 0 \end{cases} \forall i = 1, 2, \dots, n \tag{A2}$$

where *n* is the number of criteria; {*g*1, *g*2, . . . , *gn*} is the set of criteria; [*g* ∗ *i* , *gi*∗] is the evaluation scale of criterion *<sup>i</sup>*, with *i* = 1, 2, . . . , *n* and *gi*∗, *g* ∗ *i* the worst and best level of criterion *i*, respectively; *u<sup>i</sup>* is the marginal value function of criterion *i*.

On the basis of the additive Model (A1) and (A2), the value of each alternative *a* ∈ *A<sup>R</sup>* may be expressed as:

$$
\mu[\mathbf{g}(a)] = \sum\_{i=1}^{n} u\_i[g\_i(a)] - \sigma^+(a) - \sigma^-(a), \tag{A3}
$$

where *σ* <sup>+</sup>, *σ* − are the overestimation and underestimation errors, respectively.

In addition, linear interpolation is used in order to estimate the corresponding marginal value functions in a piecewise linear form. More specifically, for each criterion *i*, the interval [*g* ∗ *i* , *gi*∗] is first cut into (*α<sup>i</sup>* − 1) equal intervals, where the points *g j i* are given by the following formula:

$$\mathbf{g}\_{i}^{j} = \mathbf{g}\_{i\*} + \frac{j-1}{\mathbf{a}\_{i}-1} (\mathbf{g}\_{i}^{\*} - \mathbf{g}\_{i\*}) \; \forall j = 1, 2, \dots, \mathbf{a}\_{i} \tag{A4}$$

Then, the marginal value of each action *a* ∈ *AR*, for which *gi*(*a*) ∈ h *g j i* , *g j*+1 *i* i is approximated by the following linear interpolation:

$$u\_{i}[\mathcal{g}\_{i}(a)] = u\left(\mathcal{g}\_{i}^{j}\right) + \frac{\mathcal{g}\_{i}(a) - \mathcal{g}\_{i}^{j}}{\mathcal{g}\_{i}^{j+1} - \mathcal{g}\_{i}^{j}} \left[ u\left(\mathcal{g}\_{i}^{j+1}\right) - u\left(\mathcal{g}\_{i}^{j}\right) \right] \tag{A5}$$

Furthermore, the set of reference actions *A<sup>R</sup>* = {*a*1, *a*2, . . . , *am*} is rearranged so that *a*<sup>1</sup> is the action with the best ranking, *a<sup>m</sup>* is the action with the worst ranking, and for each pair of consecutive actions (*a<sup>k</sup>* , *ak*+<sup>1</sup> ), either *a<sup>k</sup> ak*+<sup>1</sup> (preference) or *a<sup>k</sup>* ∼ *ak*+<sup>1</sup> (indifference) holds, thus if

$$
\Delta(a\_{k'}a\_{k+1}) = \mu[\mathbf{g}(a\_k)] - \mu[\mathbf{g}(a\_{k+1})],\tag{A6}
$$

one of the following holds:

$$\begin{cases} \Delta(a\_{k\prime}a\_{k+1}) \ge \delta \text{ if } a\_k \succ a\_{k+1} \\ \Delta(a\_{k\prime}a\_{k+1}) = 0 \text{ if } a\_k \sim a\_{k+1} \end{cases} \tag{A7}$$

where *δ* is a small positive number, which, however, allows the equivalence discrimination of two actions, which are successive in the ranking.

A final important modification of the UTASTAR method concerns the monotonicity constraints of the criteria that are taken into account through the following transformations:

$$w\_{\mathbf{i}\mathbf{j}} = u\_{\mathbf{i}} \left( \mathbf{g}\_{\mathbf{i}}^{j+1} \right) - u\_{\mathbf{i}} \left( \mathbf{g}\_{\mathbf{i}}^{j} \right) \ge 0 \\ \forall \mathbf{i} = 1, 2, \dots, n \text{ and } \mathbf{j} = 1, 2, \dots, a\_{\mathbf{i}} - 1,\tag{A8}$$

which allow the replacement of the monotonicity conditions for *u<sup>i</sup>* with non-negative constraints for the variables *wij*.

Based on the above, given the ranking over a reference set *A<sup>R</sup>* of alternative actions *a* ∈ *AR*, the UTASTAR method can be implemented via the following four steps:

1. The global value of all reference actions *u*[**g**(*a<sup>k</sup>* )], *k* = 1, 2, . . . , *m*, is first expressed in terms of the marginal values *ui*(*gi*), and then in terms of the variables *wij*, according to (A8), through the following relationships:

$$\begin{cases} u\_i(\mathbf{g}\_i^1) = 0 \,\forall i = 1, 2, \dots, n\\ u\_i(\mathbf{g}\_i^j) = \sum\_{t=1}^{j-1} w\_{ij} \,\forall i = 1, 2, \dots, n \text{ and } \forall j = 2, 3, \dots, a\_i - 1 \end{cases} \tag{A9}$$

2. For each pair of actions, which are consecutive in the given ranking, error terms are introduced using the following relationship:

$$\Delta(a\_k, a\_{k+1}) = u[\mathbf{g}(a\_k)] - \sigma^+(a\_k) + \sigma^-(a\_k) - u[\mathbf{g}(a\_{k+1})] + \sigma^+(a\_{k+1}) - \sigma^-(a\_{k+1}) \tag{A10}$$

3. The following linear programming problem is solved:

$$\begin{aligned} \min z &= \sum\_{k=1}^{m} \left[ \sigma^{+}(a\_{k}) + \sigma^{-}(a\_{k}) \right] \\ \text{subject to } &\Delta(a\_{k}, a\_{k+1}) \ge \delta \text{ if } a\_{k} \succ a\_{k+1} \\ &\Delta(a\_{k}, a\_{k+1}) = 0 \text{ if } a\_{k} \sim a\_{k+1} \\ &\sum\_{i=1}^{n} \sum\_{j=1}^{a\_{i}-1} w\_{ij} = 1 \\ &w\_{ij} \ge 0, \sigma^{+}(a\_{k}) \ge 0, \sigma^{-}(a\_{k}) \ge 0 \,\forall i, j, k \end{aligned} \tag{A11}$$

4. The existence of multiple or near optimal solutions of the Problem (A11) is examined (stability analysis), and the mean additive value function of those (near) optimal solutions is found, which maximize the objective functions:

$$\mu\_i(\mathcal{g}\_i^\*) = \sum\_{j=1}^{\alpha\_i - 1} w\_{it} \,\forall i = 1, 2, \dots, n \tag{A12}$$

on the polyhedron of the constraints of the Problem (A11), bounded by the following additional constraint:

$$\sum\_{k=0}^{m} \left[ \sigma^{+}(a\_k) + \sigma^{-}(a\_k) \right] \le z \ast + \varepsilon,\tag{A13}$$

where *z*∗ is the optimal value of Problem (A11) and *ε* is a very small positive number.

#### **Appendix B. The Multi-Objective Decision Model of the Application Example**

This Appendix provides an overview of the mathematical model of the considered multi-objective problem. The details of the model can be found in [5].

*Appendix B.1. Parameters and Decision Variables*

**Table A1.** Doors-related parameters and data.


**Table A2.** Windows-related parameters and data.



**Table A3.** Walls-related parameters and data.

**Table A4.** Ceilings-related parameters and data.


**Table A5.** Floors-related parameters and data.




**Table A6.** Heating-only systems' parameters and data.


**Table A7.** Cooling-only systems' parameters and data.


**Table A8.** Domestic hot water (DHW) supply-only systems' parameters and data.




**Table A9.** Combined heating–cooling systems' parameters and data.


**Table A10.** Combined heating–DHW systems' parameters and data.


**Table A11.** Solar collectors' parameters and data.



**Table A12.** Fuel and emissions-related parameters and data <sup>1</sup> .

<sup>1</sup> Parameter values have been adopted from [5].



1 It is assumed that the building is located in the wider area of Athens, Greece [5].

**Table A14.** Parameters and data describing comfort-related user preferences and foreseen operational conditions of the building.


<sup>1</sup> Rough estimates assuming 4 inhabitants in the building [5].

**Table A15.** Decision variables.



*Appendix B.2. Multi-Objective Decision Model*

Minimize

$$g\_1(\mathbf{x}) = \frac{\mathbf{Q}^{\mathrm{HD}} SE H\_{el}}{n\_{el}} + \sum\_{f \text{uel} = 1}^{\mathrm{FLL}} \left( \mathbf{Q}^{\mathrm{HD}} SE H\_{\mathrm{nel}, f \mathrm{nd}} \right) + \frac{\mathbf{Q}^{\mathrm{CD}} SE \mathbf{C}\_{el}}{n\_{el}} + \frac{\mathbf{Q}^{\mathrm{WD}} SE \mathbf{W}\_{\mathrm{el}}}{n\_{\mathrm{d}}} + \sum\_{f \text{ud} = 1}^{\mathrm{FLL}} \left( \mathbf{Q}^{\mathrm{WD}} SE \mathbf{W}\_{\mathrm{nel}, f \mathrm{nd}} \right)$$
 
$$g\_2(\mathbf{x}) = \left( \mathbf{Q}^{\mathrm{HD}} SE H\_{\mathrm{d}} + \mathbf{Q}^{\mathrm{CD}} SE \mathbf{C}\_{\mathrm{d}} + \mathbf{Q}^{\mathrm{WD}} SE \mathbf{W}\_{\mathrm{d}} \right) \mathbf{F}\_{\mathrm{Station}} + \sum\_{f \text{ud} = 1}^{\mathrm{FLL}} \left( \mathbf{Q}^{\mathrm{HD}} SE H\_{\mathrm{nel}, f \mathrm{nd}} + \mathbf{Q}^{\mathrm{WD}} SE W\_{\mathrm{nel}, f \mathrm{nd}} \right) \frac{\mathbf{F}\_{\mathrm{fnd}}}{\mathrm{LDF}\_{\mathrm{fnd}}}$$

*g*3(**x**) = *COSTDOR* + *COSTWIN* + *COSTWAL* + *COSTCEIL* + *COSTFLO* + *COSTHS* + *COSTCS* + *COSTWS* + *COSTHCS* + *COSTHWS* + *COSTSLC*

Subject to

*Q HD* = 12 ∑ *n*=1 *Q HD n Q CD* = 12 ∑ *n*=1 *Q CD n Q WD* = 12 ∑ *n*=1 (*WSnDQdDHW*,*n*) *SEHel* = *EHI* ∑ *ehi*=1 *EHJehi* ∑ *ehj*=1 *x EH ehi*,*ehj e EH ehi*,*ehj* ! + *EHCI* ∑ *ehci*=1 *EHCJehci* ∑ *ehcj*=1 *x EHC ehci*,*ehcj e EHC ehci*,*ehcj* ! + *EHWI* ∑ *ehwi*=1 *EHW Jehwi* ∑ *ehwj*=1 *x EHW ehwi*,*ehwj e EHW ehwi*,*ehwj* ! *SEHnel*, *f uel* = *NEHI* ∑ *nehi*=1 *NEHJnehi* ∑ *nehj*=1 *x NEH nehi*,*nehjFUNEH nehi*,*nehj*, *f uel e NEH nehi*,*nehj* + *NEHWI* ∑ *nehwi*=1 *NEHW Jnehwi* ∑ *nehwj*=1 *x NEHW nehwi*,*nehwjFUNEHW nehwi*,*nehwj*, *f uel e NEHW nehwi*,*nehwj* ∀ *f uel* ∈ {1, . . . , *FUEL*}

