Prediction Of Energy Consumption Of Large-scale Civil Buildings In Hong Kong Based On Machine Learning

With the acceleration of urbanization in China,the contribution of building energy consumption to total energy consumption becomes more and more important.With the advent of big data era,building data collection and energy consumption model building is the primary issue of building energy consumption research.In this paper,web crawler was used to collect data,Thirteen machine learning regression algorithms were used to build the energy consumption model,and the model was filtered by the performance evaluation parameters.Taking 1923 commercial buildings in Hong Kong as samples,this paper used Python programming language to collect and preprocess 43 parameter data including physical parameters,usage parameters and environmental parameters.Using 13 kinds of regression model of building energy consumption of machine learning training,including linear regression ? ridge regressor ? SVR ? Lasso ?Elastic Net?linear SVR?ada Boost?bagging?XGBoost?random forest regressor?extra trees regressor ? MLP regressor and KNN regressor,based on 13 used regression model to investigate the construction of physical parameters and environmental parameters for the effect of building energy consumption.Commercial building energy consumption model in the early stage of the study,on the basis of this paper is trying to train complete regression model for Hong Kong40716 of the civil building energy consumption prediction,combining the geographical position,population density,land use and other conditions,the fixed number of year of the building energy consumption and population density,building,land use,such as index of correlation analysis,the main factors that influence the civil building energy consumption.It has certain guiding significance to municipal planning and building planning.The results show that the four regression models of bagging,XGBoost,random forest regressor and extra trees regressor are the most accurate for the prediction of building energy consumption.At the same time,it was found that the parameters with the largest impact on the prediction of building energyconsumption were successively the building square footage,whether it was an independent building,wind environment parameters with a height of 500 meters,tenant status,point solar radiation,building height and number of floors.For the samples containing the data of these seven parameters,building energy consumption could be predicted more accurately.The parameter that has the greatest impact on building energy consumption is the building square footage.In civil buildings in Hong Kong,with the increase of building square footage and building height,the average EUI value increases,and the EUI value distribution becomes more and more narrow and stable.For buildings with low building square footage and building height,the average EUI value is small,but the EUI value varies greatly from building to building,especially for buildings with building square footage less than10 * 105m2 and building height less than50 m.It is necessary to further analyze the characteristics of this kind of buildings with low building square footage and building height,reveal the fundamental reasons for the great differences in EUI,and put forward measures or strategies to reduce EUI.High-energy buildings are mainly found in areas with high population density and high building density on Hong Kong Island,as well as in areas with low population density and low building density in the New Territories and Kowloon.