M-TAEDA:多变量水质参数时序数据异常事件检测算法

发布时间：2018-03-04 07:28

本文选题：无线传感器网络　切入点：异常事件检测　出处：《计算机应用》2017年01期 　论文类型：期刊论文

【摘要】：在供水管网中部署传感器网络实时获取多个水质参数时间序列数据,当供水管网发生污染时,高效准确地检测水质异常是一个重要问题。提出多变量水质参数时间异常事件检测算法(M-TAEDA),利用BP模型分析多变量水质参数的时序数据,确定可能离群点;结合贝叶斯序贯分析独立更新每个参数的事件概率,预测单个传感器节点检测的异常概率;将单变量的事件概率融合为统一多变量事件概率,融合判断异常事件。实验结果表明:BP模型模拟多变量水质参数进行预测可以达到90%精确度;与单变量参数时间异常事件检测算法(S-TAEDA)相比,M-TAEDA可以提高异常检出率约40%,降低误报率约45%。
[Abstract]:The sensor network is deployed in the water supply network to obtain the time series data of multiple water quality parameters in real time. When the water supply network is polluted, It is an important problem to detect water quality anomalies efficiently and accurately. An algorithm for detecting multivariable water quality parameters is presented in this paper. BP model is used to analyze the time series data of multivariate water quality parameters to determine the possible outliers. Combined with Bayesian sequential analysis to independently update the event probability of each parameter, the anomaly probability of single sensor node detection is predicted, and the single variable event probability is fused into a unified multivariable event probability. The experimental results show that the multivariable water quality parameters can be predicted with 90% accuracy by using the component BP model. Compared with the single variable parameter time anomaly detection algorithm (S-TAEDAA), M-TAEDA can increase the abnormal detection rate by about 40% and reduce the false alarm rate by about 45%.
【作者单位】：河海大学计算机与信息学院;
【基金】：国家自然科学基金资助项目(U1301252) 国家科技支撑计划项目(2013BAB06B04) 国家重点研发计划项目(2016YFC0400910) 中国华能集团公司总部科技项目(HNKJ13-H17-04) 云南省科技计划项目(2014GA007) 中央高校基本科研业务费专项资金资助项目(2015B22214)~~
【分类号】：X832;TP212.9;TN929.5

【相似文献】