Application of the Typology Approach for Energy Renovation Planning of Public Buildings’ Stocks at the Local Level: A Case Study in Greece

Abstract

According to the latest energy efficiency European directive (EED 2023/1791/EU), the expected energy renovation rate of at least 3% of the buildings’ floor area each year towards nearly zero-energy buildings (nZEBs) is extended to include public buildings not only of the central government (as per the first EED 2012/27/EU) but also of regional and local authorities. This poses a great challenge, especially for Municipalities that often manage large building stocks with high energy demands. In response to this challenge, this paper presents the application of the so-called “typology approach” for conducting public buildings’ energy renovation plans at the local level. A computational survey is initially introduced to decide the optimal set of building-stock clustering criteria among all possible combinations, involving the minimization of the RMSE index regarding the primary energy consumption of each building. For a representative building from each identified typology, the key performance indicators (KPIs) are computed for alternative energy-upgrading scenarios. Exploiting the IMPULSE Interreg-MED project tools, the KPIs from each representative building are at first extrapolated to all buildings of the examined stock and, finally, a gradual energy renovation plan is automatically produced based on user-defined decision parameters including the required annual renovation rate. The methodology is applied for the case of the Municipality of Hersonissos in Greece. For the specific 44-buildings’ stock it was found that the optimal clustering set included four criteria, building use, construction year, heating, and a cooling system, leading to 15 building typologies. Finally, assuming a 7% renovation rate per year, a 12-year gradual renovation (nZEB transformation) plan is obtained foreseeing an 85% CO₂ emissions’ reduction.

1. Introduction

The European Commission in its so-called EU Green Deal (EGD) [1] raised the bar regarding the energy transition to the ambitious goal of reducing greenhouse gas emissions by at least 55% in the year 2030 compared to the 1990 levels. Considering the high energy share of the building sector, the EU energy policies pay high attention to the energy-efficiency management and planning of building stocks as a key component contributing to the aforementioned target. For more than a decade, with the energy efficiency directive EED 2012/27/EU [2], it has been acknowledged that the public sector shall constitute an exemplary role in the promotion of energy efficiency projects. To tackle the so-far low buildings’ energy renovation rates (less than 1% of the national building stock is renovated each year [3]), several policy measures have been introduced, including, among others, the enhancement of the public buildings’ exemplary role as per the latest 2023/1791/EU directive on energy efficiency [4]. The latter extends the measure of at least 3% of the floor area of public buildings being renovated each year beyond the ones of the central government, as initially envisaged in the directive 2012/27/EU, including public buildings of regional and local authorities. This means that regions and municipalities are expected to conduct plans and policy measures in adaptation to this new engagement and accelerate project implementation of energy-efficiency improvement in their public building stocks.

The above requirement poses great challenges in planning energy efficiency projects for the public building stock especially at the local level, i.e., for buildings owned or managed by regions and municipalities. Indeed, local authorities often manage a large stock of buildings of diverse uses, e.g., offices, schools, cultural centers, libraries, social houses, and high energy demands. At the same time, budget restrictions hinder the realization of building renovation projects at a massive scale. Such conditions necessitate effective decision-making strategies regarding projects’ prioritization and gradual renovation planning [5,6]. In parallel, to balance all technical, budgetary, social, and regulatory constraints and trade-offs, informed decision making, involving the quantification of key performance indicators (KPIs) for the alternative renovation scenarios, remains of paramount importance [7]. In the context of building-stock energy planning, this means that for each building and potential renovation scenario, the set of KPIs, such as the primary and final energy consumption and saving, the investment cost, the payback period, etc., should be available or estimated, thus providing the possibility to rank the buildings and projects on the basis of user-defined terms, e.g., the highest energy saving under least cost.

Following the bottom-up approach to produce the necessary KPI database, the archetype-based approach (or typology-based) is preferred over the building-by-building one especially for large building stocks [8,9]. Although the building-by-building approach provides more reliable building-stock modelling with a high granularity level which is the key to identifying energy-saving and renovation potential at high resolution, for large building stocks it is still very demanding in terms of computational resources and desk-research work due to the need for building energy simulation for each building individually as well as for big data acquisition regarding the detailed technical features for each building [10]. On the other hand, provided that a valid classification system is employed in stock processing, the typology-based approach is considered more efficient in engineering practice. A classification system initially provides building-stock clustering into representative typologies. The necessary KPI database is then produced simply by scaling-up energy analysis results from a strategically selected (or statistically formulated) representative building of each typology to all buildings of the typology; hence, the requirement for detailed technical data collection and energy calculations is significantly reduced as it refers only to the selected representative buildings [10,11]. Pristerà et al. [9] also argue that the typology-based approach is more efficient as it simply relies on the scale-up of KPIs from the representative (archetype) buildings to the whole building stock based on the buildings’ number or floor area. This means that the KPIs (normalized per sq.m.) of a representative building of each typology/cluster can be extrapolated to all buildings of the typology/cluster, thus providing the KPI database for the whole building stock without the need to analyze each building of the stock in depth one-by-one as in the building-by-building approach [9]. In conclusion, the typology approach ensures less computational effort and overcomes significant technical and administrative barriers compared to the building-by-building one, such as the requirement for collecting a high amount of detailed information about the whole building stock [12,13]. An additional crucial component in building-stock energy planning is the prioritization of projects and the development of the optimal roadmap of renovation projects (gradual renovation of the building stock) taking into account the energy and financial performance indicators as well as the recommended annual renovation rate, particularly for public bodies’ buildings [14]. This is interpreted as the determination of specific projects to be implemented annually [8].

A practical typology-based methodology for conducting gradual-renovation projects for public building stocks, particularly at the local administration level, was developed in the framework of the IMPULSE Interreg-MED project [15]. The project introduced easy-to-use Excel-based tools that automatically extrapolate the KPI database for the whole building stock, for the existing situation, and for various renovation scenarios and automatically produce a gradual renovation roadmap based on user-defined terms including the annual renovation rate. The tools are successfully applied in ref. [8] for a case study regarding the development of a gradual renovation roadmap for a building stock of 10 public buildings of the Municipality of Aigialeia in Greece. Furthermore, the IMPULSE tools have been included in training sessions and have been recommended for further exploitation to conduct local energy efficiency planning [16]. An important prerequisite in the valid application of such tools remains the reliable identification of building-stock typologies/clusters. Following the bottom-up typology-based approach and focusing on non-residential buildings, Gangolells et al. [13] argued that the clustering technique should identify the most relevant and influential buildings’ characteristics beyond the energy intensity such as the geometry, occupancy, and energy systems. They applied the k-means clustering method to process a sample of more than 6000 objectives of the Spanish office buildings stock towards the identification of the optimal building typologies and respective representative buildings and suggested them as appropriate for further planning procedures, e.g., the cost-optimality assessment. A critical component in the context of the typology-based planning approach is the determination of the most relevant clustering criteria, i.e., the most influential characteristics on building energy performance. For a stock of 56 office buildings in Singapore, Deb and Lee [17] employed the k-means algorithm and concluded that the key variables of the highest influence on energy consumption (for the examined stock) are the gross floor area, the energy consumption not attributed to air conditioning, and the nominal capacity and efficiency of chillers. They suggested that the extracted clusters can be used to benchmark offices in terms of ex-ante conditions and the energy-saving potential. The identification of the optimal set of classification criteria for buildings’ benchmarking purposes has been addressed in numerous studies by means of employing the typology approach based on machine learning algorithms exploiting multi-dimensional domains of building features [18,19,20,21]. Such methods, when unsupervised, often allow buildings with different key features to be included in the same typology simply because of the proximity among the (selected) dependent variable (usually the energy use intensity). Taking into account that in common planning practices the same renovation interventions are assumed for buildings of the same typology since they are fairly represented by one archetype or ambassador building [8,22,23,24,25], an additional limitation for any (supervised) classification algorithm would be to ensure that buildings of the same cluster have similar key features, which are related to the energy interventions to be tested in a further step. For instance, if the replacement of an oil burner is included in the set of interventions tested for a typology, then all buildings of the same typology should have an oil burner for heating purposes in their existing situation. In conclusion, the main challenge in building-stock energy efficiency planning is to ensure both the representativeness of the building stock by the identified typologies and the retrofit relevance of key building features.

