Optimal operational strategies for a day-ahead electricity market in the presence of market power using multi-objective evolutionary algorithms.

摘要

This dissertation introduces a novel approach for optimally operating a day-ahead electricity market not only by economically dispatching the generation resources but also by minimizing the influences of market manipulation attempts by the individual generator-owning companies while ensuring that the power system constraints are not violated. Since economic operation of the market conflicts with the individual profit maximization tactics such as market manipulation by generator-owning companies, a methodology that is capable of simultaneously optimizing these two competing objectives has to be selected. Although numerous previous studies have been undertaken on the economic operation of day-ahead markets and other independent studies have been conducted on the mitigation of market power, the operation of a day-ahead electricity market considering these two conflicting objectives simultaneously has not been undertaken previously. These facts provided the incentive and the novelty for this study.; A literature survey revealed that many of the traditional solution algorithms convert multi-objective functions into either a single-objective function using weighting schemas or undertake optimization of one function at a time. Hence, these approaches do not truly optimize the multi-objectives concurrently. Due to these inherent deficiencies of the traditional algorithms, the use of alternative non-traditional solution algorithms for such problems has become popular and widely used. Of these, multi-objective evolutionary algorithms (MOEA) have received wide acceptance due to their solution quality and robustness. In the present research, three distinct algorithms were considered: a non-dominated sorting genetic algorithm II (NSGA II), a multi-objective tabu search algorithm (MOTS) and a hybrid of multi-objective tabu search and genetic algorithm (MOTS/GA). The accuracy and quality of the results from these algorithms for applications similar to the problem investigated here reinforced the selection of these algorithms. The results obtained from each of the three algorithms used in the evaluations are very comparable. Thus one could safely conclude that the results obtained are valid. Three distinct test power systems operating under different conditions were studied for evaluating the suitability of each of these algorithms. The test cases included scenarios in which the power system was unconstrained as well as constrained. Repeated simulations carried out for the same test case with varying starting points provided evidence that the algorithms and the solutions were robust.; Influences of different market concentrations on the optimal economic dispatch are evidenced by the pareto-optimal-fronts obtained for each test case studied. Results obtained from a traditional linear programming (LP) based solution algorithm that is used at present by many market operators are also presented for comparison. Very high market-concentration-indices were found for each solution from the LP algorithm. This suggests the need to use a formal method for mitigating market concentration. Operating the market at industry-recommended threshold levels of market concentration for selecting an optimal operational point is presented for all test cases studied. Given that a solution-set instead of a single operating point is found from the multi-objective optimization methods, additional flexibility to select any operational point based on the preference of those operating the market clearly is an added benefit of using multi-objective optimization methods. However, in order to help the market operator, a more logical fuzzy decision criterion was tested for selecting a suitable operating point. The results show that the optimal operating point chosen using the fuzzy decision criterion provides a higher economic benefit to the market, although at a slightly increased market concentration.; Since the main objective of this research was to simultaneously optimize