Research On Method Of Predicting SOC Of Power Battery For Electric Bus Based On Sparse Sampling Data

The development of electric vehicles(EV)is of great significance for maintaining national energy security,reducing vehicle emissions and ensuring sustainable social development.State of Charge(SOC)is an important parameter of electric vehicle power battery.It plays a guiding role in preventing battery overcharge and overdischarge problems and predicting driving mileage.It is the basic basis for battery status assessment and energy management control strategy.Therefore,it is of great significance to accurately predict the SOC value of electric vehicle power battery.With the development of vehicle networking and data monitoring cloud platform technology,data-driven method has become an important tool for vehicle condition monitoring and early warning.In order to realize the prediction of SOC of pure electric bus power battery,a prediction algorithm of SOC of electric bus battery driven by sparse sampling data is proposed.In this paper,two electric buses of a bus company are taken as research objects.Based on the sparse sampled battery data stored on the data monitoring cloud platform with 30 seconds as sampling period,the establishment process of the proposed prediction algorithm is divided into the following aspects:(1)Firstly,the internal structure,working principle and related parameters of lithium iron phosphate batteries are described,and the system architecture and significance of remote data monitoring cloud platform for pure electric vehicles are systematically introduced.Sparse sampling data source with 30 s sampling period is analyzed,which provides a basis for the establishment of the following algorithm.(2)Based on the sparse sampling data stored on the cloud platform,the operation process of the power battery of electric bus and the influencing factors of its SOC change are analyzed.The total voltage,current,temperature mean of battery module and SOC value of historical time are selected as predictive variables,and the training data set and test data set are constructed.(3)The training data set is trained with support vector machine(SVM),SOC prediction model is established,and Bayesian optimization algorithm(BOA)is used to optimize the parameters of the model,a single step prediction method based on sparse sampling data for electric bus battery SOC is proposed;And an independent prediction method based on the sparse sampling data for electric bus battery SOC is further proposed by redividing the training data set,which can get rid of the dependence on the real SOC value in the SOC long term prediction process,and the prediction method can predict SOC value independently,which only needs to calculate the SOC prediction value of the second sampling time according to the true SOC value of first sampling time,and then replace the true SOC history value with the SOC prediction value as one of the prediction variables in the subsequent prediction process.The experimental results show that the maximum absolute error of the SOC single step prediction method is only 1.82%,and the maximum absolute error of the SOC independent prediction method is only 5.89%,and both of them have high prediction accuracy.(4)The robustness of SOC prediction model is validated by selecting No.2 electric bus which has different running routes and different ambient temperatures as the research object.The experimental results show that the SOC prediction model has high robustness.