Evolutionary algorithms-based optimal preventive maintenance scheduling of power systems generators

Abstract

This thesis addresses the optimal Generators preventive-Maintenance Scheduling (GMS) problem in electric power systems that includes several machines. This problem can be solved using a variety of ways, such as metaheuristic methods and mathematical programming. The problem is formulated as a mathematical optimization model using mathematical programming techniques, and the best solution is then found using algorithms. Simulating the maintenance schedule allows you to assess its effectiveness while modeling the equipment and its failure behavior. Metaheuristic methods entail creating maintenance schedules utilizing generalizations or subject-matter expertise. The primary objective of this thesis is to contribute to the performance improvement of a discrete evolutionary algorithm for a reliable and extremely accurate optimization of the discrete objective functions in order to address the issue of the best preventive maintenance scheduling of power systems generators. For planning the generator preventative maintenance, a modern metaheuristic algorithm named "the Discrete Mayfly Optimization (DMFO)" has been designed. This algorithm was proposed as an innovative swarm intelligence optimization algorithm in 2020, it combines the advantages of several existing optimization algorithms. his algorithm has been used in several applications including industrial optimization, ensemble forecasting system, and photovoltaic systems. A First-Bit Flip and Shift (FBFS) strategy for binary vectors, which is a process of manipulating binary vectors, has been first proposed to improve the performance of evolutionary algorithms. The FBFS strategy is a local search strategy that performs small changes to the obtained solutions to help evolutionary algorithms in local optimization and avoiding them from getting stuck in local optima. The proposed technique has been evaluated on a 21-unit test power system with a peak power load demand of 4739 MW in three cases where the total number of the workers available per week is limited. The improved algorithm showed at the end its effectiveness to find a solution for the GMS problem where the Sum of Squares of the Reserves (SSR) of generation is minimized. The results are compared to previous works that used other metaheuristic techniques in order to evaluate the performance of the proposed FBFS-DMFO algorithm and its search process in solving power system GMS problem.