Fast and Robust Estimation of Lithium-ion Batteries State of Health Using Ensemble Learning

Extreme learning machine (ELM) has attracted attention in battery SOH estimation due to its advantages such as fast operation, straightforward solution, and less computational complexity. However, the relatively low accuracy and poor stability are still problems. To achieve high accuracy and good generalization performance, a bagging-based ELM is proposed in this paper, which combines ELM with bagging technology. Bagging is used to reconstruct the dataset so that multiple base-level ELMs can be trained. In addition, the input voltage sequence is extracted from the partial charging curve, and its length and starting points are optimized. In order to illustrate the performance of the proposed algorithms, both self-validation and mutual validation are used. Finally, experiments are performed to verify the effectiveness of the proposed method. Results reveal that the proposed method improves the accuracy of the traditional ELM method by 40% in the case of self-validation. Even in the mutual validation where traditional ELM cannot accurately estimate the SOH, the proposed method still maintains a high estimation accuracy.