Optimization techniques based on Evolutionary Algorithms for problems of energy system management

Abstract

Over the years, progress in the industry and the living standards of people have increased electricity demand. Moreover, generation from Renewable Energy Sources (RES) has been deployed to increase the sustainability of power sectors. Due to the recent developments in power systems, system operators and planners have to manage complex decision-making processes to ensure the reliability and quality of power supply at minimal costs over different time horizons. Given the complexity of the efficient management of power systems, the application of advanced optimization tools has become essential for decision-makers in their effort to make optimal decisions. In this context, this dissertation focuses on developing and evaluating optimization tools, which aim at facilitating decision-making in two important problems associated with power system management. The first one is the Generation Expansion Planning (GEP) problem, in which the optimum generating capacity additions to an energy sector over a long term planning horizon are sought to meet the anticipated increase in electricity demand. The second one, is the Short-term Generation Scheduling (STGS) problem, in which the optimal operating schedule of a given set of generators in a power system is determined to efficiently meet the expected load. A model for GEP of a semi-liberalized energy market is developed. In this model electricity producers are grouped by type of generating technology. The maximization of the power system’s Net Present Value (NPV) is the model’s objective. The evolution of the average annual System Marginal Price (SMP) and its interaction with the structure of the power sector is simulated. Moreover, energy objectives set by policy makers to promote capacity additions in RES are modelled. The optimal annual capacity orders and the load intensity factors are estimated. The model may allow for an assessment of the impact of energy policies both on the structure of the generating mix and the evolution of the SMP, assisting policy makers during indicative energy planning. A hybrid method based on Evolution Strategies and Interior Point Algorithm is developed to optimize the problem. Regarding the STGS problem, three variants are examined. The first one is the conventional form of the model. In the second one, the system’s reliability is considered. Both models attempt to minimize the operation cost in power systems comprising thermal generators. The last one is a multi-objective model, developed in this dissertation, which considers the emissions of the system as an additional objective. Specifically, it concerns power systems comprising hydro plants and RES besides thermal generators. The model may allow system operators to determine the optimal reserve capacity, considering the unavailability of the units as well as uncertainties related to load and wind power forecasting. A set of solutions is obtained which minimize both operation cost and emissions, each of which corresponds to an optimal operating schedule of the generating units of the power system. The optimization of all variants of the STGS is implemented using a real-coded Differential Evolution. Moreover, a two-step function is included to determine the operating states (on/off) of the generators. Heuristic repair mechanisms are developed and included within the optimization method to facilitate the obtainment of feasible solutions. Two techniques are proposed allowing an efficient integration of information of the Priority List within the optimization procedure. Moreover, a novel mutation operator and a local search technique are developed to enhance the performance of the single- and multi-objective Differential Evolution, respectively. The GEP model has been examined on an indicative case and several important conclusions have been reached. The proposed hybrid method has consistently obtained solutions of higher NPV values compared to other optimization methods. This, suggests that the derived capacity orders may increase the producer’s probabilities for higher yields. Regarding the examined test case, the future structure of the power sector may be affected due to the integration of energy objectives. Capacity orders in on-shore wind turbines and concentrated solar power are proposed to meet short-term energy targets. Moreover, part of the installed capacity of lignite-fired generators will be replaced by RES as a result of long-term energy objectives. Furthermore, meeting the energy targets, might result in a slight reduction of the SMP in the long run, when present values are considered. Important conclusions have been derived regarding the optimization of the STGS problem. The proposed mutation operator and the procedure for efficiently integrating information of the Priority List significantly enhance the algorithm’s performance. The method for the single objective problems has performed competitively, yielding in some cases generating schedules with approximately 0.8% lower operation cost compared to the previously best reported results, in reduced computational time. Moreover, the developed multi-objective algorithm has efficiently optimized the hydro-thermal-wind STGS model with economic/environmental objectives, deriving sets of solutions that approximate the Pareto fronts of the problem. Therefore, it may potentially assist system operators or generating companies towards efficiently scheduling their generating units. Furthermore, the results reveal that increased load forecasting error and unit’s unavailability may require the scheduling of higher reserve capacity, resulting in operating schedules with increased operation cost and emissions. This, however, might not be the case for increased wind power uncertainty as demonstrated by the results of the method. The main contributions of the thesis regarding the GEP model are the following: i) a stochastic optimization procedure without recourse is developed to determine the best capacity orders towards the optimal compliance with energy objectives, ii) a relaxation factor has been applied on the equality constraints of the energy objectives and its effect on the future generating mix and the optimization procedure is examined and iii) a hybrid algorithm based on Evolution Strategies and Interior Point Algorithm is developed to optimize the GEP model. Regarding the STGS problem, the main contributions are: i) a method based on a real-coded DE and a two-step function is proposed to optimize the examined variants of the mixed-integer STGS problem, ii) a mutation strategy and a local search technique are developed and combined with the real coded DE to enhance its performance, iii) a new multi-objective formulation of the problem considering system’s cost and emissions is developed and may assist system operators to determine the spinning reserve in power systems comprising several generating technologies.