Speaker
Description
This paper investigates short-term energy produc-
tion forecasting in Poland using three distinct predictive mod-
els: LightGBM, an ARIMA-based model, and a Long Short-
Term Memory (LSTM) network. Leveraging historical data
from Poland’s energy sector, our study evaluates each model’s
performance in terms of accuracy, robustness, and computational
efficiency. The LightGBM model employs gradient boosting
techniques to capture non-linear relationships, while the ARIMA
approach provides a classical linear autoregressive approach.
Meanwhile, the LSTM network exploits its recurrent architecture
to model complex temporal dependencies inherent in the time-
series data. Comparative analysis based on metrics such as
RMSE and MAE demonstrates that although all models exhibit
competitive forecasting abilities, the LSTM model exhibits a
modest performance advantage over the other approaches within
the examined forecast scenario. However, both the LightGBM
and ARIMA models offer advantages in terms of reduced com-
putational overhead and ease of implementation. Furthermore,
I analyze ensemble models in search of the most accurate
predictions. The insights derived from this analysis aim to assist
policymakers and energy sector stakeholders in making informed
decisions regarding energy distribution and operational planning
in Poland.
Index Terms—energy forecasting, time series, machine learn-
ing, LightGBM, LSTM, ARIMA
