基于离群数据挖掘的多点电力负荷预测方法  

Multi-point Power Load Forecasting Method Based on Outlier Data Mining

作　　者：胡佳佳 杨洪耕[1] HU Jia-jia;YANG Hong-geng(School of Sichuan University,Electrical Engineering,Chengdu Sichuan 610000,China)

机构地区：[1]四川大学电气工程学院,四川成都610000

出　　处：《计算机仿真》2021年第12期66-69,93,共5页Computer Simulation

摘　　要：针对传统多点电力负荷预测方法未进行数据校正问题,提出离群数据挖掘的多点电力负荷预测方法。通过计算离群数据点与其邻近点的距离,进行离群数据的挖掘;依据离散傅里叶转换数据集合,实现离散数据点集合求取;采用储存间距获取数据点与其邻近点的距离;节点与整体单位设定单层或多层节点,根据整体辅助节点负荷构建多点预测模型;利用信息熵准则选取聚类的代表性点与合并点,通过负荷特征曲线达成数据校正,从而实现多点电力负荷的预测。仿真结果表明,提出的电力负荷预测方法不仅具有精准预测的有效性,噪声去除效果最佳,而且预测效率较高。Traditionally, the forecasting method ignores the data correction. Therefore, a method of forecasting multi-point power loads based on outlier data mining was presented. By calculating the distance between outlier data points and their adjacent points, the outlier data was mined. According to the discrete Fourier transform data set, the set of discrete data points was obtained. The distance between data points and adjacent points was obtained by the storage distance. The single-layer or multi-layer nodes were established for nodes and overall units. According to the whole load of auxiliary nodes, the multi-point prediction model was built. Moreover, the information entropy criterion was used to select the representative points and consolidation points of the cluster. Furthermore, the data correction was achieved through the load feature curve. Finally, the prediction of multi-point power load was completed. Simulation results show that the proposed method not only has the accurate prediction but also has the best denoising effect.

关 键 词：离群数据挖掘 多点电力 负荷预测 信息熵 离散傅里叶转换 

分 类 号：TM715[电气工程—电力系统及自动化]