PTSNet: A Parallel Transient Simulator for Water Transport Networks based on vectorization and distributed computing

Modeling transient flow in water transport networks (WTNs) is characterized by hyperbolic partial differential equations. Existing commercial and open-source software for transient modeling in WTNs have limitations, such as lack of scalability and compatibility with high-performance computers, diffi...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Environmental modelling & software : with environment data news Ročník 158; s. 105554
Hlavní autoři: Riaño-Briceño, Gerardo, Hodges, Ben R., Sela, Lina
Médium: Journal Article
Jazyk:angličtina
Vydáno: Elsevier Ltd 01.12.2022
Témata:
High-performance computing
Message Passing Interface
Networked water systems
Parallel computing
Transient flow
Transient flow
Parallel computing
High-performance computing
Networked water systems
Message Passing Interface
ISSN:1364-8152, 1873-6726
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:Modeling transient flow in water transport networks (WTNs) is characterized by hyperbolic partial differential equations. Existing commercial and open-source software for transient modeling in WTNs have limitations, such as lack of scalability and compatibility with high-performance computers, difficulty to systematically execute simulations and analyze results. This work proposes a novel open-source Python package that relies on vectorization and distributed computing to overcome the limitations of existing software. The proposed library, the Parallel Transient Simulator for Water Networks (PTSNet), surpasses in computational performance existing modeling tools and incorporates novel analytics functionalities. PTSNet has been tested on WTNs composed of tens, hundreds, and thousands of hydraulic elements using a personal and a supercomputer, running from tens to hundreds of processors. We show through rigorous analyses that PTSNet is scalable, accurate, and significantly speeds up simulations with sufficiently dense numerical grids. •The new method relies on vectorized and distributed parallel computing.•The proposed method surpasses performance of existing modeling tools.•Open-source Python package for modeling water transport systems is developed.
ISSN:1364-8152
1873-6726
DOI:10.1016/j.envsoft.2022.105554