As stated by Droutsa et al. [12], although the typology approach has been extensively exploited for residential building stocks, the same is not true regarding the non-residential ones. Furthermore, while significant progress has been encountered in building clustering methods and applications, most studies are either focused on the residential sector or limited to the application of the typology approach for benchmarking purposes. The present study suggests a holistic bottom-up typology-based methodology starting from the reliable clustering of the building stock to the development of a gradual renovation plan, particularly suitable for public buildings at the local level, in response to the latest European energy-efficiency policies. The methodology is demonstrated by means of a case study for the public buildings’ stock of Municipality of Hersonissos in Greece and is unfolded as follows:

• Initially, the Greek policy environment is briefly presented concerning public building energy efficiency planning;
• A clustering method is introduced, aiming at the determination of the optimal buildings’ classification criteria and the identification of representative typologies, under special constraints considering the retrofit relevance of key building features and reasonable computational effort. The RMSE index is used as the objective function on the basis of the available primary energy consumption for each building;
• The IMPULSE Interreg-MED project tools are then used for the examined building stock and typologies towards the fast production of the KPI database for the whole building stock and for alternative retrofit scenarios and, finally, towards a reliable gradual renovation plan based on user-defined conditions including the desired annual renovation rate.

Through the elaboration of the case study for a stock of 44 buildings, it was found that the optimal set of clustering criteria includes four features, namely the building use, construction year, the heating, and the cooling system. The clustering algorithm led to 15 building typologies. Assuming a 7% renovation rate per year, a 12-year gradual renovation (nZEB transformation) plan is obtained foreseeing an 85% CO₂ emissions reduction. The proposed work efficiently tackles the most important planning pillars from building-stock classification to the production of a renovation plan since it demonstrates easy-to-use tools for automatic extrapolation of KPI databases and fast development of renovation roadmaps in computer environments within reasonable resources. It becomes evident that the methodology is particularly useful for public building energy efficiency plans since it includes the renovation rate as a decision-making parameter (among others) in response to the directive 2023/1791/EU.

2. Materials and Methods

2.1. The Greek Policy Context

In focus of the exemplary role of public buildings specifically at the local administration level, so far, the directive 2018/2002/EU [26] has been transposed into the Greek legislation (L.4843/2021), which enforces the obligation for regions and municipalities to conduct and monitor energy renovation plans (ERB plans) for their public building stocks. To facilitate this obligation, the Hellenic Ministry of the Environment and Energy (MINENV) has released a specific template/guide for conducting the ERB plans [27]. As presented in a previous publication [8], the MINENV guide requests extensive KPI databases (energy, environmental, and cost state and impact indicators) for the whole building stock, for the existing situation, and for at least two retrofitting scenarios, namely (a) the minimum energy performance requirements scenario (MEPR) and (b) the nZEB scenario. In addition, a renovation roadmap is requested, including prioritized projects, in a time horizon of at least 4 years towards the achievement of a feasible user-defined goal in terms of energy savings. Considering the latest directive 2023/1791/EU, once it is transposed to the Greek framework, it is expected to introduce an additional planning parameter to the anticipated renovation roadmap that is the renovation rate, i.e., at least 3% of the floor area of the public building stock to be renovated each year. Hence, this measure should be also adopted in the planning procedure.

2.2. The Case Study Considered

The case study considered herein refers to the public building stock of the Municipality of Hersonissos, in Crete, Greece. The Municipality is located at the north seafront of the Regional Unit of Heraklion. As established by L.3852/2010 (Program “Kallikratis”), it consisted of four Municipal Units, specifically Hersonissos, Episkopi, Gouves, and Malia. According to the Greek energy performance of buildings directive (EPBD), the Municipality of Hersonissos belongs to the “A” climate zone, i.e., the warmest among the four climate zones of Greece. Based on recordings from the Hellenic National Meteorological Service, the winter period lasts from December to March with average temperatures ranging between 12 °C and 14 °C, while the summer period lasts from mid-May to mid-September with average temperatures ranging between 20.5 and 26.5 °C. The coldest and wettest month of the year is January, with an average temperature of 12.1 °C, an average rainfall of 91 mm, and an average of 16 rainy days. The hottest and driest month is July, with an average temperature of 26.4 °C and negligible rainfall.

The Municipality of Hersonissos is very motivated to promote climate-change mitigation and adaptation actions. As an active signatory of the “Covenant of Mayors” initiative, it has developed a targeted Sustainable Energy and Climate Action Plan (SECAP) rolling out specific actions throughout the local policy sectors (including public buildings) to reduce its CO₂ emissions by at least 40% by the year 2030 in relation to the year 2018, which is adopted as the baseline year. The current work reflects the emissions reduction target for the municipal buildings sector as well as the outlining of specific projects to be implemented in the following years. The outcomes of the work are included in the ERB plan of the municipality in the framework of the ongoing legislation.

The building stock included in the analysis herein consists of 44 municipal buildings with a useful area above 250 m² (which is the minimum limit for buildings included in the ERB plan according to Greek law). Out of the aforementioned buildings, 32 are educational, 9 are office buildings, 2 are multi-purpose buildings, and 1 is a rural medical facility. The spatial distribution of the buildings and building uses across the municipal units are depicted in Figure 1. Following the same numbering of the latter, the examined building stock is tabulated in Appendix A, enlisting the values of key building features and the total primary energy consumption found from available energy performance certificates (EPCs) that are used in the clustering process below.

2.3. Building-Stock Clustering Approach

As mentioned in Section 1, for the formulation of a reliable energy renovation plan, building-stock energy modeling is a key aspect to producing the necessary KPI database in order to identify the energy-saving potential and to finally develop the appropriate renovation scenarios and roadmaps. This poses great challenges when focusing on building stocks of public authorities, especially municipalities. The latter own or manage a large building stock consisting of diverse uses and of large proportion of old buildings, often lacking detailed technical information and energy audits. Especially in Greece where there is still a very limited extent of operating energy management systems in the public sector, energy-consumption data are also very difficult to gather. In such an environment, building-stock energy planning indicates the necessity for conducting building-stock energy audits, data acquisition, and energy modeling to assess retrofit scenarios from scratch, i.e., for each building individually. This building-by-building approach would require high amounts of resources in terms of human effort, services/expertise cost, and computer power. This hinders the on-time informed decision-making in prioritizing energy investments at the municipal level. Specifically in Greece, such an approach causes delays in compliance with ongoing legislation requirements (refer to Section 2.1) regarding the ERB plans of municipal and regional authorities.

To respond to the aforementioned challenges, the bottom-up clustering approach is employed in the present study taking advantage of limiting the detailed data collection and energy simulations only to the representative buildings of building-stock typologies and then extrapolating the simulated indicators to all buildings of each identified typology; hence accelerating the planning process. The approach followed refers to the classification of the building stock based on common values/options/ranges of the key building features. Considering that depending on the number of the building features and corresponding values adopted there might be many possible combinations of eligible criteria and eventual alternative typologies, a structured method is required that identifies the optimal set of classification criteria and the most representative typologies. In this paper, a supervised clustering approach is proposed consisting of the following components:

• At first, based on the scientific experience regarding energy-related building typologies, the most relevant options are assigned to each building feature.
• Secondly, a clustering algorithm is developed in a computer code, aiming at grouping buildings with the same building features. The algorithm tests all possible combinations of building features and retrieves the optimal set of clustering criteria and final typologies, which correspond to the minimum value of the RMSE index based on the available primary energy consumption of each object/building. The algorithm encompasses two additional constraints: (a) a minimum set of specific building features ensuring (as far as possible) the retrofit relevance, i.e., ensuring the suitability of interventions to both the representative building of each typology and the rest buildings of the typology and (b) a maximum number of typologies.

