Spatial-temporal Information Based Charging Station Load Forecasting And V2G Pricing Strategy

In recent years,with the opening up of the electricity market and deepening of the energy internet,a large number of distributed energy equipment participate in the operation of the electricity market.As one of the special load types,electric vehicles（EVs）have random charging uncertainties in temporal,spatial and behavior.Disorderly access of large-scale EVs to the power grid may lead to excessive fluctuation of the grid load and endanger the grid operation security.Charging stations are the main places where EVs can be connected to the power grid.Forecasting the load variation of charging stations can effectively avoid any potential power grid risks.At the same time,based on Vehicle to Grid（V2G）technology,EVs can discharge to the Grid to assist adjusting the load change,while reasonable V2 G incentive measures can optimize the load distribution of EVs so as to guarantee the stability of the Grid.This paper mainly studies the charging station load forecasting and V2 G pricing strategy,aiming to reduce the adverse impact of large-scale EVs access to the grid and promote the coordinated deployment of EVs and the grid.The main contributions of this thesis are summarized as follows:1.A multi charging station collaborative load forecasting model（MCSCLF）based on graph neural network is proposed.According to the temporal and spatial load information of the multiple charging stations,the graph structure learning method based on metric learning is introduced to investigate the implicit correlation among the multiple charging stations and generate graph structure,then the discrete load sequence data can be converted into graph data so as to carry out the collaborative load prediction of the multiple charging stations via graph neural network.A feature extraction module is designed based on graph convolutional neural network and temporal convolutional network to extract the spatial load features.In order to realize the joint training with the graph structure learning module and the spatial-temporal feature extraction module.Further,a joint loss function is constructed to complete the end-to-end load prediction.2.A V2 G pricing strategy based on market trading model and two-level auction mechanism is designed to encourage EVs to assembly participate in the grid load regulation.The trading models of the grid,aggregators and EVs are established respectively to describe the behaviors of each market entity and analyze their operational costs and benefits.A two-level auction mechanism from the grid and aggregators,and aggregators and EVs is constructed,while EVs can be orderly scheduled charging and discharging for load regulation participation in the grid.Finally,MOEA/D,a heuristic method is used to design the optimal electricity price to achieve balanced profit distribution among the grid,aggregators and EVs.3.In order to ensure the security and fairness of the V2 G market transactions,we have built a V2 G trading system based on hierarchical blockchain architecture,considering the advantages of the blockchain platform,such as imtamability and traceability.Bottom-up pattern is applied,including the data blockchain layer,operation blockchain layer and transaction blockchain layer,to achieve the data access,intelligent contract operation and transaction execution functions,so as to realize the automated operation of the V2 G market on the blockchain.Finally,the developed MCSCLF model,V2 G pricing strategy and V2 G trading system are verified with simulation experiments.The dataset is constructed based on the load data of the public charging stations in Shenzhen,and LSTM and MLP are selected as the benchmark models for comparation.A simulation experiment platform of V2 G trading system on block chain is designed via cellular automata for grid load variation comparison in the trading process.The results demonstrate that the designed V2 G pricing strategy can effectively stimulate orderly EV charging and discharging,achieving the objective of peak cutting and valley filling,thus the efficiency of the whole V2 G trading process can be guaranteed.