With the continuous increase in the scale of new energy grid integration, the importance of energy storage in the generation, transmission, distribution and use of new power systems has become increasingly prominent. Among them, lithium-ion batteries have become one of the mainstream energy storage systems in the new power system due to their advantages of high energy density, fast conversion rate, convenient deployment and low cost. The state of health (SOH) of lithium-ion batteries can effectively reflect the aging degree of batteries, and in actual operation, replacing low-SOH batteries according to the set threshold is one of the effective measures to ensure the safe and stable operation of battery energy storage systems. Therefore, SOH evaluation is one of the important functions of the battery management system (BMS).
In order to solve the problems of insufficient accuracy, high complexity and low interpretability of existing evaluation models, Gu Juping, Jiang Ling, Zhang Xinsong and other scholars from the School of Information Science and Technology of Nantong University, the School of Electronics and Information Engineering of Suzhou University of Science and Technology, and the School of Electrical Engineering of Nantong University proposed a method for the evaluation of lithium-ion battery SOH and the analysis of influencing factors based on feature extraction.
Figure 1 Battery health feature extraction process
Firstly, they proposed two new health features to quantify the similarity between the initial cyclic charging voltage curve and the current cyclic charging voltage curve, namely the dynamic time-adjusted distance feature and the Wasserstein distance feature. Secondly, the CatBoost method was used to evaluate the battery SOH, and the SHAP method was introduced to analyze the influence of each health feature on the evaluation results and the coupling relationship between the features. Finally, the researchers used multiple battery data in the University of Maryland's battery aging dataset for experimental verification.
Fig.2 Lithium-ion battery SOH evaluation and analysis process
Experimental results show that compared with the existing methods, the proposed SOH evaluation methods have higher accuracy, with an average error of less than 2.2%, and can quantitatively analyze the influencing factors of SOH.
However, the researchers also pointed out that since the battery data was obtained under fixed charging conditions, the health assessment needs of batteries in specific application scenarios were not considered. There are significant differences in the operating conditions of lithium-ion batteries in different energy storage systems, such as vehicle energy storage systems and energy storage power stations, and follow-up research will formulate appropriate evaluation methods for lithium-ion battery SOH according to the needs of different application scenarios, so as to realize the accurate and efficient operation of the evaluation model in multiple application scenarios.
This work was published in the 19th issue of Transactions of China Electrotechnical Society in 2023, with the title of "Lithium-ion battery health state assessment and influencing factor analysis based on feature extraction". This project is supported by the National Natural Science Foundation of China Smart Grid Joint Fund, the National Natural Science Foundation of China, and the Key R&D Program of Jiangsu Province (Industrial Prospect and Key Core Technologies).