A Decision Framework for Public-Private Partnership Selection in Sustainable Energy: Integration AROMAN and C5.0 Algorithm

AROMAN was chosen for the performance ranking phase for two specific reasons that are important for this study. First, its dual normalization process (combining linear and vector normalization) is particularly robust for handling decision matrices with criteria measured on very different scales, such as V6: Total Investment (million dollars) and V3: Political Risk (an index value between 0-1). This ensures that the outcome is not disproportionately affected by the scale of a single criterion [19, 20]. Second, it achieves this robust ranking with fewer computational steps compared to some other MCDM methods, making it an efficient tool for analyzing the 127 projects in our dataset [21].

Upon examining the latest literature, AROMAN has been effectively applied to rank various options, demonstrating its versatility across different domains, such as evaluating digital economic development [22], and assessing countries based on the macroeconomic performance indicators [23]. Table 2 outlines the procedure of the AROMAN method:

As shown in Table 2, before starting the decision-making process, a decision matrix should be created using existing input data. The second stage of the method involves normalizing the input data. During this stage, the data are scaled to a range from 0 to 1 to ensure comparability. Normalization is performed using two different methods: first, linear normalization, and then, vector-based normalization. After the normalization process, which is performed in two stages, the total average normalized matrix ($\mathrm{t}_{\mathrm{ij}}{ }^{\text {norm}}$) values are calculated using Eq. (3). Here, $\mathrm{t}_{\mathrm{ij}}{ }^{\text {norm}}$ represents the combined and averaged normalized matrix, and the parameter $\beta$ represents a weighting factor varying between 0 and 1. Multiplying each element in the total average normalized decision matrix by the corresponding weight values results in the creation of the weighted decision matrix. In the fifth step, the weighted normalized values for benefit and cost criteria are gathered using Eq. (5). The $\lambda$ parameter in the formula indicates the degree of influence of the criterion type.

Table 2. Procedure of the AROMAN method

There is no single PPP model that can satisfy all requirements regarding a project's technical, financial, and locational conditions. The most suitable model should be selected considering the country's political and legal position, PPP market's development level, financial and technical features of the project, and sectors in question [25]. In a PPP model, the private sector takes risks and assumes the uncertainty of the profit-oriented return of the investment, while the public has concerns such as compliance with complex legislation, regulation and sharing of authority, political ideas, the democratic decision-making process, minimizing risk, and maximizing social benefit. These projects are used to benefit from the expertise and financing capacity of the private sector, without jeopardizing the quality of the services provided, amidst increasing financial constraints.

Natural gas and coal sources were added into the analysis to examine non-renewable technology. Türkiye's non-renewable energy resources are limited and inadequate; energy imports take the highest share of total imports. In non-renewable technologies, capacity and foreign investor participation have been identified as the determining factors in the selection of. Below 775 MW capacities, a Merchant model is suitable if there is no foreign investor, while BOT is preferred if there is a foreign investor. In capacities above 775 MW, the BOO model is effective. While the BOO model is applied in cases where regulatory quality risk is high or low, the Merchant model is suitable in cases of medium risk. This situation shows that in large-scale and risky projects; it is preferred to keep the ownership and operating responsibility under a single entity for the long term. The participation of foreign investors in projects creates an important dynamic in the energy sector in Türkiye.

Türkiye's long coastline and sea areas, especially in the Aegean and Mediterranean regions, have high wind potential. These regions can contribute to Türkiye's clean energy goals by providing favorable conditions for offshore wind renewable technology. The total investment amount in wind energy projects is a determining factor in selecting PPP models. The BROT model is preferred for investments below 16.70 million US dollars, and the BOO model is preferred for those above 16.70 million US dollars. The financial structure of wind energy projects plays a critical role in investors' decision-making processes. High initial costs and long payback periods are among the factors that affect investors' risk assessments. In this context, real option models used in wind energy investments help investors make more effective decisions under uncertainties [26]. In particular, the economic feasibility of wind energy projects increases the interest of investors in these projects, which contributes to the growth of the wind energy sector [27]. This situation shows that due to the high investment cost in large-scale wind energy projects, it is preferable that ownership and operating responsibilities remain with the investor in the long term.

