Short-term forecasting of electricity prices using mixed models

This chapter focuses on short-term forecasting and offers a simple but accurate method and an automatic tool to compute 1-day-ahead forecasts of electricity prices. It introduces two new models, each of which deals with the 24-hourly time series of electricity prices instead of a complete one. The chapter obtains the appropriate length of the time series used to build the forecasting autoregressive integrated moving average (ARIMA) models. It presents some previous work on short-term forecasting of electricity prices and reviews the theoretical basics of time series analysis. The chapter considers a study of forecasting errors from two different points of view: descriptive and applying locally weighted regression (LOESS). It also presents the fundamentals of the design of experiments (DOE) as well as its application to develop mixed models, selecting the preferred combination of "Model" and "Length" in terms of prediction error.