数据驱动的钠离子电池健康状态评估方法研究

The state of health（SOH） estimation for sodium-ion batteries is crucial for their safe and efficient applications, which is also a key to large-scale energy storage implementations. However, sodium-ion batteries exhibit usage-induced degradation with unclear mechanisms and are sensitive to operating conditions and environmental factors, posing a challenge to the accurate SOH estimation. In this paper, a data-driven SOH estimation method for sodium-ion batteries is proposed. The charging data is correlated with capacity degradation, and variance filtering, grey relational analysis and recursive feature elimination are integrated for feature selection. In addition, four machine learning methods including multiple linear regression, support vector machine, Gaussian process regression and error back propagation neural network are applied to formulate the corresponding estimation methods. Test results reveal that the root mean square errors for the four methods are all less than 1.6%, with Gaussian process regression showing an error rate below 0.8%, indicating a precise SOH estimation for sodium-ion batteries.