A New ODE-Based Julia Implementation of the Anaerobic Digestion Model No. 1 Greatly Outperforms Existing DAE-Based Java and Python Implementations

The Anaerobic Digestion Model 1 is the quasi-industry standard for modelling anaerobic digestion, and it has seen several new implementations in recent years. It is assumed that these implementations would give the same results; however, a thorough comparison of these implementations has never been...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Processes Ročník 11; číslo 7; s. 1899
Hlavní autoři: Allen, Courtney, Mazanko, Alexandra, Abdehagh, Niloofar, Eberl, Hermann J.
Médium: Journal Article
Jazyk:angličtina
Vydáno: Basel MDPI AG 01.07.2023
Témata:
Anaerobic digestion
Chemical reactions
Computer simulation
Differential equations
Industry standards
Java
Java (Computer program language)
Numerical analysis
Ordinary differential equations
Python
Python (Programming language)
Reproducibility
Canada
ISSN:2227-9717, 2227-9717
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:The Anaerobic Digestion Model 1 is the quasi-industry standard for modelling anaerobic digestion, and it has seen several new implementations in recent years. It is assumed that these implementations would give the same results; however, a thorough comparison of these implementations has never been reported. This paper considers four different implementations of ADM1: one in Julia, one in Java, and two in Python. The Julia code is a de novo implementation of the ODE formulation of ADM1 that is reported here for the first time. The existing Java and Python codes implement the more common DAE formulation. Therefore, this paper also examines how DAE implementations compare to ODE implementations in terms of computational speed as well as solutions returned. As expected, the ODE and DAE forms both return comparable solutions. However, contrary to popular belief, the Julia ODE implementation is faster than the DAE implementations, namely by one to three orders of magnitude of compute time, depending on the simulation scenario and the reference implementation used for comparison.
Bibliografie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:2227-9717
2227-9717
DOI:10.3390/pr11071899