2.3.1. Definition of Clustering Criteria

Since the current classification approach focuses on building-stock energy assessment, the classification criteria selected are building features that greatly influence energy performance. For non-residential buildings at the national level, the principal criteria are the building use, age, and location (climate zone) [12]. However, when working at the local level, especially in the framework of energy renovation planning, additional criteria should be included to better represent the heterogeneity of the building stock, such as the construction type and envelope characteristics (insulation level and window types), the heating, and the cooling system [10,13]. Taking into account the building-stock characteristics available from energy audits and EPCs (refer to Appendix A) and the retrofit-relevance principle, the following initial set of classification criteria are adopted for the considered case study: use, age, No. of floors, heating system, cooling system, construction type, insulation level, and type of windows. The alternative options assigned to each criterion and respective justification are provided in

Table 1. Initial set of classification criteria and alternative options assigned.

Classification Criterion	Assigned Options (for Classification Purposes)	Justification/Explanation
Building use	Office; Educational; Multi-purpose building; Rural medical center	- According to the examined building stock; - Based on the building-use definitions of the Greek EPBD; - Buildings of the same use category have the same operation properties (schedule, occupancy, etc.) according to the EPBD.
Construction year	Before 1980; From 1980 to 2010; After 2010	- Based on the milestone years when a regulatory framework regarding minimum energy performance requirements issued in Greece 1.
No. of floors (including ground floor)	1; 1 to 3	- The examined building stock includes buildings with up to 3 floors only; - It is related to the extent of the impact of possible roof interventions on the overall building energy performance.
Heating system	Oil boiler; Local heat pump (Split unit); Central heat pump; Not exist	- Based on the available energy audits reports or EPCs for the examined building stock.
Cooling system	Local heat pump (Split unit); Central heat pump; Not exist	- Based on the available energy audits reports or EPCs for the examined building stock.
Construction type	Heavy construction (bricks and reinforced concrete); Very heavy construction (stonework)	- Based on the available energy audits reports or EPCs for the examined building stock.
Insulation applied	Yes; No; Insufficient according to the TIR 2; Insufficient according to the EPBD 3	- Based on the categories of thermal insulation of the Greek EPBD; - Based on the available energy audits reports or EPCs for the examined building stock.
Window type	Single-glazed aluminum frames; Double-glazed aluminum frames; Single-glazed wooden frames; Single-glazed aluminum and double-glazed wooden frames; Double-glazed aluminum and single-glazed wooden frames	- Based on the available energy audits reports or EPCs for the examined building stock; - For buildings constructed before 2010, it is assumed that existing window types do not meet the EPBD minimum requirements.

¹ In Greece, the first thermal insulation regulation (TIR) was issued in 1979 with the presidential decree 362D/4-7-1979, imposing minimum requirements regarding thermal insulation; hence, for buildings constructed before 1980, it is fairly assumed that they have no thermal insulation. In 2010, the Greek EPBD (TOTEE KENAK 20701-1) was issued, which determines the minimum energy performance requirements for new buildings regarding thermal insulation and energy systems; hence, for buildings constructed in the period 1980–2010 and for those constructed after 2010, thermal insulation properties of the ‘79 thermal-insulation regulation and the Greek EPBD properties, respectively, are fairly assumed (unless otherwise indicated by the energy audit); ² The Greek Thermal Insulation Regulation of the Presidential Decree PD 362D/4.7.1979; ³ The Greek technical energy performance of buildings directive TOTEE KENAK 20701-1/2017.

.

2.3.2. The Clustering Algorithm Applied

The proposed clustering aims at the identification of the optimal set of classification criteria and at the extraction of the eventual building typologies. The procedure refers to ensuring the best representativeness of the whole building stock by fewer buildings, i.e., that the buildings included in the same typology are very similar; thus, they have similar energy performance. In agreement with previous studies [13,18], a modified root-mean-square-error (RMSE) index is used on the basis of the total primary energy consumption of each building of the examined stock (refer to Appendix A). Specifically, the weighted average of RMSE of each borne typology for each possible criteria combination (Table 1) is used as the objective function, i.e., when minimized it corresponds to the best clustering. For each combination of criteria, it is expressed as follows:

$${\overline{RMSE}}_w=\frac{\sum _{i=1}^n{N}_i·{RMSE}_i}{\sum _{i=1}^n{{\rm N}}_i}$$

(1)

where $w$ means “weighted”, $i$ is the $i-$th typology, $N_i$ is the number of individual buildings of the $i-$th typology, $n$ is the total number of typologies, and the ${RMSE}_i$ is the RMSE of the $i$-th typology, which is calculated as follows:

$${RMSE}_i=\sqrt{{\sum }_{k=1}^{N_i}\frac{{\left({\overline{PE}}_i-{PE}_{i,k}\right)}^2}{N_i}}$$

(2)

where $k$ is the $k-$th building of the $i-$th typology, ${PE}_{i,k}$ is the normalized (per sq.m.) primary energy consumption of the corresponding $k-$th building of the $i-$th typology, and ${\overline{PE}}_i$ is simply the numerical average of the available normalized (per sq.m.) primary energy consumption of buildings of the $i-$th typology, i.e.,

$${\overline{PE}}_i=\frac{{\sum }_{k=1}^{N_i}{PE}_{i,k}}{N_i}$$

(3)

Taking the above (Equation (3)) as the target parameter is based on the assumption that buildings of the same typology have similar building characteristics and consequently should have similar normalized primary energy consumption. It is obvious from the above equations that if each building was its own typology, the average RMSE would be equal to 0; thus, if left unsupervised, the clustering algorithm would return the one-building/one-typology solution as the optimal one. This implies that to continue with the next steps of the energy planning process, the alternative retrofit scenarios would be simulated for each building of the stock one-by-one; which requires excessive effort, time, and resources for building-stock energy modeling purposes. Consequently, to balance both the energy calculations cost effectiveness and the fidelity of calculations, the following constraints are incorporated in the classification procedure:

• Considering the specific building stock and the retrofit relevance of building features, the following key criteria should definitely be included in the optimal set of clustering criteria: building use, construction year, heating system, and cooling system;
• A maximum limit of 15 building typologies is set, which is also in compliance with the capabilities of the IMPULSE tools [8] employed later in the planning procedure.

The problem is solved in the open-source programming language Python v.3.11.0.

2.4. Typologies’ Manipulation towards an ERB Plan

First of all, from each typology extracted by the above clustering process, a representative building is selected as the ambassador building, which reflects the energy performance of all buildings of the same typology. For each ambassador building, the energy performance is calculated for the following renovation scenarios (the specific interventions tested for each typology are presented in Section 3) [27]:

• MEPR: Each upgrading intervention included in the renovation scenario meets the minimum energy performance requirements according to the Greek EPBD;
• nZEB: The interventions included achieve a nearly zero-energy building. According to the Greek policy context (Ministerial Decision 5447B/2018), an existing building is considered nZEB if it belongs to the B+ energy class as per the Greek EPBD.

The scenarios are numerically constructed upon the computer .xml executable files developed in the national calculation software KENAK v.1.31 (which incorporates the quasi-steady calculation method) [28], being available for the existing situation of the buildings from already conducted energy audit campaigns. For the existing situation and for each of the retrofit scenarios, the energy analysis yields the KPIs’ results of all energy-consumption components (primary and final energy consumption per end-use and per carrier) as well as the CO₂ emissions per energy carrier. To account for cost-performance assessment, the cost indicators, i.e., the investment cost and energy-related cost (energy expenditure) are estimated by adopting typical unit costs found in ref. [27]. As far as the photovoltaics measure is concerned, it is treated as grid-connected under the net-metering mechanism while the energy cost reduction due to the PV is estimated following the process and electrical energy unit cost found in ref. [29].

