A data-driven framework for performance prediction and parameter optimization of a proton exchange membrane fuel cell. (1st November 2022)

• Record Type: Journal Article
• Authors: Li, Hong-Wei; Qiao, Bin-Xin; Liu, Jun-Nan; Yang, Yue; Fan, Wenxuan; Lu, Guo-Long
• Is Part Of: Energy conversion and management. Volume 271(2022)
• Publication Date: 2022-11-01
• Subjects: PEMFC -- Machine learning -- Performance prediction -- Parameter optimization -- Kernel extreme learning machine; Direct energy conversion -- Periodicals; Energy storage -- Periodicals; Energy transfer -- Periodicals; Énergie -- Conversion directe -- Périodiques; Direct energy conversion; Periodicals; 621.3105
• DOI: 10.1016/j.enconman.2022.116338
• Languages: English
• ISSNs: 0196-8904

Abstract

Highlights: A data-driven framework for performance prediction and optimization is proposed. Different data-driven models are compared with the proposed model. The proposed model as a surrogate model is coupled to the optimization framework. Structure parameters and operating conditions are optimized simultaneously. Abstract: The optimization of structure and operating conditions for enhancing the performance of proton exchange membrane fuel cells have attracted much attention. High-precision modeling and optimization methods are significant factors for efficient optimization. Machine learning and intelligent algorithms have potent capabilities that make efficient and robust optimization possible. A framework is proposed in this study that combines the state-of-the-art meta -heuristic algorithm and machine learning method for the performance prediction and parameter optimization of proton exchange membrane fuel cells. Firstly, a three-dimensional model is developed as a data source for the framework. Then, a prediction model is created based on the Kernel Extreme Learning Machine and it is integrated into the Improved Gray Wolf Optimizer as the surrogate model. Different data-driven models are compared with the proposed surrogate model Finally, the optimal combination of the structural and operating parameters that maximize the power density is obtained based on the above optimization framework. The results show that the proposed model outperforms other models, where the mean absolute percentage error and coefficient of determination on the testing set are 0.68 % and 0.999, respectively. The maximum power density obtained is 0.704 W/cm- 2, with a relative error of only 0.325 % to the three-dimensional model results. This work provides a useful guide for the optimization of the fuel cells.