State of health estimation of lithium-ion batteries based on feature optimization and data-driven models

With the widespread application of lithium-ion batteries in electric vehicles, accurately estimating their state of health (SOH) has become a key focus of research. This paper explores various feature optimization methods and data-driven models with different structures, and constructs various SOH estimation models suitable for lithium-ion batteries. Based on battery testing data, multiple features are extracted from voltage and temperature to characterize the battery aging process. To reduce information redundancy among features, filtering methods, Principal Component Analysis (PCA), and Multi-dimensional Scaling (MDS) are applied for optimization, aiming to maximize feature information utilization. This paper compares four common and structurally different data-driven models: linear regression (LR), Gaussian process regression (GPR), support vector regression (SVR), and long short-term memory (LSTM) networks. The effectiveness of each model is validated using multi-feature inputs, and a multi-dimensional assessment of feature selection and data-driven model performance in SOH estimation is conducted, the average absolute error of all models under 60 % training set conditions is 0.8 %. The average absolute error of estimating the four batteries using the fused PCA features as input and the GPR model is less than 1.2 %. At the same time, using the optimized features as input reduces the average training time by 46.63 % compared to using multiple features as input. In summary, the combination of PCA features and GPR models has good performance in both estimation accuracy and computational efficiency for different batteries.