State-of-Health Estimation of Li-Ion Batteries Using Multiple Linear Regression and Optimized Feature Extraction Based on Principal Component Analysis

Abstract

The State-of-Health (SoH) of a battery is an important attribute that needs to be monitored in a Battery Management System (BMS). The Indirect Health Indicators (IHIs) based SoH estimation is most predominant for online applications. In this paper, Indirect Health Indicators (IHIs), which reflect the battery capacity degradation are extracted from the terminal voltage and current during the process of charging and discharging of the batteries till the End of Life (EOL). Further, using the Principal Component Analysis (PCA) technique, the optimized features are selected from all the extracted IHIs. These optimized features are selected as the inputs to the Multiple Linear Regression (MLR) for the final SoH estimation. The proposed work has been validated with two datasets, where extensive validation has been performed considering extracted features (Without PCA) and transformed features (With PCA). The result shows that the Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) are 0.00144 and 0.00277 respectively. Also, the training execution time has been reduced by 1.13ms.