Inflation GDP deflator, capacity, and CO₂ emission levels have a complex relationship that affects the selection of hydropower renewable technology. While the BOO model was preferred in low inflation periods (below 6.28%) and low capacities (below 47.40 MW), in contrast, the Merchant model was applied at high capacities. In high inflation periods (above 6.28%), different models are applied according to emission levels and capacity. While BROT and BOT models were applied at low CO₂ emission levels (below 355.69 tonnes and 301.44 tonnes) and low capacities (below 15 MW and 33 MW), respectively, the BOO model was preferred at high capacities and high CO₂ emission levels. Environmental impacts of hydroelectric projects, economic conditions, and perceptions of local people are also effective in model selection. This shows that environmental impacts and economic conditions of hydroelectric projects play an important role in model selection. Policymakers should be sensitive to cultural variation for creating green infrastructure PPPs [28].

In the study, other renewable technologies included waste, biomass, solar, geothermal and biogas alternatives. In other renewable technologies, GDP deflator and capacity were the main factors affecting model selection. In cases where the inflation GDP deflator was below 12.50%, the Merchant model was considered appropriate for capacities below 55 MW, and the BROT model was considered appropriate for capacities above 55 MW. In high inflation periods (above 12.50%), the BOO model is preferred. In high inflation periods, the preference for the BOO model stems from investors' desire to retain ownership and operational responsibility. Since high inflation increases investment risk, investors generally prefer less risky and more controllable models.

A central and perhaps contrary to common sense finding of this study is the high-performance ranking of non-renewable energy projects, despite the overall investment trend toward renewable energies. As seen in the AROMAN results (Table 3), the best-performing projects under the BOO and BOT models are predominantly based on non-renewable sources such as natural gas and coal. This raises a critical governance question: Does Türkiye’s PPP framework implicitly reward financial bank lending eligibility over environmental sustainability, as observed in historical data? The analysis suggests this may be the case for several reasons. First, non-renewable projects, particularly large-scale thermal power plants, have historically offered more predictable and stable cash flows, making them highly attractive to private investors and lenders. Their established technologies and continuous operation (unlike intermittent renewable energies) reduce operational and market risks, making them more bankable. The high rankings of projects like InterGen Gebze Adapazarı İzmir and Enerjisa Tufanbeyli Coal Plant, known for their efficiency and reliability, support this view.

Secondly, these projects often benefit from strong state support mechanisms or strategic importance, further mitigating investment risk. For example, the success of the Zonguldak Coal-Fired Power Plant is dependent on the use of domestic coal reserves in line with national energy import dependency goals. This highlights a potential conflict within public policy itself: the goal of energy security through domestic fossil fuels may conflict with the goal of environmental sustainability.

This finding suggests that the governance framework may need to be adjusted for Türkiye to successfully steer its PPP program toward achieving sustainable development goals. Simply attracting private investment is not enough; the framework must be designed to make sustainable projects more or equally financeable than their fossil fuel counterparts. This could include establishing more robust and long-term price guarantees for renewable energies, developing risk mitigation tools specifically for green projects, or giving higher weight to sustainability criteria in the project selection and tendering process.

The sample included 127 projects in the analysis: sixty were BOO (47%), fifteen were BROT (12%), twenty-five were BOT (20%), and twenty-seven were Merchants (21%). In all PPP models, the investment preference focused on generating energy based on sustainable technology. In sustainable technology preferences, the intensity was also wind energy and hydropower plants. BOO and BROT models were prominent in the study of wind energy. In hydro energy, the highest intensity was BOT, Merchant, BROT, and BOO models. The prominent models in energy project investments with unsustainable technology content were the BOT (20%), Merchant (22%), and BOO (10%) models. AROMAN analysis’ results can be discussed for PPP models in the following:

When the density distribution of sixty BOO model project was examined, forty-two (70%) were wind energy projects, ten (17%) were hydropower plants, three were coal (6%), two were natural gas (4%), one was solar (1%), one was geothermal (1%), and one was biomass (1%) energy project. In other words, 90% of BOO projects produce electricity with sustainable technology while 10% produce electricity with unsustainable technology. According to the order of success, the first three projects were BOO projects using unsustainable technology. Only one in the first five projects was using sustainable technology, four of the projects in the first ten were using unsustainable technology, and six of them were using sustainable technology. Two of the first three projects had foreign partners. Only one of the projects in the first ten was domestic, and two were 100% foreign investments. The project, which ranked first, achieved a 99.8% availability rate at the Adapazarı plant, which was seven points above the industry average and a world record for Class F gas turbine technology. Taking this performance into account, Power Magazine selected the Adapazarı plant as the power plant of the year among the world's natural gas power plants in 2011. The main source of the project, which ranks second, was imported coal. Over time, the project integrated sustainable technologies into its structure. The project, which ranks third, was Türkiye's first and only domestic coal power plant. Zonguldak is Türkiye's coal city. It is distinguished from its counterparts by being in a location with the richest coal reserves in Türkiye, such as Zonguldak, and by its location on the seashore. Thermal power plants contribute to the balance of production by providing energy in every month of the year. Since it provides electricity production according to renewable resources, it forms the backbone of the electricity system [29]. Therefore, although 90% of the BOO projects included in the analysis are based on sustainable technology, the top three in terms of success were projects that generate electricity from coal and natural gas-based power plants utilizing the country's unsustainable underground resources and were foreign capital-intensive.

