State of health estimation for lithium-ion batteries based on temperature prediction and gated recurrent unit neural network

<p>Accurate state of health estimation for lithium-ion batteries is crucial to ensure the safety and reliability of electric vehicles. This study presents an accurate state of health estimation method based on temperature prediction and gated <a href="https://www.sciencedirect.com/topics/engineering/recurrent" target="_blank">recurrent</a> <a href="https://www.sciencedirect.com/topics/engineering/neural-unit" target="_blank">unit neural</a> network. First, the <a href="https://www.sciencedirect.com/topics/engineering/extreme-learning-machine" target="_blank">extreme learning machine</a> method is leveraged to forecast the entire temperature variation during the constant current charging process based on randomly discontinuous short-term charging data. Next, a <a href="https://www.sciencedirect.com/topics/engineering/finite-difference-method" target="_blank">finite difference method</a> is employed to calculate the raw differential temperature variation, which is then smoothed by the <a href="https://www.sciencedirect.com/topics/engineering/kalman-filter" target="_blank">Kalman filter</a>. On this basis, multi-dimensional health features are extracted from the differential temperature curves to reflect battery degradation from multiple perspectives, and six strong correlated features are selected by the <a href="https://www.sciencedirect.com/topics/engineering/pearsons-linear-correlation-coefficient" target="_blank">Pearson</a> correlation coefficient method. After preparing all the related health features, the gated recurrent unit neural network is exploited to predict state of health. The feasibility of the developed method is verified by comparing with other classic approaches in terms of accuracy and reliability. The experimental results demonstrate that the proposed method can effectively lead to the error of state of health within 2.28% based on only partial random and discontinuous charging data, justifying its anticipated prediction performance. </p>