Research Output
Estimation of Remaining Useful Life in Lithium-Ion Batteries using Bidirectional Long-Short Term Memory
Comparison of Regression Methods for Estimation of State-of-Health in Lithium-Ion Batteries
1 - 2 of 2
Publication
Lithium-ion batteries are a type of rechargeable battery known for their high energy capacity and extended lifespan. Although lithium-ion battery technology is advancing rapidly, these batteries have a limited operational lifespan and their energy storage capability decreases with time and usage. This is where Remaining Useful Life (RUL) calculations become essential for battery maintenance planning. This study aims to employ the Bidirectional Long-Short Term Memory (BiLSTM) technique to predict the RUL of Li-ion batteries and compare it with the Long-Short Term Memory (LSTM) method to determine the most effective approach. The training data included batteries B0005, B0006, B0007, B0018, B0025, B0026, B0027, B0028, and B0055 for experiments. The BiLSTM approach consistently outperformed the LSTM method for each battery. The best results were achieved with battery B0005 using BiLSTM, with RMSE, MSE, MAE, and MAPE values of 0.01612, 0.00026, 0.00971, and 0.00684, respectively, indicating that the BiLSTM method is capable of accurately estimating the RUL of lithium-ion batteries.
Publication
Lithium-ion batteries are a type of rechargeable battery with a high energy density and an extended cycle life. The development of lithium-ion batteries is very rapid, but lithium-ion batteries have a limited lifespan and their energy storage capacity decreases with time and use. Therefore, the State of Health (SoH) of lithium-ion batteries is crucial when planning battery maintenance. The purpose of this study is to compare regression techniques for estimating the health of Li-ion batteries. XGBoost, Support Vector Regression (SVR), Random Forest Regression, Linear Regression, Gradient Boosting Regression, and Decision Tree Regression are the regression methods utilized in this investigation. All types of batteries from NASA's Prognostics Data Repository were utilized in the investigation. Support Vector Regression (SVR) yields the most accurate results compared to other techniques. The SVR technique yields RMSE, MSE, MAE, and MAPE values of 0.0226, 0.0005, 0.0208, and 0.0264, respectively. This indicates that the SVR method is capable of accurately estimating the SoH of a lithium-ion battery.