Reducing System Wide Event Detection False Positive Alerts by Using Reversed Hydraulic Simulation

摘要

The EPA's Water Quality Event Detection Challenge evaluated five event detection systems (EDS) with real world water quality monitoring data. The findings are that, on average, there are about one false positive alert per day per monitoring station. Other publications supports these findings. A real world municipal water distribution system (WDS) may consist of tens of monitoring stations resulting in a large number of false positive alerts thus making the system wide EDS unpractical from the WDS operator point of view. This paper presents a new approach for reducing false positive alerts by using alerts raised from more than one monitoring station at different times. First, the methodology uses the reversed hydraulic simulation method which allows to identify the possible time-locations pairs where contamination injection could explain each monitor's. Then, using the super-position method, the time-location pairs where contamination injection could explain all monitors alerts are found. In the next step a tracer injection simulation is performed for each time-location pair in order to validate that the tracer should not have triggered an alert in a non-alerting monitoring station or did not arrive before the alert for an alerting monitoring station. If such time-location pairs are found the alerts are suspected to be false positive. The method is demonstrated on a small water distribution system model.