Power Load Forecasting And Application In The Optimal Configuration Of Distributed Energy Storage

With the integration of clean energy such as wind and solar,the power grid is facing challenges in the balance of energy supply and demand.And the rapidly increasing charging load of electric vehicles further threatens the stability of the distribution network.Therefore,the research and application of distributed energy storage has gradually gained attention of the industry.In recent years,these complex changes in power grid attributes have brought new issues to load forecasting,and they have also brought opportunities for improving the demand response capability of distribution networks through load forecasting.Aiming at the background of the current large-scale access of electric vehicles,this thesis studies a load forecasting method that consider the access of electric vehicles,and optimizes the configuration of distributed energy storage in the system based on the prediction results.In this thesis,the Monte Carlo simulation method is used to establish the charging(discharging)load model of electric vehicles.And then the charging(discharging)load characteristic curves of electric vehicles under the disorder charging,orderly charging,and V2 G methods are obtained.This thesis discusses the influence of these modes on the grid load characteristics and load forecast results.Subsequently,a CNN-LSTM joint prediction method considering the impact of electric vehicle charging load is proposed.This method is constructed based on the characteristics of Convolutional Neural Network and Long Short-Term Memory Network.It separately predicts electric vehicle charging load and conventional load,and superimposes the total predicted load value.Compared with other methods,it is verified that the CNN-LSTM prediction method proposed in this thesis has better prediction accuracy.And this method has good applicability to the scenarios where a large number of electric vehicles are connected.Subsequently,this thesis established the charge and discharge model of distributed energy storage equipment,and simulated the charge and discharge behavior of electric vehicles when participating in V2 G,and then analyzed the impact of the above behavior on the grid load characteristics.What’s more,this thesis discussed the applicability of CNN-LSTM prediction method in the case when electric vehicles participate in V2 G discharge.On this basis,combined with the results of power load forecasting,this thesis proposes a distributed energy storage optimization configuration algorithm aimed at minimizing the peak-valley difference of the system.First,the principle of selecting points and determining capacity for distributed energy storage devices was formulated.The standard deviation of network loss sensitivity was used as the location index;then combined with the previous CNN-LSTM joint prediction method,the traditional load of the system and the charging(discharging)load of electric vehicles were predicted.Furtherly,aimed at minimizing the peak-valley difference of load,this thesis used the PSO algorithm to optimize the charging and discharging power of energy storage,and determines the appropriate capacity for energy storage devices connected in the four scenarios where there is no electric vehicle access in the distribution network,electric vehicles are disorderly connected,electric vehicles are orderly connected,and electric vehicles participating in V2 G dispatching.The peak-valley difference before and after the energy storage devices were connected,network loss and other system parameters were compared.Finally,this thesis verified the operation status of energy storage in the actual system,and compared the investment costs and expected benefits of energy storage in the four scenarios.