Research On Dynamic Air Conditioning Load Forecasting Method For Office Buildings

Accurate load forecasting is conducive to formulating a reasonable system operation strategy in advance according to load changes.It is the basis for optimizing the operation of HVAC systems,and provides an important basis for improving indoor thermal comfort and reducing building energy consumption.Therefore,this study mainly studies the dynamic thermal load forecasting model of buildings based on various meteorological factors and historical loads.Firstly,this paper utilizes the eQUEST energy simulation software to deeply study the the influence of meteorological factors on load,such as dry bulb temperature,humidity content,solar radiation intensity and wind speed,based on the sensitivity analysis method,and exploits the sensitivity coefficient as the evaluation index to screen out Meteorological factors that have a greater impact on the load.In order to reduce the error of the prediction model,this paper also considers the influence of building thermal inertia and building internal disturbance on the load.Therefore,historical meteorological parameters and historical load are introduced as the input variables of the model,and then the correlation analysis method is used to further optimize the input of load forecasting model.Secondly,based on BP and RBF neural network algorithms,four building dynamic thermal load forecasting models with different prediction time steps are established.And the combination of regression model and daily load factor is also adopetd to predict the daily average heat load and hourly heat load of the building.The selected impact factors are used as input variables of BP and RBF neural networks.The training samples adopt actual meteorological parameters and actual building loads.The average absolute relative error,absolute relative error and root mean square error are used to verify the accuracy of the load forecasting model.Finally,the variation of the prediction error of the typical winter model in winter and summer is analyzed,and the model correction coefficient is introduced to improve the overall accuracy of the model.The results show that:(1)BP neural network heating load prediction accuracy is higher,the model relative absolute error is 7.9%;(2)RBF neural network cooling load prediction accuracy is better,the model relative absolute error is 7.3%;(3)The office building format has a great influence on the accuracy of the forecasting model.