Research On The Load Characteristic And Load Prediction Of Office Buildings In Tianjin

Reducing the heating and cooling energy consumption of office buildings caneffectively relieve the energy problem in our country as office buildings energyconsumption is huge. Currently, the HVAC systems often work at low efficiency andthe rough operation adjustment can’t meet the demand of building energyconservation. Load prediction is one of the key factors in the success of energy-savingmeasures. Therefore, the cooling and heating prediction models were established inthis study based on the monitoring data of several office buildings in Tianjin and thenrelated case was studied. The purpose of this study is to establish a simple andeffective load prediction model for office buildings and realize the goal of optimaloperation for HVAC system.In this paper, the load evaluation index was first established and the loadcharacteristic was analyzed. Then TRNSYS software was used to analyze thecomposition of load and the factors of cooling and heating load were screened. Forcooling load, the factors are dry-bulb temperature, wet-bulb temperature and buildingoccupancy rate. Dry-bulb temperature and building occupancy rate are the mainfactors of heating load.In general, building load can be decomposed into three parts: meteorological termwhich changed with outdoor meteorological parameters, internal term which changedwith building occupancy rate, random error term. It can be obtained from themonitoring data that:(1) The correlation coefficient of daily average heat load andaverage temperature is higher;(2) Daily average load is closely related to the holidayeffect;(3) The use of residual error correction can efficiently eliminate the influenceof random error. Therefore, multivariate nonlinear regression with residual errorcorrection model was established to predict the daily average heat load. The inputs ofmodel were daily average temperature and the holiday effect. Then the hourly heatload can be got by the daily distribution coefficient. Case study showed the predictionmodel has high prediction accuracy.People’s habit has an impact on the cooling load. Research showed that the cumulative effect of high temperature has a significant effect on daily average coolingload for office buildings. The multiple linear regression with residual error correctionmodel was used to predict the daily average cooling load. The inputs of the modelwere daily maximum dry-bulb temperature, maximum wet-bulb temperature, dailyaverage temperature and high temperature cumulative effect. Case studies showed thatthe prediction accuracy of the model can be controlled at10%. Then the hourlycooling load was predicted by the ARX model and the accuracy can be controlled at12%. BP neural network was used in this study to predict the next1hour cooling loadand results indicated the accuracy can be controlled at8%.