Doctoral Thesis Proposal: DEMAND MODELLING AND CALIBRATION FOR HYDRAULIC NETWORK MODELS

Water distribution network models are used by water companies in a wide range ofapplications. A good calibration of these models is required in order to increase theconfidence of the applications’ results. The aim of this doctoral thesis is to developan adaptive water distribution model which both calibrates its parameters and discernsbetween faults and system evolution. In previous projects, nodal demands were themajor uncertainty within the model parameters. A demand calibration methodologywas developed during the master project. The results obtained were promising, althoughthe work done fulfils only a small part of the whole application. In order to accomplishthe remaining tasks, further work must be done. First, system identifiability will beperformed in order to determine the number of required sensors that make the systemobservable. The identifiability study will lead to sampling design methodologies andnetwork reduction (skeletonization). Once the model is identifiable, two calibrationtechniques based on non-linear least squares and artificial intelligence techniques willbe studied and adapted for the final application. A methodology for distinguishingbetween faults and parameter evolution will be developed too. All the subprocesses willbe assembled in an open source software which combines the simulating engine fromEPANET with the computational power from MATLAB, becoming a full calibrationand monitoring application for water networks. Finally, at least two real scenarios willbe monitored through the proposed application.This thesis proposal sets the basis for the thesis development, presenting the work doneon the subject, organising the future tasks and proposing a working plan.