*SECel* = *ECI* ∑ *eci*=1 *ECJeci* ∑ *ecj*=1 *x EC eci*,*ecj e EC eci*,*ecj* ! + *EHCI* ∑ *ehci*=1 *EHCJehci* ∑ *ehcj*=1 *x EHC ehci*,*ehcj e EHC ehci*,*ehcj* ! *SEWel* = *EWI* ∑ *ewi*=1 *EW Jewi* ∑ *ewj*=1 *x EW ewi*,*ewj e EW ewi*,*ewj* ! + *EHWI* ∑ *ehwi*=1 *EHW Jehwi* ∑ *ehwj*=1 *x EHW ehwi*,*ehwj e EHW ehwi*,*ehwj* ! *SEWnel*, *f uel* = *NEWI* ∑ *newi*=1 *NEW Jnewi* ∑ *newj*=1 *x NEW newi*,*newjFUNEW newi*,*newj*, *f uel e NEW newi*,*newj* + *NEHWI* ∑ *nehwi*=1 *NEHW Jnehwi* ∑ *nehwj*=1 *x NEHW nehwi*,*nehwjFUNEHW nehwi*,*nehwj*, *f uel e NEHW nehwi*,*nehwj* ∀ *f uel* ∈ {1, . . . , *FUEL*} *Q HD <sup>n</sup>* = *HS<sup>n</sup> BLC*(*θIH* − *θE*,*n*)*T<sup>n</sup>* + *ρaircairVair*(*θIH* − *θE*,*n*) − *QAINHGT<sup>n</sup>* − *WN* ∑ *wn*=1 *A WIN wn FF*,*wnFS*,*wnFCM*,*wn ISL*,*wn*,*<sup>n</sup> S* ∑ *s*=1 *Ts* ∑ *t*=1 *x WIN st g WIN st* , if positive 0, else *Q CD <sup>n</sup>* = *CS<sup>n</sup> WN* ∑ *wn*=1 *A WIN wn FF*,*wnFS*,*wnFCM*,*wn ISL*,*wn*,*<sup>n</sup> S* ∑ *s*=1 *Ts* ∑ *t*=1 *x WIN st g WIN st* +*QAINHGT<sup>n</sup>* − *BLC*(*θIC* − *θE*,*n*)*T<sup>n</sup>* − *ρaircairVair*(*θIC* − *θE*,*n*) , if positive 0, else *BLC* = *DR* ∑ *dr*=1 *A DOR dr b DOR dr V* ∑ *v*=1 *x DOR <sup>v</sup> UDOR v* + *WN* ∑ *wn*=1 *A WIN wn b WIN wn S* ∑ *s*=1 *Ts* ∑ *t*=1 *x WIN st UWIN st* + *WL* ∑ *wl*=1 (*AWAL wl <sup>b</sup>WAL wl* ) *W* ∑ *w*=1 *xWAL w KWLw* ∑ *kwl*=1 *d dWAL w*,*kwl kkmWAL <sup>w</sup>*,*kwl* ! + *Yw* ∑ *y*=1 *x dWAL wy Cwy* ∑ *c*=1 *xmWAL wyc kmWAL wyc* !!!! + *CE* ∑ *ce*=1 (*A CEIL ce b CEIL ce* ) *D* ∑ *d*=1 *x CEIL d KCLd* ∑ *kcl*=1 *d dCEIL d*,*kcl kkmCEIL <sup>d</sup>*,*kcl* ! + *Fd* ∑ *f*=1 *x dCEIL d f Ad f* ∑ *a*=1 *xmCEIL d f a kmCEIL d f a* ! + *FL* ∑ *f l*=1 *A FLO f l b FLO f l H* ∑ *h*=1 *x FLO h KFLh* ∑ *k f l*=1 *d dFLO h*,*k f l kkmFLO <sup>h</sup>*,*k f l* ! + *Eh* ∑ *e*=1 *x dFLO he Ghe* ∑ *g*=1 *xmFLO heg kmFLO heg* !!!! *DQDHW*,*<sup>n</sup>* = *Qdhwu* − *ASLC ISL*,*SLC*,*<sup>n</sup> FS*,*SLC U* ∑ *u*=1 *Bu* ∑ *b*=1 (*x SLC ub e SLC ub* ) 106 , if *Qdhwu* ≥ *ASLC ISL*,*SLC*,*<sup>n</sup> FS*,*SLC U* ∑ *u*=1 *Bu* ∑ *b*=1 (*x SLC ub e SLC ub* ) 106 0, else *QdSLC*,*<sup>n</sup>* = *ASLC ISL*,*SLC*,*nFS*,*SLC U* ∑ *u*=1 *Bu* ∑ *b*=1 *x SLC ub e SLC ub* 10<sup>6</sup> *COSTDOR* = *DR* ∑ *dr*=1 *A DOR dr <sup>V</sup>* ∑ *v*=1 *x DOR <sup>v</sup> C DOR v COSTWIN* = *WN* ∑ *wn*=1 *A WIN st <sup>S</sup>* ∑ *s*=1 *Ts* ∑ *t*=1 *x WIN st C WIN st COSTWAL* = *WL* ∑ *wl*=1 *A WAL wl <sup>W</sup>* ∑ *w*=1 *x WAL w KWLw* ∑ *kwl*=1 *d dWAL <sup>w</sup>*,*kwl CKmWAL <sup>w</sup>*,*kwl* + *Yw* ∑ *y*=1 *x dWAL wy Cwy* ∑ *c*=1 *x mWAL wyc C mWAL wyc* !!! *COSTCEIL* = *CE* ∑ *ce*=1 *A CEIL ce <sup>D</sup>* ∑ *d*=1 *x CEIL d KCL<sup>d</sup>* ∑ *kcl*=1 *d dCEIL <sup>d</sup>*,*kcl CKmCEIL <sup>d</sup>*,*kcl* + *Fd* ∑ *f*=1 *x dCEIL d f Ad f* ∑ *a*=1 *x mCEIL d f a C mCEIL d f a* 

*COSTFLO* = *FL* ∑ *f l*=1 *A FLO f l <sup>H</sup>* ∑ *h*=1 *x FLO h KFL<sup>h</sup>* ∑ *k f l*=1 *d dFLO <sup>h</sup>*,*k f l CKmFLO <sup>h</sup>*,*k f l* + *Eh* ∑ *e*=1 *x dFLO he Ghe* ∑ *g*=1 *x mFLO heg C mFLO heg* !!! *COSTHS* = *EHI* ∑ *ehi*=1 *EHJehi* ∑ *ehj*=1 *x EH ehi*,*ehjCSTEH ehi*,*ehj* + *NEHI* ∑ *nehi*=1 *NEHJnehi* ∑ *nehj*=1 *x NEH nehi*,*nehjCSTNEH nehi*,*nehj COSTCS* = *ECI* ∑ *eci*=1 *ECJeci* ∑ *ecj*=1 *x EC eci*,*ecjCSTEC eci*,*ecj COSTWS* = *EWI* ∑ *ewi*=1 *EW Jehi* ∑ *ewj*=1 *x EW ewi*,*ewjCSTEW ewi*,*ewj* + *NEWI* ∑ *newi*=1 *NEW Jnehi* ∑ *newj*=1 *x NEW newi*,*newjCSTNEW newi*,*newj COSTHCS* = *EHCI* ∑ *ehci*=1 *EHCJehci* ∑ *ehcj*=1 *x EHC ehci*,*ehcjCSTEHC ehci*,*ehcj COSTHWS* = *EHWI* ∑ *ehwi*=1 *EHW Jehwi* ∑ *ehwj*=1 *x EHW ehwi*,*ehwjCSTEHW ehwi*,*ehwj* + *NEHWI* ∑ *nehwi*=1 *NEHW Jnehwi* ∑ *nehwj*=1 *x NEHW nehwi*,*nehwjCSTNEHW nehwi*,*nehwj COSTSLC* = *ASLC U* ∑ *u*=1 *Bu* ∑ *b*=1 *x SLC ub CSTSLC ub* ∑ *V v*=1 *x DOR <sup>v</sup>* = 1 ∑ *S <sup>s</sup>*=<sup>1</sup> ∑ *Ts t*=1 *x WIN st* = 1 ∑ *W w*=1 *x WAL <sup>w</sup>* = 1 ∑ *Cwy c*=1 *x mWAL wyc* = *x WAL <sup>w</sup>* , ∀(*y* = 1, . . . ,*Yw*, ∀*w* = 1, . . . , *W*) *x dWAL wy* ∈ h 0, *d WAL* max,*wy*<sup>i</sup> , ∀(*y* = 1, . . . ,*Yw*, ∀*w* = 1, . . . , *W*) ∑ *D d*=1 *x CEIL <sup>d</sup>* = 1 ∑ *Ad f a*=1 *x mCEIL d f a* = *x CEIL d* , ∀(*f* = 1, . . . , *F<sup>d</sup>* , ∀*d* = 1, . . . , *D*) *x dCEIL d f* ∈ h 0, *d CEIL* max,*d f* <sup>i</sup> , ∀(*f* = 1, . . . , *F<sup>d</sup>* , ∀*d* = 1, . . . , *D*) ∑ *H h*=1 *x FLO <sup>h</sup>* = 1 ∑ *Ghe g*=1 *x mFLO heg* = *x FLO h* , ∀(*e* = 1, . . . , *E<sup>h</sup>* , ∀*h* = 1, . . . , *H*) *x dFLO he* ∈ h 0, *d FLO* max,*he*<sup>i</sup> , ∀(*e* = 1, . . . , *E<sup>h</sup>* , ∀*h* = 1, . . . , *H*) *EHI* ∑ *ehi*=1 *EHJehi* ∑ *ehj*=1 *x EH ehi*,*ehj* + *EHCI* ∑ *ehci*=1 *EHCJehci* ∑ *ehcj*=1 *x EHC ehci*,*ehcj* + *EHWI* ∑ *ehwi*=1 *EHW Jehwi* ∑ *ehwj*=1 *x EHW ehwi*,*ehwj* + *NEHI* ∑ *nehi*=1 *NEHJnehi* ∑ *nehj*=1 *x NEH nehi*,*nehj* + *NEHWI* ∑ *nehwi*=1 *NEHW Jnehwi* ∑ *nehwj*=1 *x NEHW nehwi*,*nehwj* = 1 *ECI* ∑ *eci*=1 *ECJeci* ∑ *ecj*=1 *x EC eci*,*ecj* + *EHCI* ∑ *ehci*=1 *EHCJehci* ∑ *ehcj*=1 *x EHC ehci*,*ehcj* = 1 *EWI* ∑ *ewi*=1 *EW Jehi* ∑ *ewj*=1 *x EW ewi*,*ewj* + *EHWI* ∑ *ehwi*=1 *EHW Jehwi* ∑ *ehwj*=1 *x EHW ehwi*,*ehwj* + *NEWI* ∑ *newi*=1 *NEW Jnewi* ∑ *newj*=1 *x NEW newi*,*newj* + *NEHWI* ∑ *nehwi*=1 *NEHW Jnehwi* ∑ *nehwj*=1 *x NEHW nehwi*,*nehwj* = 1 *U* ∑ *u*=1 *Bu* ∑ *b*=1 *x SLC ub* ≤ 1

#### **References**


**Konstantinos Ioannou 1,\* and Dimitrios Myronidis <sup>2</sup>**


**Abstract:** The number of solar photovoltaic (PV) arrays in Greece has increased rapidly during the recent years. As a result, there is an increasing need for high quality updated information regarding the status of PV farms. This information includes the number of PV farms, power capacity and the energy generated. However, access to this data is obsolete, mainly due to the fact that there is a difficulty tracking PV investment status (from licensing to investment completion and energy production). This article presents a novel approach, which uses free access high resolution satellite imagery and a deep learning algorithm (a convolutional neural network—CNN) for the automatic detection of PV farms. Furthermore, in an effort to create an algorithm capable of generalizing better, all the current locations with installed PV farms (data provided from the Greek Energy Regulator Authority) in the Greek Territory (131,957 km<sup>2</sup> ) were used. According to our knowledge this is the first time such an algorithm is used in order to determine the existence of PV farms and the results showed satisfying accuracy.

**Keywords:** PV farms; deep learning; satellite imagery; CNN; automatic detection

**Citation:** Ioannou, K.; Myronidis, D. Automatic Detection of Photovoltaic Farms Using Satellite Imagery and Convolutional Neural Networks. *Sustainability* **2021**, *13*, 5323. https://doi.org/10.3390/su13095323

Academic Editor: Maria Malvoni

Received: 15 April 2021 Accepted: 7 May 2021 Published: 10 May 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

**1. Introduction**

During the last three decades mankind is witnessing an evolution in the energy sector as we notice a shift in energy production methods, from the usage of fossil fuels (petroleum, natural gas, coal, etc.) to more environmentally friendly methods. This is caused mainly due to the fact that a significant portion of the world's carbon dioxide production is a result of fossil fuels used for energy production [1–3].

However, as electricity consumption plays an important role for modern societies (and its usage cannot be reduced) other forms of energy production must be used in order to satisfy current and future energy demands [3–7].

Renewable energy methods can be considered as a viable solution for energy production and the reduction of CO<sup>2</sup> emissions. These methods include the usage of sustainable sources based on wind, water, biomass, solar and geothermal energy for energy production which are in general called renewable energy sources (RES) [8].

The exploitation of solar energy is considered as one of the most common types of RES. Solar panels are used for transforming energy from indecent sunlight, to electricity using solar cells based on the photovoltaic effect, thus they are also called photovoltaic (PV) panels [9]. Nowadays, massive arrays of PV panels (in the form of solar or PV farms) are used for energy production throughout the world. These farms energy production capability ranges from 1 to 2000 MW, in the case of mega projects covering thousands of hectares [10].

In Europe, PV farms account for 13% of the total RES production. Furthermore, solar power is the fastest-growing source: in 2008, it accounted for 1%. This means that the

growth in electricity from solar power has been dramatic, rising from 7.4 TWh in 2008 to 125.7 TWh in 2019 [11].

In Greece, data provided by the Regulatory Authority for Energy (RAE) indicate that currently there are 9791 PV potential installations (farms) in a variety of stages (licensed investments, licensed installations, licensed production or under evaluation), currently producing 715.6 MW of electric energy.

The variety of the existing stages of PV farms is making difficult to track the infiltration of PV to the Greek market as in many cases the time period from the initial evaluation of the energy production license to production can be years. Financial difficulties, public reaction against the investment as well as technical difficulties can pause the entire installation process.

In this work we investigate a new method of collecting installed PV information which is potentially cheaper and faster than existing methods. The proposed approach uses an algorithm which can automatically detect the existing PV farms based on high resolution free to use satellite imagery, current RAE data for training and deep learning techniques. The entire methodology can be divided in two separate steps.

The first step involves the association of the data provided by RAE with satellite images. For the implementation of this step, we used an algorithm for automatically annotating the images and matching RAE data with satellite images in order to create two datasets. A high-resolution dataset and a low-resolution dataset.

The second step involve the usage of the output produced in the first step in order to train a deep learning (DL) algorithm to automatically detect the PV farm's locations. The algorithm apart from the determination of the locations can also help scientists to extract other information. As it is basically a data unaware algorithm, it can also provide information such as the effect of land use in the selection of PV farm locations, the effect of micrometeorology to the installation locations etc.

The proposed approach offers a series of benefits when compared with other data analysis methods. First it allows the scalability of the produced results as well as the automatic improvement of the data collection. Usage of higher resolution images will provide the user with better results. Thus, the user is free to use data which originate from a variety of sources even from Google Earth, with the best results however, to be expected with data from paid services such as LandSat [12,13].

Additionally, the implementation of the approach using a computer algorithm allows the automation of the process. The entire procedure is easy to use and can be executed multiple times in order to monitor the installation rate. The produced information can also help scientists to predict the level of energy produced as well as help the Government to initiate programs related with RES adoption and provide a valuable tool to enhance the decision-making process regarding the determination of potential installation sites [2,14]. Finally, the presented methodology can be easily adapted in order to monitor other types of RES and reproduced in other regions.

#### **2. Literature Review**

Computer applications, sensor networks as well as the Internet of Things are responsible for the creation of enormous amounts of data [15]. For this reason, new and innovative techniques must be applied in order to perform sufficient analysis of the accumulated data. Deep Learning is a part of machine learning (ML) methods based on the usage of artificial neural networks with representation learning (supervised, semi-supervised or unsupervised learning) [16].

