Enhancing battery health estimation using model selection criteria-based genetic programming

The reliability and safety of lithium-ion batteries due to the complex interaction of degradation mechanisms lead to battery aging and faults with substantial hazards. This will increase the difficulty in precisely estimating the state of health (SOH) to ensure efficient management. To overcome SOH complexity, this work investigates the application of genetic programming (GP) to identify battery degradation and forecast SOH. GP is powerful but faces the challenges of creating accurate and robust models that can handle the nonlinear and dynamic nature by balancing model complexity. Additionally, GP's adaptability to battery usage and sensitivity to parameter selection must be carefully considered. Despite these challenges, GP can create sophisticated, data-driven models, making it a promising SOH estimation tool. Henceforth, a model selection criterion genetic programming (MSC-GP) approach has been proposed to address these issues. The investigation evaluates the effect of objective functions (OFs) on algorithm performance through rigorous key statistical metrics. Furthermore, it demonstrates the significant influence that the choice of OFs has on the model's performance, emphasizing the algorithm's potential for accurate battery health assessment. The results unequivocally show that the MSC-GP algorithm is more effective at recognizing the aging state of lithium-ion batteries compared to artificial neural network (ANN) and Gaussian progress regression (GPR). Although the initial findings are encouraging, additional research is required to tackle the multifaceted deprivation associated with accurately predicting battery life.