Adjustment of a Time Series model to predict flow in the Funil / MG hydroelectric plant

Abstract

The installation of dams to the constructions of hydroelectric plants in Brazil started from the late 19th century. It is a clean and safe energy model of generation. One of the factors that more influence the operation of the hydroelectric plants is the  flow rate of water in the reservoir. There is a minimum volume of water that must be damned to be able to start the turbine and generate electricity. Predicting the values ​​of this flow is extremely important for the planning of hydroelectric systems .In view of the above, this work will adjust a Time Series model to the series of average monthly water flow in the Funil / MG hydroelectric plant, located in Rio Grande, in the southern region of the state of Minas Gerais, on the border between the municipalities of Lavras and Perdões, with the verification of trend, seasonality, interventions and make predictions. The data used in this work will be obtained from the National Electric System Operator (ONS). The analyzes will be performed by the software R and Gretl, and the methodology used to adjust the time series model will be that of Box & Jenkins. From the models suggested and appropriate to the series, the information criteria of Akaike and Hanna-Quinn were used, as well as the mean square error to select the best model. The adjusted model for making the predictions was SARIMA (2, 0, 1) × (0, 1, 1)₁₂.