Comparing the Forecasting Performances of Linear Models for Electricity\n Prices with High RES Penetration

Abstract

This paper compares alternative univariate versus multivariate models,\nfrequentist versus Bayesian autoregressive and vector autoregressive\nspecifications, for hourly day-ahead electricity prices, both with and without\nrenewable energy sources. The accuracy of point and density forecasts are\ninspected in four main European markets (Germany, Denmark, Italy and Spain)\ncharacterized by different levels of renewable energy power generation. Our\nresults show that the Bayesian VAR specifications with exogenous variables\ndominate other multivariate and univariate specifications, in terms of both\npoint and density forecasting.\n