Dealing with Data Missing and Outlier to Calibrate Nodal Water Demands in Water Distribution Systems

Model parameters of the water distribution system (WDS) such as nodal water demands, should be carefully calibrated by measurements. However, the inconvenience of data missing and outliers is a common feature of real-time measurements, and significantly reduces the robustness and usability of WDS models. Hence, dealing with these uncertainties in WDSs is still a challenge that needs to be tackled. This paper developed an approach to detect and compensate for the missing data and outliers in real-time. Then the compensated data are fused by a Bayesian method to calibrate the nodal water demand. The developed approach is validated by using a simple network and a realistic network. The results demonstrate that the developed approach can effectively improve the robustness of the calibration algorithm in the presence of data missing and outliers.