饱和负荷预测中的多级聚类分析和改进Logistic模型

摘要

针对常用的Logistic负荷预测模型所需历史数据较多且预测结果不够准确的问题,提出了基于多级聚类分析的空间饱和负荷预测方法.首先,采用基于密度的改进K-均值聚类算法,按照变电站的位置分布对预测区域进行网格划分,生成Ι级元胞;接着,将对应于Ι级元胞的建设用地按其功能进行细分,生成Ⅱ级元胞;之后,运用一种改进的Logistic模型求取待预测Ⅱ级元胞的饱和负荷值;最后,将所提方法应用于山东省某市区电网的饱和负荷预测中.仿真结果表明所提方法预测精度较高,且适用于历史数据不够充分的新区.%Considering that the widely used Logistic model for predicting load requires a great volume of historical data and it cannot obtain accurate forecasting results,a multilevel clustering analysis based approach is presented for spatial saturated load forecasting. First,an improved density-based K-means clustering algorithm is employed,and a mesh structure is obtained for the area to be forecasted by considering the locations of substations. In this way ,the first-level cells can be generated. Then,the construction land corresponding to the first-level cells is further divided according to functions,and the second-level cells are thus generated. The value of saturated load in each second-level cell is further forecasted by an improved Logistic model. Finally,the proposed method is employed to forecast the saturated load in an urban power network in Shandong Province. Simulation results show that the proposed method can obtain accurate fore-casting results,and it is also applicable to new areas with insufficient historical data.