---

Selection of features in reinforcement learning applied to energy consumption forecast in buildings according to different contexts

Forecasting techniques can help optimize the energy management of buildings by finding load consumption patterns. These patterns are important for deciding which forecasting technique gives more accurate predictions in different contexts. The paper reinforcement learning model uses Multi-armed Bandit algorithm to decide the best consumption forecast algorithm for each period and context, to improve, in the long-term, the forecast accuracy. The reinforcement learning model was tested using two approaches with different exploration options: upper confidence bound, and greedy.

Publication of the scientific paper "Selection of features in reinforcement learning applied to energy consumption forecast in buildings according to different contexts" in Energy Reports journal (IF: 4.937). D. Ramos, P. Faria, L. Gomes, P. Campos, Z. Vale, “Selection of features in reinforcement learning applied to energy consumption forecast in buildings according to different contexts”, Energy Reports, Volume 8, Supplement 3, 2022, Pages 423-429. DOI: 10.1016/j.egyr.2022.01.047

Open publication