Title: TriXAI twin: a threefold explainability approach for battery health monitoring in electric vehicles

Authors: P.G. Parvati; Mehbooba P. Shareef

Addresses: Department of Computer Science and Engineering, Rajagiri School of Engineering and Technology, Kerala, 682022, India ' Department of Computer Science and Engineering, Rajagiri School of Engineering and Technology, Kerala, 682022, India

Abstract: When electric vehicles are redefining modern transportation, the ability to monitor and predict battery health in real-time becomes critically important for ensuring safety, reliability and energy efficiency. This paper introduces a deep learning framework based on a long short-term memory (LSTM) for accurate prediction of state of charge (SoC) and state of health (SoH) of lithium-ion batteries using NASA battery dataset. However the opacity of deep learning(DL) models often limits their adoption in safety critical applications like battery management. To overcome the interpretability challenges of DL models, a hybrid explainability framework combining Shapley additive explanations (SHAP), local interpretable model-agnostic explanations (LIME), and layer-wise relevance propagation (LRP) is introduced, which enables both global and local feature attribution. This unified framework not only delivers high predictive accuracy but also unveils temporal and contextual significance of features driving the model's decisions, making it suitable for reliable decision making in EV enabled battery systems.

Keywords: BMS; battery management system; LSTM; long short-term memory; XAI; explainable artificial intelligence; LIME; local interpretable model-agnostic explanations; SHAP; Shapley additive explanations; LRP; layer-wise relevance propagation; DT; SoC; state of charge; SoH; state of health.

DOI: 10.1504/IJVSMT.2026.155792

International Journal of Vehicle Systems Modelling and Testing, 2026 Vol.20 No.3, pp.245 - 264

Received: 14 May 2025
Accepted: 08 Aug 2025
Published online: 14 Aug 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article