Combining offline and online machine learning to estimate state of health of lithium-ion batteries

Paper i proceeding, 2022

This article reports a new state of health (SOH) estimation method for lithium-ion batteries using machine learning. Practical problems with cell inconsistency and online implementability are addressed using a proposed individualized estimation scheme that blends a model migration method with ensemble learning. A set of candidate models, based on slope-bias correction (SBC) and radial basis function neural networks (RBFNNs), are first trained offline by choosing a single-point feature on the incremental capacity curve as the model input. For online operation, the prediction errors due to cell inconsistency in the target new cell are next mitigated by a proposed modified random forest regression (mRFR) for high adaptability. The results show that compared to prevailing methods, the proposed SBC-RBFNN-mRFR-based scheme can achieve considerably high SOH estimation accuracy with only a small amount of early data and online measurements are needed for practical operation.