Participants are required to submit their results in the format specified in the template_results.txt file. The file will include the locations and times of each assumed leakage events. The start time of a leakage is specified in the ISO 8601 time format YYYY-MM-DD hh:mm, as illustrated in the template results file. The location of the leakage is specified by the link ID, as defined in the EPANET model of the network “L-Town.inp”. Participants can specify any number of leaks.
Evaluation of participant results will follow an economic approach, i.e., the water utility of L-Town will calculate a specific reward (or cost) for the application of the submitted leakage detection and isolation results.
A correct detection is considered one that is near to the start time of a leak (e.g., within Tmax=1, a week of the start of the leak) and inside a pre-defined pipe length radius around the leak location (e.g. Xmax=300 meters).
Detections will be rewarded depending on the proximity to the start time and location. For example, a detection which is near to the start time of a leak (e.g., within 2 hours from the leak start time), and near to the location (e.g., 50 meters shortest distance from the center of the pipe), will receive a higher score in comparison with detections which detect the events with delay (e.g., after 2 days), and at greater distances (e.g., 150 meters).
Detections before the start time or above a certain time delay after the start time of a leak will not be rewarded. Leak localization is considered correct and will be rewarded if the given point is inside the pre-defined pipe length radius. To clarify, if leak localization is correct but it is done outside the accepted time delay, the detection will not be rewarded. Moreover, false detections will be penalized, i.e., detections in which a leak does not exist at the specific time and pre-defined radius, as they cause financial loss to the water utility due to lost person-hours.
An example of a detection scoring function given the rules described above is illustrated in Figure 1. You can download the template here.
|# Team name|
# Participant Name(s) and Emails
# linkID, startTime
p1, 2019-02-03 10:00
p23, 2019-05-09 02:05
p234, 2019-08-10 03:00
Figure 1 – Illustrative example of the detection scoring function