Spatial Load Forecasting Considering The Integration Of Electric Vehicles Into Power Grid

The spatial distribution of power load is the basis of power network planning,which is of great reference significance to the setting of power points,the determination of transmission corridors,the selection of transmission lines and related equipment.In the traditional sense,the system load forecasting is to predict the regional load size from the perspective of geographical macro,but the spatial distribution of power load cannot be obtained.Therefore,the spatial load forecasting based on the spatial and temporal dimension distribution of power load emerges at the historic moment.Land use simulation method and load density index method are two common spatial load forecasting methods.Existing models often believe that land use decision or load density can be obtained directly.However,with the increasingly accurate requirements of power grid planning,it is difficult to predict the spatial distribution of load in accordance with simplified data such as urban land planning data and historical load density index to meet the requirements of accurate quantification of load.At present,space load forecasting appears a new situation.Under the guidance of relevant policies,the electric car ownership increased year by year,as a temporal and spatial dimensions have strong randomness of power load,along with the electric car ownership has increased,charging load errors occurring in the spatial load forecasting results,therefore,it is necessary in the spatial load forecasting on the basis of considering the electric car charging load distribution of space and time.Under the above background,this paper divides the research on spatial load prediction considering the grid-connection of electric vehicles into three problems:land use decision problem,load density index acquisition problem,EV charging load spatial distribution prediction problem and so on.On the problem of land use decision,an improved cellular automata algorithm is proposed to simulate the land development trend.The geographical attribute and non-spatial attribute of the cell(cellular)are taken into account in the spatial partition to form the generalized distance-space clustering algorithm.In order to quantify the difference of regional development,the concept of spatial clustering regional development speed was proposed.C5.0 decision tree algorithm was used to process cellular state data,and the transition rules of cellular automata were obtained to form an improved model.Aiming at the problem of obtaining the classified load density,this paper adopts the radial function neural network model of the improved principal component analysis method.Through the improved algorithm,the existing algorithm can not fully reflect the nonlinear coupling between the influencing factors and the variation rate of the parameters themselves.The dimensionality reduction effect of the parameters is more obvious.In the face of different cellular load densities of the same kind,this paper puts forward the coordination coefficient of load density index among the cells to quantify the difference of spatial load densities.In terms of the spatiotemporal distribution prediction of EV charging load,in order to reflect the distribution of charging load on the spatial scale,this paper,based on the simulation results of land use,analyzes the parking demand of various types of land to form the parking generation rate,and then analyzes the driving law of EV and the change of charging state in the driving process,and fitting the probability distribution of the sunrise,ending time,driving distance and other behaviors of EV.Monte carlo model was used to simulate EV driving and charging process,and the spatial and temporal distribution of charging load was obtained.In order to avoid the time inconsistency when the charging load is superimposed with the cellular load,a simultaneous rate acquisition method is proposed in this paper.Through numerical simulation,the prediction results can be achieved with high accuracy.