Head-dependent modelling and optimisation of water distribution systems

  • Alemtsehay Gebremeskel Seyoum

Student thesis: Doctoral Thesis

Abstract

The construction, operation and maintenance of water distribution systems (WDSs) involve a huge capital investment and it is essential to design and manage them in a cost effective way. The optimisation approaches (e.g. evolutionary algorithms) for optimal design or operation of WDSs require simulation models to evaluate the hydraulic and/or water quality performances of solutions to the problem. An accurate performance assessment of solutions (feasible and infeasible) is crucial to guide the search towards the optimal solution efficiently. Conventional models are demand-driven and consequently they are incapable of simulating pressure-deficient (infeasible) solutions accurately. When simulating a pressure-deficient network, results produced by demand-driven analysis model are highly unreliable and misleading. This thesis is concerned with the development and application of a new integrated head-dependent hydraulic and water quality model for pressure-deficient network modelling and optimisation of real-world systems. The original and novel aspects of the work carried out in this research are stated next. A new pressure dependent analysis (PDA) model has been developed for pressure-deficient network modelling. The model is an enhanced version of the pressure-dependent extension of EPANET (EPANET-PDX) that has an embedded logistic nodal head-flow function. The novelty of the proposed PDA is the formulation of a new, more efficient and robust implementation of the line search and backtracking procedure that greatly enhances computational properties for highly pressure-deficient networks; and increases the algorithm's consistency over a wider range of operating conditions. The model performed consistently well when simulating hypothetical and real-life networks under normal and pressure deficient conditions. Comparison between results generated by the new model and the original EPANET-PDX demonstrated that the two models produce identical hydraulic results. From a numerical perspective, a significant reduction in numbers of iterations to complete a simulation has been obtained for all pressure operating conditions. Also, for extremely pressure deficient conditions a significant reduction in computational time has been achieved in comparison to the original EPANET-PDX. Water quality modelling under pressure-deficient conditions is addressed for the first time in this thesis. Hydraulic and water quality analyses based on two water supply zones in the UK were conducted for a range of simulated operating conditions including normal and subnormal pressure and pipe closures using PDA. It is shown that operating conditions with subnormal pressures, if severe and protracted, can lead to spatial and temporal distributions of the water age and concentrations of chlorine and disinfection by-products that are significantly different from operating conditions in which the pressure is satisfactory .A new parallel model for the solution of multi-objective WDSs optimisation problems is developed. The model utilises a multi-objective genetic algorithm that has an embedded PDA hydraulic simulator. The PDA takes into account the pressure dependency of the nodal flows and thus avoids the need for penalties to address violations of the nodal pressure constraints. A controller-worker approach is implemented to parallelise the optimisation process in which a single controller processor executes the routine operation of the algorithm and employs the workers to carry out fitness evaluation. A real-life network that comprises multiple sources, multiple demand categories, many fire flows and extended period simulation is used to demonstrate the effectiveness of the model. Results show that the algorithm is stable and finds optimal and near-optimal solutions reliably and efficiently.
Date of Award10 Sep 2015
Original languageEnglish
Awarding Institution
  • University Of Strathclyde
SupervisorTiku Tanyimboh (Supervisor) & Rebecca Lunn (Supervisor)

Cite this

'