REAL-TIME ESTIMATION OF BATTERY ENERGY EFFICIENCY BASED ON OPERATING CONDITIONS USING REGRESSION MODELING

• Main Author: Rivaldi, Safin
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/85645
• Institution: Institut Teknologi Bandung

The evaluation of energy use in MG relies heavily on the energy efficiency of the Battery Energi Storage System (BESS), which is an important parameter in measuring system performance, especially in BESS scheduling algorithms. However, its value varies depending on the operating conditions and is not easily measured directly in real time. This research focuses on analyzing and estimating the battery energy efficiency in real time. First, the battery energy efficiency under operating conditions is analyzed using the Round-trip Efficiency (RTE) calculation on the ITB Energy Management Laboratory SBPE. Therefore, this research focuses on Valve Regulated Lead Acid (VRLA) batteries installed in the MG system of ITB Energy Management Laboratory. Furthermore, the correlation between battery energy efficiency and parameters such as C-rate, State of Charge (SoC), and temperature is analyzed. The battery energy efficiency estimation model was built using the Support Vector Regression (SVR) method and three-degree polynomial regression. The estimation results from these two methods were then compared with the actual battery energy efficiency calculation to evaluate the accuracy of the developed model. The results showed that battery energy efficiency was identified using the RTE method with 1531 detected energy efficiency values. Correlation analysis revealed that RMS C-rate and SoC gradient have the most influence on energy efficiency, with correlation coefficients of -0.8385 and -0.8094, respectively. Correlation modeling using Support Vector Regression (SVR) and three-degree polynomial regression showed that polynomial regression was more stable, with R^2 = 0.7208 and RMSE = 0.0608 on the validation data. Modeling estimates compared to actual data (RVP: MAE = 0.133, RMSE = 0.159; polynomial: MAE = 0.085, RMSE = 0.117), the three-degree polynomial regression model provides more accurate predictions compared to SVR.