Probabilistic forecasting of hourly electricity prices in the medium-term using spatial interpolation techniques

In the context of competitive electricity markets, medium-term price forecasting is essential for market stakeholders. However, very little research has been conducted in this field, in contrast to short-term price forecasting. Previous studies of electricity price forecasting have tackled the problem of medium-term prediction using fundamental market equilibrium models with daily data, or at most, averages of groups of hours. Similarly, the limitations of point forecasts are recognized widely, but the literature dealing with probabilistic forecasts is sparse. In this study, a novel methodology for the medium-term hourly forecasting of electricity prices is proposed. This methodology is unique in the sense that it also attempts to perform punctual and probabilistic hourly predictions simultaneously. The approach consists of a nested combination of several modeling stages. The first stage consists of generating multiple scenarios of uncertain variables. In a second stage, a market equilibrium model that incorporates Monte Carlo simulation and a new definition of load levels is executed for a reduced combination of the scenarios generated. The application of spatial interpolation techniques allows us to estimate numerous feasible realizations of electricity prices from only several hundred executions of the fundamental market equilibrium model without any loss of accuracy. The efficiency of the proposed methodology is verified in a real-size electricity system that is characterized by complex price dynamics: the Spanish market.