Advanced Method for Unit Commitment Problem Using Genetic Algorithm and Mathematical Programming

Abstract

This paper presents an advanced method for unit commitment problem using a Genetic Algorithm and the Lagrangian Relaxation Method. Lagrangian Relaxation Method efficiently produces an optimal schedule for large-scale problem with limited constraints, but some coupling constraints, for example prohibition of simultaneous unit start-up/shut-down at same plant, can't be considered. Genetic Algorithm can easily include the complicated constraints by introduction of a penalty function. This method utilizes both advantages of the Genetic Algorithm and the Lagrangian Relaxation Method. Moreover, the introduction of heuristics simplifies genetic string manipulation which improves optimization efficiency. These include addition of limitation of modifying the schedule, adaptation of fitness function, and limitation of the crossover point based on the string expression. Numerical results have shown that the method is effective in solving the practical unit commitment problem.