Enhanced Internal Short Circuit Diagnosis of Lithium-Ion Batteries Using an Equivalent Circuit Model and Machine Learning algorithms

Diagnosing and detecting internal short circuits in lithium-ion batteries is a key issue to prevent thermal runaway failures and ensure overall safety. This study focuses on utilizing an Equivalent Circuit Model (ECM) and machine learning techniques to detect ISC faults in lithium-ion batteries. The impact of battery aging is considered by adjusting capacity and resistance values in the battery simulation. First, an ECM of a lithium-ion battery pack is developed in the MATLAB/SIMULINK platform, incorporating faulty and normal battery cells with varying aging levels. Subsequently, the output data from the simulated model is utilized as input for the Random Forest (RF) model under different operational scenarios. Classification results of the proposed model demonstrate an impressive accuracy rate of $96 \%$. Moreover, by comparing the performance of the RF model with two other machine learning techniques, based on various matrices, a comprehensive analysis has been conducted. The RF algorithm outperformed other methods in battery fault detection, resulting in enhanced safety and reliability in energy storage systems.