Cellular estimation of distribution algorithm designed to solve the energy resource management problem under uncertainty

Abstract

The Energy Resource Management (ERM) can be modeled as a Mixed-Integer Non-Linear Problem whose aim is to maximize profits generally using smart grid capabilities more than importing energy from external markets. Due to this, many resources and customers are involved in optimization, making ERM a complex problem. Moreover, when the inherent uncertainty of weather conditions, load forecast, electric vehicles planned trips, or market prices is considered, deterministic approaches might fail in obtaining optimal solutions to the problem. In this context, evolutionary algorithms are a useful tool to find effective near-optimal solutions. In fact, to design and test evolutionary algorithms to solve the ERM problem under uncertainty, the research community has developed a simulation framework. In this paper, we propose the Cellular Univariate Marginal Distribution Algorithm with Normal-Cauchy distribution (CUMDANCauchy) to address the ERM problem in uncertain environments. CUMDANCauchy uses a univariate estimation of the product of Normal and Cauchy distributions over each feature, and produces new individuals not only by the sampling of the learned distributions but also using neighborhoods of individuals from a ring cellular structure. The experiments performed over two case studies show that: CUMDANCauchy is as competitive as the previous dominant class of algorithms in terms of the global fitness achieved; its convergence behavior is among the best in comparison with the other tested algorithms; its running time is similar to the algorithm with the best global fitness achieved in the first case study, and it is the fastest algorithm in the second one.