Conference Paper
Proceedings of eSim 2024: 13th Conference of IBPSA-Canada
An inquiry into time-series forecasting methods to quantify on-peak electricity demand savings through direct load control
An inquiry into time-series forecasting methods to quantify on-peak electricity demand savings through direct load control
Burak Gunay, Brodie Hobson, Andre Markus, Farzeen RizviCarleton UniversityAbstract: Direct load control (DLC) for HVAC systems are automated temporary adjustments to the sequences of operation that aim to reduce on-peak electricity demand. DLC algorithms represent a major cost-saving opportunity, particularly for large commercial buildings with automation and control infrastructure to shift and reduce HVAC loads on demand. While DLC can become a mainstream business for energy services companies, methods that can reliably estimate the impact of DLC algorithms on the electricity demand during peak events are needed for their use in performance contracts. Simply put, we need methods that can answer the question: What would happen to the electricity demand if a DLC algorithm was not implemented? This paper compares six forecasting methods in terms of their ability to predict the baseline electricity profiles during historical peak periods in Ontario. To this end, an EnergyPlus model of an office building in Ottawa, Canada, was used to generate hourly cooling electricity use data with and without DLC. The results indicate that a motif-learning and rule extraction model outperformed four other SARIMA configurations and Meta’s Prophet, and achieved an average demand reduction prediction error of less than 30%. Keywords: Demand response; Load forecasting; Baseline modelling; Direct load controlPaper:![]()
