Online Condition Monitoring and Health Management of Batteries Based on Digital Twin Technologies

Student thesis: Doctoral Thesis


In recent years, lithium-ion batteries (LIBs) have been extensively selected as the power source for battery energy storage systems (BESS) and electric vehicles (EVs) due to their merits, including environmental friendliness and long lifetime. Therefore, a reliable battery management system (BMS) is essential to guarantee the reliability and safety of electrical systems. The strongly correlated state of charge (SOC) and state of health (SOH) are two fundamental functions in BMS. However, accurate and efficient SOC and SOH estimations for real-world applications are still challenging tasks due to the nonlinear characteristics of electrochemical LIBs. Digital twin (DT) technology is an effective method to describe the complex characteristics of the battery by transferring the physical entity to the digital model. Therefore, this research project aims to investigate more advanced DT-based battery SOC and SOH estimation technologies for different scenarios in practice.
Firstly, this project develops novel model parameter identification and SOC estimation methods for LIBs in BEES. The model parameter can be accurately identified based on the particle swarm optimization (PSO) algorithm and dynamic voltage responses in the practical constant current (CC) discharging process. The improved hybrid method can minimize system errors for accurate SOC estimation by combining the merits of the extended Kalman filter (EKF), ampere-hour counting (AHC) method, and PSO. The effectiveness of the developed methods is finally validated by typical CC tests operated at 25℃ and 5℃.
Furthermore, this research summarises and extracts influential health indicators from CC charging voltages for battery SOH estimation based on the Oxford Degradation dataset. It is found that the required charging times in the ranges of 2.8-4.2 V, 3.0-4.2 V and 3.2-4.2 V are highly correlated with the battery SOH. The results reveal the Gaussian process regression model can achieve robust SOH prediction with 𝑅2 close to 1 and lower MAE and MAX.
To enhance the performance of the electrical variables based SOC/SOH estimation technologies, a novel active acoustic emission (AE) sensing technology is proposed to consider the changes in battery material properties during charging, discharging, and cycling. Analytic and experimental studies have found that AE signals in the frequency band 270-300 kHz with 7𝑓0 can successfully achieve the simultaneous estimation of battery SOC and SOH at any stage. This finding can be included in the model-based/data-driven approaches or designed as a standalone method for fast SOC/SOH estimation.
Date of Award8 Aug 2022
Original languageEnglish
SupervisorAndrew Ball (Main Supervisor), Fengshou Gu (Co-Supervisor) & Ann Smith (Co-Supervisor)

Cite this