Research On Prediction And Control Of Electric Vehicle Charging Load In Smart Grid

With the issue of sustainable energy development and depletion of fossil energy.In recent years,due to the policy promotion of electric vehicles(EVs)and its environmental protection characteristics,the scale of EVs has developed rapidly.However,the randomness and complexity of the large-scale charging of EVs into the grid will increase the burden on the power grid and affect the stable operation of the power system.By means of analyzing the historical charging load data of EVs,this paper establishes a data mining model to predict the EV charging load.Through the prediction information of the EV and Vehicle-to-Grid(V2G)technology,this paper completes the orderly interaction between EV and smart grid.Then,make the EV charging power under control.Otherwise,the data source is the real EV charging load data collected by the EV monitoring platform in a city in the southwestern region.Firstly,this paper proposes an EV charging demand prediction based on time series decomposition and ensemble learning methods.The original charging load time series are decomposed additively by empirical modal decomposition method.In order to reduce the prediction error caused by all the decomposed intrinsic mode functions(IMFs),the fuzzy entropy method is used to calculate the fuzzy entropy value for all the IMFs,and conbines the IMFs with the same complexity calculation value.The combined components are divided into high-frequency and medium-frequency subsequences according to the richness of the characteristic frequency information.Long short-term memory(LSTM)is the base learner for high-frequency IMFs,and support vector regression(SVR)for the medium-frequency IMFs.In addition,the charging demand of EVs is also related to the temperature and humidity data and the time lag data of the EV charging demand.Features of the EV charging demand are input into the fully connected neural network of the Stacking model.Compared with other classical algorithms,the validity and accuracy of the proposed model is verified.Secondly,a group interactive control strategy for EVs based on V2 G technology is proposed.The EV's energy boundary model is used to manage the EVs in groups.With the EVs in different group,the control and dispatch center calculates the EV's chargedischarge load capacity according to the EV charging load forecast and the residential area conventional load forecast.And issues an EV charge-discharge instruction based on the calculated value.EVs responding to charge and discharge scheduling plans in priority.Then,the allocated power of an EV must meet its own charge and discharge constraints and the sum of response power cannot be higher than the load dispatch value.Finally,the optimal charge and discharge schedule of EVs is established with the objective function of minimizing the fluctuation of grid load and the minimum cost of EV charging.This method can effectively reduce the fluctuation impact of EV disorderly charging on the grid and reduce user's charging fees.