Paper Title: SHAP-based Explainable Model-in-the-Loop for Digital Twins in Battery Management Systems
Conference Name: The 34th Joint Conference on Communication and Information (JCCI 2024)At: Busan, South Korea
Abstract: Battery management systems have advanced to digital twin technology for more robust state estimation and monitoring. Most of these digital twin solutions use model-in-the-loop solutions backed by artificial intelligence (AI) algorithms. However, even though AI improves the predictive capabilities of digital twins, they pose a challenge, due to their opaque nature, thus necessitating the development of explainable models. This study explores the use of an explainable AI (XAI) tool, SHapley Additive exPlanations (SHAP), for explaining the predictions of a deep neural network (DNN) and a long-short-term memory (LSTM) model for state of charge (SoC) estimation. Results highlight the role of XAI in decision-making and the selection of the best model. Index Terms-digital twin, battery management systems, ex-plainable AI