基于数据驱动的办公建筑电耗预测方法

Power Consumption Prediction Method of Office Building Based on Data Driven Model

对建筑电耗的有效预测能够为用能诊断、运行优化及区域能源管理提供有效的支撑。基于支持向量回归,建立复合的数据驱动模型来预测典型公共建筑的逐时耗电量。该模型的核心为根据电耗数据,利用关联挖掘并结合模糊聚类划分电耗类型,采用支持向量回归模型和小波分解技术实现对建筑电耗的非线性平稳预测和修正。实例应用结果表明:关联挖掘与聚类算法的结合可以实现对建筑电耗特征的清晰归类,小波分解的引入能够增强支持向量回归模型对于弱规律数据预测的泛化能力。

The dynamic prediction of building energy consumption can provide effective support for energy diagnosis,operation optimization and regional energy management. Based on support vector regression,a composite data-driven model is established to predict the hourly power consumption of typical public buildings. Using the method of association mining and fuzzy clustering to classify the power consumption data,further support vector regression model and wavelet decomposition technology are used to realize the nonlinear prediction and correction of building power consumption,which is the core of the model. The application shows that the combination of association mining and clustering algorithm can achieve a clear classification of building electricity consumption characteristics,and the introduction of wavelet decomposition can enhance the generalization ability of support vector regression model for weak regular data prediction.