Probabilistyczne prognozowanie cen energii elektrycznej z wykorzystaniem 
regresji izotonicznych
[Probabilistic forecasting of electricity prices with isotonic regressions]

Arkadiusz Lipiecki (Doctoral School - Management, KSZiRO, PWr, Wroclaw)


Forecasting of electricity prices is an essential tool for supporting 
decision-making of electricity markets' participants. However, simple 
point forecasts provide limited information. Hence probabilistic 
forecasts, represented as distributions of future prices, gain 
increasingly more attention, as they allow to assess and mitigate the 
risks connected with high volatility of electricity prices. In my talk I 
will discuss models for probabilistic forecasting of day-ahead electricity 
prices that rely on point forecasts as input. This approach allows to 
borrow the complexity of expert point forecasting models and build 
relatively simple probabilistic forecasting models. In particular, I will 
focus on the application of stochastic order and discuss Isotonic 
Distributional Regression, which has not been previously studied on 
electricity markets. Additionally, I will introduce an isotonic version of 
Quantile Regression Averaging - a well-established forecast combination 
method. To thoroughly assess the accuracy of the models, they are tested 
on four day-ahead markets during two periods of qualitatively different 
price dynamics - COVID-19 pandemic and Russian invasion of Ukraine.