SOC Estimation System For Lithium-ion Power Batteries Of Electric Vehicles

As the world’s energy and environment continue to deteriorate,electric vehicles as new energy projects have also received more and more attention.As a power source for electric vehicles,batteries are an important part of electric vehicles.The estimation-of battery state of charge(SOC)is the core function of electric vehicle battery management.Therefore,this thesis mainly studies the method of estimating SOC.The main contents are as follows:This thesis first briefly introduces the electric vehicle battery and SOC,explains the significance of predicting SOC,and analyzes the factors affecting SOC and the advantages and disadvantages of the traditional method of estimating SOC.In order to facilitate the measurement of the data,the vehicle simulation software ADVISOR was used.At the same time,the battery model of ADVISOR was improved in small capacity and low voltage.A new battery model was established and the battery data was better measured.The model then compares the simulation results of the new battery model with the actual test data to verify the authenticity of the model.Combining the advantages of Kalman filter and neural network,this thesis proposes a dynamic Kalman neural network(KDNN)algorithm based on genetic optimization.The algorithm fits the relationship between the input vector and the output SOC through the neural network,and updates the neural network weight through Kalman filtering.At the same time,the initial weight is updated by the genetic algorithm during initialization.It is proved by experiments that the algorithm combines the advantages of Kalman filter and neural network and the prediction effect is better than the former two.At the same time,in order to estimate the poor stability of SOC for KDNN,this thesis proposes an integrated support vector regression(FCDE-SVR)algorithm based on feature clustering based on data-driven theory.By analyzing the data,the algorithm extracts the data clusters with different feature parts in the data,and removes the data with insignificant features,and then uses each support vector regression as a sub-learning machine for each cluster,and then determines each cluster according to the distance.Weight,and then integrate the learning of each sub-learning machine.Then the experimental results show that the algorithm prediction effect is better than the integrated learning algorithm and single SVR.These two methods can be applied to the automotive battery management system(BMS)and the intelligent battery service system based on the vehicle networking technology.This thesis also developed an electric vehicle battery SOC real-time estimation system through C#.The system can adjust the algorithm in real time through the user interface and monitor real-time battery parameters.