A Novel Hybrid Method Based on the Sliding Window Method for the Estimation of the State of Health of the Proton Exchange Membrane Fuel Cell

To study the state of health (SOH) of the proton exchange membrane fuel cell (PEMFC), a novel hybrid method combining the advantages of both the model-based and data-driven methods is proposed. Firstly, the model-based method is proposed based on the voltage degradation model to estimate the variation trend, and three parameters reflecting the performance degradation are selected. Secondly, the data-driven (long short-term memory (LSTM)) method is presented to estimate the variation fluctuation. Moreover, the core step of the hybrid method is returning the results of the LSTM method to the power degradation model as the “observation” and modifying related parameters to improve the estimation accuracy. Finally, the sliding window method is applied to solve the problem of the data increase with the increase of the operating time. The results show that the power estimation is better than the current estimation for the SOH estimation. The estimation accuracy of the hybrid method dependent on the model accuracy, the amount of experimental data, and the data preprocessing is higher than that of the model-based method. The power estimation accuracy by the hybrid method of the estimation window length 5 h, 10 h, and 20 h are 99.2%, 98.68%, and 96.87%, respectively.