(A) pipeline network design for single-source and multi-terminal distribution systems

Abstract

This thesis deals with the problems of designing pipeline distribution networks which have a single-source and multiple demand centers. Pipeline network design involves the determination of pipeline route (or network configuration), pump capacity for each pumping station and pipe sizes for each line-section. Two such types of design problems are formulated : one is the case, where junctions could be located anywhere and the other, where a discrete number of possible sites for junctions are already given. A two-stage heuristic algorithm is developed for the former problem. Using the proposed NLP model, the optimal network configuration is obtained in the first stage, under some initial condition. In the second stage, for practical implementation pipe sizes and pump capacities are readjusted using the proposed LP model within the obtained network configuration in the second stage. Zangwill``s DP algorithm for concave-cost network flow problem is shown to be applied effectively for the latter problem. Computational results for a real life situation are presented. Most of design problems for pipeline networks could be solved using the proposed algorithm within reasonable time and also be applicable to other large scale networks.