Research On Short-term Forecasting Of Multiple Loads In A Regional Integrated Energy System

District-level Integrated Energy System(DIES)is the inevitable product of the development of the times and plays an important role in the modern energy system.Due to the multi energy flow coupling and multi system integration,the load characteristics of DIES are changeable,the influencing factors are complex and the nonlinearity is stronger.In order to keep the dynamic balance between energy supply and demand of DIES,more accurate short-term load forecasting of DIES is needed.According to the characteristics of DIES,this dissertation studies the short-term load forecasting of DIES.Firstly,the characteristics of DIES load are analyzed,and the grey relation analysis(GRA)method is used to quantitatively analyze the correlation coupling between multiple loads and meteorological factors.According to the analysis results,a short-term forecasting model of DIES multiple loads based on the combination of GRA Method and LSTM is constructed.The model uses the advantages of long short term memory(LSTM)neural network in processing samples and nonlinear data with long interval or delay in time series to make short-term forecasting of DIES multiple loads.The analysis results of an example show that the dies multiple load forecasting model based on the combination of GRA Method and LSTM has better forecasting accuracy.The LSTM network model has three gating structures and retention time characteristic information,which results in large amount of calculation and slow decline of error.In order to improve the prediction level of the model,this dissertation improves the LSTM model,and constructs the DIES multivariate short-term load forecasting model based on PSO-Att-ALSTM.The attention layer and dropout layer are added to the LSTM prediction model.The attention mechanism can give different weights to the hidden layer of the model.The dropout layer can regularize the model.Particle swarm optimization(PSO)is used to optimize the parameters of the prediction model,and Adam algorithm is used to optimize the weights and thresholds of the updated model.The simulation results show that the proposed PSO-Att-ALSTM prediction model can effectively improve the prediction accuracy.Considering the coupling and complexity between multiple loads,the DIES short-term multiple load forecasting model based on the combination of CNN and PSO-Att-ALSTM is constructed.The new model complements the advantages of CNN and LSTM,and the CNN model is used to extract the feature information,reduce the dimension and decouple the input sample data.LSTM model is connected with CNN,and its function is to forecast multiple loads and fully mine the correlation information between time series.Through the analysis results of the example,the prediction model with CNN has better prediction accuracy,which proves the feasibility of the model,and provides some reference for the future research of DIES multiple load forecasting.