管网水质多指标动态关联异常检测方法
发布时间：2018-08-04 09:47
【摘要】：为了提高城市供水管网水质监测系统的污染检测能力,利用污染物所引起多个指标变化之间的关联特性,提出基于多常规水质指标动态关联分析的水质异常检测方法.应用动态时间规整算法(DTW)衡量多个常规水质指标时间序列间的动态距离,刻画各指标波动的相似程度和动态关联特性.利用D-S证据理论融合各指标单独的异常概率,将融合后得到的供水管网水质异常概率与所设定的多指标融合异常概率阈值进行比较,作出当前时刻水质是否存在水质异常的综合判断.依托课题组模拟供水管网实验系统,设计不同浓度的硫酸铜和铁氰化钾污染物的注入实验,利用在线监测的pH值、浊度、余氯、溶解氧等8种常规水质指标进行动态关联分析和水质异常检测,方法的可行性和异常检测性能通过受试者工作特征曲线(ROC)进行验证.
[Abstract]:In order to improve the pollution detection ability of the water quality monitoring system of urban water supply network, a water quality anomaly detection method based on dynamic correlation analysis of multiple conventional water quality indexes is proposed. Dynamic time warping algorithm (DTW) is used to measure the dynamic distance between time series of several conventional water quality indexes and to characterize the similarity degree and dynamic correlation characteristics of each index fluctuation. The D-S evidence theory is used to fuse the abnormal probability of each index, and the abnormal probability of water quality of water supply network is compared with the threshold value of multi-index fusion anomaly probability. Make a comprehensive judgment on whether the water quality is abnormal at the present time. Based on the experimental system of water supply network simulated by the research group, the injection experiment of different concentrations of copper sulfate and potassium ferricyanide was designed, and the pH value, turbidity, residual chlorine of on-line monitoring were used. Eight kinds of conventional water quality indexes, such as dissolved oxygen, were analyzed by dynamic correlation analysis and water quality anomaly detection. The feasibility and performance of the method were verified by the operating characteristic curve (ROC) of the subjects.
【作者单位】： 浙江大学工业控制技术国家重点实验室;
【基金】：国家自然科学基金资助项目(61573313;U1509208) 浙江省科技厅公益资助项目(2014C33025) 浙江省重点研发计划资助项目(2015C03G2010034)
【分类号】：X832
，
本文编号：2163474
[Abstract]:In order to improve the pollution detection ability of the water quality monitoring system of urban water supply network, a water quality anomaly detection method based on dynamic correlation analysis of multiple conventional water quality indexes is proposed. Dynamic time warping algorithm (DTW) is used to measure the dynamic distance between time series of several conventional water quality indexes and to characterize the similarity degree and dynamic correlation characteristics of each index fluctuation. The D-S evidence theory is used to fuse the abnormal probability of each index, and the abnormal probability of water quality of water supply network is compared with the threshold value of multi-index fusion anomaly probability. Make a comprehensive judgment on whether the water quality is abnormal at the present time. Based on the experimental system of water supply network simulated by the research group, the injection experiment of different concentrations of copper sulfate and potassium ferricyanide was designed, and the pH value, turbidity, residual chlorine of on-line monitoring were used. Eight kinds of conventional water quality indexes, such as dissolved oxygen, were analyzed by dynamic correlation analysis and water quality anomaly detection. The feasibility and performance of the method were verified by the operating characteristic curve (ROC) of the subjects.
【作者单位】： 浙江大学工业控制技术国家重点实验室;
【基金】：国家自然科学基金资助项目(61573313;U1509208) 浙江省科技厅公益资助项目(2014C33025) 浙江省重点研发计划资助项目(2015C03G2010034)
【分类号】：X832
，
本文编号：2163474
本文链接：https://www.wllwen.com/kejilunwen/huanjinggongchenglunwen/2163474.html
最近更新
教材专著
