Application of random number generators in genetic algorithms to improve rainfall-runoff modelling

0480548 - ÚH 2018 RIV NL eng J - Journal Article
Chlumecký, M. - Buchtele, Josef - Richta, K.
Application of random number generators in genetic algorithms to improve rainfall-runoff modelling.
Journal of Hydrology. Roč. 553, October (2017), s. 350-355. ISSN 0022-1694. E-ISSN 1879-2707
Institutional support: RVO:67985874
Keywords : genetic algorithm * optimisation * rainfall-runoff modeling * random generator
OECD category: Hydrology
Impact factor: 3.727, year: 2017
https://www.sciencedirect.com/science/article/pii/S0022169417305516

The efficient calibration of rainfall-runoff models is a difficult issue, even for experienced hydrologists. Therefore, fast and high-quality model calibration is a valuable improvement. This paper describes a novel methodology and software for the optimisation of a rainfall-runoff modelling using a genetic algorithm (GA) with a newly prepared concept of a random number generator (HRNG), which is the core of the optimisation. The GA estimates model parameters using evolutionary principles, which requires a quality number generator. The new HRNG generates random numbers based on hydrological information and it provides better numbers compared to pure software generators. The GA enhances the model calibration very well and the goal is to optimise the calibration of the model with a minimum of user interaction. This article focuses on improving the internal structure of the GA, which is shielded from the user. The results that we obtained indicate that the HRNG provides a stable trend in the output quality of the model, despite various configurations of the GA. In contrast to previous research, the HRNG speeds up the calibration of the model and offers an improvement of rainfall-runoff modelling.
Permanent Link: http://hdl.handle.net/11104/0276310