Research On Electric Vehicle Charge And Discharge Scheduling Considering Uncertainty Forecasting

The penetration rate of Electric Vehicles(EVs)has greatly increased with the strong encouragement of relevant policies of various countries.With the development of smart grid,Vehicle to Grid(V2G)technology enable EVs to be used as energy storage devices to discharge to the grid and then use an orderly EVs’ charging and discharging scheduling strategy to adjust power peaks and stabilize grid loads.Some factors that need to be considered in the scheduling strategy are uncertain,and the prediction of uncertainty is the prerequisite for the application of the EVs’ scheduling strategy.This thesis focuses on the forecasting method and scheduling strategies required by the application of EVs’ scheduling.Firstly,an agent operation mode and model considering V2 G is constructed,and a scheduling framework based on game and demand response is adopted for EVs in the area served by the agent.The simulation analyzes the influence of EVs charging and discharging on regional load and the influence of load forecasting deviation on scheduling results under the model and scheduling framework.Secondly,aiming at the advantages and disadvantages of the deep learning model for predicting uncertain factors in smart grid,a prediction model that can consider the gap time period is designed based on neural translation network,thus improving the prediction accuracy and expanding the application range.The experimental results of local marginal electricity price show that the prediction effect of the designed model is better than that of the mainstream deep learning time series prediction model in four evaluation metrics under different prediction periods.The ratio of training parameters increasing with the increase of prediction scale is the smallest,and the ability to perceive load peaks and spikes is better.The experimental results of residential load forecasting show that the deviation of forecasting is within the acceptable range,and it is feasible to use the forecasting values for scheduling.Finally,based on the forecast of residents’ basic load,the charging and discharging scheduling research of EVs was carried out.To solve the problem that EVs users need to consider additional cost factors in actual charging and discharging,the cost factors such as charging and discharging efficiency,battery aging and user satisfaction are taken into account in the user’s objective function.This thesis constructs a multi-player non-cooperative game with the objective of minimizing the comprehensive cost of users’ charge and discharge,transforms the game problem into an optimization problem,and gives centralized and distributed solution methods.Firstly,the users’ objective function with or without considering various cost factors is simulated and analyzed,and then a large penetration of EVs are simulated by using the comprehensive objective function.Simulation results show that the scheduling strategy can not only reduce the comprehensive cost of each user,but also stabilize the load and reduce the cost of agent.