Calibration of water distribution system hydraulic models.
A number of mathematical models are used nowadays to describe behaviour of the reallife
water distribution system (WDS). It is a well known fact that, to have any
meaningful use, any WDS mathematical model must be calibrated first. Here,
calibration is defined as process in which a number of WDS model parameters are
adjusted until the model mimics behaviour of the real WDS as closely as possible. In
this thesis, WDS mathematical models that are used to model water quantity aspect only
are analysed. Three hydraulic models considered here are: (1) steady-state flow model,
(2) quasi-steady flow (extended period simulation) model and (3) unsteady flow model.
The calibration problem analysed here is formulated as a constrained optimisation
problem of weighted least square type with the objective defined in a way that enables
effective incorporation of prior information on calibration parameters. WDS calibration
problem is then analysed in detail, including special issues of identifiability, uniqueness
and stability of the problem solution. A list of diagnostic and other statistics and
analysis is presented to improve existing calibration approaches by providing partial
insight into the calibration process. Calibration of WDS hydraulic models is further
improved by the development of new hybrid optimisation method.
Being closely related to calibration, the problem of sampling design for calibration of
WDS hydraulic models is also addressed here. First, sampling design is formulated as a
constrained two-objective optimisation problem. Then, two novel models are developed
to solve it. The first model is based on standard, single-objective Genetic Algorithms
(SOGA). The second model is based on multi-objective Genetic Algorithms (MOGA).
Finally, all novel methodologies presented here are verified successfully on multiple
case studies that involve both artificial and real-life WDS. At the end, relevant
conclusions are drawn and suggestions for further research work are made.