Predictive models of reservoir storage-yield-reliability functions: Inter-comparison of regression and multi-layer perceptron artificial neural network paradigms

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    1 Citation (Scopus)

    Abstract

    In this study, functions for predicting the total (within-year plus over-year) reservoir capacity have been developed using: first classical multiple regression, and secondly artificial neural networks (ANNs). The basis of the models is the storage-yield-reliability (S-Y-R) analysis of 18 international rivers using the sequent-peak algorithm(SPA). The results showed that the regression model performed better than the ANN model. The relative superiority of the regression model was attributed to its use of the over-year capacity as an independent variable. In contrast, the ANNs use basic variables as inputs and thus offer more flexibility than the regression model, particularly at ungauged sites. Copyright © 2007 IAHS Press.

    Original languageEnglish
    Title of host publicationIAHS-AISH Publication - Quantification and Reduction of Predictive Uncertainty for Sustainable Water Resources Management
    Pages479-486
    Number of pages8
    Edition313
    Publication statusPublished - 2007
    Event24th General Assembly of the International Union of Geodesy and Geophysics 2007 - Perugia, Italy
    Duration: 2 Jul 200713 Jul 2007

    Conference

    Conference24th General Assembly of the International Union of Geodesy and Geophysics 2007
    Abbreviated titleIUGG 2007
    Country/TerritoryItaly
    CityPerugia
    Period2/07/0713/07/07

    Keywords

    • Artificial neural networks
    • Multiple regression
    • Over-year capacity
    • Sequent peak algorithm
    • Storage-yield-reliability
    • Within
    • Year capacity

    Fingerprint

    Dive into the research topics of 'Predictive models of reservoir storage-yield-reliability functions: Inter-comparison of regression and multi-layer perceptron artificial neural network paradigms'. Together they form a unique fingerprint.

    Cite this