A Fusion Intelligent Degradation Interval Prediction Method Based on Hybrid Health Indicator for Proton Exchange Membrane Fuel Cells

Abstract

The degradation state is a critical specification reflecting the health of proton exchange membrane fuel cells (PEMFC). Uncertainty factors associated with PEMFC make traditional data-driven method difficult to accurately predict the decay behavior of fuel cells. To address the problem, a fusion interval prediction method based on hybrid health indicators and Quantile Regression (QR)-Bidirectional Gate Recurrent Unit (BiGRU) is proposed to quantify the uncertainty of PEMFC decay. First, a hybrid health indicator is developed to characterize the decay behavior of PEMFCs under static and dynamic operating conditions. Secondly, a QR-BiGRU interval prediction model is proposed to forecast the fluctuation range of PEMFC degradation through constructing intervals of different quantities. Third, the sparrow search algorithm is utilized to search for the optimal hyperparameters of the QR-BiGRU and to reduce interval prediction model complexity. Finally, the upper and lower bounds of the prediction intervals are obtained by linearly weighting the combination of the predicted values of different QRs to realize the multi-step interval prediction of PEMFC decay. The performance of the fusion interval prediction method for quantifying decay uncertainty is verified based on real PEMFC durability experimental data. The prediction error of the fusion interval prediction method static and dynamic operating conditions is only 0.13% and 0.18%. Compared with three popular prediction methods, the fusion interval prediction method can accurately quantify the uncertainty of degradation, which is important for the energy management of PEMFC.