State of Health Prediction for Battery Energy Storage Systems under Random Walk Operations

This study addresses the challenge of predicting State of Health ($\mathbf{S o H}$) and capacity degradation in Battery Energy Storage Systems (BESS) that experience random walk phenomena due to frequent control algorithm adjustments. Through meticulous feature engineering on diverse operational cycle data, this research identifies and utilizes critical predictors influencing battery behavior. We employ a Gaussian Process Regression model, renowned for its effectiveness in capturing complex, nonlinear interactions characteristic of BESS under dynamic conditions, which consistently achieves accurate prediction with R-squared value of 0.99. The results affirm the potential of advanced machine learning techniques to enhance the management and reliability of BESS, particularly in applications where control algorithms frequently induce battery cycling akin to random walk, thereby complicating SoH and capacity degradation assessments.