Parameter optimization of PEMFC model with improved multi-strategy adaptive differential evolution

• Authors: Wenyin Gong;Zhihua Cai
• Venue: Engineering Applications of Artificial Intelligence
• Year: 2014

Parameter optimization of proton exchange membrane fuel cell (PEMFC) model has received considerable attention recently. In order to estimate the unknown parameters of PEMFC model faster and obtain more accurate solutions, in this paper, an improved multi-strategy adaptive differential evolution (DE) is presented for the parameter optimization problems of PEMFC model. The approach is referred to as rank-MADE, for short. In rank-MADE, the multiple mutation strategies of DE are adaptively selected to avoid choosing a suitable strategy for a specific problem by trial-and-error method. Furthermore, the ranking-based vector selection technique is employed in different mutation strategies to accelerate the process of parameter optimization of PEMFC model. In order to verify the performance of rank-MADE, it is applied to estimate the parameters of the Ballard Mark V FC, the SR-12 Modular PEM Generator, the BCS 500-W stack, the Temasek FC, and the WNS-FC model. In addition, rank-MADE is compared with other advanced DE variants and other evolutionary algorithms (EAs). Experimental results show that rank-MADE is able to provide higher quality of solutions, faster convergence speed, and higher success rate compared with other DE variants. Additionally, the V-I characteristics obtained by rank-MADE agree well with the experimental data in all cases. Therefore, rank-MADE can be an effective alternative in the field of other complex parameter optimization problems of fuel cell models.