Essentially DL is a methodology where many classifiers work together, and it is based on linear regression followed by activation functions. DL foundation relay on the same traditional statistical linear regression approach. The only difference is that there are many neural nodes in deep learning instead of only one node (in the case of linear regression). These nodes are known as neural network, and one classifier (a node) is known as perceptron. The network is organized in layers and each layer can have many

hundreds or even thousands of nodes. Layers which are situated between the input and output layers constitute the hidden layer and accordingly the nodes which constitute this layer are known as hidden nodes. In contrary with traditional machine learning classifiers where the user must write complex hypothesis, in deep neural network applications the hypothesis is generated by the network itself, making it a powerful tool for learning nonlinear relationships effectively [16].

ML can be divided into two development phases, shallow learning (SL) and deep learning. The most widely spread SL methods include logistic regression, support vector machine (SVM) and Gaussian mixture models [17–26]. SL main disadvantage is that it cannot handle complex real-world problems such as voice and image recognition [16]. On the contrary DL specializes in solving problems such as image classification, voice recognition etc. For example, image classification of 1000 kinds of images provided a classification error rate of 3.5% which is higher than the accuracy of ordinary people [27].

Various DL algorithms were used for disease determination. Quiroz and Alferez [28] used DL image recognition of legacy blueberries in the rooting stage, planted in smart farms in Chile. For this reason, they used a convolutional neural network (CNN) to detect the presence of trays with living blueberry plants, the presence of trays without living plants and the absence of trays. The model produced results with 86% accuracy, 86% precision, 88% recall and 86% F<sup>1</sup> score.

Other researchers used DL for apple pathology image recognition and diagnosis [29]. For this reason, they trained a CNN that obtained a recall rate of 98.4% using error back propagation analysis of sampled elements. In the study of Liu et al. [30], DL was used for the identification of citrus cancer based on the AlexNet model, with an optimized network structure which could reduce the network parameters while maintaining the same level of accuracy. The results from the application showed that the recognition accuracy reached 98%. In the study of Amara et al. [31], DL was used for detecting two well-known banana diseases. For this reason, they used a deep CNN based on the LeNet architecture, with the results accuracy at 85.9%, precision accuracy 86.7%, recall 85.9% and F<sup>1</sup> score 86.3%.

DL was also used for other types of image recognition. Huang et al. [32] used DL for determining crack and leakage defects on metro shield tunnels which produced very good results with an identification error of 0.8%. Yang et al. [33] used a DL algorithm (in this case a modified AlexNet model) was used in order to determine wind turbine blade damage on images taken from an unmanned aerial vehicle. The model provided better results (97.1% average accuracy) when compared to the unmodified AlexNet model and support vector machine models. In [34], a DL approach was proposed for the classification of road surface conditions. For this, they used a CNN network and created a new activation function based on the rectified linear unit function. Their results showed a classification accuracy of 94.89% on the road state database. DL were also used to perform breast cancer classification. A new method called BDR-CNN-CGN was used to perform classification of breast cancer types, the results showed improved detection rates (accuracy 96.10%) compared to other neural network models [35]. A CNN was also used in order to perform COVID-19 diagnosis. The proposed CNN employed several new techniques such as rank-based average pooling and multiple-way data augmentation. Among the eight proposed models, the model named FGCNet performed better with performance percentage higher than 97% [36]. Finally, Malog et al. [37], used high resolution satellite imagery and deep forest algorithm in order to detect roof top installed photovoltaic arrays. Their data included imagery from an area of 135 Km<sup>2</sup> and the results showed 99.9% pixel-based detection accuracy and 90% object-based detection accuracy. Table 1 presents an overview of the aforementioned literature.

**Table 1.** State of the Art.


#### **3. Materials and Methods**

For the creation of the image data sets we used data provided by RAE as well as, data which are available from Apple Maps. Apple Maps is a free map service based on satellite data which are provided from DigitalGlobe. RAE data included a series of polygons (in Shape file form) which included all PV farm investments in Greece (Figure 1). The data were categorized depending on the status of the investment in:


Each shape file was at first converted to GeoJSON format. GeoJSON is a geospatial data interchange format compatible with the GNU/General Public License (GPL) guidelines, based on JavaScript Object Notation (JSON). It defines several types of JSON objects and the manner in which they are combined to represent data about geographic features, their properties, and their spatial extents. GeoJSON uses a geographic coordinate reference system, World Geodetic System 1984, and units of decimal degrees [38].

A special PYTHON algorithm was written in order to match the polygons with base map data. The algorithm used a GNU/GPL library called jimutmap in order to read each polygon in GeoJSON form and create an image file. Thus, concluding the first step of the methodology. Jimutmap allows the user to select different zoom levels when annotating the data and create images of different resolutions.

**Figure 1.** Map of Greece with PV farms in various phases (Basemap from Hellenic Cadastre). **Figure 1.** Map of Greece with PV farms in various phases (Basemap from Hellenic Cadastre).

Each shape file was at first converted to GeoJSON format. GeoJSON is a geospatial data interchange format compatible with the GNU/General Public License (GPL) guidelines, based on JavaScript Object Notation (JSON). It defines several types of JSON objects and the manner in which they are combined to represent data about geographic features, their properties, and their spatial extents. GeoJSON uses a geographic coordinate refer-In Figure 2, we can easily observe that the library user, can easily select the zoom level value, using the zoom variable, and thus determine the resolution of the images created (higher zoom level creates images with lower resolution). This is due to the fact that satellite imagery provided by free services has limited resolution. Additionally, the library allows the usage of multiple core threads in order to perform quicker the required annotations. *Sustainability* **2021**, *13*, x FOR PEER REVIEW 6 of 15

ence system, World Geodetic System 1984, and units of decimal degrees [38].

```
A special PYTHON algorithm was written in order to match the polygons with base 
map data. The algorithm used a GNU/GPL library called jimutmap in order to read each 
polygon in GeoJSON form and create an image file. Thus, concluding the first step of the 
methodology. Jimutmap allows the user to select different zoom levels when annotating 
the data and create images of different resolutions. 
    In Figure 2, we can easily observe that the library user, can easily select the zoom 
level value, using the zoom variable, and thus determine the resolution of the images cre-
ated (higher zoom level creates images with lower resolution). This is due to the fact that 
satellite imagery provided by free services has limited resolution. Additionally, the library
```
allows the usage of multiple core threads in order to perform quicker the required anno-**Figure 2.** Code snippet from jimutmap library. **Figure 2.** Code snippet from jimutmap library.

tations. The second step included training a convolutional neural network to automatically detect the PV farm's locations. The CNN was developed using Google Collaboratory or Google Colab (GC) for short. GC is a product from Google Research allowing users to write and execute arbitrary PYTHON code using their browser, and is especially well suited to machine learning, data analysis. Additionally, it provides access to advanced cloud resources including the ability for the user to use graphics processor units (GPU's) The second step included training a convolutional neural network to automatically detect the PV farm's locations. The CNN was developed using Google Collaboratory or Google Colab (GC) for short. GC is a product from Google Research allowing users to write and execute arbitrary PYTHON code using their browser, and is especially well suited to machine learning, data analysis. Additionally, it provides access to advanced cloud resources including the ability for the user to use graphics processor units (GPU's) and

(which are installed on all personal computers using the x64 architecture), GPU's are specialized electronic circuits designed to accelerate the creation and manipulation of images. Their highly parallel structure makes them more efficient than general-purpose (CPUs) for algorithms that process large blocks of data in parallel [39]. TPU's are artificial intelligence accelerator application-specific integrated circuits (ASICs) developed by Google specifically for neural network machine learning using TensorFlow a free and open-

Convolutional neural networks (CNN) are inspired by the cat's cortex and were first proposed in the 1980s [41]. A CNN has similar structure with other multilayer neural networks, and it is comprised of layers. Each layer is composed of a number of two-dimensional planes and each plane has independent neurons. Sparse connections are used between layers, meaning that the neuron in each feature map only connects to the neurons in a small area in the upper map, in contrast with the traditional neural networks. The CNN structure depends mainly in the shared weight, the local experience field and the

The following figure (Figure 3) presents the layout of a CNN. In this case the network is comprised from an input layer, four hidden layers and an output layer. This network was created for performing image processing. In more detail image recognition of characters written by hand. In this case the input layer is made up using 28 × 28 sensory nodes. This layer receives the images which have been approximately centered and normalized in terms of size. Afterwards the computational layouts alternate between convolution and

• The first hidden layer is responsible for the convolution. This layer consists of four feature maps, with each feature map consisting of 24 × 24 neurons. Each neuron is

• The second hidden layer is responsible for subsampling and local averaging. Like the previous layer, it also consists of four feature maps, but each feature map is now made up of 12 × 12 neurons. Each neuron has a receptive field of size 2 × 2, a trainable coefficient, a trainable bias, and a sigmoid activation function. The trainable coefficient

• The third hidden layer is responsible for the second convolution. It consists of 12 feature maps, with each feature map consisting of 8 × 8 neurons. Each neuron in this

source software library for machine learning [40].

sub-collector to ensure the invariance of input data [42].

assigned a receptive field of 5 × 5 size.

and bias control the operating point of the neuron.

*3.1. Convolutional Neural Networks* 

subsampling as follows:

tensor processing units (TPU's). Unlike normal central processor units—CPU's (which are installed on all personal computers using the x64 architecture), GPU's are specialized electronic circuits designed to accelerate the creation and manipulation of images. Their highly parallel structure makes them more efficient than general-purpose (CPUs) for algorithms that process large blocks of data in parallel [39]. TPU's are artificial intelligence accelerator application-specific integrated circuits (ASICs) developed by Google specifically for neural network machine learning using TensorFlow a free and open-source software library for machine learning [40].

#### *3.1. Convolutional Neural Networks*

Convolutional neural networks (CNN) are inspired by the cat's cortex and were first proposed in the 1980s [41]. A CNN has similar structure with other multilayer neural networks, and it is comprised of layers. Each layer is composed of a number of twodimensional planes and each plane has independent neurons. Sparse connections are used between layers, meaning that the neuron in each feature map only connects to the neurons in a small area in the upper map, in contrast with the traditional neural networks. The CNN structure depends mainly in the shared weight, the local experience field and the sub-collector to ensure the invariance of input data [42].

The following figure (Figure 3) presents the layout of a CNN. In this case the network is comprised from an input layer, four hidden layers and an output layer. This network was created for performing image processing. In more detail image recognition of characters written by hand. In this case the input layer is made up using 28 × 28 sensory nodes. This layer receives the images which have been approximately centered and normalized in terms of size. Afterwards the computational layouts alternate between convolution and subsampling as follows:


The result of the previously described processes is the application of a bipyramidal effect. This means that with each convolutional or subsampling layer, the number of features maps is increased while the spatial resolution is reduced, compared to the corresponding previous layer.

CNN's first usage was for the identification of handwritten checks in banks, but they were incapable of recognizing large images. For this reason, ref. [43] developed LeNet-5 which was a classical model of convolutional neural network with low error rates (only 0.9% on the MNIST data-set).

**Figure 3.** Convolutional Network for image processing. **Figure 3.** Convolutional Network for image processing.

The result of the previously described processes is the application of a bipyramidal effect. This means that with each convolutional or subsampling layer, the number of features maps is increased while the spatial resolution is reduced, compared to the corresponding previous layer. CNN's first usage was for the identification of handwritten checks in banks, but they The main bottleneck on the application of CNN is the long training time due to many hidden nodes on the networks. However, weight sharing which is a characteristic of the CNN allows parallel processing of weights if the proper infrastructure exists. Today as modern graphics processor units (GPU's) support parallel computing the application of CNN's is easier. In [44], a GPU algorithm was used in order to solve the ImageNet problem.

hidden layer may have synaptic connections from several feature maps in the previous hidden layer. Otherwise in operates in a manner similar to the first convolutional

• The fourth hidden layer is responsible for performing a second subsampling and local averaging. It consists of 12 feature maps, but with each feature map in this case consisting of 4 × 4 neurons. Otherwise, it operates in a manner similar to the first sampling

• Finally, the output layer is responsible for the final stage of convolution. This layer consists of 26 neurons, with each neuron assigned to one of 26 possible characters. As

before each neuron is assigned a receptive field of size 4 × 4 [42].

were incapable of recognizing large images. For this reason, [43] developed LeNet-5 which was a classical model of convolutional neural network with low error rates (only 0.9% on the MNIST data-set). The main bottleneck on the application of CNN is the long training time due to many The CNN implemented for automatically detecting PV farms was based on Keras 2.3.0, a deep learning application programing interface written in PYTHON 3.7, running on top of the machine learning platform TensorFlow 2.4.1 supported by Google Colab. Keras was developed with a focus on enabling fast experimentation.

#### hidden nodes on the networks. However, weight sharing which is a characteristic of the CNN allows parallel processing of weights if the proper infrastructure exists. Today as *3.2. Building the Model*

layer.

layer.

modern graphics processor units (GPU's) support parallel computing the application of CNN's is easier. In [44], a GPU algorithm was used in order to solve the ImageNet problem. The CNN implemented for automatically detecting PV farms was based on Keras 2.3.0, a deep learning application programing interface written in PYTHON 3.7, running on top of the machine learning platform TensorFlow 2.4.1 supported by Google Colab. Keras supports various image classification models (Xception, ResNet, MobileNet, VGG, etc.). In this study we used the InceptionV3 model mainly because it performs significantly better than the other Keras Supported models [45]. The images that will be used were randomly divided in two categories, Training Images used for training and validating the model and evaluation images used for determining the network performance against new, unseen, images.

Keras was developed with a focus on enabling fast experimentation. *3.2. Building the Model*  Keras supports various image classification models (Xception, ResNet, MobileNet, VGG, etc.). In this study we used the InceptionV3 model mainly because it performs sig-Before presenting the images to the network we perform a series of augmentations which will ensure that our model would never use twice the exact same picture thus, the model will try to overfit on the training data. For this reason, we used the image data preprocessing function of Keras. This function has a series of arguments for manipulating the training image datasets. The following arguments were used for the manipulation:

	- Height shift range, shifts the image along the X axis;
	- Width shift range, shifts the image along the Y axis;
	- Horizontal flip, flips the image across the X axis;
	- Vertical flip, flips the image across the Y axis;
	- Validation split, determines the fraction of images reserved from the training dataset for model validation;
	- Zoom range, determines the zoom factor;
	- Brightness range, modifies the image brightness level;
	- Rescale, determines if the image is rescaled to specific dimensions;

Continuing, we must determine the training epochs as well as the image batch size. Epochs refers to the number of times the network is trained through the entire dataset, whereas batch size determines the number of samples processed each time (before the model is updated).

