Paper Conference

Proceedings of eSim 2020: 11th Conference of IBPSA-Canada


Forecasting the electric demand of an HVAC System with deep learning-based techniques

Jason Runge, Radu Zmeureanu
Centre for Net-Zero Energy Buildings Studies, Department of Building, Civil and Environmental Engineering, Concordia University

Abstract: This paper presents the development of deep learning-based (DL) models for the forecasting of electric demand of a heating, ventilation and air conditioning (HVAC) system during the summer operation, which includes the fans, pumps, chillers, and cooling towers. The paper reviews the recent applications of DL models for the performance forecasting of HVAC systems. Models are built and applied to a case study of an institutional building with synthetic data obtained from a calibrated eQuest simulation. The models use hourly time step data and provide a forecast horizon of six-hours ahead. Multiple long short-term memory (LSTM) models are coupled together in parallel into a homogenous ensemble to provide a combined output forecast. The proposed DL models, with several hidden layers, are compared with a simple forecasting (SFA) approach, and with a feed-forward neural network (FFNN) with a single hidden layer.