摘要
传统能耗异常检测方法大多使用固定阈值比较法、缺乏自适应能力等问题。为提高企业能耗异常检测精度,将使用基于PSO-BP神经网络的预测方法获取分工序各批次正常情况的预测能耗,根据误差的概率分布动态设置能耗的置信区间,并依据实际能耗值是否落在预测能耗置信区间范围内来判定是否出现能耗异常检测现象,从而构建了铜管能耗异常检测模型。通过对比使用不同方法构建的异常检测模型的漏检率和误检率,证明此模型能很好地减少漏检、抑制误检。
The traditional energy consumption most anomaly detection method using a fixed threshold comparison method,lack of adaptive ability problems.In order to improve the enterprise energy consumption of anomaly detection accuracy,this paper will use the PSO- BP neural network based prediction method for the division of sequence forecast energy consumption of each batch normal,according to the error probability distribution dynamic set the confidence interval of energy consumption,and on the basis of the actual energy consumption value is fall within the scope of the forecasting confidence interval of energy consumption to determine whether any anomaly detection of energy consumption phenomenon,so as to build the energy consumption of copper anomaly detection model.By comparing with different methods to build the anomaly detection model of miss rate and false detection rate,it is concluded that this model can reduce leak and inhibit mistakenly identified.
作者
代德宇
何江涛
吴中元
DAI De-yu HE Jiang-tao WU Zhong-yuan(College of Mechanical and Electronic Engineering , Guangdong University of Technology , Guangzhou 510006, China)
出处
《机电工程技术》
2016年第9期128-132,共5页
Mechanical & Electrical Engineering Technology
关键词
异常检测
优化BP网络
铜管
模型
anomaly detection
optimize BP network
well Copper pipe
model
作者简介
代德宇，男，1990年生，湖北仙桃人，硕士研究生。研究领域：智能制造、低碳制造。
