Ajuste do modelo SMAP via técnicas de inteligência computacional

Abstract

The hydroelectric production planning depends considerably on river flow predictions, which are made from mathematical, stochastic and hydrological models, mostly incorporated into computer systems. Increasing the accuracy of water flow forecasting is an important way to help making strategic decisions on energy production and storage in hydroelectric power plants, increasing their efficiency and reducing thermal complementation costs. This work presents a comparative study between three optimization techniques based on computational intelligence aiming to optimize the water flow forecast computed by the SMAP (Soil Moisture Accounting Procedure) model applied for three Brazilian sub-basins (Emborcação, Nova Ponte and Corumbá). At first, a convexity study was carried out to justify the use of computational intelligence techniques to solve the problem. In a second moment, a comparative analysis was performed between Bat Algorithm (BA) – official optimization model technique used by the electric sector to perform SMAP model optimization – Grey Wolf Optimization (GWO) and Salp Swarm Algorithm (SSA). The comparative study of these three algorithms involves the analysis of solutions’ variability, convergence characteristics and computational processing time. The results encourage the use of GWO in the SMAP model optimization.