Multi-objective optimization of generation expansion planning considering the diffusion of renewable energy

Abstract

In today's world, energy planning plays a critical role in ensuring sustainable and efficient use of energy. The planning process is essential for maximizing the effective use of limited national resources and reducing the dependence on energy imports. It contributes to the energy security and economic stability. It is also a key factor in reducing the environmental impacts and mitigating the climate change, which has led to a notable increase in renewable energy investments over the past few years. However, as investments in renewable energy increase, the additional electricity costs are inevitably passed on to end users due to the feed-in tariffs required to support these investments. This doctoral study aims to create a roadmap for Turkey for the year 2030 by representing and simultaneously optimizing investors, central decision-makers, as and end user views in objective functions. This approach not only addresses the need for a robust energy infrastructure capable of managing the variability inherent in renewable energy production and demand fluctuations but also emphasizes the necessity of strategic investments in renewable technologies to achieve both stability and feasibility in our energy systems. This doctoral study begins by exploring financial scenario based renewable energy investment trends in Turkey for the year 2030. To address this, an optimization model was developed that minimizes the initial investment cost and the levelized cost of energy plan, while examining the annual distribution of investments. The analysis included a baseline scenario tracking investment trends from past years, an optimistic scenario with annual expenditure values 20% higher than the baseline, and a pessimistic scenario with values 20% lower. According to the results, by 2023, all scenarios indicate that the installed capacity targets for solar photovoltaic renewable energy plants will be met. For wind energy, targets will only be achieved in the optimistic scenario, while for biomass and hydroelectric power plants, targets will not be met under any scenario. Furthermore, it has been observed that in all scenarios, at least 30% of the electricity generated was from renewable sources, achieving this particular target as well. With this study, a new method has been developed to select a solution adapted from multi-criteria decision-making (MCDM) techniques upon obtaining a nondominated set of solutions which allows for the evaluation and scoring of solutions under several criteria grouped into technological, economic, environmental, and socio-political categories. Within the set of optimal solutions, the solution with the highest score is recommended to decision makers as the preferred outcome. Furthermore, the model predicts the optimal timing for future investments in offshore wind farms in Turkey, which are currently not under operation. While offering a solution from a non-dominated set of solutions using multi-criteria decision-making techniques, popular distance-based methods such as VIKOR, TOPSIS, and CODAS were applied in the study. During the application of these methods, a weakness was identified in the distance-based MCDM techniques. These methods assume that the similarity of an intuitionistic fuzzy set to the reference point increases as its distance from it decreases. However, the validity of this assumption was found to be debatable, prompting an innovative shift in the doctoral study towards developing a Hypervolume-based approach as an alternative to the traditional distance-based (geometric) MCDM methods. This method was demonstrated to be effective both as a metric for ranking intuitionistic fuzzy sets and as a robust MCDM technique, laying the groundwork for a novel approach in this domain of multi-criteria assessment of energy resources research. Subsequently, it was recognized that the issues identified with the distance-based MCDM techniques were not limited to intuitionistic fuzzy sets alone; similar challenges arise with other sets when ranking according to the distance from a negative or positive ideal solution. Instead of questioning the robustness and reliability of the methods, we prefer to interpret that different rankings can be obtained depending on whether we view the problem from a positive or negative perspective. Indeed, isn't this akin to how we often make choices or address problems in everyday life, by considering situations from various positive or negative angles which affects our decisions? Next, the research further delves into the energy market by introducing a day-ahead market model that incorporates hourly dispatch to adeptly handle the uncertainties in renewable energy production and energy demand. The initial deterministic model, accounting for hourly dispatch, integrates the perspectives of investors, end-users, and central decision-makers, with objective functions focused on investment payback period, average electricity generation cost, and total investment costs. In addressing the deterministic problem, renewable energy production and demand projections for 2030 are modeled using generative adversarial network structures, with each season distinctly represented by representative weeks. The decision variables have been selected as the investment amounts for renewable energy sources and the appropriate feed-in tariff values. The capacities of conventional resources are calculated within the day-ahead market model and are not treated as decision variables. When the calculated capacity for conventional resources surpasses the existing capacity, it is evaluated as new investment; conversely, if it is less, no new investment is deemed necessary. As a result of these methodological decisions, the complexity of the problem and, consequently, the solution time have been significantly reduced. Furthermore, a scenario-based robust optimization model is created by using the scenarios generated by generative adversarial neural networks. This model employs a robust approach to optimization, seeking to make decisions that perform well under the most adverse conditions anticipated within the defined scenarios, to calculate objective functions including the minimization of the payback period of investments, average electricity generation cost, and total investment costs. A notable distinction between the deterministic and robust results was that decisions on wind energy investments in the robust model were more conservative compared to the deterministic outcomes, while investments in solar photovoltaic facilities increased relative to the deterministic model. If installation costs for solar PV panels reach to the price levels in the Energy Information Administration's new policies scenarios by 2030, considering that the remaining power plants are already near their economically feasible limits, it appears that solar PV will be the predominantly installed power plant compared to wind. As variability in wind energy production increases with additional scenarios, the ability of thermal sources to adapt to the remaining demand becomes more challenging, which in turn impacts the unit electricity price for end-users. Consequently, in the robust model, new investments in wind energy are less favored due to this increased variability. This underscores the variability and risk management inherent in robust approaches, highlighting their potential to adapt to uncertainties in renewable energy frameworks.