Fault detection of non-residential water meters

Abstract

City West Water (CWW) provides water services to the CBD and western suburbs of Melbourne, Australia. The water industry constantly faces the problem of non-revenue water, i.e. water lost for various reasons including bursts, leaks, and metering inaccuracy. Effective asset maintenance is a vital part of ensuring water companies can both reduce non-revenue water and continue to provide services to their customers. To fairly bill customers for water consumption, a water meter is located at each customer's property to measure the volume of water being supplied from the water main to the property. However, these meters can read inaccurately over time and, if inaccurate, most of the time record less than the actual consumption. This results in under billing customers for the water they consume, contributing to CWW's non-revenue water loss. Despite the small number of non-residential (business) customers serviced by CWW (9% of customers), these customers account for over 50% of total water consumption each year. Therefore it is vital to ensure the meters servicing these customers are maintained correctly. Most water meters are read only quarterly so consumption history data is quite sparse. Without removing a meter from service, or at least physically running tests at the property itself, it is uncertain whether or not these meters are performing with satisfactory accuracy. Previous replacement programs replace a meter after it has recorded up to a certain volume of water, but for meters that under record volumes, this level may never be reached and thus faulty meters remain in service indefinitely. For this reason, it is important to monitor these water meters by other means. Unlike residential meters, there was no existing data from non-residential meter tests, so it was not possible to probabilistically determine which meter characteristics (model, age, size etc.) typically lead to faulty water meters. Instead MATLAB was used to develop an algorithm which was given the name Checkmate Lite. The purpose of Checkmate Lite is to assist CWW choose which meters servicing non-residential customers may need to replaced. Checkmate Lite uses an algorithm that attempts to identify anomalies in a meter's consumption history. These meters are flagged by Checkmate Lite. The user can then interactively review the properties of these meters and their consumption histories to decide whether or not any of the meters should be replaced. With no non-residential meter test results, the current accuracy of the algorithm is unknown, hence it is being used to assist in decision making, and not make the final decision itself.