2.2.4. Data Representativeness and Quality

The quality of collected data was assessed by means of the Ecoinvent data quality system [43]. Five indicators (i.e., reliability, completeness, temporal correlation, geographical correlation, further technological correlation) were assessed using a score from the best quality (score 1 corresponding to a verified measured data) to the worst (score 5 corresponding to not qualified/or estimate data). The complete description of indicators and scores is reported in Table 2.



<sup>1</sup> Each cell in the matrix indicates a quality characteristic of inventory data. After the analyst has selected, for each item of the inventory, an appropriate cell during the Monte Carlo procedure, the software keeps track of such choice indicating the position (1 to 5) of the selected quality characteristics in each R, C, T, G, F line of the matrix itself.

After a score was assigned to each data indicator for all material and energy inputs included in the inventory, the Ecoinvent data quality system calculates a corresponding numerical value of uncertainty, assigning a specific geometric standard deviation to a log-normal distribution (see the Supplementary Materials for standard deviation values). The propagation of uncertainty throughout the model was then calculated by means of the Monte Carlo analysis (i.e., 10,000 runs), obtaining a final standard deviation on the results in each impact category.

#### 2.2.5. Important Assumptions

Transport of assembled machinery to the geothermal plant site was excluded, because of the limited distance between the plant and the production site (i.e., 150 km). However, the transport of the semi-products and raw material was included using the background processes in the Ecoinvent database.

For small steel parts, an aggregated mass value was provided by EGP. This quantity is supposed to cover all the steel used for general parts in the commissioning phase. Thus, it was equally divided between the six components: AMIS, gas intercooler, gas compressor, condenser, evaporative tower and turbine.

During the maintenance of the power plant, 10% of the steel content of the steam turbine rotor was assumed to be substituted with new steel every four years.

Drilling wastes spent mineral oil and sorbent were considered to be sent to landfill. According to the information supplied, no additional treatment processes were considered.

Data on direct emissions from the power plant's stack was taken from Ferrara et al. [44]. These emissions were modelled as output flows "emission to air, low population density".

#### 2.2.6. Allocation Procedure

The Bagnore 3 and 4 power plant is a multifunctional system, since it produces both electricity and thermal energy. In this study, an exergy-based allocation procedure was chosen to deal with such multifunctionality as a proper allocation method, according to the ILCD Handbook [41]. The exergy allocation method accounts for the quality (i.e., exergy content, ability to do work) of the two energy products (i.e., electricity and heat) generated by the power plant. Thus, 95% of total impacts were allocated to the electricity produced. The complete procedure to calculate allocation coefficients for electricity and heat is reported in detail in the Supplementary Materials, in the sheet "Allocation".

#### 2.2.7. Life Cycle Impact Assessment Method

The ILCD 2011 Midpoint+ method v1.0.9 was adopted for translating the emissions and resources use into environmental impacts, which were quantified during the inventory phase. The impact categories acidification potential (AC); climate change (CC); freshwater ecotoxicity (EC); freshwater eutrophication (FEP); human toxicity, cancer effects (HTc); human toxicity, non-cancer effects (HTnc); ionizing radiation human Health effect (IRHH); land use (LU); marine eutrophication (MEP); mineral, fossil and ren resource depletion (MFRD); ozone depletion (ODP); particulate matter (PM); photochemical ozone formation (POF); terrestrial eutrophication (TE); water resource depletion (WRD) were included in the analysis. The ionizing radiation E (interim) impact category was excluded, due to its incomplete development [45]. The normalization step was performed by applying the reference values of the "EU27 2010, equal weighting" set. According to the latest development in European guidelines of the ILCD method [45], the discussion related to the toxicity categories was excluded from the analysis. All the calculations and modelling were performed using the open-source software OpenLCA version 1.10 and LCIA package v2.0.4 [46].

