Evolutionary algorithms for switch allocation in power distribution networks with distributed generation

Abstract

Switch allocation in power distribution networks is a complex combinatorial problem that impacts both operational costs and system reliability. This paper addresses the Switch Allocation Problem (SAP) under the presence of Distributed Generation (DG), using Genetic Algorithms (GA) and Memetic Algorithms (MA). The proposed methods consider different types of switches (manual and remotely controlled) and optimize their number and placement without predefined limits. Extensive experiments on three real-world Brazilian networks, ranging from 2895 to 5523 nodes, demonstrate cost reductions of up to 61% (GA) and 57% (MA) over the base case. Additionally, the MA consistently outperforms the GA, achieving further cost savings of up to 4.9%. The integration of DG contributed to an additional 3.6% reduction in total costs, while ensuring regulatory compliance for reliability indices such as SAIDI. These results confirm the effectiveness of evolutionary algorithms in optimizing switch allocation for modern smart grids.