Study On Method Of Air-Conditioning Load Prediction For Large Commercial Buildings

As the HVAC system in commercial buildings generally consumes more energy, the load prediction in large commercial buildings was explored in this paper in order to optimizate the system operation strategies and estimate the quantity of energy storage in energy storage system. The significant factors influencing air-conditioning load were analyzed, and the load prediction model was built based on weather forecast and internal occupancy density.Firstly, taking Tianjin as an example, the feature models of large commercial buildings were built based on survey data of standards and 30 commercial buildings in Tianjin. Using the orthogonal experiment, the internal occupancy desity was found to have a significant impact on the air-conditioning load in commercial buildings. Secondly, the delayed effect of meteorological parameter and internal occupancy density on air-conditioning load was analyzed to determine the input parameters of the prediction model. Thirdly, the stepwise regression was applied to remove the non-significant factors to simplify the model. Finally, the applicability and the accuracy of some predicition methods were analyzed. It was found that a multiple linear feedback regression achieved a promising accuracy to forecast hourly cooling load. The paper proposed a modified seasonal exponential smoothing model with inputs to forecast the hourly heating load, and the highly accurate forecast result confirmed that the method was applicable.Through the study, the occupancy density was found to be a significant factor affecting air-conditioning load. The fresh air volume and other internal loads generally have a more prominent impact on air-conditioning load than other factors regarding building construction features.That adding occupancy density to the prediction model improved the accuracy of the prediction model. The prediction model was applied to air-conditioning load forecast in a commercial building in Tianjin. The result indicated that a multiple linear feedback regression provided a highly accurate forecast of the hourly cooling load, producing MRE and RMSE values of 7.90% and 316.8, respectively. The maxRE and minRE for daily cooling load are respectively 5.14% and 0.108%. The modified seasonal exponential smoothing model with inputs to forecast heating load was also accurate, giving the MRE and RMSE values of 10.05% and 262.0 for hourly values, respectively. The max RE and minRE for daily heating load were 14.5% and 0.506%, respectively. Both the two methods resulted in promising forecast accuracy. Therefore the load prediction model based on weather forecast and internal occupancy density was proved to be accurate and applicable.