#### **3. Life Cycle Inventory Analysis**

One main objective of this research was to present the most detailed and accurate life cycle inventory based on primary data for a state-of-the-art flash geothermal power plant. The collection of information was performed with the intent of obtaining the highest level of detail in terms of LCA requirements [41]. The inventory of materials and energy input and output flows was collected for each of the separated components, and based only on primary data. To our knowledge, the data inventory built in the present work represents the first of a kind LCI available in the state-of-the-art literature for geothermal power plants based on flash technology. The resulting inventory is presented in its extended version in the Supplementary Materials (sheet "Inventory").

Currently, the most referred LCI available in the literature concerning flash technology is the one published by Karlsdottir et al. [35]. As much as this inventory is quite comprehensive and detailed, and it has often been employed for geothermal system modelling in LCA studies ([24] and references therein), it fails by not providing primary data and accounting for all the life cycle stages of the energy generation system. The present work aims to provide an improved inventory for the flash technology, which could potentially be used in conjunction with the work by Karlsdottir et al. [20,35] for geothermal system modelling in future LCA studies of geothermal power plants.

Table 3 reports the main differences between the present work (right side) and the work by Karlsdottir et al. [35] (left side). Regarding data accuracy, the inventory presented in this work is entirely based on primary data coming from the EGP Company that has executed the activities. Only a

few assumptions were made based on expert knowledge, for instance, concerning power building material requirements. In this case, the primary data used for modelling of Bagnore 4 was also used also for Bagnore 3, as suggested by the power plant operator EGP, employing a scaling factor. Even though this can be considered as an estimation, it is still based on primary data and, more importantly, on the expert judgment of the operator. As a result, when considering data quality, a large part of the indicators scored between 1 and 3 (see Table 2).

**Table 3.** Main differences between the currently available life cycle inventories for flash technology.


On the other hand, Karlsdottir et al. [35] includes a higher component's specificity, for example, steel grades are provided, as are mass weight for smaller equipment parts. However, these data are mainly based on secondary data and authors' assumptions. The data coverage featured in this paper is higher, compared to the one in Karlsdottir et al. [35]. Specifically, the present work considers all the regular maintenance activities, for example, lubricating oil substitution and regular maintenance operation of machinery, EoL treatments of wastes and wells closure operations, previously never considered.

In this paper, the same approach used in Parisi et al. [37] was adopted. Such an approach relies on a statistical analysis of all the compounds emitted during power generation from geothermal exploitation. The only difference compared to the work of Parisi et al. [37] is represented by the emission values that have been updated with the most recent ones provided by the regional environmental agency [47].

Table 4 reports the main energy and material inputs for the commissioning phase related to the functional unit. Diesel consumption is primarily associated with the wells drilling process with a specific consumption of about 12 GJ/m. Concerning material use, Portland cement and steel represent the most used materials, accounting for about 70% of the total weight of equipment used in this stage. Portland cement is employed in the casing of wells and power station buildings, whereas steel is partitioned among casing, pipelines and machinery. Depending on the application, different steel grades can be used.

The material input for maintenance activities are reported in Table 5. The maintenance stage represents the planned activities required to keep the power plant in operation. Extraordinary maintenance activities are hence omitted. The maintenance activities that result in the highest material consumption are those related to the substitution of the spent Hg absorber (Selenium), the lubricating oil replacement, as well as the steel and polyvinyl chloride (PVC) replacement for power plant machinery. In this case, a substitution of 10% of the total weight of the steam turbine rotor every four years was considered.


**Table 4.** Main material and energy inputs employed in the commissioning phase. The cut-off is set at 2% of the total mass, to reduce the number of inputs reported. Complete information can be found in the Supplementary Materials.

**Table 5.** Main material and energy inputs employed in the maintenance phase. The cut-off is set 2% of the total mass, to reduce the number of inputs reported. Complete information can be found in the Supplementary Materials.


The operational stage considers the atmospheric emissions, due to geothermal fluid exploitation and the material input needed by the NH3 abatement system. As shown in Table 6, the emission of CO2 and methane (CH4) dominates the environmental emission profile of the Bagnore power plant system. In contrast, the H2SO4 is by far the most used material during the operational phase.

