基于融合模型的锂离子电池荷电状态在线估计

针对锂离子电池荷电状态（Stage of charge，SOC）在线估计精度不高，等效电路模型法估计精度与模型复杂度相矛盾的问题，本文对扩展卡尔曼滤波算法进行了改进，并以电池工作电压、电流为输入，对应等效电路模型法的SOC估计误差为输出，采用极限学习机算法，建立基于输入输出数据的SOC估计误差预测模型，采用物理–数据融合方法，基于误差预测模型，建立了等效电路模型法结合极限学习机的锂离子电池SOC在线估计模型。仿真结果表明，改进扩展卡尔曼滤波算法提高了算法的估计精度，而物理–数据融合的锂离子电池SOC在线估计模型减小了由电压、电流测量所引入的估计误差，克服了等效电路模型法估计精度与模型复杂度之间相矛盾的问题，进一步提高了SOC的估计精度，满足估计误差不超过5%的应用需求。 

Online estimation of the state of charge of a lithium-ion battery based on the fusion model

In the context of the global response to environmental pollution and climate change, countries have begun to pay attention to energy system reform and economic development to ensure low carbon transition. Among them, the development of low carbon transportation has become an important aspect of green transportation system construction. The development of electric vehicle technology can effectively reduce energy consumption and environmental pollution. However, with the recent reports of new energy vehicle safety accidents at home and abroad, the safety of lithium-ion batteries has attracted increasing attention from the industry. To prevent overcharging and overdischarging from affecting battery life and safety during use, a complete battery management system is required to control and manage a lithium-ion battery. The state of charge (SOC) used to reflect the remaining capacity of a battery is one of the key parameters. Therefore, an accurate SOC value is of significance to the safety of lithium-ion battery use and the safety performance of new energy vehicles. The low online estimation accuracy of the SOC of lithium-ion batteries and the estimation accuracy of the equivalent circuit model method are inconsistent with the model complexity. This study improved the extended Kalman filtering (EKF) algorithm and established a SOC estimation error prediction model based on the extreme learning machine (ELM) algorithm, which used the operating voltage and current of the battery as input and the SOC estimation error of the equivalent circuit model method as the output. On the basis of the physical data fusion method and the error prediction model, the online estimation model of the lithium-ion battery SOC based on the equivalent circuit model method combined with the ELM was established. The simulation results showed that the improved EKF algorithm enhances the estimation precision of the algorithm. Moreover, the physical data fusion model reduces the estimation error introduced by voltage and current measurements, overcomes the contradiction between the estimation accuracy and complexity of the equivalent circuit model method, improves the estimation accuracy of the SOC, and meets the application requirement that the estimation error must be less than 5%.