Research On Optimizing Integrated Intelligent Analysis Methods Of Building Energy Consumption

In the composition of social terminal energy consumption in China,building energy consumption accounts for more than 33% of the total energy consumption in cities and towns.Therefore,in order to achieve the goal of a conservation-oriented society,it is very important to reduce building energy consumption.The analysis of historical energy consumption data is an important way to save energy.The integrated intelligent analysis method of building energy consumption can be used to analyze building energy consumption data comprehensively and accurately from many aspects.IIT(Integration Intelligent Technology)method is one of the best integrated intelligent analysis methods for building energy consumption at present,but there are still some problems in IIT method,such as inadequate data preprocessing,limitations of clustering algorithm,slow convergence speed and local optimum of energy consumption prediction algorithm,and incomplete knowledge discovery.Therefore,in order to improve the effectiveness of building energy-saving decision-making,this paper optimized IIT method and tried to proposed an integrated intelligent analysis method of building energy consumption,IABEC(Integrated Analysis of Building Energy Consumption),which has better performance and more comprehensive knowledge discovery.As follows:(1)According to the characteristics of large numerical differences among different attributes of building energy consumption data,the improved Robust Scaler algorithm is used to normalize the building energy consumption data.Through theoretical analysis of the main clustering algorithms,the GMM clustering algorithm is selected to add energy consumption pattern labels to the data.(2)On the basis of the functions of IIT building energy consumption integrated intelligent analysis method,the fault prediction function is added.Through theoretical analysis of the mainstream fault prediction algorithm and comparative experiments of fault prediction accuracy,XGBoost is selected as the fault prediction algorithm.(3)According to the experimental results in reference [8],BP neural network is selected as the initial algorithm of building energy consumption prediction.In order to train the prediction network model faster and improve the accuracy of prediction,A method of synthesizing gradient descent method and a new improved conjugate gradientmethod is used to replace the gradient descent method to improve the BP neural network algorithm.(4)Integrated the optimized RobustScaler feature processing algorithm,GMM clustering algorithm,XGBoost fault prediction algorithm,LOF outlier analysis algorithm and gray relation analysis algorithm optimized from literature [8],and the improved BP neural network algorithm in this paper,IABEC,a new intelligent analysis method for building energy consumption,is proposed.(5)Based on the data set of building energy consumption from the National Renewable Energy Laboratory Research Support Agency and the historical data collected from a university building energy consumption monitoring platform,the following experiments were carried out respectively:1)The experiments carried out by using IABEC method,IIT method and single outlier analysis method in order to analyze outliers,and the experimental results show that the accuracy of IABEC method is 16.8-44.8% higher than that of using IIT method or single outlier analysis method,respectively.2)Experiments on energy consumption prediction using both IABEC method and IIT methods shows that the accuracy of energy consumption prediction by using IABEC method is improved by 13-15.7% compared with that by using IIT method,and the number of iterations is decreased by 24-34.2%.3)The energy fault prediction experiment was carried out using IABEC method and a single fault prediction method.The experimental result shows that the accuracy of the fault prediction is improved by 2.9% and the average absolute error is reduced by11.0% compared with the single fault prediction method.(6)Design and implement an integrated intelligent analysis system for building energy consumption.Its main functions are to use IABEC method to achieve energy consumption pattern recognition,outlier detection,energy consumption prediction and fault prediction of energy equipment.