<p>Accurately perceiving the state of health (SoH) is beneficial for the lithium-ion batteries’ security and steady operation. However, the actual usage environment of batteries is complex, resulting in a host of noise and outliers in the raw data, which poses great adversity to SoH estimation. Therefore, a wolf pack algorithm (WPA) optimized decision tree (DT) method based on indirect Health features is introduced in this paper to estimate the SoH of batteries. Firstly, the multiple indirect health features extracted in this paper can uncover deeper health information and effectively decrease the interference of noise and adverse data, which could be measured directly by sensors. Secondly, considering factors such as algorithm convergence time and computational cost, the WPA demonstrates its excellence performance and it is used to optimize the hyperparameters of the decision tree. Compared with deep learning, the combination of WPA and DT does not require a large amount of data, so the computational cost is lower. In addition, the inherent robustness of WPA, together with the interpretability of decision trees, Helps improve the generalization and accuracy of the estimation. Finally, contrast experiments are utilized to prove the validity of the introduced indirect health features and SoH estimation methods. The results illustrate that using the introduced indirect health features and SoH estimation method can significantly enhance the SoH estimation accuracy. The experiments used the first 900 and first 130 cycles from MIT and NASA dataset, and the mean absolute error (MAE) and maximum absolute error (MaxAE) on two datasets are both within 2% and 5%, respectively.</p>