Fifteen BROT model projects were used in the analysis. When the density distribution of the fifteen BROT projects was examined, six (40%) were wind energy projects, four (27%) were hydro power plants, four (27%) were geothermal, and one was natural gas (6%). In other words, 94% of the BROT projects generate electricity with sustainable technology, while 6% produce electricity with unsustainable technology. According to the ranking of success, only one project, which produces electricity with unsustainable technology (natural gas), with a share of 6% in the total project, was in the first place. All the remaining systems generate electricity with sustainable technology. While 53% of the projects were domestic projects, 47% consisted of projects in which foreign capital was a partner or entirely foreign capital. The first project is Türkiye’s first natural gas combined cycle power plant. The second project is located on Yeşilırmak in Amasya. The third project is located on the Karasu/Fırat River in Erzincan. It is a local project. IC Enterra Renewable Energy, which began its work to enhance efficiency at the Bağıştaş-1 Dam and Hydroelectric Power Plant (HES), put into operation in 2015 in Erzincan, achieved a 25% reduction in domestic consumption, thereby increasing production efficiency. With steps such as the use of efficient equipment, LED conversion project in lighting, and transition to automation, the domestic consumption of the plant was reduced by 25%, and a significant efficiency increase was achieved in production [30].

Twenty-five BOT model projects were used in the analysis. When the density distribution in terms of project type of the twenty-five BOT projects was examined, seventeen (68%) were hydropower projects, two (8%) were wind energy, one was waste (4%), four were natural gas (16%), and one was coal (4%). In other words, 80% of the BOT projects produce electricity with sustainable technology, while 20% produce electricity with unsustainable technology. 52% of the projects were foreign-capital investments and 48% were domestic investments. In the ranking of success, it was observed that all the first thirteen projects were investments with foreign partners. One of the most important advantages of using this model is that advanced technology will enter the country if foreign investors realize projects requiring advanced technology. Through the model, the government will reduce the financial pressure required for realizing the infrastructure investments, and this will have positive effects on the budget. In total, 80% of BOT projects were based on sustainable technology, while two of the top three projects (67%) and four of the top five (80%) projects in the ranking of success, use unsustainable technology. Many developing economies are still dependent on fossil fuels, especially coal and oil, and more investment in renewable energy sources is needed to reach net zero emissions globally. In Türkiye, the share of fossil energy sources in energy generation is high, which contributes to economic growth. Fossil energy sources cause not only damage to the environment but also an increase in energy costs due to imports from abroad. In this context, it is important for Türkiye to choose renewable energy to achieve sustainable growth [31]. The first project was located in the Birecik district of Şanlıurfa on the Euphrates River. It is the 6th largest hydroelectric power plant in Türkiye. The second project is award-winning. In 2014, the power plant was awarded first place in the category of natural-gas-fired thermal power plants at the ICCI Energy Awards. In 2015, it received an award in the same category. The power plant has also won environmental awards. The Turkish Healthy Cities Association presented it with its Environmentally Friendly Facility Award in 2014. The third project generates electricity with coal. It is Türkiye's 16th largest lignite thermal power plant. Local coal is used in the facility.

Twenty-seven Merchant model projects were used in the analysis. When the density distribution of the twenty-seven Merchant projects was examined, nine (33%) were hydro projects, six (22%) were natural gas, seven (26%) were wind, two were geothermal (7%), one was solar (4%), one was biogas (4%), and one was biomass (4%) projects. In other words, 78% of Merchant projects produce electricity with sustainable technology, while 22% generate electricity with unsustainable technology. In the success ranking, two of the top three projects and three of the top five projects use sustainable technology. While 67% of the projects were domestic capital, 33% were financed by foreign investors. Although the number of projects involving foreign investors was low, the top five projects in the success ranking were foreign capital projects. Also, two of the projects in the first three were established as hydropower plants.