Árvore de decisão e simulação estocástica para análise de riscos em sistemas de reservatórios superficiais.

Abstract

Some of the main challenges faced in the water resources management have been the difficulty in measuring the risks that reservoir systems are subject, given the various possibilities of hydrological scenarios. Risk analysis as a tool to support decision-making has used methodologies that contribute to a better allocation of water for multiple uses and objectives, with emphasis on mathematical-computational modeling, including technical-operational and legal aspects, which are important in defining efficient strategies for the proper use of water while minimizing system failures. This work aimed to apply the decision tree model and stochastic simulation for the risk analysis of a surface reservoir system. The methodology was based on the use of synthetic precipitation and flow data generation, which were later used in water balance simulations, where the risk indicators were calculated and the decision tree model was applied, using the reliability of the system as a dependent variable. The research was applied to the region of the upper course of the Piranhas-PB River, with emphasis on the Engenheiro Ávidos – São Gonçalo reservoir system. The obtained results made it possible to make a diagnosis of the operation and the identification of critical points for not meeting the required flows by the demands of the system. It was possible to establish a relationship between the reliability and the allocation of flows, as well as the precipitation and the volume of the reservoirs, indicating patterns in the hydrological conditions that led to the worst and best percentages of reliability. The decision trees achieved satisfactory performance and robustness in predicting the risk for the demands and operating scenarios, which is important for identifying the best decision-making in the planning of water resources.