In InceptionV3 we have the capability to use predefined training weights using the imagenet or initialize them randomly. Imagenet is an image database which is organized according to the WordNet hierarchy in which each node of the hierarchy is depicted by thousands of images [46]. The usage of this database is proven to significantly increase a CNN's performance [47]. Figure 4 displays the entire workflow of the model applied. *Sustainability* **2021**, *13*, x FOR PEER REVIEW 9 of 15

**Table 2.** Image per dataset.

#### **4. Results 4. Results**

The PYTHON algorithm used for extracting the images of PV farms created 570 images files. Of them, 220 where high-resolution images (approximately 1 MB each) and 350 where low-resolution (approximately 16 KB each). These images where divided randomly in Training and Evaluation datasets as show on Table 2. The PYTHON algorithm used for extracting the images of PV farms created 570 images files. Of them, 220 where high-resolution images (approximately 1 MB each) and 350 where low-resolution (approximately 16 KB each). These images where divided randomly in Training and Evaluation datasets as show on Table 2.

Low Res 250 100

Following that, the datasets where augmented using the image data processing func-

tion. The parameters used in this function are presented in Table 3.


**Table 2.** Image per dataset.

Following that, the datasets where augmented using the image data processing function. The parameters used in this function are presented in Table 3.



Next, the images were imported to Keras and the InceptionV3 algorithm was applied, for 15, 20 and 25 epochs with a batch size of 15 using the ImageNet pre-trained weights. Batch size number was selected mainly because the number of the images used for training and validation is rather small. Generally, we use larger batch sizes when we have large datasets. The selection on the number of training epochs is based on the produced results (there is no guideline regarding the train period of a neural network). This means that if we notice overfitting in the results (meaning that the network cannot generalize properly), then we reduce training epochs.

Table 4 includes the results taken from the three training sessions applied. The results show the percentage of correct prediction using the training dataset and the validation dataset. From the table it is evident that the applied model does not provide better results when trained for more than 20 epochs, as it can also be seen in the graphical representation of the results in Figure 5. From Figure 5 it is also obvious, that the model performs erratically during the last validation session with large fluctuations during the validation of the model. This means that the model must have overfitted during training for 25 epochs.

**Table 4.** Training and validation results.


*Sustainability* **2021**, *13*, x FOR PEER REVIEW 11 of 15

**Figure 5.** Model train and accuracy results for 15 epochs (**a**), 20 epochs (**b**) and 25 epochs (**c**). **Figure 5.** Model train and accuracy results for 15 epochs (**a**), 20 epochs (**b**) and 25 epochs (**c**).

Additionally, from the same figure it is also obvious that the model performs better when trained for 15 epochs (although the training performance in this session is slightly smaller compared to the performance during the next training session). As it can be seen in model accuracy section of the diagrams, the model validation line follows more closely the training line. Generally, models with a smaller curve fluctuation during accuracy elevation have better training convergence. Furthermore, model training is better when the two curves (train and validation) are closer. After training completion, the model is also tested against new data which were not used during train and validation sessions. The produced evaluation results are shown on Table 5 and Figure 6 which also prove that the model trained for 15 epochs provides the best overall predictions. Additionally, from the same figure it is also obvious that the model performs betterwhen trained for 15 epochs (although the training performance in this session is slightly smaller compared to the performance during the next training session). As it can be seen in model accuracy section of the diagrams, the model validation line follows more closely thetraining line. Generally, models with a smaller curve fluctuation during accuracy elevation have better training convergence. Furthermore, model training is better when the two curves (train and validation) are closer. After training completion, the model is also tested against new data which were not used during train and validation sessions. The produced evaluation results are shown on Table 5 and Figure 6 which also prove that the model trained for 15 epochs provides the best overall predictions.

On Table 5, Pv 1 refers to the high-resolution images' dataset, whereas Pv 2 refers to the low-resolution images' dataset. Precision is the ability of the classifier not to label as positive a sample that is negative. Or in other terms, precision is the number of correct

Precision = |ሼ௩௧ ௗ௨௧௦ሽ∩ሼ௧௩ௗ ௗ௨௧௦ሽ|

Recall = |ሼ௩௧ ௗ௨௧௦ሽ∩ሼ௧௩ௗ ௗ௨௧௦ሽ|

Recall is the ability of the classifier to find all the positive samples. Or in other terms,




Weighted Avg 0.53 0.52 0.52 245 0.49 0.49 0.49 245 0.51 0.51 0.51 245

results divided by the number of all returned results.

recall is the fraction of relevant documents that are successfully retrieved.

and recall:

the dataset.

1.

epochs in both datasets (high and low resolution).

**Figure 6.** Confusion matrices for 15 epochs (**a**), 20 epochs (**b**) and 25 epochs (**c**). **Figure 6.** Confusion matrices for 15 epochs (**a**), 20 epochs (**b**) and 25 epochs (**c**).

**5. Discussion**  For most researchers, terms such as deep learning and machine learning seem interchangeable concerning the world of artificial intelligence. However, this approach is mistaken. Deep learning is a specialized subset of machine learning which, in turn, is a spe-On Table 5, Pv 1 refers to the high-resolution images' dataset, whereas Pv 2 refers to the low-resolution images' dataset. Precision is the ability of the classifier not to label as positive a sample that is negative. Or in other terms, precision is the number of correct results divided by the number of all returned results.

F1 score is a measure of the test's accuracy. It is the harmonic mean of the precision

The worst value for this measure is 0 whereas the best is obtained when it equals to

Accuracy is the weighted arithmetic mean of Precision and Inverse Precision (weighted by Bias) as well as the weighted arithmetic mean of Recall and Inverse Recall (weighted by Prevalence). Inverse Precision and Inverse Recall are simply the Precision and Recall of the inverse problem where positive and negative labels are exchanged.

Macro Average, computes the F1 for each label and returns the average without considering the proportion for each label (in our case high- and low-resolution PV images) in the dataset. Weighted Average computes F1 for each label (in our case high- and low-resolution PV images) and returns the average considering the proportion of each label to the dataset. Finally, support is the number of occurrences of the given class (or label) in

The results on Table 4 indicate that the trained model produce's better results for 15

 = ଵܨ

Higher accuracy values demonstrate better model performance.

2

(3) ଵ݊ି݅ݏ݁ܿ݅ݎ + ଵ݈݈݁ܿܽିݎ

$$\text{Precision} = \frac{|\{relevant\, documents\}\cap \{retrieved\, documents\}\|}{|\{retrieved\, documents\}\|}\tag{1}$$

data and the need for substantial computing power for its usage. However, the application of deep learning algorithms nowadays is a necessity. The evolvement of Internet of Things has created multiple devices capable of collecting a va-Recall is the ability of the classifier to find all the positive samples. Or in other terms, recall is the fraction of relevant documents that are successfully retrieved.

$$\text{Recall} = \frac{|\{relevant\, documents\} \cap \{retrieved\, documents\}|}{|\{relevant\, documents\}|} \tag{2}$$

rithms have no requirement for human intervention as the nested layers in the neural networks put data through hierarchies of different concepts, and eventually learn through their own errors. Therefore, the usage of DL algorithms can greatly help toward the pro-F<sup>1</sup> score is a measure of the test's accuracy. It is the harmonic mean of the precision and recall:

$$F\_1 = \frac{2}{recall^{-1} + precision^{-1}}\tag{3}$$

processing. Thus, they can (if trained properly) used for solving many problems, including image detection and classification. The worst value for this measure is 0 whereas the best is obtained when it equals to 1. Accuracy is the weighted arithmetic mean of Precision and Inverse Precision (weighted by Bias) as well as the weighted arithmetic mean of Recall and Inverse Recall (weighted by Prevalence). Inverse Precision and Inverse Recall are simply the Precision and Recall of the inverse problem where positive and negative labels are exchanged. Higher accuracy values demonstrate better model performance.

Macro Average, computes the F<sup>1</sup> for each label and returns the average without considering the proportion for each label (in our case high- and low-resolution PV images) in the dataset. Weighted Average computes F<sup>1</sup> for each label (in our case high- and lowresolution PV images) and returns the average considering the proportion of each label to the dataset. Finally, support is the number of occurrences of the given class (or label) in the dataset.

The results on Table 4 indicate that the trained model produce's better results for 15 epochs in both datasets (high and low resolution).

#### **5. Discussion**

For most researchers, terms such as deep learning and machine learning seem interchangeable concerning the world of artificial intelligence. However, this approach is mistaken. Deep learning is a specialized subset of machine learning which, in turn, is a specialized subset of artificial intelligence. Deep learning describes algorithms that analyze data with a structure which is similar to how a human would draw to a conclusion. The

only drawback in the application of DL is the requirement of incredibly vast amounts of data and the need for substantial computing power for its usage.

However, the application of deep learning algorithms nowadays is a necessity. The evolvement of Internet of Things has created multiple devices capable of collecting a variety of unstructured data, ranging from simple arithmetic values to images from satellites. Therefore, the need arises to evaluate this data and extract useful patterns. DL algorithms have no requirement for human intervention as the nested layers in the neural networks put data through hierarchies of different concepts, and eventually learn through their own errors. Therefore, the usage of DL algorithms can greatly help toward the process of collected data, mainly because these algorithms ignore the data types which are processing. Thus, they can (if trained properly) used for solving many problems, including image detection and classification.

This study presents a novel approach towards the problem of automatic recognition of PV farms. The recognition is based on the usage of satellite imagery and image classification techniques which until recently were used for other purposes (face recognition, flora and fauna species recognition, etc.). According to our research it is the first time that neural networks (in particular a CNN) was used for the automatic detection of PV farms. From the literature review we conducted, the only similar research used a CNN for the determination of small rooftop installed PV arrays, however we did not find any other similar research, which indicates that our approach is pioneering.

Furthermore, another novelty of our approach is that the used dataset's as well as the software (libraries, functions algorithms) used for the implementation of this research are freely available to the researchers, thus making our methodology easily replicable.

The results showed that (even though the original dataset was rather small) we can expect correct identification accuracy reaching 60% when using high resolution imagery and lower results in case we use lower resolution. From the confusion matrixes we can determine that for 15 epochs 127 correct identifications were performed, 125 correct identifications were performed for 20 epochs and 125 were also recognized correctly for 25 epochs.

However, the identification results can be further improved if we use larger datasets. Additionally, the results showed that, increase in the number of training epochs does not provide significant improvements. Table 5 presents the results showing that 15 training epochs can be considered adequate for the dataset used.

Finally, the application of the algorithm also proven that high resolution images perform significantly better even in smaller datasets compared to low resolution imagery. This result was not expected because we believed that increasing the number of low-res input data could compensate for the lower resolutions, mainly due to the fact that input data are characterized by a specific geometry.

The approach presented in this work can also be applied in the recognition of other types of RES, if trained properly. It can also be used in other cases where automatic image recognition is necessary. The results could be improved by using images provided from paid services (and therefore high resolution) and by using larger datasets. Further improvements can be achieved if the user performs some kind of image pre-processing on the dataset (edge detection, color corrections, etc.), or deeper networks (more hidden layers).

#### **6. Conclusions**

Image recognition can provide a valuable tool for monitoring the adaption rate of renewable energy sources. Modern deep learning methods are unaware of the processing data and therefore can be easily used in order to recognize the various forms of RES (wind turbines, PV panels, hydroelectric stations, etc.). However, there is a need for large datasets in order to train properly the algorithms. The existence of various satellite imagery services allows the user to collect these data in a variety of resolutions and create datasets which contain images of RES forms in a variety of installation environments, various angles, different weather and time. Therefore, it is possible to create a tool which will be capable of identifying them with increased accuracy.

This paper examined a first approach towards this goal. The dataset is based on the usage of PV farms in Greece and the results proved to be adequate given the size of the training dataset. As the years pass and more installations complete the algorithm can be trained again in order to increase its efficiency. Furthermore, advancements in computer technology and DL algorithms can also help towards this goal.

Finally, the combination of these algorithms with other types of software capable of calculating the annual solar energy output can help local and regional authorities to plan their energy policy. The methodology can also be used from the national authorities in an attempt to continuously monitor current RES status, determine the investment/adoption rate of RES in the various regions and regional units, and act as an overall tool for the application of national policy.

**Author Contributions:** Conceptualization, K.I.; methodology, K.I.; software, K.I.; validation, K.I. and D.M.; formal analysis, K.I.; resources, K.I. and D.M.; data curation, K.I.; writing—original draft preparation, K.I. and D.M. Both authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data sharing not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Review* **Social Acceptance of Carbon Capture and Storage (CCS) from Industrial Applications**

**Katja Witte**

**Citation:** Witte, K. Social Acceptance of Carbon Capture and Storage (CCS) from Industrial Applications. *Sustainability* **2021**, *13*, 12278. https://doi.org/10.3390/su132112278

Academic Editors: Georgios Tsantopoulos and Evangelia Karasmanaki

Received: 25 September 2021 Accepted: 3 November 2021 Published: 7 November 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Wuppertal Institute for Climate, Environment and Energy, Division Future Energy and Industry Systems, Doeppersberg 19, 42103 Wuppertal, Germany; katja.witte@wupperinst.org; Tel.: +49-202-2492-218

**Abstract:** To limit global warming, the use of carbon capture and storage technologies (CCS) is considered to be of major importance. In addition to the technical–economic, ecological and political aspects, the question of social acceptance is a decisive factor for the implementation of such lowcarbon technologies. This study is the first literature review addressing the acceptance of industrial CCS (iCCS). In contrast to electricity generation, the technical options for large-scale reduction of CO<sup>2</sup> emissions in the energy-intensive industry sector are not sufficient to achieve the targeted GHG neutrality in the industrial sector without the use of CCS. Therefore, it will be crucial to determine which factors influence the acceptance of iCCS and how these findings can be used for policy and industry decision-making processes. The results show that there has been limited research on the acceptance of iCCS. In addition, the study highlights some important differences between the acceptance of iCCS and CCS. Due to the technical diversity of future iCCS applications, future acceptance research must be able to better address the complexity of the research subject.