Thereinafter, the calculated KPIs of the ambassador buildings are inserted in the IMPULSE Interreg-MED project Excel-based tool “KPIs’ processor” (available for direct download in ref. [30]) following the procedure extensively described in ref. [8]. Based on the assumption that buildings of the same typology have the same normalized KPIs’ values per sq.m. of floor area, the KPIs’ processor extrapolates the state and impact KPIs from the ambassador buildings to the whole building stock for the existing situation and for the studied renovation scenarios. Hence, the tool produces the state and impact of the KPI database for the whole building stock and for all the renovation scenarios.

Finally, again following the extensively described procedure in ref. [8], the Excel-based tool “KPIs’ processor PLUGIN” (available for direct download in ref. [31]) is used for the prioritization of retrofit scenarios and the eventual production of a gradual renovation plan based on user-defined preferences. The decision-making scheme is formulated based on the following municipal–authority directions:

• Desired annual renovation rate: at least 3% of the total building-stock floor area;
• Desired duration of foreseen projects’ implementation: 12 years;
• A reasonable investment cost not exceeding EUR 1 million annually should be ensured;
• Priority should be given to projects achieving the highest possible reduction in energy consumption per carrier, i.e., electricity and heating oil;
• Projects leading to nZEB transformation are the top priority;
• Priority should be given to projects regarding educational buildings;
• A reliable and manageable plan is desired with no more than three projects each year within the first 6 years of implementation.

The above decision-making considerations are compiled in the “KPIs’ processor PLUGIN” tool as presented in

Table 2. Compilation of the decision-making scheme in the IMPULSE project tool “KPIs’ processor PLUGIN”.

Decision-Making Parameter	Decision-Making Target	Treatment in the IMPULSE PLUGIN Tool
Desired annual renovation rate	At least 3% of the total building-stock floor area	Determined through sensitivity analysis, starting from a renovation rate equal to 3% as initial input condition.
Desired duration of foreseen projects’ implementation	Up to 12 years	Output of the PLUGIN tool. Treated as a convergence criterion of the sensitivity analysis.
Annual investment cost	Not exceeding EUR 1 million	Output of the PLUGIN tool. Treated as a convergence criterion of the sensitivity analysis.
Priority to projects with the highest possible reduction in energy consumption per carrier	Maximum reduction in electricity and oil consumption	The indicators electricity savings and savings of fossil-fuel consumption are included in the set of the decision-making KPIs. The weights of the KPIs are determined after convergence of a sensitivity analysis.
Buildings’ energy performance after renovation	nZEB	Only the nZEB renovation scenarios are included in the decision-making scheme.
Building types of top priority	Educational buildings	The following is included as a convergence criterion in the sensitivity analysis: % of school projects within the first 6 years >60%.
No. of projects	Not more than 3 projects each year within the first 6 years of implementation	The following is included as a convergence criterion in the sensitivity analysis: No. of years within the first 6 yrs of implementation with up to 3 projects >4.

.

3. Results

3.1. Identification of Building-Stock Representative Typologies

In the first step, the proposed clustering algorithm is applied following the procedure described in Section 2.3.2. The main target here is to determine the optimal set of clustering criteria that ensure typologies’ representativeness of the whole building stock under the specified constraints, i.e., the specific four criteria (building use, construction year, heating system, and cooling system) should be included in the clustering set and a maximum limit of 15 typologies should be retained. To ensure the best understanding of the algorithm operation, the results are presented in a constraints-free fashion, i.e., showcasing all the 255 possible classification combinations when processing all the alternative options of Table 1 for the data of the considered case study (refer to Appendix A). The results of the objective function (Equation (1)) and the extracted number of building-stock typologies are presented in Figure 2 for the various combinations of classification criteria. As expected, the ${\overline{RMSE}}_w$ index is reduced as the number of criteria increases accompanied with an increased number of typologies. In the same graphs, the solutions that respect the limitation of the specific four criteria included in the set of classification criteria are highlighted in a red color. It becomes obvious that only one eligible solution emerges that also meets the maximum threshold of the number of typologies. It is observed that the solution obtained refers to only one optimal set of classification criteria which, in fact, includes precisely the least four desired ones yielding precisely 15 typologies.

The 15 building typologies, which correspond to the aforementioned optimal classification (refer to the dashed lines in Figure 2), accompanied by the respective classification criteria, are presented in

Table 3. The identified building typologies, their characteristics and selected ambassador buildings.

Public Building Typology	Classification Criteria				Buildings 1
	Building Type/Use	Construction Year	Heating System	Cooling System
PBT1	Office	Before 1980	Local heat pump (Split unit)	Local heat pump (Split unit)	City Hall; Building 202; Municipal Store of Episkopi; Building 154
PBT2	Office	1980–2010	Local heat pump (Split unit)	Local heat pump (Split unit)	Municipal Store of Hersonissos Port; Municipal Store of Malia; Fire Brigade Building; Former office Building of Municipal Water and Sewerage Company of Hersonissos
PBT3	Office	1980–2010	Oil boiler	Local heat pump (Split unit)	Former Courthouse
PBT4	Education	Before 1980	Local heat pump (Split unit)	Local heat pump (Split unit)	Computer room of Gouves senior high School (688); Gournes Elementary School (“Eirini”—Building 207); Gournes Elementary School
PBT5	Education	Before 1980	Central heat pump	Central heat pump	Episkopi Elementary School
PBT6	Education	Before 1980	Oil boiler	Local heat pump (Split unit)	Gouves senior high School (156); Gouves Junior High School (156 and 158); 2nd Vocational Training Institute of Heraklion (Building 303); School of Fine Arts (Building 301); Music School (Building 308); Gouves Elementary School; Elia Elementary School; Kokkini-Chani Elementary School; Hersonissos Port Kindergarten; 1st Elementary School of Malia (Building 1); Mochos Elementary School; Malia Kindergarten (One Building—Old)
PBT7	Education	Before 1980	Oil boiler	Not exist	Kindergarten and Elementary School of Gonies
PBT8	Education	Before 1980	Not exist	Not exist	Music School (Building 302)
PBT9	Education	1980–2010	Central heat pump	Central heat pump	Episkopi Kindergarten
PBT10	Education	1980–2010	Oil boiler	Local heat pump (Split unit)	Episkopi Junior and Senior High School; Hersonissos Port Junior and Senior High School; Malia Junior and Senior High School -Building A; Hersonissos Port Elementary School; Elementary School of Hersonissos old village; 2nd Elementary School of Malia; 1st Elementary School of Malia (Building 2); Malia Kindergarten; Hersonissos Elementary School
PBT11	Education	1980–2010	Oil boiler	Not exist	Malia Junior and Senior High School-Building B; Mochos Junior High School- High School
PBT12	Education	After 2010	Central heat pump	Central heat pump	New School Complex of Primary Education in Kokkini-Chani (Elementary School–Kindergarten)
PBT13	Education	After 2010	Oil boiler	Local heat pump (Split unit)	Mochos Kindergarten
PBT14	Rural Medical Facility	1980–2010	Local heat pump (Split unit)	Local heat pump (Split unit)	Cultural Center of Hersonissos Port
PBT15	Multi-purpose building	Before 1980	Local heat pump (Split unit)	Local heat pump (Split unit)	Library—Multi-purpose building (310); Avdou Elementary School (New-Building)

¹ The selected ambassador buildings are marked in bold.

. According to the approach suggested herein (refer to Section 2.4), from each typology an ambassador building is selected for which the energy assessment of renovation scenarios is conducted in the next step towards the calculation of the required key performance indicators. The ambassador buildings of the identified typologies are indicated in bold in Table 3. They are selected based on the amount of available technical information meaning that buildings with the highest amount of technical data (e.g., technical studies, architectural drawings, energy audits, etc.) are preferred as ambassadors since the more extensive the available information the higher the fidelity of the energy assessments and estimated impacts of the renovation scenarios.

