DEVELOPMENT OF REMAINING USEFUL LIFE PREDICTION ON VRLA BATTERY USING SUPPORT VECTOR REGRESSION

• Main Author: Maulana Achsan, Beny
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/39649
• Institution: Institut Teknologi Bandung

Battery management system (BMS) is essential to be applied on battery usage as protection system during operation. One of the important features in battery management system is a diagnosis feature, consist of estimation and prediction system. This research was conducted on a smart microgrid system of Energy Management Laboratory, Engineering Physics Study Program - ITB. The battery modul used in this research is a VRLA battery with 200 Ah capacity. This research focused on state of charge (SOC) estimation, state of health (SOH) estimation, remaining cycle life (RCL) predition and remaining useful life (RUL) prediction using support vector regression (SVR) method, and optimum parameters recomendation for BMS to extend remaining useful life of the battery. The optimum parameters of SVR were determined by using grid search. Furthermore, the results of SOC estimation and SOH estimation were validated by determining the coefficient of determination (R2) and the root mean square error (RMSE), while the results of RCL and RUL prediction were validated by determining the absolut percentage error (APE). The value of R2 represents how close the data to the fitted regression line with range 0 to 1, where the larger the R2 mention the estimation results of the model can follow the pattern well. The RMSE value represents the error generated by the estimation model where the smaller the RMSE value, the better the estimation results. The APE value represents the absolute percent error of the prediction model to the actual data, where the smaller the APE value, the higher the accuracy of the prediction. The results show that the SOC estimation has R2 of 0.9991 with 0.0024 of RMSE, while the SOH estimation has R2 of 0.9081 with 0.0108 of RMSE. These estimation have a good result because the value of R2 is close to 1. Meanwhile, the result of APE is 3.95 % in RCL prediction and 4,16 % in RUL prediction, which also have a good result because the value of APE is less than 5 %. The optimum parameters recomendation to extend remaining useful life of battery are C10 charging current rate, maximum SOC of 0.8, and minimum SOC of 0.4.