A Comparative Study of the Kalman Filter and the LSTM Network for the Remaining Useful Life Prediction of SOFC

The solid oxide fuel cell (SOFC) system is complicated because the characteristics of gas, heat, and electricity are intricately coupled. During operation system, problems such as frequent failures a decrease in stack’s performance have caused SOFC to work less well greatly shortened SOFC’s practical life. As such, it essential accurately forecast its remaining useful life (RUL) make last longer cut down on economic losses. In this study, both model-based data-driven prediction methods used predictions about RUL SOFC. First, linear degradation model established by introducing resistance index health status. Using Kalman filtering (KF) method, status evaluated online. results state estimation indicated that KF algorithm accurate enough provide good basis for prediction. Then, long short-term memory (LSTM) network-recursive (data-driven) method presented prognostics. multi-step-ahead recursive strategy updating network with actual test data improves accuracy. Finally, comparison made between LSTM approach suggested experiments indicate more suitable than algorithm.