An arc restoration scheduling problem for pipeline networks in post-disaster management

碩士 === 國立成功大學 === 資訊管理研究所 === 104 === Pipeline networks that ship flows of gas, water, electricity, or packets are important for supporting our daily livings. Suppose arcs of a pipeline network are damaged by disasters and the limited resources (equipment, manpower, and time) required to restore eac...

Full description

Bibliographic Details
Main Authors: Ping-ChengLin, 林秉錚
Other Authors: I-Lin Wang
Format: Others
Language:zh-TW
Published: 2016
Online Access:http://ndltd.ncl.edu.tw/handle/59228317424221001527
Description
Summary:碩士 === 國立成功大學 === 資訊管理研究所 === 104 === Pipeline networks that ship flows of gas, water, electricity, or packets are important for supporting our daily livings. Suppose arcs of a pipeline network are damaged by disasters and the limited resources (equipment, manpower, and time) required to restore each damaged arc have been estimated. We investigate the problem of when and who to restore which arcs such that the flows over pipelines become accessible for people among all arcs at minimum total waiting time in the post-disaster management. We first reduce the original network into an equivalent but smaller one where only damaged arcs are considered. Then, we propose two integer programs (Branch-and-Cut, Multicommodity Network Flows frameworks) for this special resource constrained project scheduling problem. Several valid inequalities are proposed to effectively shorten the computational time for integer programs. We also explain how to deal with more general cases where heterogeneous teams as well as their collaboration are considered for calculating an optimal network restoration schedule. Three fast heuristics are designed: Sequential Segment Heuristic (SSH), Greedy Tree method (GT), and Greedy Connected Component method (GCC), where SSH solves partitioned segments sequentially, and both GT and GCC first determine an arc restoration sequence in different greedy fashions and then assign tasks to different teams by a First-Come First-Served principle. We propose three variants of Genetic Algorithms (GA). Computational experiments indicate our heuristics could calculate solutions within 2% optimality gaps in seconds for large-scale networks. The GAs, on the other hand, perform a little worse with optimality gaps up to 5%.