Browsing by Author "Dolling, OR"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Artificial neural networks for streamflow prediction
    (TAYLOR & FRANCIS LTD, 2002) Dolling, OR; Varas, EA
    This paper presents monthly streamflow prediction using artificial neural networks (ANN) on mountain watersheds. The procedure addresses the selection of input variables, the definition of model architecture and the strategy of the learning process. Results show that spring and summer monthly streamflows can be adequately represented, improving the results of calculations obtained using other methods. Better streamflow prediction methods should have significant benefits for the optimal use of water resources for irrigation and hydroelectric energy generation.
  • Thumbnail Image
    Item
    Decision support model for operation of multi-purpose water resources systems
    (TAYLOR & FRANCIS LTD, 2005) Dolling, OR; Varas, EA
    Models to search for optimal operation rules of complex water resources systems generally represent the physical system in a fixed static form, being difficult to incorporate changes in water offer, water demand and system structure. This paper presents a decision support procedure that integrates continuous simulation, artificial neural networks, and optimization to produce decision rules in watershed management for multiple purpose complex water resources systems. The system uses physical indexes to evaluate the compliance of targets for the different purposes of the system, such as occurrence of failure (frequency), resilience (duration and capacity of recovery of a state of failure) and vulnerability (severity or magnitude of the failure). It also introduces a global indicator of the behavior of the system, which combines, with user selected weights, the previous indexes in a measure of global effectiveness. The methodology was applied to the San Juan River Basin, Argentina, and results show conclusively the usefulness of simulation in the study of alternatives of water resources systems with multiple uses and the feasibility of using neural networks to encapsulate the behavior of simulation models. The encapsulated model and parametric operation rules can be included in a dynamic optimization process to search for optimal operation policies.