Research On Prediction Of Electric Vehicle Power Battery SOC And Remaining Driving Mileage Based On SK-PSO-RBF

Electric vehicles have the characteristics of low pollution and low noise,which leads to the development of the automobile industry in the future,and The progress of battery management system technology seriously restricts the development of electric vehicles.The SOC?state of charge?of battery is one of the key parameters of the BMS?battery management system?and it is an important signal parameter of the electric vehicle control unit,and it is also closely related to the remaining mileage of the electric vehicle.For ordinary vehicles drivers,the prediction of the remaining mileage of a pure electric vehicle is one of the most concerned parameters,which can effectively eliminate the driver's mileage anxiety caused by the fear that the remaining mileage can not reach the destination.Therefore,it is one of the key research topics that the battery SOC is predicted accurately,stably and rapidly,and the remaining mileage is to predicted based on its SOC in the pure electric vehicles industry.The prediction of the SOC and remaining mileage of pure electric vehicles are characterized by large errors,non-linear relationships,large calculations,complex theoretical models,slow response speeds,high hardware requirements,poor adaptability,and poor practicability.The existing mainstream real-time online prediction methods have the disadvantages of complex dynamic data collection,slow system response,high hardware requirements,and large prediction errors.Therefore,in this study,A unique prediction method is proposed by simulating the idea of stock market forecasting:The dynamic reference values of SOC and the remaining mileage was obtained by pre-training and measurement the pure electric car under the dynamic private custom automobile conditions?a special condition defined in this study,seen section 4.1 for details?,in the prediction research through that the fluctuation factors,for example,the terminal voltage,terminal current,temperature,load,and speed are only considered,but the Stable factors are not considered.In the prediction,the model is corrected using the fluctuation factor correction model to obtain the real-time prediction value,thereby reducing the complexity of the model algorithm,reducing the calculation amount,and improving the real-time performance..In this study,a pure electric vehicle developed by an electric car company was taken as the research object to simulate the dynamic private custom automobile conditions,and The data,such as voltage?U_b?,terminal current?I_b?,temperature?T?,load?L?,vehicle speed?v?,road gradient,vehicle energy consumption and mileage?S?and other related parameters,were collected,then the 1500 sets of the data were selected randomly as the predicted values and the test values,and in order to reduce the error,all the data were normalized.The core algorithm model of this research is as follows:?1?It was collected,tested and trained under the dynamic private custom automotive conditions with the the stock market forecasting idea.?2?to solve the prediction difficulties of the large error,poor adaptability and complex mathematical modeling,this paper optimizes the parameters?individual influence factor c1,social influence factor c2?of the particle swarm optimization clustering algorithm using Sudoku to calculate three parameter combinations of the radial basis function?RBF?neural network:the number hidden layer nodes q,the center vector c_i and standardization constant?_i of the Gauss function.then,it selects best combination according to the error and clustering efficiency,and updates the parameters according to the conditions timely to make the prediction result have the dynamical adaptability.?3?The SK-PSO-RBF-SOC prediction model was established with the input of the terminal voltage,the current,the temperature,and the output of the remaining mileage to get the predictions.?4?The SK-PSO-RBF-S prediction model was established with the input of the terminal voltage,the current,the temperature,the load and the output of the remaining mileage to get the Sre.?5?Then the dynamic adaptive standard SOC_N and S_N were defined in the dynamic private custom automotive conditions.?6?Finally,the S_finalinal was obtained with the mathematical models of non-essential energy consumption and driving range-speed correction.The results show that the maximum relative SOC error is 3.8%,and the maximum relative S_finalinal error is 4.2%,The maximum absolute S_finalinal error is 8.295 kilometers,compared with the test values by the pure electric vehicle,which is obviously improved compared with the other current methods.