Thermal Unit Maintenance Scheduling Using GA Combined with SA

抄録

This paper develops a new algorithm for the large scale and long term thermal unit maintenance scheduling problems including maintenance classes. This method is based on the genetic algorithm (GA) combined with the simulated annealing Method (SA). The method takes maintenance class into consideration and minimizes the weighted sum of costs and variance of reserve powers. The proposed method presents a new genetic operation that finds the local optimum faster than the simple genetic algorithm and introduces efficient encoding/ decoding technique. The Boltzmann's acceptance probability of simulated annealing method is included in the algorithm as a criterion for the survival of individuals during the evolution process.
The aim of this study is to reduce the computing time of simulated annealing based methods and make the solution be more accurate than that of the simple genetic algorithm. Numerical results on real scale thermal unit maintenance scheduling which covered several consecutive years are demonstrated, and the scheduling results are compared with those through the simple genetic algorithm or through the simulated annealing method.