An Overview of Artificial Intelligence and Machine Learning Approaches for Building Energy Analysis, Characterization, Control, and Grid Support Services Provision

Increasing penetrations of variable renewable energy sources like wind and solar photovoltaic (PV) systems are challenging power system stability worldwide. Leveraging demand‐side behavior is becoming more popular to help overcome contemporary issues concerning balancing electricity generation and demand. As significant energy users with the potential to act as electricity producers through renewable energy sources, buildings are attractive assets for contributing to power system control from energy efficiency and demand response perspectives. Meanwhile, the proliferation of “smarter” buildings equipped with network‐connected sensors and devices using Internet of Things (IoT) platforms produces significant data volumes that lend themselves to novel artificial intelligence (AI) and machine learning (ML) methods that we can apply across the suite of demand response design steps. This paper reviews the application of AI and ML methods across these steps, which include building energy analysis and auditing, modeling and predicting building load demand, detecting and classifying building energy and power flexibility, implementing flexible building load control, and participating in demand response and other ancillary service markets. Throughout the paper, we comprehensively analyze the application of various AI and ML methods, highlighting their effectiveness and limitations. We also identify emerging pertinent challenges of interest to practitioners and researchers examining the implementation of such approaches for building demand response provision. This article is categorized under: Cities and Transportation > Buildings Energy and Power Systems > Energy Infrastructure Energy and Power Systems > Energy Management