用于负荷预测的层次聚类和双向夹逼结合的多层次聚类法

Load Forecasting by Multi-Hierarchy Clustering Combining Hierarchy Clustering with Approaching Algorithm in Two Directions

针对传统聚类方法对负荷曲线形状的相似性重视不足的问题,提出了一种基于相似性原理的新的聚类方法——层次聚类和双向夹逼相结合的多层次聚类方法,该方法可以同时衡量负荷曲线形状的趋势相似性和形状相似性。分别采用该方法与传统的基于欧氏距离的层次聚类方法以某省2005年负荷数据为历史数据进行预测,结果表明本文提出方法对负荷曲线形态细节以及气候因素与负荷之间的复杂相关性具有较强的识别能力。

In order to overcome the defect of traditional clustering method which has no regard for the similarity of the shapes of load curves, a new similarity principle based clustering method, i.e., multi-hierarchy clustering, in which the hierarchy clustering combines with approaching algorithm in two directions is proposed, The proposed method can measure the tendency similarity and shape similarity of load curves simultaneously. By use of both the proposed method and traditional Euclidean distance based hierarchy clustering and taking the load data of a certain provincial power network in the year of 2005 as historical data, the load are forecasted respectively by these two methods. Forecasting results show that the proposed method possesses better ability to recognize the details of load curve shapes and the complex correlativity between climatic factors and loads.