Research On Real-time Charging Of Electric Vehicle Based On Data Driving

With the popularization of electric vehicles in the world,its charging load has become an important part of the power grid load.The prediction of electric vehicle’s charging load is the basis of siting and sizing of the charging station,demand response,analysis of the influence of electric vehicle access to power grid,charging control strategy and so on.The charging behavior is random and uncertain.If electric vehicles are randomly added to the power grid for charging without guidance,the load of grid will fluctuate greatly,which will the power grid more difficult to schedule and have a serious impact on power system construction and charging operation.Therefore,this dissertation proposes an improved algorithm for charging load prediction and charging scheduling schemes respectively.The specific work is as follows:1.To solve the problem of load forecasting,a method load forecasting based on trip chain is proposed.The method uses three-layer machine learning algorithm to simulate the travel behavior,then calculate the electricity consumption per kilometer under different weather and road conditions by using fuzzy logic,finally the charging load forecast of electric vehicles in the region is generated.There is a strong association between users’ trip chain and travel activities,and the previous trip will affect the subsequent trip,so the first layer uses LSTM to predict the type of trip chain.In the second layer,BP neural network is used to predict the start and end time of the trip.Considering that the traditional BP neural network model is easy to fall into local optimum when solving nonlinear problems,BAS is introduced to help BP optimize the global search and speed up the solution.Civen the different driving habits of users,Random Forest(RF)algorithm is used as the third layer of the model,and multiple decision trees are constructed to predict the driving distance of users.Experiments show that the difference between the predicted load and the actual load is2.35%,the difference between actual load and predicted load by MC algorithm is 5%,which show that the proposed algorithm is more accurate.2.In order to solve the problem of charging scheduling,an orderly charging method based on deep reinforcement learning algorithm A3 C is proposed.The demand response model based on consumer psychology theory is designed by the prediction method of users’ travel proposed previously.The charging scheduling and pricing strategy are optimized with the goal of maximizing the revenue of charging stations.To avoid of the strong correlation caused by experience replay,use multiprocessing to implement the effect that the model interact with multiple environments through which can improve the convergence of the algorithm.The algorithm deal with the varying state spaces caused by random EV arrivals and fix them to five by approximating the state function and policy function with feature function based on the features of the underlying physical system.The experimental results show that the load varias under disordered charging,Sarsa algorithm and A3 C algorithm are 24.41 MW,9.46 MW and 1.16 MW respectively,which proves the advantages of the proposed method in reducing load variance and peak cutting and valley filling.3.In order to display the data more directly,a charging demand forecasting and scheduling system based on Springboot,Vue and other frameworks is proposed.Considering that My Sql can’t cope with a large number of users’ simultaneous requests,the non-relational data Redis is introduced as a cache to record the data sent by users.Aiming at the display of load forecasting,the interactive regional map is drawn by Map Box.The time-varying regional charging demand,charging demand heat map and charging demand of charging stations are shown at the same time.For the demonstration of charging scheduling,the charging station uses Web Socket to receive the charging power and charging price customized by the system.In order to show the demand satisfaction of users who has arrived in the station,use Echarts to draw the chart.