**Table 6.** Main material input and direct atmospheric emissions from the operational phase. The cut-off is set 2% of the total mass to reduce the number of inputs to be reported. Complete information can be found in the Supplementary Materials.


Table 7 provides information on energy and materials inputs for the decommissioning phase. The assumption is that all the drilled wells will undergo a closure process when the plant runs out its lifetime. This approach was adopted more to test the influence of the EoL processes of wells than to represent a real option. The Bagnore power plant system is managed in a sustainable way, ensuring a constant productivity without depletion of the resource. However, since a lifetime must be set in LCA, this work has considered the unlikely option that the wells will be closed after the given lifespan to account for the EoL process.


**Table 7.** Main material and energy input employed in the well closure phase. The cut-off is set 2% of the total mass, to reduce the number of inputs reported. Complete information can be found in the Supplementary Materials.

#### **4. Results**

*4.1. Life Cycle Assessment of the Bagnore Power Plant System*

Figure 2 reports the percentage of contribution of commissioning, operational, maintenance, decommissioning and EoL phases of the Bagnore power plant system, to the total impacts for all the categories included in the ILCD method. The potential impacts on the 15 categories that were considered are essentially determined by the commissioning and operational phases, which contribute for more than 90% on the total impacts in each category. In more detail, the operational phase contributes for 80%–90% of the overall potential impacts on the AC, CC, HTnc, MEP, PM, POF and TE categories, and about 70% to WRD. These impacts are mainly linked with direct emissions to the air of NH3, CO2 and CH4. The emission of NH3 determines the impact on AC (i.e., 96% of the total impact), MEP (84% of the total impact), TE (99% of the total impact) and PM (86% of the total impact). In contrast, the impact on the CC category is shared between the CO2 (i.e., 57% of the total impact) and CH4 (i.e., 42% of the total impact) emissions. The 75% of the total impact on POF is determined by CH4.

**Figure 2.** Percentage of contribution of commissioning, operation, maintenance, decommissioning and EoL phases to the total impact in all the assessed impact categories.

The commissioning phase is responsible for more than 80% of the total potential impacts on EC, FEP and MFRD. The copper requirement during the building construction process is the main contributor to such impacts. The commissioning phase contributes about 60% to 70% to the IRHH, LU and ODP categories, for which the deep well construction process shows the highest contribution.

Subsequently, the decommissioning and EoL phases show a negligible contribution to all the considered impact categories.

The characterized results of each impact category were divided by a selected reference value, in order to better understand the magnitude of the results of impact category indicators, and to bring all the results on the same normalized scale (see the Supplementary Materials for normalization values). After the normalization step, the CC, TE and AC categories, in this order, had the highest impacts among all the selected impact categories (Figure 3). The impacts from geothermal electricity production were compared with the impacts from the average Italian energy mix [48], to give a reference system and interpret the magnitude of the geothermal eco-profile. The Ecoinvent version 3.5 employed for the analysis is based on the Italian electric energy mix by 2014. The share consisted of 60% arising from fossils (coal, gas, oil) and import (mostly nuclear). RES represents 40% of the total, with 18% generated by hydro, 7% by photovoltaics, 5% wind, 6% biofuel, 2% waste and 2% geothermal.

**Figure 3.** Normalized results for the production of 1 kWh of electric energy from the Bagnore power plant system (blue) and from the average Italian electricity mix (light grey). Climate change (CC), terrestrial eutrophication (TE), acidification potential (AC) and particulate matter (PM) have been identified as the categories with the highest impact.

As shown in Figure 3, all the impacts caused by the average Italian electricity mix are higher than those of geothermal energy production, with the exception of climate change, due to the emissions of CO*<sup>2</sup>* and CH*<sup>4</sup>* that are intrinsic to the geothermal resource exploitation activities.

The impacts on the CC, TE and AC categories for geothermal electricity production are determined almost exclusively by emissions to the air during the operational phase (i.e., NH3, CO2 and CH4). As shown in Figure 3 and Table 8, all the impacts caused by the average operational phase are mainly related to the geothermal fluid composition, and can thus be considered site-dependent.

