Long-term degradation prediction for proton exchange membrane fuel cell(PEMFC)contributes to the reduction of durability test time or cost and provides a basis for maintenance strategy.Aiming at the super-parameter issue,a data-driven method combining optimization algorithm and reservoir computing is proposed.Using available durability test datasets,with stack voltage as the degradation indicator,the sparrow search algorithm(SSA)is employed to optimize the reservoir size,leakage rate and regularization coefficient of the echo state network(ESN)to construct the prediction model.Training sets comprising 30％,40％,50％and 60％of original data are used to train the model,the long-term degradation prediction performance of the model is validated under different training set proportions.When the training set proportion is 30％,the long-term prediction root mean squared error(RMSE)of the proposed method can reach 0.008 3 under the static condition and 0.035 9 under the quasi-dynamic condition.
维普
万方数据