Short-term electricity price forecasting - evaluation of selected hybrid models
|Author:||Bc. Štěpán Svoboda|
|Year:||2017 - summer|
|Leaders:|| prof. PhDr. Ladislav Krištoufek Ph.D.
|Work type:|| Bachelors
|Awards and prizes:|
|Abstract:||In this thesis a thorough study of the previous literature and the division and special
aspects of EPF was carried out. Then the evaluation and comparison of several models
was done – the ARIMA, SVR, SVRARIMA and PSF model. This comparison was done
on the intra-day Nord Pool market, which is quite unique as almost all short-term EPF
is carried out on the day-ahead market. Our results are robust as the modeling was done
on 100 test periods and we have tested the difference in predictive accuracy using the
modified DM test. Our conclusion is the PSF model is inadequate in our intra-day setting
and the overall ARIMA model seems to outperform the SVR and SVRARIMA model
somewhat. The dominance of ARIMA is not very strong and a further investigation of the
causes of these results can better illuminate the strengths and weaknesses of these models.