MACHINE LEARNING STUDY OF THERMAL MANAGEMENT OF A BATTERY PACK IN A CONVERGED CHANNEL

Battery thermal management systems (BTMS) play a role in ensuring the efficient operation of modern high-performance batteries. To ensure performance and minimize maintenance costs it is important to predict the future capacity and remaining useful life of batteries. This research introduces an approach that combines computational fluid dynamics (CFD) and machine learning techniques to design and optimize BTMS. Specifically, we focus on using Gaussian Progress Regression (GPR) along with kernel functions to train and test the BTMS models. By conducting CFD simulations we analyze the behavior of battery packs considering factors such as transverse and longitudinal distances, channel convergence angle, Reynold number, and battery pack confinement within the system. These simulations provide insights into Nusselt number and friction factor under several operating conditions. Based on these simulations, we construct a dataset containing 4000 data points which were used to train and test a machine learning technique based on GPR algorithm. Additionally, kernel functions enhance modeling capabilities by offering representations of underlying data patterns. Prediction results from the GPR algorithm along with the Exponential kernel function showed the best and most accurate results in predicting the Nusselt number and friction factor. The proposed method provides a framework for understanding battery behavior under the battery influential factors ultimately leading to improve the system's operational efficiency.