Artificial intelligence for forecasting heating and cooling demands

AI adds intelligence to the data and communications among the different components of IoT. One particularly promising use of AI is developing predictive energy consumption models for buildings that factor in the buildings’ thermal characteristics and architecture, meteorological conditions, solar irradiance, wind velocity and direction, outdoor and indoor temperatures and user behaviour. Recent research shows that novel data-driven AI- or artificial neural network–based approaches can improve the accuracy of these predictions and enable predicting short-term fluctuations, a capability essential for control applications (Bünning et al., 2020; Petrichenko et al., 2017; Saloux and Candanedo, 2018).

By forecasting heating and cooling demands better, and then controlling loads to optimise their operation, AI enables greater flexibility in the energy system and facilitates increased use of variable renewable energy sources. Data-driven approaches using AI techniques such as deep learning also add significant energy savings. For example, the majority of buildings are mostly empty (more than 60% of the time), yet they maintain air temperatures of 21-23°C inside. AI methods can detect when a building is completely empty and then lower temperature setpoints in winter or raise them in summer during those periods. Further, AI can help size heating and cooling storage facilities better.