A penalty-free multi-objective evolutionary optimization approach for the design and rehabilitation of water distribution systems

• Main Author: Siew, Calvin Yew Ming
• Published: University of Strathclyde 2011
• Subjects: 624
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.675236

As a result of the increasing emphasis placed on water companies to conform to the stringent performance standards in supplying demands within a constrained financial budget, the application of optimization has inevitably become an integral part of managing a water distribution system (WDS) right from the initial phase of designing a new system to the latter stage of the network where rehabilitation and upgrading works are a necessity. This also includes the on-going operation of the WDS in particular the minimization of energy costs related to pumping and storage. This thesis is concerned with the development and application of a new multi-objective genetic algorithm in optimizing the design, operation and long term rehabilitation and upgrading of the WDS.The novelty and originality of the work done as part of this research are presented next. A seamless, augmented version of the renowned EPANET 2 with pressure dependent analysis (PDA) functionality has been developed. It integrates within the hydraulic engine a continuous nodal pressure-flow function coupled with a line search and backtracking procedure which greatly enhances the algorithm’s overall convergence rate and robustness. The hydraulic simulator is termed “EPANET-PDX” (pressure-dependent extension) herein and is capable of effectively modelling networks under pressure deficient situations which the demand driven analysis based EPANET 2 fails to accurately analyse. In terms of computational efficiency, the performance of EPANET-PDX compares very favourably to EPANET 2. Simulations of real life networks consisting of multiple sources, pipes, valves and pumps were successfully executed with no convergence complications. The simulator depicts excellent modelling performance while analysing both normal and abnormal operating conditions of the WDSs. The accuracy of the generated PDA results has been explicitly validated and verified. An optimization model for the optimal design and upgrading of WDS involving both the operation of multiple pumps and the sizing and location of multiple tanks is developed. The model couples a new boundary convergent multi-objective genetic algorithm to the highly efficient EPANET-PDX simulator which, inherently,automatically accounts for the node pressure constraints as well as the conservation of mass and energy. With accurate PDA, the direct application of the standard extended period simulation enables pump scheduling and tank sizing and siting to be seamlessly incorporated into the optimization without the need for any extraneous methodology or manual intervention. The significant advantage of this model is that it eliminates the need for ad-hoc penalty functions, additional “boundary search” parameters, or special constraint handling procedures. No operator intervention, parameter calibration and trial runs are required. Conceptually, the approach is straightforward and probably the simplest hitherto. The model is applied to several benchmark networks yielding superior results in terms of the initial network construction cost and the number of hydraulic simulations required. The above-mentioned optimization model is extended to form a module for the optimal long term design, upgrading and rehabilitation of WDSs. The multi-criteria problem is set up in a multi-objective frame work i.e. to minimize the capital cost,rehabilitation and upgrading costs, whilst maximizing the network hydraulic performance. A straightforward approach for incorporating reliability measures without further complicating the optimization formulation is utilised and its robustness validated. The effect of deterioration of both the structural integrity and hydraulic capacity of pipes over time is explicitly modelled. The model automatically determines the most cost effective strategy which includes the identification of pipes to be upgraded, the upgrading or rehabilitation options and the timing for the upgrade to be implemented. A real life network in Wobulenzi (Uganda) is used to demonstrate the effectiveness of the model. Results obtained demonstrated major improvements over previous work using the classical linear programming.