Previsão de vazões de afluência para o setor elétrico por meio de modelos lineares e não lineares

Abstract

The theme of this research is the use of prediction models of Integrated Autoregressive and type of moving average- ARIMA, along with the Autoregressive models of Conditional Heterokedastic-ARCH. The first class of models is used to describe the level and the second, the volatility of the series. Has, as main objective of this research, predict and analyze the variability of the flow of inputs of the Jaguari River, by means of linear and non-linear mathematical models, in order to assist in the management of water resources of the river and for power generation of pinch Furnas do Secret. This dissertation consists of two scientific articles that characterize the hydrological behavior of the Jaguari River, using to this end, mathematical models that provide predictions of the flows and the measurement of the periods considered atypical for the time series. In the first article, have chosen the model SARMA (1,0,1)(1,0,1)12-ARCH (1) able to represent the average and the variability of flows of the Jaguari River in m3s in 1942 to 2006 period. In the second article, we use mathematical models of forecasting, along with implementation of intervention analysis, investigation of hydrological behavior of flows in monthly periods January 1970 to December 2010. To do so, mathematical models adopted: Holt-Winters exponential smoothing (AEHW), ARIMA, ARCH models, and intervention analysis, concluding that the model SARMA (1,0,0)(2,0,0)12-ARCH(1) was selected to represent the average hydrological behavior and the variability of flows, in order to make forecasts, taking into account the selection criteria AIC and MAD. The ARCH model presented a degree of persistence is smaller than one, indicating that the flows, in a short time, will return to his usual level. Intervention analysis, increases were observed in the flows of 15,68 m3/s and 18,47 m3/s, during periods of November 1994 and December of 2009, possibly caused by climatic phenomena. However, it appears that the joint ARIMA modeling-ARCH and the incorporation of intervention analysis come to aid in the energy planning of PCH, in order to allow for great levels of power generation in the future.