On the contrary, the commissioning phase is common to all flash technologies, and Figure 4 shows the contribution of processes within this phase. The processes considered in the commissioning phase are clustered in drilling, drilling waste (disposal), equipment and pipelines. The drilling process includes, in addition to the drilling activities themselves, the construction of the well pads. In contrast, equipment includes all the materials and energy needed to realize the components present in the power plants and the power plants building itself. The pipelines construction process is separated from the others, because they are structures connecting wells and power plants.


**Table 8.** Contribution of LC phases to the most impacting categories: AC, CC, PM and terrestrial eutrophication (TE). Impacts are reported as person equivalent (PE) per functional unit (FU). Complete information can be found in the Supplementary Materials.

**Figure 4.** Percentage of contribution from drilling, drilling waste disposal, equipment and pipelines to the commissioning phase.

The hotspot analysis results show that the potential impacts of the commissioning phase are fairly divided among equipment and drilling processes. Building construction and the production of metals (i.e., copper) determine the impact of the equipment. Emissions from the combustion of diesel used to drive the drilling rig are the most responsible for the impact during drilling. Pipelines generally give a contribution of around 10% of the total impacts in all categories, except for CC and WRD. Drilling waste disposal has a negligible impact.

#### *4.2. Uncertainty Analysis of Results*

Figure 5 reports the uncertainty values (MIN, MAX and standard deviation) related to the average potential impact for the categories CC, TE, AC and PM, which were previously identified as having the highest impact. The uncertainty associated with the results was calculated following the procedure described in Section 2.2.4.

The calculated uncertainty of results for the identified categories is low and ranging between 2–3%. The impact of these categories is exclusively determined by airborne emissions (primary data) during the operational phase. The good quality of data for airborne emissions (low score in all indicators, see Table 2) results in a low uncertainty of the final LCA results. In those cases where the impact is determined by other stages, with different levels of quality of primary data, the final uncertainty is generally higher and, in some cases, up to a standard deviation around the mean value of 58%.

**Figure 5.** Characterized impact results per kWh of electricity produced for the categories CC, TE, AC and PM. Bars represent the standard deviation around the average impact values, whereas red dots refer to MIN and MAX values.

In Table 9, the uncertainty related to the impacts for all categories is reported together with the overall score for each data quality indicator used to calculate the impacts. Generally, a low overall data quality (high scores in Table 2) corresponds to a relatively high standard deviation (>10%). The uncertainty of results is not exclusively related to the inventory data, but also to the secondary (background) data and their relative uncertainty as specified in the Ecoinvent database.

**Table 9.** Uncertainty analysis for each impact category results and data quality indicator score. (R) Reliability; (C) Completeness; (T) Temporal correlation; (G) Geographical correlation; (F) Further technological correlation.


<sup>1</sup> numbers in columns R, C, T, G, F refer to specific scores within the EcoInvent uncertainty matrix (Table 2).2 molc unit indicates a mole of charge (molc) per unit of mass emitted.

#### **5. Discussion**

The LCA results show that direct emissions to the atmosphere released during the commissioning and operational phases are the dominant impact for the Bagnore system. For the commissioning phase, as the emissions of CO2 associated with the combustion of diesel used to drive the drilling rig are the

principal factors responsible for the environmental impact, the eco-profile would certainly improve in the future by changing the drilling technology. Unfortunately, so far, the initiatives promoted by operators to employ an electric rig, directly powered by the medium-voltage network, aiming for a simplification of the process and a reduction of impact and costs, have been unsuccessful. The main difficulties arose with the medium-voltage network connections, and for authorization procedures, which look quite complex due to safety requirements.

However, the applicability of such a system looks only suitable for the consolidated stations with several wells.

