基于模型校准的建筑冷负荷短期预测模型人工内扰特征变量获取方法
Acquisition method of artificial internal disturbance feature variables for short-term prediction model of building cooling load based on model calibration
摘要：
准确的短期建筑冷负荷预测对于建筑供能系统的运行优化具有重要意义。数据驱动模型因在挖掘建筑实际负荷特性、提高预测精度方面具有较大的优势而得到广泛应用。然而，内扰特征变量的缺失严重影响着数据驱动负荷预测模型的预测效果。为此，本文提出了一种利用模型校准技术从冷负荷时间序列中反向挖掘内扰相关数据信息的方法。案例研究结果表明，利用该方法获得的人工内扰特征变量数据对使用人工神经网络模型的短期建筑冷负荷预测效果的提升具有显著作用。相比于完全缺失内扰特征变量的预测模型，预测误差可降低11.46%，相比于使用日历信息作为内扰特征变量的预测模型，预测误差可降低6.51%。
Abstract：
Accurate prediction of short-term building cooling load is of great significance to the operation optimization of building energy supply systems. Data-driven models have been widely used due to their advantages in mining the actual building load characteristics and improving the prediction accuracy. However, the absence of internal disturbance feature variables seriously affects the prediction effect of data-driven load prediction models. Therefore, this paper proposes a method to mine internal disturbance data from cooling load time series using model calibration techniques. The case study results show that the data of artificial internal disturbance feature variables obtained by this method can significantly improve the prediction effect of short-term building cooling load using the artificial neural network model. The prediction error can be reduced by 11.46% compared to the prediction model without internal disturbance feature variables, and by 6.51% compared to the prediction model using calendar information as internal disturbance feature variables.
Keywords:loadprediction;modelcalibration;internaldisturbanceschedule;data-drivenmodel;internaldisturbancefeaturevariable