Following the order of buildings presented in Appendix A, their distribution into the 15 typologies and the energy consumption patterns are illustrated in the parallel coordinates plot of Figure 3. In the figure, the color of each line indicates the typology to which each building is allocated according to the clustering process. It is depicted that the most populated typologies are those of educational use followed by the office-use ones. An almost equal number of buildings are constructed before 1980 and within the period from 1980 to 2010, with only two being constructed after the year 2010. In terms of the heating system, most educational buildings are served by a conventional oil boiler, while almost all offices are by local heat pumps. In terms of the cooling system, the local heat pump prevails. With respect to the classification criteria, it is shown that educational buildings are mostly less energy consuming, especially compared to office buildings, which is anticipated to be due to the longer periods of non-operation annually. In general, buildings constructed before 1980 present higher energy consumption as they are thermally unprotected. Concerning the heating and cooling system, it is seen that buildings served by a central heat pump correspond to a relatively lower energy consumption compared to buildings served by other systems.

3.2. Building-Stock Energy Performance Assessment and Production of the KPI Database

Based on the available technical information and recommendations from the conducted energy audits, the feasible interventions constituting the considered renovation scenarios MEPR and nZEB for each ambassador building (also for each typology PBT) are presented in

Table 4. Description of renovation scenarios considered for each PBT and interventions’ quantities foreseen for the respective ambassador building.

Intervention 1	PBT1	PBT2	PBT3	PBT4	PBT5	PBT6	PBT7	PBT8	PBT9	PBT10	PBT11	PBT12	PBT13	PBT14	PBT15
Scenario 1: MEPR (Major Retrofit)
WTI-EPS7	490 m2	1102 m2	534 m2	249 m2	458 m2	666 m2	340 m2	897 m2	251 m2	−	−	−	−	551 m2	359 m2
RTI-EPS7	683 m2	673 m2	450 m2	377 m2	316 m2	1005 m2	265 m2	1362 m2	157 m2	−	−	−	−	371 m2	403 m2
WR-DG-LE	214 m2	226 m2	76 m2	59 m2	67 m2	251 m2	65 m2	207 m2	43 m2	542 m2	83 m2	−	−	134 m2	37 m2
CHP-H	−	−	−	52 kW	−	122 kW	50 kW	−	−	291 kW	80 kW	−	50 kW	−	−
CHP-HC	−	196 kW	73 kW	−	−	−	−	184 kW	−	−	−	−	−	91 kW	−
LHP	−	−	−	−	−	−	38 kW	−	−	155 kW	69 kW	−	−	−	37 kW
LED	4.0 kW	7.8 kW	3.0 kW	1.0 kW	0.7 kW	3.2 kW	0.7 kW	4.1 kW	0.8 kW	11.2 kW	2.7 kW	−	0.8 kW	3.6 kW	0.9 kW
Scenario 2: nZEB (Deep retrofit, built on top of Scenario 1)
Scenario 1	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	-	✓	✓	✓
PV	31 kWp	72 kWp	16 kWp	4 kWp	8 kWp	9 kWp	3 kWp	8 kWp	3 kWp	45 kWp	7 kWp	55 kWp	2 kWp	38 kWp	9 kWp

¹ Nomenclature of interventions: WTI-EPS7: Wall thermal insulation with graphite-based EPS 7 cm; RTI-EPS7: Roof thermal insulation with graphite-based EPS 7 cm; WR-DG-LE: Windows replacement with low-e double-glazed 4–18-4 mm, aluminum frame with thermal break; CHP-H: Installation of a central heat pump of COP 4.0 only for heating purpose; CHP-HC: Installation of a central heat pump of COP 4.0/EER 3.8; LHP: Installation of local heat pump; LED: Fixtures’ replacement with LED; PV: Grid-connected PV/net metering.

. As tabulated, Scenario 2 (Deep retrofit/nZEB) is built on top of the interventions of Scenario 1 (Major retrofit/MEPR) in accordance with the “aggregative principle” required for the correct application of the IMPULSE tools later on [8]. It should be mentioned, however, that the latter is not true for the case of the PBT12 since it refers to a solitary typology including a recently constructed building, which already meets the EPBD minimum energy performance requirements; thus, only the nZEB Scenario is considered.

This is followed by a numerical energy analysis, using the official national energy calculation software KENAK v.1.31, for each ambassador building and for each retrofit scenario. The results of the energy analysis regarding important state and impact KPIs are provided in

Table 5. KPIs for the PBTs’ ambassador buildings in the existing situation and for the renovation scenarios.

PBT	KPIs—Existing Situation 1					KPIs—Scenario 1 (MEPR) 1							KPIs—Scenario 2 (nZEB) 1
	FECe (kWh/m2.a)	HOFCe (kWh/m2.a)	REGe (kWh/m2.a)	t-CO2,e (kg/m2.a)	t-ECe (EUR/m2.a)	FECSc1 (kWh/m2.a)	HOFCSc1 (kWh/m2.a)	REGSc1 (kWh/m2.a)	t-CO2,Sc1 (kg/m2.a)	t-ECSc1 (EUR/m2.a)	n-ICSc1 (EUR/m2)	SPPSc1( yrs)	FECSc2 (kWh/m2.a)	HOFCSc2 (kWh/m2.a)	REGSc2 (kWh/m2.a)	t-CO2,Sc2 (kg/m2.a)	t-ECSc2 (EUR/m2.a)	n-ICSc2 (EUR/m2)	SPPSc2 (yrs)
PBT1	98.8	0.0	0.0	97.7	23.7	51.8	0.0	0.0	51.2	12.4	12,034	13.7	2.2	0.0	49.7	2.2	0.5	511	8.9
PBT2	80.4	0.0	0.0	79.5	19.3	62.3	0.0	0.0	61.6	15.0	27,601	35.6	2.2	0.0	68.4	2.2	0.5	974	11.5
PBT3	101.0	42.8	0.0	111.1	30.1	40.8	0.0	0.0	40.3	9.8	6483	9.2	2.0	0.0	38.8	2.0	0.5	317	7.6
PBT4	41.1	0.0	0.0	40.9	9.9	15.3	0.0	0.0	15.1	3.7	1421	29.3	4.8	0.0	9.7	4.7	1.2	445	22.6
PBT5	42.9	0.0	0.0	42.4	10.3	29.8	0.0	0.0	29.5	7.2	2963	45.1	4.9	0.0	20.3	4.8	1.2	487	18.8
PBT6	35.0	34.6	0.0	43.7	11.8	14.6	0.0	0.0	14.4	3.5	3608	24.0	5.4	0.0	9.5	14.4	1.3	1342	20.3
PBT7	27.2	111.4	0.0	56.3	17.4	20.1	0.0	0.0	19.9	4.8	1331	21.4	4.5	0.0	10.8	4.5	1.1	298	17.7
PBT8	42.8	0.0	0.0	42.3	10.3	10.3	0.0	0.0	10.1	2.5	3733	23.7	2.0	0.0	8.3	2.0	0.5	726	19.7
PBT9	30.9	0.0	0.0	30.6	6.6	15.5	0.0	0.0	15.3	3.3	1038	36.5	5.1	0.0	9.0	5.0	1.1	341	24.3
PBT10	30.3	8.7	0.0	32.3	8.1	19.3	0.0	0.0	19.1	4.6	17,345	23.0	5.3	0.0	12.7	5.2	1.3	4763	14.5
PBT11	29.9	20.0	0.0	34.9	8.9	15.7	0.0	0.0	15.5	3.7	2519	20.0	4.9	0.0	12.6	4.8	1.2	786	15.5
PBT12	26.2	0.0	0.0	25.9	6.3	26.2	0.0	0.0	25.9	6.3	27,834	0.0	6.2	0.0	20.0	6.2	1.5	6611	4.1
PBT13	18.8	23.0	0.0	24.7	7.7	14.1	0.0	0.0	14.0	3.4	931	15.5	2.0	0.0	12.1	2.0	0.5	132	10.9
PBT14	106.6	9.4	0.0	108	23.1	67.8	0.0	0.0	67.1	14.2	14,598	16.3	10.0	0.0	65.0	9.9	2.1	2153	9.7
PBT15	94.8	0.0	0.0	93.8	22.8	36.9	0.0	0.0	36.5	8.9	3567	10.6	14.1	0.0	35.4	13.9	3.4	1363	9.5

¹ Nomenclature of KPIs: Subscripts: e: existing situation, Sc1: Scenario 1, Sc2: Scenario 2. FEC: Annual final electricity consumption; HOFC: Annual heating oil final consumption; REG: Renewable energy generation; t-CO₂: Annual total CO₂ emissions; t-EC: Annual total energy cost; n-IC: normalized per sq.m. investment cost; SPP: Simple payback period.