The potential environmental impacts generated during the operational phase are mainly linked with airborne emissions. The comparison with the Italian energy mix makes it possible to highlight the differences in the environmental performances, which are in favor of geothermal energy exploitation for all the environmental impact categories, with the exception of climate change. This outcome is due to the significant contribution given to the average Italian electricity mix from RES like hydro, photovoltaics and wind energy, whose CO2 emission contributions in the atmosphere during the operational phase are negligible. This confirms previous evidence that geothermal energy, although renewable, is not the cleanest one, even if it performs better than any other fossil source. This finding gives a benchmark to interpret the magnitude of the power plant eco-profile. As the emissions of NH3, CH4 and CO2 during the operational phase are mainly related to the geothermal fluid composition, they can be considered site-dependent, therefore particular care should be exercised in deciding the localization of plants in the project phase. In this context, it should be mentioned that, to date, in the analysis of greenhouse gases emissions, the Intergovernmental Panel on Climate Change [49] considers the release of greenhouse gases of geothermal origin quantitatively negligible, despite the fact that this has been demonstrated not always to be true [7]. Notwithstanding the evidence that flash geothermal electricity production is contributing to CO2 emissions more than the Italian electricity mix, some intrinsic benefits connected with geothermal development should be considered. This is particularly important in the frame of a policy sensitive to environmental and social issues: (i) geothermal energy is a renewable local based energy source and not imported; (ii) a secondary, but not negligible advantage can be found in the use of thermal fluids for civil or light industry purposes in the neighboring area; (iii) electricity generated by geothermal contributes to the basic load and it is independent on atmospheric conditions. However, regarding this latter issue, we should be aware that in the future, due to discontinuity of solar and wind electricity supply, flexible power systems will be even more valuable.

The main achievement of the assessment method implemented in this work relies mainly on two aspects. Firstly, the investigated system has been selected from the latest in technological excellence in the field of flash geothermal generation in Italy. Secondly, the EGP operator granted the availability of primary data to build the LCI, as reported in the Supplementary Materials. This is noteworthy compared to the state-of-the-art LCA literature on geothermal systems, which very often uses secondary or tertiary data.

The representativeness and quality of the inventory data, presented in Section 3, should always be assessed to ensure robustness of LCA results. Significant elements of improvement in this work are represented by the level of detail for machinery and components, data quality and coverage, as well as the inclusion of the EoL as shown in Table 3.

The exergy-based allocation method chosen to address the multifunctionality represents another feature of this work: although not fully new, most LCA studies allocate according to mass, energy content or monetary value. Exergy reflects the difference in terms of energy quality among energy outputs and represents the most suitable method, from a thermodynamic point of view, for discerning the benefits of combined heat and power systems.

The uncertainty evaluation on LCA results performed with the Monte Carlo analysis shows a non-negligible dependence on the background Ecoinvent database and the LCIA method assumptions, not on foreground data. This confirms the reliability of the LCA system modelling adopted in this

work. The scientific approach employed offers a detailed insight of the research findings in agreement with the ISO 14040 and ILCD requirements. From a policy point of view, the transparency of the assessment method could support effective decision-making.

As mentioned in the Introduction, the EU has committed itself to a clean energy transition, which will contribute to fulfilling the goals of the Paris Agreement on climate change [1,5]. According to the Italian NECP targets [6], the electric generation power will be affected by an important transformation, due to the goal of the phasing out of coal generation plants by 2025, and the necessary promotion of a large contribution from RES to replace them. The maximum contribution to the growth of RES will arise particularly from the electricity sector, which, by 2030, will reach 187 TWh of generation from RES, equal to 16 Mtep. The strong penetration of technologies for renewable electrical energy production, mainly photovoltaics and wind, will make it possible to cover 55.0% of the final gross electric consumptions with RES, compared to the 40% contributed in 2014. The photovoltaics and wind capacity should triple and double, respectively, by 2030. With regards to other RES in the NECP, a limited growth of additional geothermal power from 813 to 950 MWe is foreseen, which would represent the only maintenance of the actual 2% of the Italian electric mix. This target is considered for conventional geothermal technology, with reduced direct emission limits. It arises from the awareness that, even if geothermal energy is quite suitable for replacing fossils in electricity production, the limits due to environmental impacts still hold. The possibility of providing incentives for other technologies like that with zero emissions in plants, with a total reinjection of fluids, is under evaluation. At the moment, these technologies, like geothermal at reduced environmental impact, are considered to be as innovative in the national context as offshore wind, concentrated solar power and ocean energy.

