Algoritmo dos vaga-lumes com população não homogênea aplicado ao despacho econômico de carga

Abstract

The economic load dispatch problem is a very important practical application for power systems operation. In this problem, the output power of each generator in the system, need to be adjusted in such a way that the total generation cost be minimal, and the constraints of the problem are met. Thus, an efficient solution to this problem lead to cost and emissions reduction. The use of mathematical methods for the solution of this problem has been considered, however, these methods are not applicable for non smooth or non convex cost functions. Thereby, the use of metaheuristics for the solution of economic dispatch problems with more complex effects has been shown a more appropriated alternative. The firefly algorithm is a metaheuristic that has been quite efficient in comparative studies and applications. The performance of this algorithm has been improved by the proposal of variants. In most cases, those variants are compared with the original algorithm in a set of benchmark functions. Those functions are also known as artificial problems, and are designated the present specific kinds of issues in search methods. On the other hand, in practical applications, several kinds of issues make the search for the optimum solution harder. In this work a new variant of the firefly algorithm, with non-homogenous population, has been proposed. So, each firefly has its own parameters, and hence, a different behavior. For comparison of the original method and the new variant proposed in this work, it was selected a set of ten benchmark functions, and after, the methods are applied to several economic dispatch problems. The comparison of the algorithms in a certain test case doesn’t allow a generalization. Thus, to analyse the effects of different sizes of problems, cost models and constraints, a set of twelve test systems of the economic load dispatch is selected from literature. On the selected problems are considered the valve point effect, multiple fuels, cubic cost function, transmission loss, prohibited zones, minimum/maximum outpup power limits and ramp rate limits. The variant proposed in this work achieved better results in nine of the ten test functions considered. Yet, in the economic dispatch problem, the variant has shown a better performance in large scale systems being that in ten of the twelve considered systems, a better mean cost was achieved by the non homogenous population algorithm proposed.