**Keywords:** carbon capture; acceptance; public perception; industrial applications; literature review; knowledge; awareness; communication

#### **1. Introduction**

To limit global warming to 1.5 ◦C, the use of carbon capture and storage technologies (CCS) is considered to be of major importance [1–5]. In international parlance, CCS stands for a mix of technological processes for CO<sup>2</sup> capture and storage. These are large-scale processes in which carbon dioxide (CO2) is captured from huge CO<sup>2</sup> point sources. The captured CO<sup>2</sup> is transported via pipeline, ship, or heavy transport and then either reused or injected underground into a suitable geological formation (onshore or offshore) [6].

The use of CO<sup>2</sup> capture processes is feasible both in fossil-fired power plants for electricity generation and in energy-intensive industrial processes (for example, steel or cement plants) and could enable a significant reduction in CO<sup>2</sup> emissions in these applications. According to the International Energy Agency [7], fossil-fired power plants accounted for about 42.5% of total global CO<sup>2</sup> emissions in 2013. In comparison, the share of CO<sup>2</sup> emissions caused by industrial activities was around 25%.

In recent years, the discussion around CCS has increasingly focused on its use in the context of industrial facilities (in the following, the term "industrial CCS" is referred to as iCCS). This is mainly because the technical options for the extensive reduction of CO<sup>2</sup> emissions in the area of energy-intensive industries without the use of iCCS are not sufficient to achieve the targeted GHG neutrality in the industrial sector. Ref. [4] However, what exactly distinguishes the term iCCS from the classic CCS application? Fossil fuels are an essential input to the production process of the steel, cement, lime and chemical industries, the so-called energy-intensive industries. These fuels are used in the industries for their chemical and physical properties rather than as a primary energy source for power generation, as is the case with CCS [8]. However, unlike electricity generation, it is not possible to replace fossil fuels with renewable energy sources to reduce emissions. This literature review focuses explicitly on the application of CCS to these industrial processes.

The debate to date on the commercial introduction of CCS in fossil-fired power plants (abbreviated below as CCS) has made it clear that numerous other factors are relevant in addition to purely technical and economic indicators. On the part of policymakers, there is a need for a reliable agreement and strategy on the future role of CCS, taking into account international developments around CCS as well as other technological climate protection paths. This will create planning and legal certainty for industry and society and enable the early development of CO<sup>2</sup> infrastructure.

Another essential factor, which is the focus of this publication, is the social perception of iCCS technologies and the possible assessment of their future acceptance. Previous research on CCS acceptance has made it clear that CCS technologies may meet with strong opposition, especially in regions where the applications have been tested or were intended to be deployed on a long-term, permanent basis [9,10]. For example, in Germany and the Netherlands, some projects to explore potential CO<sup>2</sup> storage formations were abandoned early, primarily due to massive opposition from local communities [11,12]. Since the early 2000s, the number of scientific publications on the acceptance of CCS has continuously increased (see also Section 3). The perception and acceptance of CCS is strongly dependent on the respective country [13] and due to the low level of knowledge about CCS [14,15], it remains difficult to make valid predictions about how specific local attitudes towards CCS might develop.

This study is the first literature review to address the acceptance of industrial CCS (iCCS). The objectives of this study are fourfold. First, it examines the extent to which iCCS acceptance has already been empirically studied. Second, an analytical framework is proposed to systematically review the existing literature. Third, factors that influence iCCS acceptance are identified and discussed based on the review. Fourth, the results on the acceptance of iCCS are compared with the acceptance of CCS in the context of fossil-fired power plants. The assumption is that the attitude of society towards iCCS differs from the attitude towards CCS along individual process steps and value chains. In this regard, first scientific findings are emerging [16,17]. It is unclear in which direction these attitude differences tend.

This study's results should not only contribute to the scientific discussion and further development of the research field, but also hopefully feed into the ongoing practical iCCS discourse in industry and politics. At the international level, there are already associations of industry players testing different technical use cases for iCCS in the form of pilot projects, for example the European Cement Research Academy (ECRA). In some industrial processes, the capture of CO<sup>2</sup> emissions is already practiced today, and currently the first projects are underway worldwide in different sectors, such as chemicals (Illinois Industrial), iron and steel (Abu Dhabi Phase 1), and hydrogen (QUEST) [18]. The results of this literature review should also provide indications of possible communication and empowerment needs on the part of the general public and at the same time enable the more technology-based scientific disciplines to place their developments on iCCS in a broader societal context.

In order to be able to better classify the present analysis, the technological component of the research object should first be explained in more detail. For a better understanding of this, Renn's classification [19] of the three areas of technology and their acceptance parameters is helpful. He distinguishes between (1) products—everyday and leisure technology; (2) technology in working life; and (3) external, large-scale and risky technology. The three technology areas differ in terms of their acceptance testing criteria. In the case of current acceptance research on carbon capture and usage (CCU), for example, the focus is often on the concrete evaluation of an end product, which can often be explained in terms of buying or not buying, manageability, long-term durability or direct physical risks (although the research approach here is also broader, for example [20–22]). In the context of the present analysis, all scientific publications dealing with acceptance research on concrete end products (e.g., mattresses, fuels) of CCU technologies were explicitly excluded. This also appears consistent with [23], who clarify that CO<sup>2</sup> utilization is often compared and contrasted with CCS; however, they are two different technology pathways so it is necessary

to address and evaluate these technologies separately. Since the subject of the present analysis is the broader society, technology area 2, which deals with technology in the workplace and thus targets "employees", can also be excluded. Following the exclusion principle, only studies dealing with iCCS as an external, large-scale and risky technology (area 3) were analyzed here. For this technology area, the test criteria of acceptability are, for example, societal interests, rights, responsibilities, and legitimacy issues. The focus of this review is therefore on technology pathways that capture CO<sup>2</sup> on a large scale and transport it for further purposes without further differentiating whether and how the CO<sup>2</sup> is further used.

This paper is structured as follows. First, Section 2 presents the selection of articles analyzed, the methodological approach and the acceptance factors for CCS already identified in the scientific literature, which are also used here as analysis dimensions. The results of the content analysis are explained in detail in Section 3. In the Discussion (Section 4), we present which of the identified acceptance factors for iCCS can be considered crucial for the further development of iCCS and which scientific implications the results induce. The conclusions in Section 5 illustrate some rough propositions for relevant groups of actors dealing with issues of societal acceptance on iCCS in the future.

#### **2. Materials, Methods and Acceptance Factors**

In order to assess the state of scientific research in the field of acceptance of industrial CCS, a content analysis of scientific articles was conducted. Only articles published in English between 2012 up to and including the end of 2020 were included. This time period was chosen because, to the best of the author's knowledge, no articles were published before 2012 that approached this topic. Thus, the chosen period of analysis seemed sufficient to generate as complete an overview as possible of the state of the scientific literature on this topic.

#### *2.1. Selection of Articles*

Articles were identified using two online databases. First, the online database of the publisher Elsevier (sciencedirect.com), a full-text database with an inventory of more than 16 million articles and book chapters [24]. Although documents from other scientific publishers are not included, Elsevier is one of the top 5 publishers in the world with over 2000 journals published [25]. Second, the online database was used through scholar.google.com. Google's search engine presents only scientific literature; that is, books or papers from professional journals [26]. Using these two most popular online databases, it was possible to generate the largest possible proportion of scientific literature on the topic of iCCS acceptance.

Only scientific papers, book and conference contributions that could be generated by keyword searches via the two online databases were included in the analysis. In addition, one master's thesis was evaluated that was identified via the online database scholar.google.com and appeared to be relevant. No other dissertations or master's or bachelor's theses were systematically searched for.

Items were identified from November 2020 to 16 January 2021. The following search terms were used to select the technology:


The technical search terms were each combined with the following acceptance-related terms:


Using a combination of search terms, between 4099 (maximum at sciencedirect.com) and 16,900 (maximum at scholar.google.com) articles were identified in the two online databases. Only articles that explicitly address the topic of industrial CCS were to be included (see Section 1 for narrowing criteria). For further identification of these articles from the existing material, the so-called PRISMA criteria were followed [27]. Based on this procedure, a complete search strategy for one of the databases used is presented below. The presentation is intended to create the prerequisite for the best possible reproducibility of the search.

The search strategy described here as an example refers to the online database scholar.google.com. As previously described, the initial selection was made according to the search terms presented above. With the search term "carbon capture industry acceptance", approximately 16,900 articles were identified on 16 January 2021 (initial access on 8 November 2020). In advance, the search of the articles was restricted to the years from 2012 to 2020 inclusive in the menu under "select period". Subsequently, the search result was sorted by relevance (an option offered by the online database in the menu). The individual short descriptions of the list of results on the homepage were read (not clicked on) and checked to see if all individual search terms were included in the respective text descriptions. This was an indication that all search terms were actually included in the respective target article. In addition, it was checked whether the keywords appeared in the desired context. If, for example, the term "industry" was linked to "coal industry" and the title also indicated that the article was exclusively about CCS as a low-carbon technology for energy generation, the article was excluded from further analysis. The matches identified in this way were further checked for accuracy of fit by reading the respective abstract or, if this did not appear to be sufficient for assessing accuracy of fit, the conclusions.

All hits identified in this way were then included in the pool for further analysis. During the course of the search, it became apparent that after approximately the fourth to fifth page of results on the homepage, the articles listed no longer appeared relevant for the analysis due to missing keywords in the short text. Additional tools from scholar.google.com were used to further identify relevant articles. The option "cited by" lists all articles in which the original hit was cited. A check of these articles was performed according to the criteria already mentioned. The option "related articles" was also used. Using these options, few additional articles could be identified. In addition, an "alert" was created, which was used to automatically notify the author via email when new articles with the given keywords appeared. This option appeared valuable in generating articles that did not appear until the end of the analysis period. To ensure that all articles published by the end of 2020 were identified, a final search query took place in mid-January 2021. The search query at sciencedirect.com followed the same procedure and selection criteria. Beyond the use of the two online databases, a few articles were identified via the references or sources of the articles already identified and read in the course of the evaluation and included in the analysis pool. Using these procedures, a total of 67 articles were identified and included in the closer analysis.

All 67 articles were then read completely. Of these, 42 articles were excluded. There were two main reasons for articles to be excluded:


Ultimately, 25 articles met the criteria to be included. It can be assumed that a large part of the relevant literature was identified.

#### *2.2. Methodical Approach*

A qualitative content analysis of 25 articles was carried out using the MAXQDA software. The software allows qualitative data and text analyses and is internationally established in the field of science. For content analysis, a deductive category system was developed (referred to as "analysis dimensions" in the following). It was derived from the previous state of attitude and acceptance research on CCS. During the coding process, some of the analysis dimensions were adapted and the possibility was left open to inductively generate new dimensions, in accordance with the approach of [28]. The individual dimensions or acceptance factors are discussed in more detail in the following subsection.

#### *2.3. Acceptance Factors from the Field of CCS*

A wealth of individual studies, results, and initial overview studies are available on the perception, attitude, and acceptance of CCS [29,30]. The first studies on the subject appeared from 2002 [31–33]. In the literature up to 2015, publications on the acceptance of CCS focus mainly on the use of the technology in the context of fossil power generation. Therefore, a considerable number of factors determining the acceptance of CCS have been proposed, many of which are commonly used to explain the acceptance of new technologies. There is not a consensus on the one model best suited to predict CCS or technology acceptance [29], although there are publications that present a technology acceptance framework [34] or provide a model approach for selected factors [20,35,36]. Most studies, as mentioned, examine the determining factors along specific research questions that can be categorized into some thematic groups. These groups of topics mainly include (a) general acceptance analyses "of the general public" in one country or in several countries; (b) analyses of real-life-projects across different groups of actors, including the local society; (c) analyses on communication and participation of CCS; and (d) analyses on specific process steps of CCS, especially storage. In recent years, since 2015, more studies have been added on the topic of CCU [20–23,37–41], which can be assigned to the abovementioned group of topics and perhaps also represent a research unit in their own right (cf. chapter 1). However, these factors have predominantly become established and are repeatedly used as a starting point for new research studies and questions. Additionally, for the analysis of the articles identified here for the topic area of industrial CCS, analysis dimensions were generated on the basis of the acceptance factors just mentioned or the state of science (cf. Table 1, here especially the factors from 1 to 8) (a similar set of influencing factors can also be found in the acceptance research on the energy transition [42]). After the initial review of the articles (relevance check), additional dimensions that seemed useful for analyzing the acceptance of industrial CCS were added (compare factors 9 to 11).


**Table 1.** Analysis dimensions of iCCS acceptance within the framework of the review.

1


#### **Table 1.** *Cont*.

It should be noted that the sources cited in the table are only a small excerpt of possible sources that have dealt with the topic. A comprehensive presentation of studies that have produced results on the respective dimensions of analysis is not intended here. Moreover, the assignment of sources is not exclusive because the respective studies often explored several categories of analysis. In this respect, relevant sources were also assigned to more than one analysis category.

> In the following, the results of the evaluated articles are presented along the acceptance factors described in Table 1. In addition to a presentation of the characteristic features, such as methodology used, year of publication and technology path, the analysis clarifies which influencing factors were assumed and investigated to explain the acceptance of industrial CCS. In Section 4 (Discussion), these results are then reflected on and classified in the context of the entire acceptance research on CCS so that first insights can be gained on whether the acceptance factors on iCCS differ from the previous ones, in which areas they differ, if any, and whether new factors have been added.

#### **3. Results**

#### *3.1. Characteristics of the Analyzed Articles*

To place the iCCS publications in the overall context of all publications on the topic of CCS acceptance, it should be mentioned in advance that until circa 2014 the number of scientific publications on the acceptance of CCS increased steadily [29]. Between 2015 and around 2018, the number of publications on the topic of CCS acceptance then remained at a lower level than in the years between 2010 and 2014 [30]. Up to this point, publications on the acceptability of CCS focused on the use of the technology in the context of fossil fuel power generation. Triggered by the Paris Agreement 2015 [2], which highlighted the urgency of limiting global warming to as close to 1.5 ◦C as possible, as well as a number of other publications [1,3–5], as described in Section 1, the discussion about CCS has continuously broadened and has more often focused on technology pathways that are not directly related to fossil energy production. Since then, there has also been an increasing number of scientific publications dealing with the acceptance of different technology paths of CCS.

