Multi-period response management to contaminated water distribution networks: dynamic programming versus genetic algorithms. Issue 3 (4th March 2018)

• Record Type: Journal Article
• Authors: Bashi-Azghadi, Seyyed Nasser; Afshar, Abbas; Afshar, Mohammad Hadi
• Is Part Of: Engineering optimization. Volume 50:Issue 3(2018)
• Page Start: 415
• Page End: 429
• Publication Date: 2018-03-04
• Subjects: Consequence management -- water distribution networks -- dynamic programming -- genetic algorithm -- simulation-optimization; Engineering design -- Periodicals; Mathematical optimization -- Periodicals; 620.0042
• DOI: 10.1080/0305215X.2017.1318871
• Languages: English
• ISSNs: 0305-215X

Abstract

ABSTRACT: Previous studies on consequence management assume that the selected response action including valve closure and/or hydrant opening remains unchanged during the entire management period. This study presents a new embedded simulation-optimization methodology for deriving time-varying operational response actions in which the network topology may change from one stage to another. Dynamic programming (DP) and genetic algorithm (GA) are used in order to minimize selected objective functions. Two networks of small and large sizes are used in order to illustrate the performance of the proposed modelling schemes if a time-dependent consequence management strategy is to be implemented. The results show that for a small number of decision variables even in large-scale networks, DP is superior in terms of accuracy and computer runtime. However, as the number of potential actions grows, DP loses its merit over the GA approach. This study clearly proves the priority of the proposed dynamic operation strategy over the commonly used static strategy.