An approach to predicpt discharge voltage of lithium-ion batteries under dynamic loading conditions

Terminal voltage is an important indicator to alarm end-of-discharge of lithium-ion batteries. Therefore, predicting the terminal voltage is helpful in preventing issues that caused by running out of power. However, the loading condition of battery is usually dynamic in real practice which greatly increases the difficulty of prediction. In this paper, we propose a novel approach to predict the terminal voltage under dynamic loading condition. This approach transforms the problem of predicting the terminal voltage into the problem of predicting the state-of-charge (SOC) of battery, using equivalent circuit model and a polynomial function. In the prediction of SOC, an accurate value of capacity is required, but it is not practical to be measured in each discharge process. Therefore, we develop an adaptive capacity method based on feature extraction in charging profile and k-nearest neighbor algorithm to timely update batterys SOC after each charge process. The whole prediction approach is tested on an open dataset, and comparison experiments demonstrate that it outperforms traditional approaches.