摘要
针对传统的电网工程数据识别与检测严重依赖人工核算,实时性和准确性难以得到保证的问题。为准确、高效地识别电网工程的异常数据,文中建立了分层数据监测系统。为了区分异常数据与正常数据,引入NIDES统计学模型,建立归一化的统计学架构。在对数据进行分类的基础上,选取5种神经网络进行分类对比实验。实验结果证明,BP神经网络分类结果最优、成本最低。通过对搭载BP神经网络分类器的异常数据识别系统进行压力测试可知,该系统可以精确识别出仅占全局数据5%~10%的异常数据,且总体性能良好。
In view of the traditional power grid engineering data identification and detection rely heavily on manual accounting,real⁃time and accuracy are difficult to be guaranteed.In order to accurately and efficiently identify the abnormal data of power grid engineering,a hierarchical data monitoring system is established.In order to distinguish abnormal data from normal data,NIDES statistical model is introduced and a normalized statistical framework is established.On the basis of data classification,five kinds of neural networks are selected for classification and comparison experiments.The experimental results show that the BP neural network classification results are the best and the cost is the lowest.Through the pressure test of the abnormal data recognition system equipped with BP neural network classifier,we can know that the system can accurately identify the abnormal data which only accounts for 5%-10%of the global data,and the overall performance is good.
作者
杨文生
王雁宇
李海清
宦晓超
YANG Wensheng;WANG Yanyu;LI Haiqing;HUAN Xiaochao(Inner Mongolia Electric Power Economic and Technological Research Institute Branch,Inner Mongolia Electric Power(Group)Co.,Ltd.,Hohhot 010020,China)
出处
《电子设计工程》
2021年第7期157-160,165,共5页
Electronic Design Engineering
基金
内蒙古电力(集团)有限责任公司科技项目(内电科信[2019]6号1)。
