电力网络DCS数据库中的过负荷数据挖掘方法研究

Research on Overload Data Mining Method in Power Network DCS Database

电力网络中的分散控制系统(distributed control systems,DCS)数据库中寄存有海量的电力数据,进行电力系统智能调度和控制。对电力网络DCS数据库中的过负荷数据的有效挖掘是实现电力网络系统过载保护的关键环节。当前对DCS数据库的过负荷数据挖掘采用基于决策树特征分类方法进行特征提取和挖掘实现,在过负荷数据序列的广域子空间中产生大量干扰噪声,挖掘算法的置信度较低。提出一种基于经验模态分解和决策树分类结合的电力网络DCS数据库中的过负荷数据挖掘方法。构建了电力网络的DCS数据库结构模型,在DCS数据库中进行数据流信号模型构建,采用经验模态分解算法对数据信号流进行固有模态时频特征提取,以此特征为基础,采用决策树分类算法实现过负荷数据的准确检测和挖掘。仿真结果表明,采用该算法能有效实现对电力网络DCS数据库中的过负荷数据的特征提取和分类挖掘,误码率较低,性能优越于传统算法。

A mass of power data is stored in the distributed control system( distributed control systems, DCS) database of the power network for the intelligent scheduling and control of the power system. The effective mining of the overload data in the DCS database is a key link in the realization of the overload protection of the electric power network system. At present, the overload date in the DCS database is mined based on the decision tree feature classification method to extract features and realize mining, and this method produces a lot noise in the wide area subspace of the overload data sequence, therefore the mining algorithm is of low confidence. To this end, a method of data mining based on combination of the empirical mode decomposition and decision tree classification is proposed in this paper. The DCS database structure model of the power network is built and the data flow signal model is built in the DCS database, and the intrinsic mode time-frequency feature are extracted in the data signal flow using the empirical mode decomposition algorithm. On the basis of the extracted feature,the accurate defection and mining of the overload data are realized using the decision tree classification algorithm. The simulation results show that the proposed algorithm can effec-tively extract and classify the overload data in the DCS database of the power network, and the error rate is low, and the performance is superior to the traditional algorithm.