Comparison between five stochastic global search algorithms for optimizing thermoelectric generator designs

Abstract

In this study, the best settings of five heuristics are determined for solving a mixed-integer non-linear multi-objective optimization problem. The algorithms treated in the article are: ant colony optimization, genetic algorithm, particle swarm optimization, differential evolution, and teaching-learning basic algorithm. The optimization problem consists in optimizing the design of a thermoelectric device, based on a model available in literature. Results showed that the inner settings can have different effects on the algorithm performance criteria depending on the algorithm. A formulation based on the weighted sum method is introduced for solving the multiobjective optimization problem with optimal settings. It was found that the five heuristic algorithms have comparable performances. Differential evolution generated the highest number of non-dominated solutions in comparison with the other algorithms.

Additional information

Funding

The authors’ work is supported by the Natural Sciences and Engineering Research Council of Canada (NSERC).