. The average values of the energy, environmental, and cost indicators are illustrated in Figure 4, presented for all ambassador buildings as well as per use and period of construction and for all studied situations, i.e., the existing condition and the two renovation scenarios. Overall, on average, it is seen that the total final energy saving, the CO₂ emissions avoidance, and the energy cost are reduced by around 50% and 90%, for the MEPR and the nZEB scenario, respectively. The overall average investment cost is approximately 140 EUR/m² for the MEPR scenario and reaches 170 EUR/m² in the nZEB scenario. Despite the higher investment in the nZEB scenario, it achieves a payback period below 20 years (contrarily to the MEPR which exceeds 25 years) due to the high energy compensation achieved by the PV integration. Focusing on building use, it is seen that the educational buildings present the least energy-saving potential, especially those constructed after the year 1980 (milestone of the first thermal insulation regulation (TIR)), which is normal since these buildings are not operating for almost 4 months a year (Easter, Christmas, and the summer holidays). As a result, the average payback period of the MEPR investment exceeds 25 years for educational buildings. The latter is improved with the PV integration in the case of the nZEB scenario, which, on the other hand, exhibits a much better economic performance for non-educational buildings with an average payback period of around 10 years. Concerning the building-use and age-wise average investment cost, it exceeds the overall average for both MEPR and nZEB in the case of non-educational buildings regardless of the period of construction and of educational ones constructed before 1980. On the contrary, it appears to be significantly lower in the case of educational buildings constructed after 1980 due to the less extensive interventions considered for MEPR and nZEB transformation.

To the direction of producing the necessary KPI database for the examined building stock, the ambassadors’ indicators of Table 5 are inserted in the IMPULSE KPIs’ processing tool (data insertion in the tool is demonstrated in ref. [8]). The tool then automatically extrapolates the KPIs from ambassador buildings to all the rest of the same typology; thus, the required database is obtained for the whole building stock. A sample of the resulting database is provided in Appendix B. The aggregated KPIs per PBT obtained again in the IMPULSE tool is provided in Figure 5, which also showcases the comparisons of impact KPIs among PBTs in a bar form. It is estimated that if all buildings are renovated under the MEPR condition, the final electricity consumption and heating oil consumption will be reduced by 864 MWh/y and by 635 MWh/y, respectively, while CO₂ emission avoidance reaches around 1000 t/y. The required investment cost to realize these MEPR projects is EUR 5.8 million corresponding to a simple payback period of 22 years.

On the other hand, with an additional budget of EUR 1 million more, projects of nZEB transformation are feasible. It is seen that the complete nZEB transformation almost doubles the impacts compared to the MEPR scenario, i.e., the final electricity consumption and the CO₂ emissions are reduced by 1840 MWh/y and by 1890 t/y, respectively, while the simple payback period is significantly reduced (in relation to the MEPR scenario) down to around 14 years. From the relative performance point of view, it is observed that typologies PBT6 and PBT10 exhibit the highest energy-saving potential but under the highest investment cost. These are followed by the typologies PBT1 and PBT2 in terms of energy savings, although with much lower investment costs and simple payback periods compared to the former typologies PBT6 and PBT10. In any case, it is demonstrated that the tool used herein facilitates the rapid generation of the required KPIs database, which can be further examined and elaborated to prioritize and plan renovation projects.

3.3. Development of the Gradual Energy Renovation Plan

To support the setting up of a reliable renovation roadmap, the IMPULSE project KPIs’ processor PLUGIN tool is used. As presented in ref. [8], the tool requires many inputs which are not always known; therefore, the planning considerations from the municipal authority reported in Section 2.4 should be interpreted appropriately in order to define inputs and formulate the decision-making scheme. In this framework and according to the compilation of Table 2, to determine the years of projects’ implementation, the PLUGIN tool is initially executed with the same weights imposed to the three selected decision-making criteria, i.e., 33.33% for the heating-oil consumption savings, the electricity savings, and the investment cost, for different annual renovation rates starting from 3% according to the latest EPBD directive. This initial sensitivity analysis determined a 7% renovation rate (corresponding to a 3266 m² floor area renovated annually) as the least acceptable one that achieves the desired duration of 12 years of implementation of deep retrofit projects towards nZEB transformation. A second sensitivity analysis is executed to define the optimal weights of the three decision-making parameters. More specifically, the tool is executed for six alternative combinations of the weights assigned to the following decision-making criteria:

• DC1: Annual savings of fossil fuel consumption;
• DC2: Annual electricity savings;
• DC3: Estimated investment cost.

The results of the sensitivity analysis regarding the considered convergence criteria of Table 2 are presented in

Table 6. Combinations of the weights tested in the IMPULSE PLUGIN tool and corresponding convergence criteria results (values that meet the convergence criteria are underlined).

Combination	Weights	Duration of Implementation (yrs)	Max Annual Investment (EUR)	Years with Up to 3 Projects in the First 6 yrs	% School Projects in the First 6 yrs
Convergence When:		≤12	≤1,000,000	>4	>60%
Comb1	DC1: 30% DC2: 10% DC3: 60%	12	928,073	2	82%
Comb2	DC1: 10% DC2: 30% DC3: 60%	12	851,927	3	52%
Comb3	DC1: 60% DC2: 10% DC3: 30%	11	906,620	4	91%
Comb4	DC1: 10% DC2: 60% DC3: 30%	12	902,326	6	43%
Comb5	DC1: 30% DC2: 60% DC3: 10%	13	896,799	6	45%
Comb6	DC1: 60% DC2: 30% DC3: 10%	12	906,621	5	65%

. It is concluded that the combination that satisfies all the preferences refers to the following weights: 60%, 30%, and 10%, assigned to the DC1, DC2, and DC3, respectively. Accordingly, the final decision-making scheme in the IMPULSE PLUGIN tool is provided in Figure 6 also taking into account that the nZEB transformation, i.e., the deep retrofit condition, is the authority’s top priority.

The renovation roadmap produced by the PLUGIN tool is presented in

Table 7. The produced renovation roadmap for the examined building stock.

