TinyML-Based Battery Health Estimation for Electric Vehicles Using Kronecker-Compressed LSTM on Embedded BLE Microcontroller

Abstract

The rapid adoption of electric vehicles (EVs) has made battery health monitoring a critical aspect of safe and efficient operation. Among several diagnostic indicators, the State-of-Health (SOH) of a lithium-ion battery represents its remaining usable capacity relative to its original design. Accurate, real-time estimation of SOH can help prevent unexpected failures, improve charging efficiency, and enable predictive maintenance—key requirements for next-generation battery management systems (BMS).