Research And Application Of Short-term Load Forecasting Based On Electricity Market

Electricity is the lifeblood of the country,which is the one of the most important basic industries,and it's inseparable with the country's economic and social development.Short-term load forecasting plays an important role in the safe dispatching of power system.The electric power market reform is the inevitable trend and requirement of the development of our country.In the open market environment,the load will be directly affected by the real-time changing electricity price,which increases the difficulty of load forecasting to some extent,and the accuracy and speed of load forecasting will directly affect the economic cost of operators in the power market.Nowadays,although China's electricity market is not yet fully open,it is of great significance to study short-term load forecasting under real-time electricity price by understanding and studying the mature operation modes in developed countries and conducting in-depth research on their operation data.Firstly,through the analysis of historical load data,it is found that the load fluctuation has obvious regularity characteristics in different time scales,which can provide reference for accurate load prediction.Then,studies the influence factors for the load fluctuation and Pearson correlation coefficient method is used respectively to the correlation of electricity price and the historical load and forecasting load are analyzed,and analyzes the real-time electricity price,the relevance of historical load and current load level,through the analysis on electricity price,historical load and forecasting load is strong correlation,so the historical load and real-time electricity price to join the two influence factors as input vector are considered in the model prediction.Considering the temporal characteristics of power load,the obvious advantages of LSTM network in dealing with the problem of sequence prediction,and the great success of LSTM network in some fields,this paper establishes a load prediction model based on LSTM network.This paper analyzes the influence of the parameters of the model,including the setting of the number of neurons,the selection of activation function,the selection of the optimizer and the formulation of network depth,on the prediction effect of the model.Aiming at other factors that affect the prediction accuracy of the model,the model is tested with different training set sizes,unsynchronized length and different input vectors,and the prediction ability of the model under various conditions is analyzed.Finally,the best scheme of the model is determinedIn order to verify the high accuracy of the LSTM model in this paper,the convolutional neural network and random forest in the depth neural algorithm are selected,and the prediction results of the three algorithms are compared.It is confirmed that the three algorithms have high prediction accuracy,and the prediction accuracy of the LSTM optimal configuration model in this paper is slightly higher than that of the other two algorithms.