Design of PEMFC Stack Intelligent Diagnosis System Based on Improved Neural Network

To address the difficulty in obtaining fault data and improve the accuracy of fault diagnosis in the operation of proton exchange membrane fuel cells (PEMFC), this paper proposes a mechanism modeling method based on operational data. This method corrects the mechanism model of the proton exchange membrane fuel cell stack and auxiliary systems by actual operational data and uses it to simulate faults under given working conditions to obtain sample data. In addition, an improved GA-BP neural network algorithm is designed for the fault diagnosis system, which serially trains and tests the simulated fault data. Simulation results show that compared with the traditional BP neural network algorithm, the improved GA-BP neural network algorithm designed in this paper increases the minimum diagnostic accuracy of a single fault to above 93.5% and improves the average diagnostic accuracy by about 4.5%. This research method has important engineering application value.