Digital Twin-Driven Estimation of State of Charge for Li-ion Battery

Under the net-zero carbon transition, lithium-ion batteries (LIB) plays a critical role in supporting the connection of more renewable power generation, increasing grid resiliency and creating more flexible energy systems. However, poor useful life and relatively high cost of batteries result in barriers that hinder the wider adoption of battery technologies e.g., renewable resources storage. Furthermore, the useful life of a battery is significantly affected by the materials composition, system design and operating conditions, hence, made the control and management of battery systems more challenging. Digitalisation and artificial intelligence (AI) offer an opportunity to establish a battery digital twin that has great potentials to improve the situational awareness of battery management systems and enable the optimal operation of battery storage units. An accurate estimation of the state of charge (SOC) can indicate the battery's status, provide valuable information for maintenance and maximise its useful life. In this paper, a digital twin-driven framework based on a hybrid model that connects LSTM (long short-term memory) and EKF (extended Kalman filter) has been proposed to estimate the SOC of a li-ion battery. LSTM provides more accurate initial SOC estimations and impedance model data to EKF. According to experimental results, the developed battery digital twin is considered less dependent on the initial SOC conditions and is deemed more robust compared to traditional means with a lower RMSE (root mean squared error).