| In recent years,the accidents of water pollution in water distribution systems occurred frequently.In order to protect the safety of drinking water,it's of indispensable significance to develop better water quality event detection algorithms to detect the abnormal fluctuation more accurately.Previously,temporal analysis-based methods for a single sensor station have shown limited performance as they fail to consider spatial information.So a detection method based on both temporal and spatial information of correlative stations is proposed in this paper.The main contents and innovative points are summarized as follows.(1)In order to decrease the missed alarms using methods based on single station temporal analysis,a contamination event detection method is proposed,in which both temporal and spatial information from correlative stations in water distribution systems are used.The spatial causal relationship between upstream and downstream stations is modelled by Bayesian Network.Logistic Regression is used to integrate spatial information and temporal information to determine the final detection result.(2)According to the transitivity of contamination event between stations,the Dynamic Time Warping(DTW)is used to measure the relevance of the water quality time sequence fluctuation.Then the result of relevance analysis are integrated into the temporal-spatial detection algorithm to avoid false alarms.The contamination events classification is also discussed using DTW distance.(3)The proposed method is evaluated and discussed using EPANET simulation and contamination injection experiments on experimental system.It is also encapsulated into algorithm package on the water quality monitoring software of this experimental system.By introducing the spatial relationship of the correlative stations and the integration of temporal and spatial information,the proposed method shows better performance in water quality event detection. |
PDF Full Text Request