Parameter extraction of PEMFC via Bayesian regularization neural network based meta-heuristic algorithms. (1st August 2021)

• Record Type: Journal Article
• Authors: Yang, Bo; Li, Danyang; Zeng, Chunyuan; Chen, Yijun; Guo, Zhengxun; Wang, Jingbo; Shu, Hongchun; Yu, Tao; Zhu, Jiawei
• Is Part Of: Energy. Volume 228(2021)
• Publication Date: 2021-08-01
• Subjects: PEMFC -- Hydrogen -- Parameter extraction -- Meta-heuristic algorithm -- Bayesian regularized neural network; Power resources -- Periodicals; Power (Mechanics) -- Periodicals; Energy consumption -- Periodicals; 333.7905
• DOI: 10.1016/j.energy.2021.120592
• Languages: English
• ISSNs: 0360-5442

Abstract

Abstract: It is essential to establish an accurate model for precise and reliable evaluation of the characteristics of proton exchange membrane fuel cell (PEMFC). However, the inherent multi-variable, multi-peak, and nonlinear features of PEMFC seriously increase the difficulty and complexity of its parameter extraction. Besides, noised data, which is inevitable in various operation conditions, usually hinders meta-heuristic algorithms (MhAs) to obtain high-quality PEMFC parameters. For the sake of solving these obstacles, a Bayesian regularized neural network (BRNN) based parameter extraction strategy of PEMFC is proposed. Furthermore, performance of the proposed approach is thoroughly evaluated and analyzed through a comprehensive comparison with several advanced MhAs under various operation conditions. Lastly, simulation results verified that BRNN based MhAs (BRNN-MhAs) can effectively extract the parameters of PEMFC with higher accuracy, faster speed, and enhanced stability. In particular, the accuracy of parameter extraction of PEMFC is growing by 34.18%. Highlights: An efficient BRNN-MhAs method is presented for PEMFC parameters extraction. Denoised data via BRNN can improve global exploration of MhAs. The method provides a more reliable fitness function for PEMFC modeling. The method can be applied to parameter extraction of more complex models.