Data-Driven Performance Degradation Prediction of PEM Fuel Cell using Bi-GRU

Proton exchange membrane fuel cell (PEMFC) is recognized as one of the clean and sustainable energy solutions for industrial, residential and commercial applications. However, reliable and and accurate prediction of PEMFC health degradation is vital for ensuring their efficient and reliable operation. The performance degradation prediction of PEMFC can be helpful in increased durability, effective utilization and contributes in risk management and maintenance. Deep learning techniques are identified as powerful techniques for prediction as they can learn about features, especially if the data involved in the prediction process is complex and large. Therefore, this study evaluates a data driven performance degradation prediction of a PEMFC using Bi-directional gated recurrent unit (Bi-GRU). Results obtained using Bi-GRU are compared with other approaches, such as,Recurrent Neural Network (RNN), Long Short Term Memory (LSTM), Bi-LSTM, Gated Recurrent Network (GRU). R2 score and other performance indices such as MSE, MAE, MAPE, and RMSE are used to verify the results. From the results, it is elucidated that Bi-GRU predicts PEMFC degradation with the highest R2 score of 0.9993, and the outcome values of MSE, MAE, MAPE, and RMSE are 0.0006, 0.0176, 0.0498, and 0.0255, respectively.