Enhanced temporal prediction of electrochemical impedance spectroscopy using long short-term memory neural networks

Large-scale applications of many electrochemical energy devices, such as batteries, fuel cells, and electrolyzers, are hindered by insufficient lifetime and durability. Traditional methods for assessing device reliability are time-consuming and resource-intensive, necessitating alternative approaches. This study presents a novel prediction framework that integrates electrochemical impedance spectroscopy (EIS) with advanced long short-term memory (LSTM) neural networks, leveraging a distribution of relaxation time (DRT)-assisted approach. Compared to the previously proposed direct EIS prediction framework, the DRT-assisted prediction framework significantly enhances the accuracy of DRT predictions and ensures the physical validity of EIS forecasts. Despite the improved prediction accuracy, forecasts extending to more distant future dates increase uncertainty, highlighting a trade-off between prediction duration and reliability. Overall, this framework provides valuable insights for optimizing the performance and lifetime predictions of electrochemical devices, offering a more robust tool for future research and application. © 2024 Elsevier Ltd