Research On Power Lithium Battery State Estimation Method Based On Data Driv

Traditional fuel cars are slowly being phased out to alleviate environmental pollution,energy depletion problems at present,replaced by new energy electric vehicles.Lithium ion batteries are widely used in the electric vehicle industry as a clean and reusable power source.A stable battery management system is beneficial for managing batteries and ensuring their safe and efficient operation.State of Charge(SOC),as a key measurement parameter in battery management systems,accurate and real-time SOC estimation is closely related to the safety and stability of lithium-ion batteries and the performance of electric vehicles.SOC cannot be measured directly by the accuracy of the sensor element,and its magnitude can often only be estimated with the help of other external state parameters.Moreover,the directly measurable state parameters such as temperature,voltage,and internal resistance in lithium-ion batteries are highly susceptible to external environmental factors,making the estimation of SOC in lithium-ion batteries more challenging.Therefore,this article conducts research on data-driven state estimation methods for lithium-ion batteries and analyzes them through experimental verification.The main research work of this article is as follows:(1)The structure and operating principle of lithium-ion batteries are analyzed,Several common battery models were compared and the potential effects of lithium-ion batteries during operation were analyzed.The voltage characteristics,discharge rate characteristics,and capacity attenuation characteristics of lithium-ion batteries were analyzed from multiple perspectives through modeling and simulation.The parameter variation curves under different conditions obtained from experiments were discussed,laying the foundation for future research.The main focus was on the temperature,voltage,and Study on SOC estimation of input relationships under current and different charge discharge rates.(2)On the basis of studying the characteristics of lithium-ion batteries,a method for generating small sample datasets was first proposed.A one-dimensional lithium-ion battery model was built using the COMSOL multi-physical simulation platform,and a charge and discharge experiment was conducted on the lithium-ion battery.The data obtained from the experiment were exported,including the potential diagram,current density analysis diagram,and performance analysis diagram of the battery at different rates,Provide data support for the estimation research of lithium battery SOC.Then,based on the small sample data set,shallow neural network fusion particle swarm optimization algorithm is used to establish the nonlinear relationship between the voltage,discharge rate and SOC of lithium ion battery,so as to achieve the purpose of predicting SOC.(3)In response to the lack of comprehensiveness when considering the redundancy of feature information in shallow learning architecture,based on the advantages of deep models in feature extraction,a lithium ion battery SOC estimation model based on deep neural networks is further proposed.Firstly,the data on lithium-ion batteries was expanded by utilizing two publicly available datasets: the CALCE dataset and the Oxford path dependent battery degradation dataset;Secondly,a Bidirectional Long Short Term Memory Neural Network(Bi LSTM)that can not only positively acquire the correlation of sequence data,but also capture the information of reverse sequence data is built.Finally,the Bi LSTM model is optimized using Bayes to find the optimal hyperparameter combination of the model.By conducting multiple comparative experiments at different temperatures and variable temperature conditions,it has been proven that the proposed method has good fitting ability and exhibits higher accuracy and adaptability in SOC estimation for batteries with different chemical compositions.Research on the construction of SOC estimation model for lithium-ion batteries is explored in this paper,and the improvement of the stability and accuracy of battery SOC estimation is achieved.