MACHINE LEARNING MODELS FOR PREDICTING ELECTRICAL AND THERMAL BEHAVIOR IN HIGH-CAPACITY LITHIUM-ION BATTERIES

The rising prominence of EVs has substantially increased the demand for high-capacity, rechargeable Li-ion batteries, necessitating advanced BMS to ensure reliable and optimized operational performance. Central to BMS functionality is the precise modeling of Li-ion battery behavior, particularly their thermal and electrical dynamics. Given the substantial variance in the nominal capacity of Li-ion pouch cells used in commercial EV battery packs, accurate estimation of heat generation in these cells is crucial for designing an effective battery thermal management system. Traditional BMS applications have typically relied on simplistic equivalent circuit models, which often fail to capture complex thermal behaviors. The advent of cloud-based BMS platforms has introduced the possibility of leveraging machine learning (ML) models, promising enhanced accuracy and reliability. This study examines the efficacy of four distinct sequential cascaded multi-stage ML regression models in predicting the depth of discharge and heat generation responses of Li-ion cells. Experimental trials were conducted on a prismatic 20-Ah Li-ion battery, subjected to cyclic loading under varying ambient temperatures and C-rates. Heat flux and surface temperature readings were diligently gathered, forming the cornerstone for model development and validation. Utilizing Python's Scikit-learn library, different models were trained and evaluated against the experimental dataset, with the coefficient of determination (R²) employed as the performance metric. Among these models, the random forest-based approach demonstrated exceptional proficiency, achieving an R2 score of 0.9994. This result underscores its superiority in accurately capturing the complex thermal and electrical dynamics of Li-ion batteries under diverse operating conditions. The findings highlight the potential of ML-driven approaches to transform BMS design, enabling a more nuanced and accurate understanding of battery behavior, thereby paving the way for safer and more efficient EV battery systems.