A Cloud-Assisted Framework for Battery Fault Diagnosis and State Estimation via ThingSpeak IoT

Abstract

Lithium-ion battery packs in electric vehicles and portable energy storage systems are susceptible to thermal accumulation during long-term charge-discharge cycling, which can lead to performance degradation, rapid capacity fade, and safety hazards. In this paper, a low-cost digital twin framework is proposed, and it consists of IoT-based real-time monitoring, active thermal management, lightweight state estimation, and rule-based fault detection including overvoltage, undervoltage, overcurrent, and overtemperature conditions in a unified ESP32-based embedded architecture. Terminal voltage, current and surface temperature are sampled at 15 second intervals using a resistive divider, Hall-effect transducer and thermistor respectively and streamed to a cloudhosted ThingSpeak dashboard. Thermal regulation is handled by a hysteresis control policy where dual DC fan assemblies are activated at 40 °C and deactivated below 35 °C to avoid rapid relay switching and long-term thermal stress.A MATLAB/Simulink digital twin, based on a Thevenin equivalent circuit, replicates electrochemical battery dynamics in parallel. To avoid the overhead of the recurrent neural network approaches, three formula-based estimators are used: Coulomb-counting state-of-charge, impedance-referenced state-of-health and cycle-regression-based remaining useful life prediction. The experimental validation for ten chargedischarge cycles shows a peak temperature reduction of 9.2 °C, a state-of-charge error below 5%, a state-of-health error below 8%, and a remaining useful life prediction error within ±20 cycles.