Electric Vehicle Charging Load Forecasting Based On Neural Network And The Influence On Distribution Network

The wide application of electric vehicles has played an important role in alleviating energy crisis and realizing sustainable development.In recent years,promoting the popularization of electric vehicles has been one of the important development strategies in China.Electric vehicles are widely used,and their charging demand is multiplied,thus bringing severe challenges to the economic and stable operation of the power grid at the same time.Large-scale grid connection of electric vehicles may lead to the reduction of power generation and transmission efficiency,overheating of lines and power transformation and distribution equipment,and deviation of electrical equipment from rated voltage.In this thesis,the charging load prediction of electric vehicles is realized effectively and quickly,and the characteristics and changing trend of charging load are analyzed,so as to further study the influence of charging load on distribution system.First of all,the factors affecting the electric vehicle charging load are analyzed from multiple angles,and the influence of users’ travel behavior on the time-space characteristics of charging load is further analyzed.The time-space characteristics of working days and rest days are counted and the corresponding probability distribution model is established.Among them,the time feature quantity includes the first trip time,driving time and parking time.Spatial features include travel destination,mileage and spatial transition probability.Then,according to the space-time characteristic model of electric vehicle charging load and considering the charging habits of users and charging modes of electric vehicle,Monte Carlo method is used to predict the charging load of electric vehicles in different charging scenarios on weekdays and rest days,and the rules are analyzed.To improve the real-time application ability of charging load forecasting,a charging load forecasting model for electric vehicles based on neural network and Monte Carlo is proposed,and the effectiveness and rapidity of the model are verified by an example simulation.Finally,based on the prediction of charging load,taking IEEE33-node distribution system as a model,and comparing the load peak-valley difference,network loss and voltage deviation of distribution network after charging load is connected,the influence of charging load on the overall and local operation characteristics and power quality of distribution network under different permeability is quantitatively analyzed.The simulation results show that with the increase of permeability,the load peak-valley difference,network loss and voltage deviation of the whole distribution network system will increase significantly.When the permeability is constant,the charging load has different influences on different functional areas of distribution network,which will cause serious local voltage drop and increase the peak-valley difference of regional load,among which the charging load has less influence on the head-end node of feeder and greater influence on the end node of feeder.