Abstract:
The introduction of the electricity market at the Brazilian Power System makes the electricity
price forecast an important area of research with many challenges to the market players because
of its volatility. The identification of the main uncertainty sources is crucial to the development
of the market. This knowledge helps in decision-making process related to energy contracting
maximizing profits and reducing the risks to the agents. Thus, the aim of this study is forecasting
electricity price from a computational chain composed of different interconnected models:
CFSv2 climate model to forecast precipitation, MGB-IPH hydrological model to compute the
inflows to reservoirs and energy programs currently in use in the Brazilian sector to define the
price for three months ahead. The results show that, in general, the proposed methodology is
able to predict and represent the streamflow behavior for the target operating weeks in the case
study for June, July and August (JJA) 2020. This performance is justified by incorporating data
from precipitation forecast to the hydrological model, improving the sensitivity of the
streamflow forecasts. Streamflow forecasts are currently the variable with the greatest impact
on the electricity price (PLD) that is calculated by the National Interconnected System (SIN).
However, when analyzing the results obtained for the forecasted electricity price, the
computational chain incorporating PREVIVAZ for the weekly streamflow forecasts presents
error propagation along the simulation chain. The PREVIVAZ, used by the National System
Operator (ONS), uses historical data and not the CFSv2 precipitation prediction. The different
approaches are analyzed and conclusions about their uses in the middle office of traders are
taken.