Online Voltage and Degradation Value Prediction of Lead Acid Battery Using Gaussian Process Regression

Abstract

Monitoring battery voltage is important to ensure a steady supply of energy. A crucial aspect to avoid failure is estimating the voltage required by the battery load. Lead acid batteries play a vital role as engine starters when the generators are activated. The generator engine requires an adequate voltage to initiate the power generation process. This article discusses three prediction models for estimating the voltage and degradation values based on data-driven methods. The machine-learning models used were Gaussian process regression (GPR), Support Vector Regression (SVR), and Random Forest. The prediction results were compared using evaluation metrics, such as the Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and R-squared (R2). The implementation of the Internet of Things (IoT) was demonstrated to collect real-time battery data using a voltage sensor and a temperature sensor as inputs for the prediction model. According to the experiment, the Random Forest algorithm provided highly accurate predictions, with an RMSE of 0.0401, MAE of 0.0241, and R-squared of 0.9651.

Keywords

• data-driven
• real-time
• internet of things
• lead acid battery