Od stopni Celsjusza do megawatogodzin: prognozowanie zapotrzebowania na 
energię elektryczną przy użyciu publicznie dostępnych prognoz pogody  
[From Celsius to megawatt hours: Forecasting electricity demand using 
publicly available weather predictions]

Dawid Linek (Faculty of Management, PWr, Wrocław)


Current weather conditions significantly impact electricity generation and 
consumption. However, it is unclear whether publicly available short- and 
medium-term weather forecasts can improve the accuracy of electricity demand 
predictions up to seven days ahead. In this study, we evaluate the predictive 
power of variables such as temperature, wind speed, cloud cover, and 
atmospheric pressure. We identify the most informative variables, discuss 
methods for processing weather predictions, and highlight key considerations 
for incorporating these variables into forecasting models.

----------


Prognozowanie cen energii elektrycznej na rynku dnia następnego: uśrednianie 
prognoz a wykrywanie punktów zmiany
[Electricity price forecasting in the day-ahead market: averaging forecasts 
vs breakpoint detection]

Piotr Zaborowski  (Doctoral School - Management, PWr, Wrocław)


The profitability of a battery energy storage system (BESS) in the day-ahead 
market is determined by the precise timing of buying (charging) and selling 
(discharging) electricity. The latter requires reliable electricity price 
forecasts for the next day. In this paper, two approaches are compared using 
regression-based models. The first averages forecasts across calibration 
windows of varying lengths, balancing short-term responsiveness with 
long-term stability. The second uses structural break detection with the 
pruned exact linear time (PELT) algorithm, calibrating models only within 
homogeneous segments. The approaches are evaluated using both statistical 
(MAE, RMSE) and economic criteria (average opportunity cost of a BESS trading 
strategy and the Sharpe ratio) on four European day-ahead markets: Germany, 
Finland, Poland, and Spain. The test period includes the COVID-19 pandemic 
and the 2022 energy crisis. The results show that averaging-based models 
outperform breakpoint-based approaches in both accuracy and profitability.