Hybrid prediction of fuel cell remaining useful life with dynamic grey relational analysis-bidirectional long short-term memory networks and extended Kalman filter

Abstract In recent years, the prediction of fuel cell performance degradation has become a research focus. To reflect the aging characteristics of proton exchange membrane fuel cells under diverse operating conditions, a health power indicator <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" overflow="scroll"> <mml:mrow> <mml:mrow> <mml:msub> <mml:mi>S</mml:mi> <mml:mi>P</mml:mi> </mml:msub> </mml:mrow> </mml:mrow> </mml:math> , based on current and voltage, has been introduced. To address the underutilisation of operational parameters, this paper develops a multi-input model by integrating dynamic grey relational analysis (DGRA) with bidirectional long short-term memory networks (Bi-LSTM). This approach effectively selects operating parameters, utilises multiple parameters efficiently, and reduces computational complexity. Additionally, to address error accumulation in multi-step prediction, a hybrid method is presented. The method combining a semi-empirical model with an extended Kalman filter is utilised to predict the overall aging trend. And it is integrated with the DGRA-Bi-LSTM networks model to enhance the accuracy of multi-step predictions. The results demonstrate that the presented method offers high accuracy and practical applicability in multi-step prediction of fuel cell aging.