Research On Methods Of Cooling And Heating Load Prediction For District Buildings

Building energy efficiency is an important part of China’s strategy of sustainabledevelopment. With the rapid development of urbanization in our county, there are more andmore buildings in form of community built district cooling and heating systems. Therefore,energy consumption of district buildings is concerned to be an important part of buildingenergy efficiency. In the process of district cooling and heating system design, we need tonearly accurately predict the aggregate cooling and heating load of district buildings inabsence of detailed information at construction planning stage, thus provide guidelines fordistrict buildings energy planning, project designing and product developing. Existing methodsof cooling and heating load prediction for buildings are mainly aimed at single buildings, butresearch methods on a group of buildings are relatively less or the prediction accuracy is nothigh. Therefore, a method of using computer simulation combined with statistical regression ispresented in this paper, which can be used to build a model to predict aggregate load of districtbuildings.This paper firstly annlyses existing methods of cooling and heating load prediction forbuildings and their limitations. Then, the measures of all kinds of factors are summarized byclassifying and analyzing the cooling and heating load influence factors of buildings and thestandard building model category is set up for district buildings. Utilizing those measures,builds all kinds of standard building model and proposes simplification principles. Accordingto the result of investigation, selectes appropriate factors and levels to design orthogonal test todetermine the sample size of standard building model. DesignBuilder is used to simulate thedynamic load of standard buildings.Based on the above analysis, obtained load are treated as prior information andinvestigative statistical datas other district buildings as sample information. Then, Bayesianregression model is established and solved to obtain posterior information which can be regardas load predictors. In addition, take a demonstrated community in a certain area as a case, thehourly cooling load of the community on typical day in summer is calculated by usingrespectively the conventional superposition model based on GFA and Bayesian regressionmodel respectively and the two predicting results are compared with measured values. Theprediction accuracy based on the two models is evaluated by three indictors: hourly relativeerror, mean square relative error, maximum error ratio.The comparison results show that the three kinds of error indictors of the prediction basedon Bayesian regression model are less than those of the conventional superposition model based on GFA, so it also shows that Bayesian regression prediction model is effective and havebetter prediction accuracy. Accordingly, during energy-use planning stage of district buildings,to obtain better prediction accuracy, the method of using computer simulation combined withstatistical regression is better than the conventional superposition method based on GFA.