Constrained Optimization Based on Ensemble Differential Evolution and Two-Level-Based Epsilon Method

• Published in: IEEE access Vol. 8; pp. 213981 - 213997
• Main Authors: Xu, Bin, Zhang, Zonghao
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 2020; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: constrained optimization; Constraints; differential evolution; ensemble; Evolutionary algorithms; Evolutionary computation; Heuristic algorithms; Linear programming; Mutation; Optimization; Population distribution; Search algorithms; Search problems; Search process; Sociology; Statistics; twolevel-based comparison
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2020.3040647

Constrained optimization problems (COPs) are common in many fields, and the search algorithm and constraint handling technique play important roles in the constrained evolution algorithms. In this article, we propose a new optimization algorithm named CETDE based on ensemble differential evolution (DE) and a two-level epsilon-constrained method. In the ensemble DE variant, some promising parameters and mutation strategies constitute the candidate pool, and each element in the pool coexists throughout the search process and competes to generate new solutions. The two-level epsilon method is proposed by incorporating a generation and a population comparison level to retain more promising solutions without degrading the solution quality. Moreover, a diversity promotion scheme is developed to improve the population distribution when the search becomes trapped in a small region. The superior performance of CETDE is validated by comparison with some state-of-the-art COEAs over two sets of artificial benchmarks and five real-world problems. The competitive results show that CETDE is an effective method for solving COPs.