Research On The Analysis And Prediction Method Of Building Energy Consumption Based On Neural Network

With the rapid development of economy and urbanization,the energy consumption of large buildings leads to the excessive emission of carbon dioxide(CO2),which makes the environmental pollution and climate warming more and more serious.Therefore,reducing the energy consumption of buildings is one of the important means to reduce CO2 emissions and environmental pollution.The main factors affecting the energy consumption of building operation are analyzed,and the construction plan is adjusted and optimized,which has important scientific research significance and a wide range of application scenarios.The factors that affect the energy consumption of buildings mainly include the building materials,the building envelope and the climate(temperature,humidity,etc.)where the buildings are located.Therefore,researchers predict and analyze the energy consumption of buildings,find out the ways to reduce the energy consumption of buildings,and provide opinions and suggestions for the construction plan,which will help to improve the building energy efficiency.This paper studies the neural network method to predict the load data and power consumption data of energy consumption of buildings.Through the prediction of heating and cooling load and the analysis of building envelope,the effectiveness of radial basis function(RBF)neural network based on affinity propagation(AP)algorithm(AP-RBF)is verified.Furthermore,in the case of large amount of data,through the prediction of energy consumption of buildings,the analysis of temperature and humidity around buildings,the validity of DDAMSGrad model is verified.The results of the application case study show that the building construction plan can be adjusted and optimized according to the results,so as to reduce the building operation energy consumption and achieve energy conservation and emission reduction.The main work of this paper is as follows:1.In view of the error caused by the hidden layer selection when the traditional artificial neural network models in different data sets,this paper proposes an AP-RBF method.AP clustering is used to process the data,and the clustering center is obtained adaptively as the hidden layer of RBF.Compared with the classical RBF network using K-means(K-means)clustering method as a way to select hidden layer nodes,AP clustering method avoids the error caused by the artificial number of clustering centers.At the same time,when AP clustering is initialized,all nodes are potential clustering centers by default,which makes the model more stable.Through the comparison with other methods,the effectiveness of the method is verified,which can provide guidance for reducing the energy consumption of building operation,and then reduce CO2 emissions.2.In view of the problem that the traditional neural network training time is long and over fitting or under fitting phenomenon is easy to occur when the data features and the amount of data increase,this paper proposes a method based on DD-AMSGrad.This method does not need to specify the number of hidden layer nodes,and also avoids the error caused by the number of hidden layer nodes.In the training process of using gradient descent method,AMSGrad method is used to update the weights,seek the optimal solution,and finally get the model adaptively.By predicting and analyzing the influence of humidity and temperature on operation energy consumption of buildings,the effectiveness of the method is verified,and effective suggestions for building location are provided.3.The prototype system of energy consumption of buildings prediction and analysis based on neural network is designed and developed.The functions of the system are described in detail.The application of the research results in energy consumption of buildings analysis is visualized,which makes energy consumption of buildings prediction and analysis more convenient and efficient.