18th AIAI 2022, 17 - 20 June 2022, Greece

Energy Load Forecasting: Investigating Mid-Term Predictions with Ensemble Learners

Charalampos M. Liapis, Aikaterini Karanikola, Sotiris Kotsiantis


  In the structure of the modern world, energy and especially electricity is a prerequisite for regularity. Thus, the requirement for accurate forecasts regarding power system loads seems self-evident. In machine learning, a time series forecasting endeavor can be treated as a regression problem. In such scenarios, ensemble methods are often used for robustness and increased accuracy of the generated predictions. This work is a comparative investigation of the use of ensemble schemes for medium-term forecasting of energy system load. The use of over 300 regression schemes is investigated, in a total of 8 different modifications of the input data, over 5 different time-frames, that is, one day, 7-day, 14-day, 21-day, and 30-day horizons, resulting in a loop of 12000 experiments. Summary tables with representative results from the corresponding Friedman rankings are presented.  

*** Title, author list and abstract as seen in the Camera-Ready version of the paper that was provided to Conference Committee. Small changes that may have occurred during processing by Springer may not appear in this window.