The articles analyzed here were published between 2012 and 2020. Table 2 illustrates the year of publication of the articles in combination with the selected technology path.


**Table 2.** Theme clusters of iCCS acceptance in combination with year of publication [13–17,30,41,61–78].

<sup>1</sup> These two articles have already been published in mid-January 2021. Due to their relevance, the author decided to include them before completing this article at the end of January. No other articles from 2021 were included in the analysis.

> As shown in Table 2, by the end of 2020, most articles on iCCS were published in 2018 and 2019 (n = 6 in each year). A slight majority of the 25 articles (n = 13) use the terminology "industrial CCS" (compare row 1 Table 2), but do not further explain which technological concept of iCCS technologies is involved in the definition or within the operationalizations. This is not surprising, as the technological applications of iCCS are highly complex along the process steps and the different value chains that may be involved.

> To address this complexity, four of the studies provided their participants with a selection of different realistic CCS technology pathways to evaluate (compare row 2 Table 2), which at least allowed for a more differentiated view according to different CO<sup>2</sup> sources, such as the evaluation of CO<sup>2</sup> capture in a chemical plant [41]. Since 2019, there has been an increase in acceptance studies investigating the impact of specific industrial CCS applications, such as from cement or steel plants or for the BECCS sector. These studies are often linked to specific project proposals, for example the ALIGN project (It is expected that further scientific publications on the acceptance of iCCUS will be published in 2021 from research projects that have been and will be funded within the framework of Horizon 2020 of the European Commission, such as the ALIGN-CCUS and STRATEGY CCUS projects) [74], and concentrate on regions with industrial clusters that are significant geologically and in terms of their industrial structure with regard to the development of iCCS and are already being scientifically researched in part (compare lines 4 to 6, Table 2).

> The analyzed articles on iCCS acceptance come from a total of 15 different countries, of which European countries represented 13—an overwhelming majority. The following European countries were involved in the preparation of the articles: United Kingdom = 7; The Netherlands and Germany = 4 each; Norway = 3; Finland and Sweden = 2 each; and Austria, Belgium, Lithuania, Portugal, Romania, Spain and Switzerland with one article each. Five of the European articles involved more than one country. As mentioned at the beginning, previous studies on the acceptance of CCS have made clear that protests and

risk perceptions on CCS have formed along exploration plans and projects, especially in Europe—particularly in the Netherlands [12] and Germany [11].

In this respect, if an iCCS strategy is to be pursued on the political level in the long term, these countries seem to have a particular interest in predicting future developments regarding the acceptance of iCCS. For Great Britain, the situation is similar; here, according to [79], 17.2% would "probably not use" or "definitely not use" CCS technologies according to a representative survey. A further three articles come from Russia and another one from the United States of America. According to [30], Russia has a special interest in the use of enhanced oil recovery (EOR) technology, which requires a lot of CO2, and therefore is considering CCS as a future option to develop this technology.

The relevant articles on the acceptance of iCCS were published in a wide range of journals. In total, the 25 articles come from 15 different journals. *The International Journal of Green-house Gas Control* accounts for 8 articles—by far the most. This is followed by the journals *Energy Procedia* and *Journal of Cleaner Production*, with 2 publications each on the topic. One of the analyzed articles is a Master's thesis, which was written at the University of Graz and cannot be assigned to any journal [73].

Different theoretical concepts and approaches were used in the articles included. Twelve of the analyzed articles on iCCS acceptance do not mention any theoretical concepts. The concept of Wüstenhagen [80] to classify three different dimensions of social acceptance is mentioned and applied in two articles. Studies that focus their analysis more on the regional or project level often include actor and communication-related approaches, such as the theory of public engagement in [68], the social licence to operate (SLO) in [14,74], the end-to-end stakeholders involvement approach in [67], the concept of procedural fairness in [62], the concept of media agenda-setting in [75], the stakeholder theory for management in [70] and the cognitive theory of shifting coalitions in [73].

In addition, the articles mention social-psychological concepts that illuminate social behavior even more against the background of cultural aspects and certain values, such as the theory of planned behavior in [30] and, in the context of the Master's thesis, the concept of the Ethical landscape of CCS, the theory of worldviews and the cultural theory to specify belief systems in [73]. Two of the analyzed articles reflect their findings on iCCS acceptance to the whole debate on energy system transformation using the just transition approach [65,78] or the multidimensional research concept as in [15].

A complete table of the analyzed articles with the categories "first author", "year of publication", "method(s) used", "country", "iCCS-related technology", and "important statement in relation to iCCS" is provided in the Appendix A (Table A1: Overview of the analyzed articles).

#### *3.2. Key Findings along the Dimensions of Analysis as well as Additional Insights*

In the following, the main results of the analyzed articles are presented along the analysis dimensions shown in Table 1.

#### 3.2.1. Perceived Benefits

The results of the studies analyzed have identified some benefits that appear to be associated with the use of iCCS and thus may have a positive impact on social acceptance. These benefits include the possibility of creating local and national value through iCCS projects [64].

For example, the municipality of Porsgrunn in Norway considers iCCS important in legitimizing industry in the region and thus sustaining related jobs in the long term [64]. Additionally, ref. [71] sum up that the potential of iCCS can protect and rejuvenate historical employment patterns and this opportunity makes iCCS an attractive option for an area. This is also important to counteract the out-migration of the local population that threatens to occur if established industries go away [64]. Beyond protecting existing jobs, ref. [71] make the argument that providing infrastructure for iCCS can also create additional employment opportunities in the region. Consistent with this, communities hosting CCS

projects would benefit economically from the jobs and revenue that the industry would provide [13].

In addition, regional clusters containing multiple capture projects can benefit from shared CO<sup>2</sup> transport and storage infrastructure to maximize value, share investment decisions and operating costs, and thus reduce development costs [78]. Thus, ref. [64] postulate benefits from mergers of larger regional clusters for iCCS (across national borders). For example, in their study, they identified the notional "Skagerrak Cluster" for the countries of Norway, Sweden, and Denmark, which identifies some key geographic features that have good conditions for establishing iCCS technology (similar to the northeast region of Scotland). The advantages come from the possibility of storing the CO<sup>2</sup> offshore, with emission sources relatively close to the sea. According to [64], the relevance of looking more closely at the Skagerrak cluster provides valuable input for evaluating acceptance and communication challenges for other iCCS clusters in the Nordic region. These benefits of iCCS overall can be linked to increasing the economic viability of both the technology itself and the region in question, these are benefits that [30,70] also highlight in their study.

However, not only is the preservation or renewal of existing economic structures identified as a benefit of iCCS, but the technologies should also serve to promote and profile municipalities and regions as environmental and technological leaders, ultimately to develop new industrial activities [64]. In this context, there is also talk of a potential image boost for iCCS industries and regions [62]. For example, refs. [75,77] argue the relevance of developing and deploying BECCS, a technology pathway discussed as an advantage for forest-rich countries such as Finland [75] and which holds the potential to establish itself as a "first mover" [77]. Without BECCS it would be a challenge to meet emission targets, but with BECCS Finland could gain advantages by saving and trading emission rights [75] (see also Section 3.2.11).

Regarding the impact of environmental effects (reduction of CO<sup>2</sup> emissions, slowing of climate change) and their classification as a benefit for the acceptance of iCCS, there are different results in the analyzed studies. Some study results suggest that attributing the benefits of iCCS to improving the regional and global environmental situation can create an advantage for the perception of acceptance [15,30,70,75]. Similarly, the results of a representative study in Canada, the USA, the UK, the NL, and Norway illustrate that iCCS can help mitigate climate change and support the economy according to the respondents in [13], which could be interpreted as a benefit for the technology. However, the same study also highlighted that framing CCS as dealing with 'waste' (in conjunction with CO<sup>2</sup> reuse) seems to be more persuasive in encouraging support than framing it in terms of climate or economic benefits. The authors of [74] critically note that the siting of new or expanded iCCS facilities is more likely to be associated with national and international benefits, for example achieving energy and climate goals and economic revenues (on this also see [70]), and that the apparent benefit to local communities may turn out to be a potential burden, for example through subjectively perceived risks. Such a perceived imbalance between (negative) local impacts and national or global benefits would pose a challenge when it comes to public response to iCCS technologies [74]. Hence, currently there is no consistent evidence from the scientific community as to whether iCCS is perceived as a mitigation option for CO2, and thus as a climate technology, and whether this has a positive or negative effect on the perception of the benefits of the technology. Moreover, such a perception is certainly also dependent on many regional factors.

For completeness, here are the five main benefits of CCS industrial projects according to [70]: (1) reduction of negative impacts on the environment, (2) contribution to socio-economic development of regions and territories, (3) attractive direction for socially responsible investments, (4) support for sustainable development of companies involved in CCS projects, (5) use of CO<sup>2</sup> for purposes such as improving oil recovery by oil and gas companies, increasing energy efficiency of industrial companies.

The analysis of perceived benefits gives the impression, as also indicated by [30] and previous studies on the benefits and risks of CCS, that benefit perception may exert a stronger influence on iCCS acceptance than risk perception.

#### 3.2.2. Perceived Risks

According to the studies analyzed, the use of iCCS technologies is associated with various societal risks that can have a negative impact on acceptance. These include perceived risks at the local level, for subsequent generations and for ecological and economic systems, but also risks for making political decisions that do not contribute to improving climate protection in the long term. The most frequently mentioned risk perceptions in the studies relate to negative health impacts, especially for people living near CO<sup>2</sup> storage and transport infrastructure [62].

The local impacts of iCCS are particularly addressed here [68], and with it the accompanying sense of unfair treatment of those who suffer disadvantages [30,74]. It is believed that iCCS could become locally entrenched as a "risky technology" in the perception of local and regional populations [15], especially if CO<sup>2</sup> storage occurs on land [77]. Hazards are expected from possible CO<sup>2</sup> leakage and seismic risks [15,75,77]. The perception would not improve even if already existing infrastructure were used [15]. The same applies to the CO<sup>2</sup> transport route; here, too, leakages and unforeseen risks are feared by the population [15]. In addition, several stakeholders in Germany expected so-called spillover effects, which occur when already existing rejections of CO pipelines are transferred to CO<sup>2</sup> pipelines on the grounds that these transport options are not sufficiently differentiated in society [15].

In this context, the fear of a lack of acceptance of responsibility on the part of politics and industry [71,77] and the societal desire to avoid uncertainties are mentioned [30], especially when it comes to long-term monitoring of CO<sup>2</sup> infrastructure, which is primarily intended to ensure the protection of future generations [71,73]. In addition to health risks from the use of iCCS, ecological risks were also mentioned in the analyzed articles [15,75,76], which can have an unfavorable impact on acceptance. For example, interventions in the ecological system through the construction of new CO<sup>2</sup> infrastructure can permanently endanger the environment [15]. In addition, one study expressed fears about the possible effects of stored CO<sup>2</sup> in the seabed [73], which could, for example, affect the fauna and flora of nearby coastal regions and lead to catastrophic consequences there [71]. At the same time, the use of iCCS technologies was interpreted as a standstill for other climate protection measures in industry that would lead to lock-in effects of unsustainable corporate practices [73]. However, the results on the perception of iCCS technologies are partly contradictory; on the other hand, there is apparently the concern that without their use, no adequate emission reductions for the climate can be achieved by energy-intensive industries [62] (which can ultimately be seen as an advantage for iCCS).

In addition to these societal risks, the studies also mentioned some personal risks that may be decisive with regard to the perception of iCCS. These include, in particular, the previously mentioned perceived health risks, which could lead to a strong rejection of iCCS technologies, especially on the part of the local population [13,30,71]. Personal risks may also be perceived in conjunction with the economic factors of iCCS. For example, the results of the analyzed studies illustrate that the factor of employment can be perceived as both a personal risk and a benefit [14,65] for people in a region in the context of iCCS. For example, one study expressed concerns that iCCS may impose costs that are then offset by, for example, lower employment levels in iCCS operations. On the other hand, the introduction of the technologies could create new areas of work and if steps were taken to retrain and employ industrial workers within the iCCS sector, this would be a benefit [71]. However, there has been an equal concern that there may be inflation of products through use with iCCS and in the long run this effect will contribute to industrial companies becoming uncompetitive in the global market and may lead to local plant closures [65].

#### 3.2.3. Preferences/Values

In the context of the studies analyzed, a variety of values and attitudes were explained that can have an influence on the acceptance of iCCS. These broadly include cultural identity, the closely related moral concepts of a society, environmental awareness, the perceived influence of iCCS on people's living conditions and attitudes toward technological developments and industry.

According to the study by [13], nationality is the strongest predictor of support for iCCS. Closely related to nationality is the cultural identity of a country. Thus, a study explained that compensation services to communities [74] must take into account the cultural as well as the social context [14,30,62]. Here, it is especially important that sacred values such as human safety are not mixed against a secular value, for example, by accommodating a hazardous facility in exchange for monetary compensation [74]. Certain normative ideas and moral values are also obviously advantageous for the development of a positive attitude towards iCCS [63,76]. Insofar as the use of iCCS can compensate for possible inequalities in society [65], for example, by allowing regions with a high proportion of energy-intensive industries to hold on to their economies to some extent or to operate them in a climate-friendly manner through iCCS, this represents an advantage for the perception of iCCS [64]. However, such perspectives do not go hand in hand with the moral notion that iCCS is interpreted as an intrusion into the subsurface "wilderness" or that BECCS is morally indefensible due to the still unclear availability of biomass, as stated in [71]. A view that, according to [71], occurs among those with strongly ecological values. According to [71], iCCS can only contribute to justice in society where a common understanding of cultural, natural and socio-economic systems prevails.

