High-precision Modeling And Multi-state Intelligent Estimation Of Lithium-ion Battery

In order to deal with the global environmental pollution and the gradual reduction of fossil energy,the new energy industry,especially the electric vehicle industry,has been paid more and more attention by governments.As the power source of electric vehicle,the research of power battery has become the top priority of the whole electric vehicle industry.Lithium-ion batteries have attracted more and more attention because of their advantages in energy density,calendar life and self-discharge rate.At present,a lot of research work has been carried out in the power battery modeling theory research and battery internal state estimation.However,in the field of power battery modeling,the accuracy of battery test data plays a decisive role in the accuracy of identified model parameters.Due to the existence of environmental noise and measurement errors,there will always be noise errors in battery test data,thus reducing the accuracy of identification of equivalent circuit model parameters.Therefore,this paper uses the output response reconstruction method to filter out the voltage noise and improve the modeling accuracy.In addition,inspired by the increasingly popular research in the field of artificial intelligence,this paper uses the long short term memory network of neural network algorithm to estimate the internal state SOC and SOH of lithium-ion battery.Therefore,this paper has done the following work:Firstly,the internal working mechanism and external characteristic parameters of the ternary lithium battery are analyzed,and the external characteristic test data of the battery are obtained based on the power battery test platform.This data provides abundant original data support for the next step of power battery modeling and internal state estimation.Secondly,the output response reconstruction method is proposed to solve the problems of the traditional low-pass filtering method,such as the difficulty in selecting the cut-off frequency and the distortion of the filtering waveform.This method reconstructs the response voltage of the battery based on the convolution principle and correlation function and other mathematical theories.Compared with the voltage after low-pass filtering,the filtering effect of the reconstructed voltage is more obvious.Then,based on the reconstructed voltage,the nonlinear least square method is adopted to identify the second-order equivalent circuit model parameters.The experimental results show that the model parameters identified based on reconstruction voltage are more accurate than those based on low-pass filter voltage.Then,using the long short term memory network algorithm and the gated recurrent unit algorithm in the artificial intelligence algorithm for reference,the SOC of the internal state of the power lithium-ion battery is estimated,and the estimation accuracy and model training time of the two algorithms are compared.The experimental results show that the accuracy of SOC estimation is almost the same between the two,but the training time of gated recurrent unit model is shorter.Therefore,the gated recurrent unit algorithm should be selected preferentially when estimating the short time-scale state of battery SOC.Finally,the long short term memory network algorithm and the gated recurrent unit algorithm are applied to the field of battery SOH estimation.The experimental results show that the estimation accuracy of the long short term memory network is slightly higher than that of the gated recurrent unit algorithm,so the long-term memory network algorithm should be used to estimate the long term-scale state of the battery SOH.