Data-driven Prediction Model Of Building Electric Energy Consumption

The construction industry plays a key role in promoting the development of the national economy and society.However,my country’s building energy consumption accounts for about 27.5% of the country’s total consumption.With the continuous acceleration of industrialization and urbanization,this proportion is still rising rapidly.Therefore,the use of building energy must be reasonably planned and managed to achieve the goal of reducing building energy consumption.Building energy consumption forecasting is the basis of energy planning,management and conservation.The current building electricity energy consumption forecasting model is mainly aimed at buildings with a fixed operating mode,which has the problem of insufficient flexibility and versatility.Therefore,based on clustering algorithm,modal decomposition algorithm and prediction algorithm,this thesis proposes a hybrid building electricity energy consumption prediction model,and verifies it with real data.At the same time,in order to provide a reliable data support for the proposed energy consumption prediction model,a comprehensive building data preprocessing model is proposed.The main research contents of this thesis are as follows:(1)This article systematically discusses the background,significance and research status of the subject research,analyzes and researches building electrical energy consumption data,points out the current shortcomings of building data preprocessing and building electrical energy consumption prediction models,and determines the main research content on this basis.(2)Analysis and research on data types of building electricity consumption.The building energy consumption data is divided into four categories based on the nature of electricity consumption.Show each part of energy consumption data graphically and analyze the change trend;based on the correlation coefficient,this thesis explores the relationship between energy consumption value and temperature and humidity,and provides reference for the subsequent formulation of relevant strategies.(3)Research on preprocessing of building electricity energy consumption data.When building data is collected,it will be affected by external factors,leading to deviations in the data.Analyze the energy consumption data of four different types of buildings,summarize four types of common abnormal data,and propose different processing methods for the four types of abnormal data: use the R language to identify zero values and continuous repeated data;The K-means algorithm is used to process discrete abnormal data,and the number of clusters is optimized;the K-Nearest Neighbor(KNN)algorithm is used to fill in missing data.(4)Research on prediction model of building energy consumption.In order to solve the problem that the existing building energy consumption forecasting model is not universal and flexible,a hybrid forecasting model of building energy consumption is proposed.Fuzzy C-means(FCM)algorithm is used to cluster the data related to architecture,and the time with similar running characteristics is clustered into one class.The Ensemble Empirical Mode Decomposition(EEMD)algorithm is used to decompose the hourly energy consumption data in each cluster to improve the linear relationship between the energy consumption data.Three prediction models are established respectively for each component obtained by modal decomposition,and the parameters are optimized.Calculate the error between forecast value and actual value,choose the best forecast method,and finally carry on the conformity.