A Unified Piecewise Modeling Framework for Battery Knee Point Detection and State of Health Estimation

The knee point in battery capacity curves signifies the onset of accelerated capacity degradation. This phenomenon leads to markedly different degradation rates and patterns before and after the knee point, affecting the accurate estimation of the battery state of health (SOH). To address this challenge, we propose a unified piecewise modeling framework for simultaneous knee point detection and SOH estimation. First, knee point detection is formulated as a binary classification problem for each cycle. A convolutional neural network (CNN) is employed to construct a classification model for knee point detection, utilizing informative features extracted from capacity-voltage(QV) curves. Subsequently, two Elastic Net-based models are respectively trained to estimate the SOH for cycles occurring before and after the knee point. Finally, the SOH for a given cycle is estimated using a weighted sum of the predictions from these two Elastic Net models, where the weights are determined by the probability of class membership output by the CNN classification model. A case study demonstrates that the proposed piecewise strategy and models achieve superior results with an average accuracy of 94.5% in identifying knee points and estimating SOH with an average root mean squared error (RMSE) of 0.014.