Enhancing Energy Consumption Prediction by Integrating Occupant Activity with Machine Learning Models

Abstract

The precision of the forecast of the power consumption of buildings is
essential for big constructions in the present day. However, many of
the models in use fail to consider the effect of people’s activities
within the building on energy consumption. To overcome this
limitation, this paper uses a synchronized data collection approach to
collect data from different sensors about occupancy activity and
power consumption. Several machine learning models are employed
with this coordinated data, and the effects of occupant behaviour on
power usage are explored. By analyzing the results of the models
generated by the two algorithms, the best ways of reaching
behaviour-sensitive power consumption prediction are determined.
Therefore, the findings establish that the additional data concerning
occupant activity provides more accurate assessments of energy
usage that can be quite beneficial for enhancing the further
development of better adaptive and more efficient building
management systems. This work also helps to fill the existing gap in
energy prediction literature wherein, unlike other fields, the human
factor is considered in machine learning models that can lead to more
accurate and more immune to distortion energy forecasting.