A fast state-of-health estimation method using single linear feature for lithium-ion batteries

Data-driven methods are commonly used for state of health (SOH) estimation, which is essential to battery energy management. However, complex machine learning models, data gathering, and feature processing hinder its further implementation. A fast SOH estimation method based on linear properties of short-time charging is proposed to overcome these challenges. Only the exceptional single linear health factor (LHF) is required for effective SOH estimation. The LHF is chosen through correlation analysis from short-term feature derived from charging curves. The processing is straightforward. To define the relationship between LHF and SOH, a linear regression model is developed. For the simplicity and effectiveness of the method, it is suitable to be implemented in online applications with low hardware requirements. Finally, experiments show that the SOH estimation method has the highest accuracy of 0.54%, and the biggest estimation error is 2.20%. Furthermore, the data from first 20% cycles of the battery are used to build the model, ensuring that the SOH estimation accuracy is comparable. It is worth noting that the time cost of data acquisition does not exceed 30 s, which is important for fast estimation. • A single-feature linear regression model is proposed to achieve SOH estimation. • Short-time linear and efficient aging features are extracted. • Accurate SOH estimation is achieved by using only the first 20% of the data. • Less than 30 s of data is required to extract features.