Can Markov-regime switching models improve power price forecasts? Evidence for German daily power prices

Nonlinear autoregressive Markov regime-switching models are intuitive and frequently proposed time series approaches for the modelling of electricity spot prices. In this paper such models are compared to an ordinary linear autoregressive model with regard to their forecast performance. The study is carried out using German daily spot prices from the European Energy Exchange in Leipzig. Four nonlinear models are used for the forecast study. The resultsof the study suggest that Markov regime-switching models provide better forecasts than linear models.