Year of Implementation		2024	2025	2026	2027	2028	2029	2030	2031	2032	2033	2034	2035
Annual absolute	Floor area retrofitted m²	4263.04	5579.64	5445.22	3906.35	4780.34	3684.64	4421.52	3630.62	3305.49	3490.26	3310.44	846.15
	Investment EUR	906,621	761,148	663,138	825,766	797,339	543,976	86,393	603,874	474,150	556,820	579,631	166,562
	Cost savings EUR/yr	44,600	44,844	54,108	67,032	56,766	43,820	21,223	55,904	27,746	30,543	47,501	7372
	Avoided CO2 kg/yr	124,907	155,245	207,668	273,960	216,192	169,331	87,104	222,272	93,100	115,889	196,281	30,631
	Energy Savings kWh/yr	413,515	482,809	636,176	806,741	652,948	489,892	256,448	627,591	295,116	349,375	558,735	87,577
Annual accumulated	Floor area retrofitted m²	4263	9843	15,288	19,194	23,975	27,659	32,081	35,711	39,017	42,507	45,818	46,664
	Investment EUR	906,621	1,667,769	2,330,907	3,156,673	3,954,012	4,497,988	4,584,381	5,188,255	5,662,405	6,219,225	6,798,857	6,965,418
	Cost savings EUR/yr	44,600	89,444	143,552	210,584	267,349	311,169	332,393	388,297	416,043	446,586	494,087	501,459
	Avoided CO2 kg/yr	124,907	280,152	487,820	761,781	977,972	1,147,303	1,234,407	1,456,679	1,549,778	1,665,667	1,861,949	1,892,579
	Energy savings kWh/yr	413,515	896,324	1,532,499	2,339,240	2,992,188	3,482,080	3,738,528	4,366,119	4,661,235	5,010,610	5,569,345	5,656,922
DEEP RETROFIT PROJECTS PER YEAR		PBT6—Music School (Building 308); PBT6—School of Fine Arts (Building 301)	PBT6—2nd Vocational Training Institute of Heraklion (Building 303); PBT10—Hersonissos Port Junior and Senior High School	PBT11—Mochos High School; PBT3—Former Courthouse; PBT10—Hersonissos Port Primary School	PBT6—Gouves senior high School; PBT2—Municipal Store of Hersonissos Port; PBT14—Cultural Center of Hersonissos Port	PBT7—K/garten and Primary School of Gonies; PBT6—Mochos Primary School; PBT10—Episkopi High School; PBT2—Municipal Store of Malia	PBT10—Elementary School of Hersonissos Old village; PBT6—1st Primary School of Malia (Building 1); PBT1—Building 202	PBT12—New School Complex of Primary Education in Kokkini-Chani	PBT1—City Hall; PBT1—Building 154; PBT10—Malia High School -Building A; PBT6-Gouves Junior High School (156 and 158)	PBT11—Malia High School -Building B; PBT6—Gouves Primary School; PBT13—Mochos Kindergarten; PBT6—Elia Primary School; PBT10—Malia Kindergarten; PBT6—Kokkini-Chani Primary School	PBT8—Music School (Building 302); PBT6—Hersonissos Port Kindergarten; PBT6—Malia Kindergarten (Old building); PBT10—1st Primary School of Malia (Building 2); PBT10—Hersonissos Primary School	PBT10—2nd Primary School of Malia; PBT1—Municipal Store of Episkopi; PBT15—Library—Venue of Multiple Use (310); PBT15—Avdou Primary School; PBT2—Former office Building of Municipal Water and Sewerage Company of Hersonissos; PBT2—Fire Brigade Building; PBT9—Episkopi Kindergarten; PBT5—Episkopi Primary School; PBT4—Computer room of Gouves High School (688)	PBT4—Gournes Secondary School; PBT4—Gournes Primary School (“Eirini”—Building 207)

. The accompanied KPIs are also presented in a graphical form in Figure 7 in their annual absolute and cumulative versions. Assuming the year 2023 as the baseline year, the implementation period lasts from 2024 to 2035. The produced 12-year renovation roadmap yields a 5657 MWh primary energy savings, an 1892 t avoidance of CO₂ emissions, and an energy-cost saving of around EUR 0.5 million, under a total investment of EUR 7 million. In agreement with the aggregated performance depicted in Figure 5, i.e., that the PBT6 and PBT10 typologies present the highest final energy saving, the tool brings forth buildings from these typologies in the early years of implementation.

4. Discussion

The present work describes a step-by-step methodology for the rapid development of reliable energy renovation plans for public building stocks. In exploiting the typology-based approach, the initial target was to determine the optimal set of building features to be used as classification criteria. To that direction, a specialized clustering algorithm was employed, which uses the typologies’ weighted average RMSE index of the primary energy consumption (available from the EPCs) as the objective function. It was concluded that the optimal set of clustering criteria for the examined stock includes the building use, construction period, the heating, and the cooling system. This result agrees with previous works focusing on Greek non-residential building stocks such as for example in ref. [8], which adopted the same building features as classification parameters, as well as with the Hellenic national typology of non-residential buildings, which is based on the building use and construction period [12]. Unlike other studies which utilize machine-learning clustering algorithms [18,19,20,21], the current study employs a straightforward procedure considering the limited magnitude of the stock (only 44 buildings) and according to specific predefined options assigned to each classification parameter (refer to Table 1). Despite the rather empirical selection of these options, they are still considered valid because they are based on the regulatory framework (e.g., the period of construction implies a certain level of insulation as per the Greek EPBD) and on the expected influence of building features’ options on building energy performance based on the national EPBD. As explained in Section 2.3, the assigned options also satisfy the retrofit-relevance principle meaning that they are compatible with the desideratum that the same energy-upgrading interventions are valid for all buildings of the same typology. Working beyond the level of buildings’ benchmarking by means of addressing the retrofit-relevance issue as well, the scientific literature confirms the presently concluded clustering criteria [24,25]. For the case-study building stock, the identified optimal set of clustering criteria (refer to Figure 2) led to 15 public building typologies (PBTs), which is perceived as adequate in the sense of balancing the reliability of building-stock energy assessments and the required computational resources for the development of the gradual renovation plan. The dominant building use is the educational one as 10 out of the 15 typologies are school buildings, followed by the office buildings numbering 3 PBTs and finally, the remaining two PBTs refer to the multi-purpose use and the medical-care use.

As explained in Section 2.4, for each selected ambassador building of each PBT, energy modeling is conducted focusing on the calculation of the energy performance for the two feasible scenarios, MEPR and nZEB, in compliance with the official Greek ERB guideline for public buildings [27]. On average, in the ambassador buildings’ sample, it was found that (refer to Figure 4) the MEPR and the nZEB scenarios led to a 41 kWh/m² and a 65.4 kWh/m² saving of the total final energy consumption, under an investment cost of 142 EUR/m² and 170 EUR/m², respectively. Concerning the dominant PBT, i.e., educational buildings, the ones constructed before 1980 (no thermal insulation) on average exhibit a 49 kWh/m² total final energy saving with an investment of 195 EUR/m² and a 62.7 kWh/m² total final energy saving with an investment of 212 EUR/m² for the MEPR and the nZEB scenario, respectively. On the other hand, regarding educational buildings constructed after 1980, they present almost half of the total final energy savings, compared to the renovation cases of buildings constructed before 1980, under an estimated investment of 74 EUR/m² and 90 EUR/m², for the MEPR and the nZEB scenario, respectively. The previous results reveal that the older educational buildings possess approximately double energy-saving potential compared to the more recently constructed ones; however, this is at the expense of a double investment cost. Concerning the non-educational buildings of the case study (offices, multi-purpose, and medical care), interestingly, they present a practically similar performance regardless of the construction period with an average energy saving ranging from 52.5 kWh/m² to 56.5 kWh/m² and from 88.7 kWh/m² to 108.7 kWh/m² in the MEPR and the nZEB scenario, respectively. The respective investment cost ranges from 150.6 EUR/m² to 162 EUR/m² and 194 EUR/m² to 214.5 EUR/m². Due to the almost continuous operation of office buildings, which stands for the majority of the non-educational ones, they do possess a considerably higher averaged energy-saving potential, especially in relation to educational buildings constructed after 1980. Similar investment costs for common MEPR and nZEB scenarios may be also found in the Greek long term renovation strategy (LTRS) [32] as well as in other research studies [33,34].

