Working zone for a least-squares support vector machine for modeling polymer electrolyte fuel cell voltage. (1st February 2021)

• Record Type: Journal Article
• Authors: Zou, Wei; Froning, Dieter; Shi, Yan; Lehnert, Werner
• Is Part Of: Applied energy. Volume 283(2021)
• Publication Date: 2021-02-01
• Subjects: Polymer electrolyte fuel cells -- Data-driven method -- Least squares support vector machine -- Statistical noise -- Model suitability; Power (Mechanics) -- Periodicals; Energy conservation -- Periodicals; Energy conversion -- Periodicals; 621.042
• DOI: 10.1016/j.apenergy.2020.116191
• Languages: English
• ISSNs: 0306-2619

Abstract

Highlights: A fast responsive and accurate model for fuel cell dynamic is presented. Statistical analyses of voltage oscillation were carried out. Effects of the system's pre-setting parameters were intensively investigated. Effects of the load variations on the model's responses were thoroughly discussed. A working zone regulating the system's pre-setting and load variations is presented. Abstract: The least squares support vector machine method has been successfully applied to modeling the transient behavior of polymer electrolyte fuel cells; this paper analyzes the credibility and definition of its reliable working zone when dealing with multiple load changes. The transient model based on the least squares support vector machine is initially established. Then, the effects of the fuel cell system's setup and exterior load behavior on the transient model are investigated. Artificial data from experimentally-validated Simulink simulations are used, by which extreme working conditions could be taken into account. We found that the fuel cell system's setup with intensive sampling brings about better model performance than that with a sparse sampling interval, as sharp peaks are well characterized when intensive sampling is applied and more information on the fuel cell system is provided to the transient model. Furthermore, the performance of the transient model is better when smoother load changes are imposed on the system, and so a large ramp time and small ramp value are preferable. A working zone for a least squares support vector machine to model polymer electrolyte fuel cell is defined, for which an absolute error is used. Based on the acceptable level of error in the fuel cell system, a set of feasible combinations of its setup and exterior load changes is regulated. Accuracy in the transient model is achieved when the fuel cell runs within the working domain.