DEVELOPMENT OF A DIGITAL TWIN-BASED BATTERY MANAGEMENT SYSTEM FOR ESTIMATING STATE OF CHARGE OF BATTERY SYSTEMS IN ELECTRIC VEHICLES

• Main Author: Solavide Gulo, Teguh
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/72880
• Institution: Institut Teknologi Bandung

The development of electric vehicles worldwide is still progressing to achieve efficient and optimal performance. The main challenges in electric vehicles lie in their batteries, which serve as the energy source. Battery capacity and lifespan are crucial factors that need improvement. Extending battery lifespan is closely related to effective monitoring. One aspect of battery monitoring is the state of charge (SoC). However, direct measurement of SoC is not feasible and relies on estimation. Estimating the battery's state of charge is difficult due to the electrochemical complexity involved in battery aging, leading to challenges in achieving the required accuracy. This difficulty arises from identifying the time-varying model parameters and collecting training datasets from physically diverse battery usage scenarios. Therefore, a battery management system (BMS) is necessary to monitor the battery. BMS is a key component in ensuring safety, durability, avoiding physical damage, and thermal runaway. However, the development of BMS continues to enhance its reliability in battery management. In general, the main challenges in BMS technology development can be divided into three aspects. Firstly, lithium battery systems are highly nonlinear, exhibiting multispatial and multitime scale aging, making accurate modeling challenging. Secondly, the internal conditions of the battery cannot be directly obtained through measurement approaches. Thirdly, battery cell inconsistencies affect packaging efficiency, increasing hidden risks associated with batteries. This research focuses on developing a Battery Management System (BMS) for state-of-charge estimation, aiming to ensure reliable operations, optimize battery systems, and provide a foundation for safety management. The problem addressed in this study is the lack of state-of-charge estimation for electric vehicles (etrikes) developed by the Energy Management Lab at ITB. This research aims to enable the application of the BMS to electric vehicles equipped with 480 lithium NMC 18650 battery cells. The development of the BMS in this study is based on a digital twin, which represents the physical object. The advantages of a digital twin-based BMS include continuous and accurate battery status monitoring supported by high computational power. Accurate degradation prediction using machine learning and early detection of system failures through big data analysis are also possible. Other benefits include enhanced system security and reliability, optimization of system design and operational strategies through the evaluation of big data from battery systems under different operating scenarios. State-of-charge estimation is performed using an artificial neural network (ANN) algorithm inspired by the functioning of the human brain. The ANN requires training data to create a model for battery state-of-charge estimation. Once the model is obtained through data training, it is implemented to estimate the state-of-charge of the battery pack in electric vehicles during operation. The final results of the research demonstrate the interconnectedness and real-time data exchange between the physical and digital objects. The accuracy of state-of-charge estimation using ANN in this study is indicated by an RMSE value of 1.43, MAE of 1.17, and MAPE of 1.35. The time required for state-of-charge estimation with a dataset of 1934 rows is 0.24 seconds. During operation, the estimation of battery state-of-charge in electric vehicles is influenced by variables such as current, voltage, and temperature. The estimated state-of-charge of the battery during operation ranges from approximately 81.47% to 90.34%.