Comparing alternative PDA solvers with EPANET. Issue 3 (21st March 2022)

• Record Type: Journal Article
• Authors: Todini, E.; Farina, A.; Gabriele, A.; Gargano, R.; Rossman, L. A.
• Is Part Of: Journal of hydroinformatics. Volume 24:Issue 3(2022)
• Page Start: 697
• Page End: 710
• Publication Date: 2022-03-21
• Subjects: active set method -- constrained emitter method -- EPANET 2.2 -- global gradient approach -- pressure-driven analysis -- water distribution networks; Hydrology -- Data processing -- Periodicals; Geographic information systems -- Periodicals; Geographic information systems; Hydrology -- Data processing; Electronic journals; Periodicals; 551.480285
• DOI: 10.2166/hydro.2022.005
• Languages: English
• ISSNs: 1464-7141

Abstract

Abstract: This work compares several alternative methods of pressure-driven analysis (PDA) implemented within the code base of the EPANET water distribution system modeling software. The resulting code includes a direct method, the original and a newly modified version of the inverse method currently used by EPANET 2.2, and two recently published methods (the global gradient approach-based PDA and the active set method). The latter four methods solve PDA problems using an inverse flow–pressure relationship, whereas the direct method uses the original relationship. The alternative methods were extensively tested, and their performance was compared over several case study water distribution networks of vastly different sizes and complexities. The results showed that all of the new inverse methods are equally efficient and reliable, whereas the direct method is less reliable by having a higher frequency of failing to converge. HIGHLIGHTS: A demand-driven approach and five alternative PDAs were incorporated into EPANET. The EPANET 2.2 barrier function algorithm was modified to improve convergence, a direct PDA and the inverse global gradient approach-based PDA and ASM approaches were also included. The code, thoroughly tested on the available water distribution network, cases up to over 150, 000 nodes, will be made available for further comparisons. Results confirm the lack of convergence of the direct approach.