Self-adaptive differential evolution with multiple strategies for dynamic optimization of chemical processes

Abstract

Dynamic optimization has become an increasingly important aspect of chemical processes in the past few decades. To solve such chemical dynamic optimization problems (DOPs) effectively, we put forward a modified differential evolution algorithm named XADE in this paper, which integrates the self-adaptive principle and multiple mutation strategies. In XADE, four mutation strategies with different characteristics are introduced instead of using a single strategy. Meanwhile, the mutation strategies and DE's two control parameters are gradually adjusted adaptively based on the knowledge learned from the previous searches in generating improved solutions. The advantageous performance of XADE is validated by comparisons with several state-of-the-art adaptive DE variants on 24 complex test instances. Experimental results show that XADE is an effective approach to solving global numerical optimization problems. Moreover, the effectiveness of XADE is validated by applying the approach to 4 real-world complex DOPs with different characteristic in the chemical engineering field.

Differential Evolution With Adaptive Guiding Mechanism Based on Heuristic Rules

“…In the historical and heuristic DE (HHDE) [37], three mutation strategies are employed to construct a candidate pool, and then the cumulative probability based on the historical experience of each strategy and the heuristic value based on the fitness value of each vector are combined together to select the suitable strategies for different vectors. Besides, there are other DE variants that are proposed with the ensemble of mutation strategies, such as, DE with ensemble of parameters and mutation strategies (EPSDE) [38], DE with a multi-layer competitive-cooperative (MLCC) [39], Ensemble of DE variants (EDEV) [11], DE with strategy adaptation mechanisms (SaM-DE) [10], DE with self-adaptive control parameters based on zoning evolution (ZEPDE) [40], Self-adaptive DE with multiple strategies (XADE) [41], dual-strategy DE with affinity propagation clustering (DSDE-APC) [42], DE with underestimation-based multimutation strategy (DE-UMS) [43], and so on.…”