A Novel Diagnosis Method of Proton Exchange Membrane Fuel Cells Based on Multi-Grained Cascade Forest and Principal Component Analysis

Fuel cell diagnosis is very important to ensure the reliability of its operation and application. The data-driven method is concerned for its simplicity and accuracy. This paper proposes a fuel cell fault diagnosis method based on multi-Grained Cascade Forest (gcForest) and principal component analysis (PCA). This method uses PCA to reduce the dimensionality of the fault data and extract appropriate features. Based on relatively simplified features, the classification algorithm of gcForest is used to diagnose the fault status of the fuel cell. Through experimental analysis, this proposed method can quickly identify the three health states of membrane drying, hydrogen leakage, and normal state. The diagnostic accuracy of this method is 99.39%, and the diagnosis period is 0.372s. Therefore, the method proposed in this paper is suitable for online fault identification of proton exchange membrane fuel cell systems with large data samples and multi-dimensional data.