Research On Public Building Energy Consumption Forecasting Method Based On Combination Model

In recent years,the proportion of energy consumption of public buildings in the total building energy consumption continues to rise,and the energy consumption per unit area is large and there is a situation of unreasonable energy utilization.Predict the energy consumption of public buildings accurately and efficiently,and help managers formulate economical and reasonable energy distribution plans,the auxiliary building intelligent system completes effective system optimization and control,which can reduce building energy consumption to a certain extent and achieve the purpose of building energy saving.However,due to the influence of factors such as personnel density,weather factors,and air conditioning on time,there are uncertainties and randomness in building energy consumption data.The traditional single prediction method can no longer meet the requirements of building energy consumption prediction.In contrast,the combined forecasting method has greatly improved the performance of building energy consumption forecasting,which has attracted more and more researchers attention.This paper proposes three combined forecasting methods to forecast building energy consumption.The main research content is carried out from the following three aspects:Aiming at the problem of building energy consumption prediction,this paper uses Sketch Up software to draw a three-dimensional model of office buildings in Xi ’an City.Energy Plus simulation software is used to simulate the energy consumption of the building,and the hourly energy consumption data set of the building in summer working conditions is obtained.On the basis of using sinusoidal function to adjust the inertia weight of particle swarm optimization algorithm,the SFPSO algorithm is established to optimize the support vector regression combination prediction model.Based on the building hourly energy consumption data set above,the SVR prediction model,PSO-SVR prediction model and SFPSO-SVR prediction model are built respectively.The results show that the proposed combined prediction model can effectively improve the prediction accuracy.Aiming at the problem of multiple correlation between noise and energy consumption influencing factors in building energy consumption data,this paper first carries out principal component analysis on energy consumption influencing factors,obtains the principal component of the maximum contribution rate,and determines the input variables of the model.Then the pre-processed building energy consumption data and principal components are taken as the input data of the long and short term memory network,and the PCA-LSTM model is constructed,and the hourly energy consumption of the building is preliminarily predicted.Finally,use the MLR algorithm to optimize the PCA-LSTM prediction results,and then build the PCA-LSTM-MLR combined prediction model to predict building energy consumption.Based on the above energy consumption simulation data set,LSTM prediction model,PCA-LSTM model and PCA-LSTM-MLR model were built respectively.The results show that compared with the LSTM model,the PCA-LSTM model and the SFPSO-SVR model,the combined model has improved energy consumption prediction accuracy by 9.27%,6.23% and2.37%,respectively,and can more accurately predict the peak energy consumption of buildings.Aiming at the periodic oscillation characteristics of monthly energy consumption of buildings,this paper designs a cumulative gray radial basis function neural network combination model based on function combination transformation.Firstly,based on the design of the combined transformation method of inverse trigonometric function and power function,the TGM(1,1)model is established.Then use the cumulative method to estimate the model parameters,and build the cumulative TGM(1,1)model.Finally,the RBF network is designed to correct the residual error of the predicted value of the cumulative TGM(1,1)model,and then build a cumulative TGM-RBF combined prediction model to predict building energy consumption.Using the monthly building energy consumption simulation data and the actual energy consumption data of the Chinese construction industry from 2009 to 2018,the GM(1,1)prediction model,cumulative TGM(1,1)model and cumulative TGM-RBF model building energy consumption were constructed respectively make predictions.The results show that the combined prediction model has higher prediction accuracy than the GM(1,1)model and the cumulative TGM(1,1)model.