Online statistical hypothesis test for leak detection in water distribution networks

This paper aims at improving the operation of water distribution networks (WDN) by developing a leak monitoring framework. To do that, an online statistical hypothesis test based on detection is proposed. The developed technique, so-called exponentially weighted reduced kernel generalized likelihood ratio (EW-ORKGLRT), addressed so that modeling phase performed using principal component analysis (KPCA) model, which capable dealing with higher computational cost. Then computed model fed to EW-ORKGLRT chart for purposes. proposed approach extends ORKGLRT method one uses exponential weights residuals in moving window. It might be able further enhance performance detecting small and moderate leaks. method’s main advantages are first required time leaks then updating KPCA according dynamic change process. evaluated compared conventional techniques simulated WDN data. selected criteria excellent rate, false alarm CPU time.