#### **6. Conclusions**

In this paper, a cradle to grave LCA of the Italian flash technology Bagnore power plants system has been performed, based on a comprehensive and accurate life cycle inventory of primary data supplied by the plant manufacturer and operator EGP. From the LCA results, it can be inferred that the potential environmental impacts are determined for more than 95% by the operational (direct emissions to air of NH3, CH4, CO2) and commissioning (CO2 emissions due to diesel combustion during drilling) phases. Maintenance, decommissioning and EoL phases show a negligible contribution to all the considered impact categories. Globally, out of the sixteen impact categories selected, climate change, ccidification, terrestrial eutrophication and particulate matter were the most affected. These outcomes imply that LCA results of electricity generation from flash technology employing a mid to high dissolved gas content fluid are primarily determined by emissions to air. Direct emissions into the atmosphere are the responsible for most of the environmental impact in the operational phase (84%). The comparison made with the life-cycle environmental impacts caused by the production process of the average Italian electricity mix showed that the balance is almost always in favor of the geothermal energy production, with the only exception being the climate change category. A further finding of this work is that, in the commissioning phase, the impact is equally divided between well drilling and equipment. It is notable that the copper requirement during the building construction process is the main contributor to impact in the commissioning phase. Accordingly, future research might explore the possibility of replacing metals and particularly copper in building the plant.

It should be noticed that the data referring to the commissioning, maintenance and EoL stages presented in this study might be used by the scientific community to evaluate the potential environmental impact of geothermal systems. On the other hand, site-specific information, such as direct environmental emissions measured during the operational phase, is exclusively valid in this specific geothermal field.

This work offers the most complete life cycle inventory for a state-of-the-art flash system conversion technology accounting for the whole life cycle of the geothermal power plant. The robustness of the results obtained here, as demonstrated by the uncertainty analysis, emphasises the need for high quality primary data for performing reliable and consistent LCA studies. This is particularly true in the geothermal sector, where the lack of primary data and precise information about the conversion technology and the geo-specificity of the reservoir for long periods prevented the possibility to get reliable results, affecting the quality of the LCA studies. In this context, the availability of primary data and open access to technical repositories become essential to reach high standards in the LCA literature concerning geothermal systems. We believe that the accurate approach presented in this paper will aid promoting the implementation of environmental assessment studies, which are essential to undertake impact minimization actions on currently operating power plants, and to improve the eco-design perspective of future installations.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/1996-1073/13/11/2839/s1, Excel file S1: energies-819511-supplementary.xls.

**Author Contributions:** Conceptualization, R.B. and M.L.P.; methodology, R.B. and M.L.P.; validation, R.B. and M.L.P.; formal analysis, L.T., N.F. and M.L.P.; investigation, L.T., N.F. and M.L.P.; data curation, L.T., N.F. and M.L.P.; writing-original draft preparation, L.T., N.F., R.B. and M.L.P.; writing-review and editing, L.T., N.F., R.B. and M.L.P.; visualization, L.T. and N.F.; supervision, R.B. and M.L.P.; project administration, R.B. and M.L.P.; funding acquisition, M.L.P. All authors have read and agreed to the published version of the manuscript.

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

**Acknowledgments:** MIUR Grant—Department of Excellence 2018–2022. L.T. and M.L.P. acknowledge the European Union's Horizon 2020 Framework Program for funding Research and Innovation under Grant agreement no. 818242 (GEOENVI) for funding. Authors would like to thank Romina Taccone, Marco Paci and Massimo Luchini (EGP) for primary data supplying. EGP and Roberto Bonciani are gratefully acknowledged for fruitful discussion. NF thanks COSVIG (Regione Toscana) for contributing to his grant. Careful reading and revising of the manuscript by Emeritus Michael Rodgers, Bowling Green State University, is gratefully acknowledged.

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