The electricity price forecast is relevant data for evaluating the financial results of operating companies, assessing new investment projects, and structuring hedging portfolios. This paper analyzes the effectiveness of spot price forecasting methods for electric power in Colombia based on Deep Learning and their ability to overcome an ARIMA-type structure. The results are input for researchers and analysts of global electricity markets who are interested in making projections of electric power and some variables that have mean reversion, high volatility, and jump characteristics. Deep Learning methods can capture the series trends in out-of-sample periods. However, they cannot beat the performance of forecasts made by applying Box-Jenkins approaches.
Trespalacios et al. (Wed,) studied this question.