基于深度学习的锂电池SOC和SOH联合估计研究
A Joint Estimation Study Based on Deep Learning of SOC and SOH for Lithium Batteries

目前，新能源的开发和利用越来越受到世界各国的关注，锂离子电池以其优良的特性逐渐成为世界上应用最为广泛的储能元件。因此，保障锂离子电池的安全可靠运行也成为当下的研究热点，然而其荷电状态(SOC)和健康状态(SOH)无法被直接测量。为了实现锂离子电池SOC和SOH的联合估算，本文分析了SOC和SOH之间的关联，并设计一种基于深度学习的锂离子电池SOC和SOH联合估算方法。本方法通过采集到的电流、电压、充放电功率、电压变化率等多个参数丰富特征值样本，利用CNN卷积神经网络估计锂电池的内阻，用于SOH的估算。其中，数据为一维数据，需要对CNN神经网络结构进行调整。进一步将估计得到的SOH与采集到的数据利用GRU-RNN神经网络联合估计得到锂电池SOC。此方法不需搭建电池模型，甚至不需要设置SOC初值，也可以快速收敛。
Nowadays, the development and utilization of new energy have attracted more and more attention in the world. Lithium-ion batteries have gradually become the most widely used energy storage component in the world because of their excellent characteristics. Therefore, ensuring the safe and reliable operation of lithiumion batteries has become a current research focus, but their State of Charge (SOC) and State of Health (SOH) cannot be directly measured. In order to realize the joint estimation of SOC and SOH for lithiumion batteries, this paper analyzes the correlation between SOC and SOH, and designs a joint estimation method based on deep learning of SOC and SOH for lithiumion batteries. This method uses CNN convolution neural network to estimate the internal resistance of lithium batteries, which is used for the estimation of SOH, through collecting multiple parameters such as current, voltage, charge-discharge power, voltage change rate and other rich eigenvalue samples. The data is one-dimensional data, and the structure of CNN neural network needs to be adjusted. Further, the estimated SOH and the collected data are combined to estimate the lithium batteries’ SOC using GRU-RNN neural network. This method does not need to build a battery model, and even can quickly converge without setting the initial value of SOC.