Performance Prediction And Optimization Of Proton Exchange Membrane Fuel Cells Based On Data-driven Models

The Proton Exchange Membrane Fuel Cell(PEMFC)is characterized by its low start-up temperature,high energy utilization efficiency,and environmentally friendly properties,making it highly suitable for application in the transportation sector.However,the performance of the PEMFC remains a key issue that hinders its development.Existing research has shown that many operational conditions greatly affect the performance of the PEMFC.To explore the impact of different operational conditions on the performance of the fuel cell stack,a commonly used method is to obtain a large amount of data through experiments,and then establish a PEMFC model based on this data to complete performance research.However,the traditional mechanism model of the fuel cell stack is complex and cumbersome due to the many physical and chemical processes involved within the stack.In recent years,data-driven models have been widely used in modeling and performance optimization of PEMFC,due to their high accuracy and adaptability.They are often combined with optimization algorithms to complete the modeling and performance optimization work of PEMFC data-driven models.In this thesis,a PEMFC data-driven model and performance optimization framework based on neural networks and optimization algorithms is constructed to complete the research related to PEMFC performance prediction and optimization.The detailed research content is introduced as follows:(1)An orthogonal experiment is designed and data analysis methods are applied to investigate the effects of various operating conditions on the performance of PEMFC.A PEMFC testing platform is established,including a testing bench and testing software.Based on the performance characteristics of PEMFC and the functions of the testing platform,the operating conditions and performance indicators to be studied are determined.Orthogonal experiments are designed and completed at multiple current density points on the testing platform.The experimental data are analyzed using range analysis and multi-objective variance analysis.The orthogonal experiment data will be used in the subsequent data-driven modeling process.(2)A voltage data-driven model is constructed and optimized using radial basis function neural network and improved grey wolf optimization algorithm,and singleobjective performance optimization of PEMFC is completed.To better solve the singleobjective optimization problem,various improvement methods are proposed for the grey wolf optimization algorithm.The voltage data-driven model is constructed using radial basis function neural network,and the model parameters are optimized based on the improved grey wolf optimization algorithm.The training effect and generalization of the voltage data-driven model are evaluated using the orthogonal experiment dataset and validation experiment dataset.The single-objective performance optimization of PEMFC is completed using the voltage data-driven model and the improved grey wolf optimization algorithm,and the effectiveness of the optimization method is demonstrated by comparing the results.(3)A multi-objective optimization framework is constructed using data-driven models and the multi-objective grey wolf optimization algorithm,upon which the multiobjective performance optimization and software design implementation of PEMFC are completed.Multiple performance optimization indicators with research significance are selected and defined.A power consumption data-driven model is constructed to replace its mechanistic model,achieving a significant improvement in computational efficiency.By coupling the data-driven model and the multi-objective grey wolf optimization algorithm,a multi-objective optimization framework is established.The PEMFC multiobjective performance optimization is completed using the multi-objective optimization framework,and the optimization results are analyzed.A PEMFC performance prediction and optimization software is developed based on the multi-objective optimization framework.Ultimately achieving convenience,visualization and systematization of the technical roadmap presented in this thesis.