1 Presentation

• 
  01:30 PM - 03:10 PM

  Short-term electricity demand forecasting

  • Fabian Bastin, presenter, Université de Montréal

  In recent years, the energy problem has emerged as one of the major issues to address due to the transition to sustainable energy sources and ever-growing demand. As we rely increasingly on electricity for heating, communications, transportation, leisure, and more, consumption patterns change but they must be known to ensure acceptable quality of service and to allow electricity producers to benefit from market opportunities. Achieving accurate short-term demand consumption forecasting is crucial. In this tutorial, we will explore several models addressing this question, starting from simple regression techniques applied to publicly available datasets. More advanced models take advantage of various information sources and the physics of the problem, resulting in parametric models that can be calibrated using constrained least-squares minimization. Although such models have been successfully employed for several decades, changes in environmental conditions and consumption habits raise difficulties in model adjustments, even though they are grounded in physical principles. The emergence of large datasets alongside new machine learning techniques such as temporal convolutional networks enables the design of new forecasting models complementing the existing ones. We will conclude by suggesting several research avenues in the field. This tutorial will also highlight the collaborative work with the supply and demand forecasting unit at Hydro-Québec.