A Model Transfer Framework for PEMFC Water Fault Diagnosis Based on Hybrid Transfer Learning Strategy

• Published in: 2024 IEEE 25th China Conference on System Simulation Technology and its Application (CCSSTA) pp. 595 - 599
• Main Authors: Gao, Shangrui, Sun, Zhendong, Li, Mince, Chen, Zonghai
• Format: Conference Proceeding
• Language: English
• Published: IEEE 21.07.2024
• Subjects: Accuracy; Data models; Faces; Fault diagnosis; Feature extraction; Fuel cells; model transfer; Protons; Systems simulation; Transfer learning; Water resources
• DOI: 10.1109/CCSSTA62096.2024.10691858

The proton exchange membrane fuel cell (PEMFC) is a "green" energy conversion device that is widely considered to be one of the best power sources for future electric vehicles and static energy systems. The task of fault diagnosis for PEMFC often faces the dilemma of data shortage, especially in terms of internal faults of the fuel cell. To solve these problems, We proposed a method for model transfer based on hybrid transfer learning to solve the problem of insufficient data sets in the target domain. The key point of this approach is to use the TrAdaBoost algorithm for transfer learning. In order to meet the algorithm's requirements for initial diagnosis accuracy in practical applications, we combine a fine-tuning model transfer strategy with this algorithm. Compared with other methods, this method has significant improvement in water fault diagnosis accuracy.