SOC Estimation Of Lithium-ion Battery Based On Model And Data-driven Method

Lithium-ion batteries have been the leading power cells with the superiorities of high energy storage density,many times of charge and discharge,low self-discharge rate,etc.As one of the key parameters of the battery management system,State of Charge（SOC）has significantly impact on the efficiency and safety of power battery packs.Given that the chemical reaction process inside the lithium-ion battery is a highly nonlinear dynamic system,obtaining an accurate SOC of the lithium battery is very challenging,so it is meaningful to research on SOC estimation methods.By researching the general lithium-ion battery model,the second-order equivalent circuit model is used to explain the battery operating characteristics.The test data of the lithium-ion battery are received by ortho-test,and are used for fitting the nexus curve between the Open Circuit Voltage（OCV）and SOC.The forgetting factor recursive least squares algorithm is applied to identify the parameters of the quadratic equivalent circuit model.Experiments show that the absolute error between the model output voltage value of the model and the actual voltage value is less than 0.07 V.The inconclusive noise value and the negative covariance of the Unscented Kalman Filter（UKF）algorithm are often involved in the estimating SOC process.Thus,this thesis coined a Square-Root Adaptive Unscented Kalman Filter（SR-AUKF）algorithm,which is adds the adaptive noise adjustment algorithm and the theory of a root average squared filter to UKF algorithm.Results display that the SR-AUKF algorithm is more precise and faster than the UKF algorithm for computing the SOC of lithium-ion batteries under diverse initial SOC error conditions.Given the equivalent circuit model is severely constrained by the battery model and the accuracy of model parameters,and the external conditions such as temperature,current and voltage are often slighted in the reckoning SOC process.Therefore,a hybrid data-driven method that combines Gated Recurrent Unit（GRU）and UKF is proposed to reckoning the SOC of lithium-ion battery.The method uses deep learning technology to learn the nonlinear relationship between the lithium-ion battery SOC and measurements,including the current,voltage,temperature.The relationship is used as the observation equation of UKF,and the SOC is estimated by the UKF to improve the accuracy and stability of algorithm estimation method.The input data is split and processed using window sliding technology to improve GRU training speed and estimation accuracy.Experimental results show that the Root Mean Square Error（RMSE）and Mean Absolute Error（MAE）of the proposed method are less than 0.51% and 0.46%at different temperatures and different working conditions,and the efficiency of the approach is verified.