A Fusion of SVR and Fully Connected Networks for Improved RUL Prediction of Lithium Ion Batteries

The purpose of this study is to evaluate the prediction of the Remaining Useful Life (RUL) of lithium-ion batteries using the Support Vector Regression (SVR) algorithm with a radial basis function (RBF) kernel and compare the outcomes with a combined model that combines SVR and Fully Connected Networks (FCN). A dataset comprising approximately 6,000 records of lithium-ion battery data were utilized, the analysis reveals significant insights into the effectiveness of SVR as a standalone predictive tool. The SVR model outcomes show that it can capture variance in the measurements, with a Mean Squared Error (MSE) of 0.2269 and a R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> score of 0.9531. Conversely, the FCN and SVR combined model, which employs a simple averaging technique for predictions, demonstrates marked improvements in accuracy, yielding an MSE of 0.0571,with an R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> score of 0.9882, showcasing its superior predictive performance. These results underscore the prospect of ensemble methods in strengthening prediction accuracy, positioning the combined SVR FCN model as a more effective approach for estimating the RUL of lithium-ion batteries. This research contributes to the ongoing efforts to improve predictive modeling techniques in battery life estimation, paving the path for innovations in battery management technologies and sustainability initiatives in energy storage solutions.