Wykorzystanie wygładzonej regresji kwantylowej do prognozowania probabilistycznego 
cen energii elektrycznej
[Smoothing Quantile Regression Averaging for probabilistic electricity price forecasting]

Bartosz Uniejewski 
Katedra Badań Operacyjnych i Inteligencji Biznesowej, PWr, Wrocław


Quantile Regression Averaging (QRA) has sparked interest in the electricity price forecasting 
community after its unprecedented success in the Global Energy Forecasting Competition 2014, 
where the top two winning teams in the price track used variants of QRA. However, recent 
studies have revealed the method’s deficiency. Firstly, the quality of forecasts significantly 
decreases when the set of regressors is larger than just a few. Secondly, interval predictions 
obtained with QRA tend to be too narrow. The problem is definitely more visible for early 
morning hours, but it does not disappear completely during the remaining hours of the day. 
Here, we introduce Smoothing Quantile Regression Averaging (SQRA), a novel approach that 
addresses the problem of too narrow interval predictions. It outperforms not only the standard 
QRA but also a number of other benchmarks.