Community-scale residential air conditioning control for effective grid management

This paper investigates the potential for coordinated control of a large number of residential air conditioning systems to achieve substantial reductions in peak electricity demand. To do so, an extensive data set including home energy audits, homeowner surveys, and electricity meter measurements from homes in Austin, Texas, USA, was used to build a simulated community of 900 homes. Based on a reduced-order modeling strategy and an economic model predictive control approach, we analyze the effects of the community of homes responding optimally to variations in wholesale market electricity prices. We find that when exposed to dynamic pricing, peak demand from residential electricity consumption is shifted to earlier in the day, and is lower than the peak where no intervention is made. We also consider centralized and decentralized strategies for minimizing the peak demand of the community. For this simulated community, we find that centralized, coordinated control of residential air conditioning systems reduces overall peak by 8.8% but increases total energy consumption by 13.3%. Decentralized control reduces overall peak by 5.7%, demonstrating that the value of information sharing for peak reduction is 3.1%. It is also shown that properly tuned penalty terms allow a penalty-based decentralized controller to approach the optimal solution obtained by a centralized controller without the requirement of information sharing.