The influence of environmental awareness on the acceptance of iCCS is still evaluated very differently. Thus, ref. [13] clarify that a high environmental awareness can lead to a low acceptance of iCCS as the technology is seen as less important for coping with climate change than other technological options [63]. Whereas BECCS technologies seem to get a better rating in [71] compared to CO<sup>2</sup> capture from further industrial processes (here certain views of environmental awareness do not seem to be in conflict with the moral risks of BECCS mentioned above). Either way, BECCS is obviously viewed positively here because it is more likely to be associated with natural processes through the use of biomass [16]. However, if iCCS technologies are placed in the larger context of addressing climate change, where the technologies are embedded as part of an overall strategy to reduce CO2, their perception as an environmentally conscious technology may change if necessary [13,65]. Here, the urgency to address climate change postulated in recent years seems to have become a helpful vehicle for improving society's perception of iCCS technologies [63]. Another step towards valuing iCCS as an environmental technology focuses on the perception of CO<sup>2</sup> as a significant resource [64] rather than a waste product (see Section 3.2.1) or iCCS as a socially desired argument to support energy-intensive industries in the context of political decarbonization intentions [53].

It remains open whether, far from being environmentally conscious, people can develop a positive perception of iCCS out of a certain technological affinity. The authors of [30] present a study in which people with a positive attitude toward gas infrastructure development are more supportive of iCCS than people without this attitude. In addition to environmental awareness and technological affinity, the perceived impact of iCCS on people's concrete living conditions is also likely to be significant in assessing acceptance [68]. For example, results from a focus group [71] illustrate people's fears that a life based on the renewable energy technology system may be very regimented and "robotic" and that this development may negatively affect previously valued lifestyles. In light of these considerations, the use of iCCS technologies is evaluated in a different context; in which through them traditional ways of life can be maintained for longer, which is evaluated as quite positive [70]. The authors of [13] also found in their study that people with energyintensive lifestyles were more likely to prefer iCCS than others because they too could maintain their lifestyles while not being accused of promoting climate change.

The general attitude of the population toward the industry could also be an indicator for the future acceptance of iCCS. This is an aspect that will be discussed in more detail in the following section, as it is very closely linked to questions of the regional affiliation of the public.

#### 3.2.4. Regional Factors

In this section, we will focus on the factors that can exclusively determine the regional characteristics and conditions for the development of iCCS acceptance (independent of other factors such as trust, knowledge, and communication, which can also influence the regional perception of iCCS). These factors on regional specificity include the specific history of an area and the regional perception of iCCS technologies in the context of other developments, such as the economic activities and geological conditions of the region.

The results of the studies analyzed suggest that despite the processes of deindustrialization in advanced capitalist economies, deeply rooted cultural narratives of industrial modernity and manufacturing employment remain powerful markers of identity and social progress [64,71]. In regions with an industrial heritage, where the local public feels connected to industry, this identity is particularly high [74]. Regional populations appreciate it when industrial actors inform them and involve them in their activities and plans to give them a sense of belonging and identity [66,74]. It is becoming apparent that people in such regions are concerned that these industries remain fully intact and are becoming sustainable [13,62].

Ref. [14] contribute to this thesis, for example, with the study of Teesside (UK). Teesside is a conurbation with a strong industrial base that residents rely on. Ref. [74] also assume that people in such regions are more positive about iCCS development than people who are less rooted in their industrial heritage. For example [66], describes that the Norcem industry began producing cement as early as 1919 and quickly became a major player in the economic life of the region. Ref. [64] emphasize the aspect of habituation. If people are used to industrial activities, especially when industry has operated in the area for decades, this has a positive effect on trust towards local industry and politics. For example, residents in northern regions are also accustomed to transporting products that are considered more dangerous than CO2, such as ammonia.

Ref. [13] assume that areas where iCCS plants are likely to be built are typically those locations where (analogous) industry already exists. Subjective familiarity with such an industry could also serve to reduce the perceived risks associated with new infrastructure, leading to greater acceptance (or tolerance) of iCCS within regions. Fundamentally, according to [74], there is a need to understand local social realities, such as understanding what a particular place means to the local public, as well as how iCCS technology can impact this meaning at an early stage of the projects.

However, refs. [15,30,67] also emphasize that past economic activities, for example, when coal mines are present in the region or there have been incidents with health impacts for local residents, can have a lasting negative effect on the implementation of new projects. For example, the explosion of a gas pipeline in Belgium in 2004 increased public concern about the perceived reliability of CO<sup>2</sup> transport [30] (see also [15] regarding the CO pipeline in Section 3.2.2).

Another crucial factor for the regional acceptance of iCCS seems to be the specific perception of actors and issues related to a (possible) project. Ref. [74] suggests that this debate is also in the literature on the so-called social license to operate (SLO): "SLO refers to the informal permission granted to industry by the local community and wider society to develop a technology; in the context of CCS, SLO has been recognized as very preliminary and fragile". The following factors are summarized for achieving an SLO by [74] and are supplemented here by the results of other studies:

• Weighing the costs and benefits to the community, based on the particular characteristics of the project (see also [13]). Here, the ability of iCCS to protect jobs was identified as one of the key benefits. These benefits can be felt even more strongly for iCCS as it

both protects employment in existing industries and provides infrastructure that can attract new investment and employment opportunities [13,66,71];


In addition to the factors already mentioned, the studies identified further aspects that may have an influence on the regional acceptance of iCCS; these include the specific economic situation and the geological conditions of a region. These have already been discussed in more detail in Section 3.2.1 on the perceived benefits of iCCS and will not be repeated here.

#### 3.2.5. Trust

In almost all analyzed studies (n = 23), the topic "trust" was treated as a crucial acceptance factor for iCCS. Ref. [74] conclude that research indicates that trust in developers and other stakeholders is a critical factor influencing public response to a development such as iCCS as a whole, as well as at the community level. Within the studies analyzed, the trust factor is predominantly discussed in the context of regional processes and stakeholders on iCCS. Some stakeholder groups enjoy more trust among the population than others. These groups include in particular (environmental) non-governmental organizations (ENGOs) and local stakeholders, for example politicians and investors, who are considered to represent local and civic interests [15]. These groups of people are thus seen as having a certain degree of integrality. Whereas [62] notes that in the context of a focus groups in Wales (United Kingdom), a distrust of both a major steel producer and the government at all levels was mentioned based on a lack of integrity and competence. According to [14], perceptions of trust in key institutions depend on the track record of those institutions in managing past industrial processes.

Local authorities seem to have a special role to play here in developing a deeper commitment, as they can act as facilitators for the deployment of iCCS [65]. The importance of the position of the municipality towards CCS projects has been shown in previous studies. In Barendrecht in the Netherlands, the local government rejected a proposed CCS project because they feared negative impacts on public health and a decline in property values [64]. Accordingly, it is important that the community, including the people who live there, feel that the continued efforts of industry to build technology like iCCS is also directed toward solutions to environmental challenges [64]. This is where community familiarity with industry relevant to CCS implementation may also be important [64]. Moreover, ref. [13] argues that subjective familiarity with such an industry may serve to reduce the perceived risks associated with new infrastructure, leading to greater acceptance of iCCS within the intended communities.

At the same time, gaining public trust is an extremely lengthy and labor-intensive process that is highly dependent on experience in the interaction between laypersons and project stakeholders [30]. It is also important to avoid violating trust as much as possible, as it can be difficult to rebuild and can also cause negative spillover effects on perceptions of other technologies and projects [14]. Distrust can have an effect in different areas, on the one hand with regard to the competence of the responsible persons (competence-based distrust), especially when it comes to the implementation of a complex infrastructure project such as iCCS technology [62]. On the other hand, distrust can also relate to procedural fairness in

the participation process (integrity-based distrust [62]; compare also the comments on sociopolitical legitimacy in Section 3.2.4). According to [74], without a more comprehensive public involvement strategy, the question remains whether this is sufficient to build a sense of trust towards the developer.

#### 3.2.6. Knowledge/Awareness

As expected, none of the studies analyzed provide any information on what the state of public perception and knowledge of iCCS technologies is. However, the results of [13] show that public awareness of CCS (without concreteness to iCCS) remains low (here for Canada, the Netherlands, Norway, the UK and the US) and this result is also in line with previous research. However, in deciding whether to accept or reject CCS, the general level of knowledge and awareness plays an important role, as illustrated by the presentations from Tcvetkov's literature review on CCS [30]. Stakeholders interviewed by [15] in the ELEGANCY project rate public knowledge about CCS as rather low and perceive that iCCS technologies are not yet present in the current public discussion due to low market penetration. The results of [61] in the context of an experiment suggest that iCCS is viewed more positively by those who claim to have more knowledge about iCCS and that they are also likely to show a higher interest in the technology. Additionally, ref. [41] found that higher information levels can fundamentally change the evaluation of CO<sup>2</sup> capture options (for example air capture or from chemical plants).

The study [64] emphasizes that the local population in Porsgrunn (Norway) is not only used to industrial activities, but is also likely to have concrete experience with iCCS activities. There is a sense that the local population is positive about the proactive approach to managing CO<sup>2</sup> emissions, and this assumes that there is some level of knowledge about iCCS locally. Beyond this level of knowledge about iCCS, ref. [77] clarified that industries also have an interest in iCCS technologies becoming more widely known. For example, to market BECCS, public knowledge of low-carbon technologies is a possible positive aspect. The reasoning is that customer demands for negative emissions make investment decisions easier for industries because they can integrate iCCS technologies as part of their sustainability strategy. According to [65], however, even key stakeholders such as trade unions and environmental organizations lack evidence-based information on the iCCS capabilities of carbon-intensive industries. Ref. [73] also assumes that environmental organizations (related to Europe) lack the necessary resources to acquire knowledge about different iCCS technology options in detail. This lack of capacity also contributes to the apparent lack of official positions on issues such as iCCS until 2018 [73].

Beyond just awareness and knowledge of iCCS, the studies address the need for contextual knowledge. For example, ref. [72] suspects that there will be a more positive perception of iCCS as people become more aware of their individual climate impacts. Thus, some of the stakeholders interviewed in the study of [15] also see a general lack of societal acceptance regarding energy technologies and large-scale infrastructure, attributed in part to a lack of knowledge. Perception of global warming issues, understanding of the role of humans in this process, and developing an objective view of the prospects of low-carbon technologies, including CCS, depend on the education of respondents [26,30]. Therefore, implementation of an educational strategy for sustainable development should be considered, which starts at school and could be part of a national "green" policy. Ref. [71] clarified in their study that with the level of knowledge about iCCS and the integration of the technologies into a higher-level thematic context, the initially perceived assessment of iCCS can change once again. If iCCS is initially interpreted as a potential threat to natural systems, subsequent presentations and scenario discussions led to a gradual shift in how participants interpreted iCCS. Similarly, ref. [62] clarifies that participants in two focus groups on the Port Talbot steel mill development acquired contextual knowledge to evaluate iCCS. For example, they express concerns that if iCCS makes steel more expensive, the Welsh steel industry could lose out to foreign competitors who continue to produce emissions-intensive steel at the lowest price. If nothing else, these findings illustrate

that awareness of iCCS does not immediately predict public acceptance of a project [30]. Ref. [66] also note that regardless of the depth of their insight and knowledge, people will acquire subjective perceptions about iCCS. Ref. [30] sees consolidating government, industry and NGO efforts as one of the key challenges to improving public perceptions of CCS.

3.2.7. Communication/Participation

The discussion of CCS communication and participation in the articles analyzed is extensive and is therefore presented in the form of a table (Table 3). Ref. [68] suggests that the CCS community is generally aware of the range of factors that influence public engagement. Whether this range changes significantly for communication about iCCS cannot be adequately answered using the available results. Ref. [74] illustrates that effective public engagement will be key to successful iCCS implementation. With this comes the need to further explore how to most effectively engage with the local public.

**Table 3.** Overview of the acceptance factor "communication/participation" of iCCS (who/what/how).


Use of digital media

The chosen order of the factors does not represent a weighting.

In this context, it seems important to mention again the aspect of [74], which emphasizes a certain flexibility in dealing with iCCS projects, as specific concerns and needs may change over time in different regions. Here, regular adjustments of the implementation strategy of iCCS projects have to be taken into account.

#### 3.2.8. Socio-Demographic Factors

The analysis of the influence of socio-demographic factors on the acceptance of iCCS from the available studies does not reveal any meaningful trend. According to [67], for example, the acceptance of iCCS among women is about three times higher than among men (in selected European countries). Additionally, according to [13], men (as well as older people and people with high incomes) showed lower support for iCCS (but only after reading the message on CCS and possible lifestyle change). In contrast, ref. [30] presents findings in which men show more tolerant perceptions of CCS risks when the economic potential is present, while women are more concerned about safety. Additionally, as mentioned earlier in the context of a country's cultural identity (see Section 3.2.3), nationality represents the strongest predictor of support for iCCS [13].

All other results on the influence of the socio-demographic factor do not explicitly refer to iCCS technologies and therefore do not find any further explanation here.

#### 3.2.9. Perceived Differences between CCS and iCCS

In the following, the question is addressed whether significant differences between the acceptance of CCS from fossil-fired power generation plants and the acceptance of iCCS from industrial processes can be derived from the results of the analyzed studies. There are a number of initial results on this, but they target different technology pathways and are therefore hardly comparable. First, ref. [30] suggests that CCS technologies received general support from respondents in a survey, but when it comes to specific options for implementation, for example as part of gas and coal-fired power plants, initial public preferences may be negated. Additionally, according to [71], focus group participants articulate more positive visions for iCCS and BECCS than for coal CCS. They affirm support for growth through iCCS in manufacturing industries, as this is highly desired by society. Additionally, ref. [15] assume that iCCS will have higher social acceptance than CCS. Beyond this more economic aspect, ref. [68] represents the need to significantly broaden the iCCS discussion to include heavy industry and processes outside of power generation. This was seen as necessary to counter the traditional arguments of environmental groups that reject CCS because of its ability to re-generate electricity. In addition, initial studies compare the acceptance of iCCS with the acceptance of gas-fired power plants. For example, ref. [16] show in their experiment that BECCS plants receive higher approval than those using conventional gas. Interestingly, as perceptions of BECCS improve, so does the willingness of one's community to accept CO<sup>2</sup> storage. Ref. [17] also found that large-scale plants converting gas to hydrogen (H2) with CCS tend to be viewed negatively by most respondents. Basically, ref. [71] assumes that fossil CCS is considered unacceptable by the local population, while other CCS options, like iCCS, remain feasible.

