Capacity Prediction of VRLA Batteries Based on Stepwise Regression Analysis

Abstract

Effective maintenance of valve-regulated lead-acid (VRLA) battery groups within substations is critical for DC system reliability. Therefore, assessing battery health and remaining capacity is essential. In this study, we focus on predicting the VRLA battery group remaining capacity and propose a prediction model based on stepwise regression. This model utilizes operational data, combining the charging and discharging characteristics of VRLA batteries. Stepwise regression selects the optimal independent variable subset, mitigating poor predictions from using all variables in multiple linear regression. Experimental results validate the accurate prediction of lead-acid battery capacity decline, offering effective forecasts. This model is expected to be widely applied in the capacity prediction of DC power systems in substations.