As far as the whole-building stock KPI database is concerned, it was produced using the IMPULSE Interreg-MED project tool “KPIs’ processor”. The aggregated results for each PBT tabulated in Figure 5 suggest that the typologies PBT6 and PBT10, both referring to educational buildings, correspond to the highest final energy saving although under higher total investment cost compared to the rest PBTs. Specifically, the PBT6 exhibits a total final energy saving of 597.5 MWh/y with a total investment of 2153 kEUR and of 697.5 MWh/y with a total investment of 2310 kEUR, in the MEPR and the nZEB scenario, respectively. On the other hand, the PBT10 presents a total final energy saving of 298 MWh/y with a total investment of 1212 kEUR and of 510 MWh/y with a total investment of 1498 kEUR, in the MEPR and the nZEB scenario, respectively. It is clear that while the older buildings of PBT6 (before 1980) give remarkably higher energy saving, they still require EUR 1 million more budget in either retrofit scenario compared to the more recently constructed buildings (after 1980). It also becomes evident that although the payback period estimated for some typologies in the MEPR scenario is prohibitive (refer for example to PBT2, PBT4, PBT5, and PBT9, for which the payback period exceeds 25 years), the MEPR renovation of all buildings in one investment/project can be attenuated within 22 years. This confirms the argument that bundled buildings’ renovation projects are more cost-efficient and, in fact, align with the policy requirement to accelerate the renovation rates of the European building stock [10]. Finally, similarly to ref. [8], which also studies public buildings’ energy upgrading in Greece, it is clearly shown that the PV integration in the nZEB scenario ensures a viable renovation investment for each PBT individually due to the acceptable simple payback periods obtained, while for the whole building stock, the nZEB transformation in one investment corresponds to a simple payback period of around 14 years.

Finally, the IMPULSE tool “KPIs’ processor PLUGIN” is used for producing the ERB plan for the examined public buildings’ stock. Following the procedure of Table 2 regarding the interpretation of the authority’s policy priorities, a two-step sensitivity analysis provided the required annual renovation rate, i.e., 7% of the floor area being renovated each year and the roadmap of specific renovation projects per year accompanied with the estimated energy, environmental, and cost performance (refer to Table 7). According to the identified optimal weights of the decision-making criteria, projects focusing on the typologies PBT6 and PBT10 are positioned in the early years due to the higher bias on fossil fuel and electricity saving. As illustrated in Figure 7, by the sixth year of implementation the cumulative energy saving is higher by 1500 MWh compared to that achieved in the last 6 years, although under higher investment cost due to the very low bias (10%) imposed at that parameter in the decision-making scheme (refer to Figure 6). It is worth noting that despite the ambitious energy saving predicted for the first 6 years, it is achieved with a relatively low number of projects, which is tolerable from a managerial point of view. In relation to the ongoing institutional framework, the proposed plan leads to an avoidance of CO₂ emissions in the municipal building sector by more than 50% by the year 2030, which is compatible with the recent obligation for Greek municipalities to achieve at least 30% GHG emission reduction by the year 2030 (compared to 2019) induced by the municipal facilities as per the national climatic law L.4936/2022.

The proposed methodology, specifically from the municipal authority point of view, is considered to be very practical and beneficial as it confronts the challenges in ERB planning. First of all, it consists of freely available and ready-to-use tools (refer to the IMPULSE Interreg-MED project tools), which perform both the extrapolation of state and impact KPIs of renovation scenarios and the automatic production of the renovation roadmap outlining specific projects per year on the basis of criteria and preferences imposed by the decision maker. Secondly, provided that a valid classification of the building stock is obtained, detailed data collection and energy calculations are needed only for the buildings selected as representatives and not for each building of the entire stock individually; thus, significantly reducing the required resources, notably by about 1/3 in the present case study (15 representative buildings out of 44 of the entire stock are analyzed in detail). Conclusively, the suggested methodology relieves the concerned public authority from extreme efforts in technical and administrative management related to data collection and acquisition of detailed technical information without prejudicing the validity of decision making and the suitability of the eventual projects’ prioritization. It should be mentioned, however, that in order to conduct the building clustering, typical data of building features (refer to Appendix A) for all buildings are still required, which nevertheless remains far more tolerable than collecting detailed data and studies for all buildings of the stock.

As far as the limitations of the research are concerned, these refer mainly to the fidelity of the produced KPI database. The latter is produced on the basis of the available energy performance certificates of the buildings; thus, they are theoretically calculated and not actual data. Ideally, the building stock energy demands and predicted impacts of the proposed plan would be more accurate if actual data were available especially as regards the state KPIs. However, since the projects’ prioritization in the planning practice is mainly based on the impact KPIs, i.e., the expected difference of performance indicators before and after the foreseen renovation scenarios—provided that the current building energy performance modeling is acceptable—the produced plan is considered acceptable at least for practical engineering purposes. It is encouraging that similar energy performance indicators are found in other studies focusing on building-stock energy assessments in the Greek context [8,12]. From a managerial and economic point of view, the location of projects should be also considered in buildings’ bundling as underlined in ref. [10]. Indeed, the organization and supervision of renovation works in close locations is far more manageable and most likely less costly due to the fewer resources required and to the principles of economies of scale. In response to the aforementioned limitations, the following aspects are recommended for future work:

• Application of the suggested approach using actual energy consumption data, e.g., collected from energy bills, comparisons with the current findings, and identification of margins for improvements;
• Execute the clustering approach taking into account the location of the buildings as well as the revised unit costs of interventions in the context of economies of scale depending on the level of scattering of renovation works.

5. Conclusions

The present study proposes a bottom-up typology-based approach for the rapid development of public buildings’ stocks energy renovation plans (ERB plans) at the local level. Acknowledging the requirement for cost-effective planning activities in view of the ambitious and emergence European and national policies, the proposed methodology takes advantage of the widely accepted typology approach towards the fast production of the ERB plan without compromising the fidelity of impacts’ prediction as a result of incorporating a comprehensive clustering process. Indeed, the latter identifies the best classification corresponding to the best accuracy in comparison with available energy performance information of the whole building stock. An additional key delivery of the suggested approach is that the required renovation rate is included in the planning process; hence, it allows for the decision-maker to examine alternative renovation roadmaps with respect to the new requirement for 3% annual renovation rate of the public building stock enforced at local and regional administration levels (refer to the latest energy efficiency directive 2023/1791/EU). The following conclusions are drawn by the presented investigation:

• In working beyond building-stock benchmarking towards the development of an ERB plan at the municipal level, retrofit-relevant building features should definitely be included as clustering criteria in the framework of the typology approach, for the studied case the following criteria apply: building use, period of construction, heating system, and cooling system;
• Bundled buildings’ energy upgrading is more cost-efficient than individual retrofitting projects per building;
• Deep retrofitting (nZEB) including RES-PV integration corresponds to a viable investment in terms of acceptable simple payback period for common municipal building typologies (offices and educational buildings);
• Imposing more bias on the final energy saving in the decision-making scheme leads to a more “aggressive” launching of the renovation plan by means of high CO₂ emissions reduction in the early years without necessarily having an increased number of projects;
• The proposed methodology is cost-effective as it ensures a rapid and rigorous development of the gradual renovation plan taking into account crucial policy priorities and using freely available tools found in official resources of previous research projects;
• For the case study in focus, it was found that with a 7% annual renovation rate, a 12-year renovation roadmap is produced, envisaging the nZEB transformation of the whole building stock and yielding a 5657 MWh primary energy saving, an 1892 t avoidance of CO₂ emissions, and an energy-cost saving of around EUR 0.5 million, under a total investment of EUR 7 million.

Finally, it may be concluded that the suggested methodology is general and presents a high level of replicability since the computer tools used are independent of local specificities, such as climate zone, regulatory framework, etc., serving only for statistical processing after being fed by KPIs from external resources which account for such specificities. Focusing on the practical implications, the methodology is considered beneficial for municipalities as it consists of freely available and ready-to-use Excel-based tools that serve to produce the building stock KPI database as well as the renovation roadmap by means of specific projects per year. In parallel, it significantly minimizes the workload and resources for the assessment of renovation scenarios and eventual planning of projects for the entire building stock, ensuring an acceptable level of accuracy of predictions for decision-making purposes in the framework of ERB planning. The approach requires only moderate skills and knowledge of building energy calculations while it retains tolerable management procedures under a minimum extent of data collection and processing. From the strategic point of view, the described approach facilitates decision- and policy-makers to develop the optimal renovation program and focus on suitable financial resources, taking into account the most appropriate renovation rate and the foreseen impacts regarding the techno-economic performance of the suggested projects.