FORECASTING THE SPOT PRICE BEHAVIOR IN THE BRAZILIAN ENERGY MARKET WITH STATISTICAL TOOLS
Palavras-chave:
Spot price, Forecasting models, Regression models, Time series forecasting, PLDResumo
Electricity in the short-term market, or spot market, is traded according to the Difference Settlement
Price PLD, whose calculation follows a complex and computationally demanding procedure. The present work
proposes the use of statistical methods to forecast PLD values and trends in order to mitigate uncertainty in the
decision making of market agents. Inputs for the proposed models are the affluent natural energy, the stored energy,
the hydroelectric, thermal and wind generations, and the demand for electric energy, publicly disclosed by the
system operator ONS, in addition to the historical series of the PLD itself, disclosed by Electricity Trading
Chamber CCEE. One week ahead forecasts for the southern Brazilian submarket under heavy demand are
proposed, based on time series and regression models for a 2015-2020 historical data. The best PLD forecast
accuracy is achieved with the simple exponential smoothing time series model, with average errors of 17.91 %
and 28.01 R$/MWh. The best trend forecast, of 64.77 %, is obtained by both time series exponential smoothing
models.