#### 3.2.10. Evaluation of iCCS for Different Process Steps

iCCS technologies encompass many different technological concepts and potential target applications. The results presented below are intended to illustrate the acceptance of iCCS along the stages of different value chains and the underlying factors. It should be mentioned at the outset that the studies analyzed did not examine in detail the possible effect of the technical feasibility of different iCCS technologies on iCCS acceptance.

The following findings are available on the CO<sup>2</sup> source and the capture process step:

• BECCS: as briefly indicated before, BECCS is preferred to fossil-based CCS. According to [76], the technological approach has reached a stage of normalization in the debate, at least in the scientific discourse, after several years of intense criticism, and has become a self-evident aspect of climate change discourse. Especially for countries with a strongly biomass-based economy, such as Finland, BECCS seems to generate benefits [75]. With reference to [71], CCS was seen as a more intuitive and natural process when linked to managed forestry and the carbon cycle. Similarly, ref. [41] presents the use of biogas plants as a source of CO<sup>2</sup> as a promising option for industry and policy makers to achieve a socially acceptable form of carbon capture. Environmental

organizations such as Greenpeace and Biofuelwatch disagree here, according to [76], emphasizing problems with agricultural production and water scarcity in the context of BECCS. This aspect is also critically addressed in the Convention on Biological Diversity from 2019 [82]. This is because significant negative impacts on biodiversity and food security are expected as a result of the extensive land use changes caused by the consistent use of bioenergy, including BECCS. It remains to be seen what effect this position can have in terms of shaping public opinion. However, ref. [13] assume that BECCS is more supported than shale gas, underground coal gasification, and the application of CCS in heavy industry.


The following findings are available on the acceptance of the CO<sup>2</sup> transport process step:


The following findings are available on the acceptance of CO<sup>2</sup> use:


The following findings are available on the acceptability of CO<sup>2</sup> storage in conjunction with iCCS:


Finally, it should be summed up here that several studies consider the acceptance of iCCS along the different process steps and value creation stages to be possible. An important approach to developing iCCS acceptance, initially primarily from an economic perspective, is the pursuit of a cluster and network approach [14,41,62,64,67,78], which is already emerging as a trend in practice (see Section 3.2.1 for a more detailed discussion).

#### 3.2.11. Regulatory/Political Aspects

This literature review also noted circumstantial evidence suggesting that a lack of regulatory frameworks, political support, and missing or complex approval processes may influence iCCS adoption.

The findings highlight a fundamental need for strong regulation and policy on iCCS, both to leverage the skills and experience of the private sector and to maintain the common good and public interest [65]. For example, a UK opinion poll cited by [62] found that a majority (74%) of adults support policies to regulate heavy industry to ensure emissions reductions in the sector. Focus group participants from a region of Scotland that has historically been closely associated with energy-intensive industry (Port Talbot steelworks) assume that there will be stricter emissions legislation for these industries in the long

term, and therefore refer to iCCS as an "inevitable" option [62]. This would imply that expectations of stricter emissions legislation in the future from national and EU levels alone can convince people that iCCS is inevitable in the future. On the other hand, the participants of this study also valued the European Union as an important partner for the implementation of iCCS technologies [60], especially by providing the necessary funding. In this context, ref. [69] also mention the funding for the development of the necessary CO<sup>2</sup> transport infrastructure.

Ref. [30] go one step further and assume that an important factor for further iCCS development is international cooperation. On the one hand, so that individual countries can embed and position their iCCS policies internationally [14], and on the other hand, international cooperation would make it possible to combine national efforts, create favorable conditions for project proposals and adopt successful experiences of other countries. Thus, it would be necessary to create a political context that can strengthen public trust due to the importance of collaborative decision-making [30]. Local and regional networks alone would be insufficient to influence national policy [14,63]. In addition, ref. [65] describe that there would be limited public communication of an iCCS project proposal if political uncertainties prevail. For this, it is also important to have political long-term strategies that create reliability, for example, regarding BECCS technology and its integration into the European Union Emissions Trading Scheme (ETS-EU) [77]. This integration would be important for Finland, for example. Without BECCS, it would be challenging to meet emissions targets, but with BECCS, Finland could gain benefits by saving and trading emissions allowances [75]. The need for ETS-EU was frequently mentioned in the analyzed studies, but mostly by industrial actors and other experts [69,75,77].

#### **4. Discussion**

The present study is the first literature review to address the acceptance of iCCS. The objective of this study was fourfold. Firstly, it is examined to what extent the acceptance of iCCS is already being empirically investigated. Secondly, an analytical framework is proposed in order to systematically review the existing literature. Thirdly, based on the review, factors influencing the acceptance of iCCS are identified and discussed. Fourthly, results for the acceptance of iCCS are compared to CCS, highlighting some important differences between the two areas of application.

First, the results show that there is still only limited research on the acceptance of iCCS. Between 2012 and 2020, 25 scientific articles were published on the subject, with very different and incomparable methodological tools and research questions.

Secondly, during the evaluation process, it became apparent that the analytical framework transferred from CCS acceptance research, with its well-established dimensions (cf. Table 1), was sufficient to systematically gather the results from the articles. The research findings of the analyzed articles could be assigned to one or more dimensions, such as findings on local aspects (as suggested by Table A1 in the Appendix A, see column "Important statement related to iCCS"). Influencing variables that emerged in the analyzed articles and initially deviated from the established factors for CCS acceptance research (for example, the employment factor) could be assigned to the existing dimensions by the author during the evaluation. Accordingly, no further factors were inductively added to the analytical framework established in Section 2. As a result, many factors explaining the acceptance of CCS seem to be decisive for the acceptance of iCCS as well. However, it became apparent that the weighting and the expressions of acceptance factors to iCCS appears to vary compared to CCS, as shown in the following. Moreover, only tentative trends for the acceptance of iCCS can be derived from the studies analyzed. It remains unclear whether iCCS applications are more likely to be accepted or rejected by society in the future. Moreover, from a scientific point of view, a methodological concept for analyzing iCCS acceptance is still lacking, even though the factors considered here already provide a good starting point for operationalizing the research subject. Given the wide

range of technological options and the resulting societal implications, this task also appears to be non-trivial.

The discussion of objectives 3 (factors influencing iCCS) and 4 (differences of CCS and iCCS) of this content analysis are now discussed in conjunction.

More specifically, acceptance at the regional level, for example, appears to depend even more significantly on the perceived societal benefits that people associate with iCCS. The potential to maintain and increase local employment through the use of iCCS applications was frequently mentioned [13,64,66,71]. This represents a difference from the debate in perceptions of the societal benefits of CCS. In sum, it appears as if the population expects the safeguarding or even increasing of economic performance in their local environment with the use of iCCS. Previous research findings illustrate that societal benefits have either the same or slightly higher explanatory power for CCS acceptance than societal risks [31,35,47,83] (see Table 1). Whether this is also valid for the acceptance of iCCS remains to be investigated.

What is clear is that both factors will also be very significant in the context of iCCS. Subjectively perceived risk associated with CO<sup>2</sup> storage has been a crucial factor in explaining local and regional resistance in the context of CCS technologies [11,16,44]. It is different from factual risk in this regard as [53] illustrated with their approach to misconceptions. The fact is that CO<sup>2</sup> pipelines are state of the art and have been operating in the United States for example since the 1970s. Additionally, no significant research and development budgets are being spent on CO<sup>2</sup> transport and the associated potential risks worldwide. In contrast, geological storage of CO<sup>2</sup> has been the subject of intensive research and development work internationally for many years, even though CO<sup>2</sup> storage is already being successfully operated in many countries [84]. Here, the exploration methods for CO<sup>2</sup> storage, the procedures for storage monitoring, the competition with other storage utilization options, the impact on geothermal energy utilization, and the theoretically possible effects on drinking water supplies are often the subject of interest [85]. In sum, the question is not so much whether CO<sup>2</sup> storage is fundamentally possible, but under what conditions it is as safe as possible. Besides these science-based facts of technical and environmental aspects, the subjectively perceived risk factor will be important in the context of iCCS acceptance, as many of the studies analyzed have made clear [13–15,30,62,65,68,71,73–77]. It seems that in this context the aspect of fair distribution of risks and benefits has to be more in focus than in the context of the CCS debate. If, in the future, the benefits associated with the use of iCCS are perceived by the population primarily at the global level in the context of climate protection and the local population gains the impression that, in contrast, they are more likely to be confronted with the disadvantages of iCCS applications, this would probably be a barrier to the development of acceptance. In relation to the perception of an equitable distribution of risks and benefits, the explicit understanding of the benefits associated with iCCS for a region therefore seems to be of importance. This starting point of an unequal distribution of risks and benefits in the context of the future deployment of iCCS offers a possible field of action, both for research and for the implementation of practical iCCS projects. The previous research approaches of possible compensation benefits in the context of CCS will be examined here for their transferability and applicability.

Furthermore, the factor trust, which was evaluated in most studies as an important tipping point for or against the acceptance of iCCS (see [13–15,30,62,64,65,74]), should be further investigated. It became clear that in the development of local iCCS projects, trust in the stakeholders involved becomes especially important when it comes to large infrastructure measures related to CO<sup>2</sup> transport [13,16,41,62]. It seems that the process step of transport has become critical to the CCS debate, even though the negative sign in the assessment of CO<sup>2</sup> pipelines does not seem to have changed. With respect to transport infrastructure, more knowledge is still needed on the acceptance of iCCS. It is unclear whether, for example, the "joint" use of infrastructure or the use of existing infrastructure by industry clusters or hubs leads to an improvement in the acceptance of iCCS. Another research question could be whether the CO<sup>2</sup> source has an influence on the acceptance

of CO<sup>2</sup> transport, for example if the source is associated with an industry that is deeply rooted in the local society and contributes to its identity.

In this context, the role of framing or a possible narrative for iCCS (and also the greenhouse gas CO2) implementation should also be further explored. The studies have illustrated that framing iCCS as a climate change mitigation technology can lead to both positive and negative acceptance tendencies [13,16,53,63–65,71]. There does not seem to be a determination yet as to whether or not iCCS technologies are perceived by society as a climate change technology. This framing was hardly conceivable in the context of the CCS debate since there were sufficient technological alternatives for sustainable generation of electricity through the use of renewable energy technologies. Anyway, it is clear that iCCS operators would benefit from such "green" framing of iCCS applications, especially in marketing potential products along the value chain. This framing approach, based on a rather economically oriented marketing strategy, would certainly fall short. Ultimately, there is an obvious need for a more overarching narrative that takes into account both the aspect of sustainability and the reduction of CO2, as well as economic issues that not only affect individual technology paths, but in sum relate to the economic viability of an entire region. As a consequence, this would mean embedding iCCS in a discourse around sustainable structural change. After all, regions with energy-intensive economic sectors are particularly affected by the challenge of structural change.

The articles analyzed have also made it clear that the factors of social "values and attitudes" can be significant for the acceptance of iCCS [13,14,30,62,64,65,71,74,76]. In this context, further research is particularly needed on the question of whether a certain environmental awareness has a positive or negative influence on the perception of iCCS. Compared to the CCS context, the clarification of this research question seems to be much more complex due to the many different possible applications of iCCS. In the context of CCS acceptance, existing studies indicate that people with high environmental awareness tend to evaluate the technology negatively [49,86]. Some authors mentioned that, triggered by the Paris Agreement, the absolute urgency of the transformation to a sustainable economy and way of life has now arrived in the perception of society. In light of this urgency, the evaluation of iCCS could also be developed in a more positive direction [63,65]. Again, there are only assumptions and no evidence-based findings yet. Interestingly, in one study, this urgency emerged as a driver of iCCS acceptance. This happens when this urgency is interpreted by society as a threat to their current lifestyles and cherished habits for everyday life, and iCCS is perceived as an option to hold on to these habits without regret [71]. This approach could also be a starting point for new research questions on the acceptance of iCCS. In addition to this urgency, another aspect could influence the perception of iCCS in the future. For example, the Global Assessment Report on Biodiversity and Ecosystem Services (IPBES) [87] does not exclude CCS (and thus iCCS) as a measure to mitigate negative impacts on biodiversity (see Glossary). Consequently, iCCS could be considered not only an option to reduce global CO<sup>2</sup> emissions, but also a generally accepted measure to avoid or limit potential negative impacts on biodiversity. If such a perception is perpetuated among individuals with a high level of environmental awareness, this aspect could be interpreted as an advantage for the use of iCCS and possibly have a positive impact on social acceptance. Whether this assumption is well-founded needs to be explored in future studies on the acceptance of iCCS.

#### **5. Conclusions**

The IEA [88] estimates that iCCS in the cement, iron and steel, and chemicals sectors will need to deliver around 28GtCO<sup>2</sup> of emission reductions between now and 2060 to meet the climate target of the Paris Agreement. To achieve these reduction goals globally, strategies for robust and timely market introduction of iCCS technologies need to be developed. For such a market introduction of iCCS, social acceptance is of particular importance in addition to technical-economic and environmental indicators, as the example of CCS has illustrated.

In the studies analyzed, a large number of indications for the design of a communication strategy were derived, largely on the basis of the findings from CCS acceptance research as well as on the basis of all the research on energy transformation (see Table 3). In view of the abovementioned abundance of requirements for such an iCCS communication, the question arises as to which institution is capable of organizing such a permanent and trust-based process and in which larger thematic context this communication can be embedded? This appears to be a difficult question to answer, especially against the background of often missing political strategies and the related regulatory frameworks on the national level. The present literature analysis shows on the one hand which starting points for the market introduction of iCCS exist so far from social science research for political and economic actors and on the other hand which research efforts are still required.

**Funding:** This research received no external funding.

**Data Availability Statement:** The articles used in the literature analysis can be accessed via the respective publishers (for more details, see the bibliography).

**Conflicts of Interest:** The author declares no conflict of interest.

#### **Appendix A**

**Table A1.** Overview of the analyzed articles.






ICQ = Information-Choice Questionnaire; <sup>2</sup> N/A = not available; <sup>3</sup> PESTEL = Political, economic, social, technological, environmental, legal.

#### **References**

1

