Federated Learning for Energy-Efficient Thermal Comfort Control Service in Smart Buildings

Abstract

Energy efficiency and occupant thermal comfort are considered as high-interest topics in a smart building. Internet of Things (IoT) technology enables smart building management and operation to improve building energy efficiency and occupant thermal comfort. In this paper, we use IoT-generated data to derive accurate thermal comfort and electricity load forecast model for smart building control. Due to privacy concerns and high accuracy targets, we take advantage of the use of edge computing and federated learning. Federated learning is a decentralized machine learning scheme that permits data volume increase and data diversity in a training model while preserving privacy. Different households contribute to the training process without revealing privacy. A deep neural network is used to model the relationship between the environmental variables, controllable building operations, and thermal comfort. The Long-Short Term Memory is used as well to forecast the energy load using previous observations of the household electrical load and real-time environmental variables. Finally, the derived thermal comfort model is used to control the smart building environment by searching for the optimal cooling set-point, which results in the desired comfort while decreasing the total energy consumed. Results demonstrate the performance of federated learning settings in terms of accuracy and prediction. The control proposition results as well in less energy consumption within a comfort zone.

Topics & Concepts