Short-term hydrothermal generation scheduling by a modified dynamic neighborhood learning based particle swarm optimization

The main objective of the short-term hydrothermal generation scheduling (SHGS) problem is to determine the optimal strategy for hydro and thermal generation in order to minimize the fuel cost of thermal plants while satisfying various operational and physical constraints. Usually, SHGS is assumed for a 1 day or a 1 week planing time horizon. It is viewed as a complex non-linear, non-convex and non-smooth optimization problem considering valve point loading (VPL) effect related to the thermal power plants, transmission loss and other constraints. In this paper, a modified dynamic neighborhood learning based particle swarm optimization (MDNLPSO) is proposed to solve the SHGS problem. In the proposed approach, the particles in swarm are grouped in a number of neighborhoods and every particle learns from any particle which exists in current neighborhood. The neighborhood memberships are changed with a refreshing operation which occurs at refreshing periods. It causes the information exchange to be made with all particles in the swarm. It is found that mentioned improvement increases both of the exploration and exploitation abilities in comparison with the conventional PSO. The presented approach is applied to three different multi-reservoir cascaded hydrothermal test systems. The results are compared with other recently proposed methods. Simulation results clearly show that the MDNLPSO method is capable of obtaining a better solution.