Direct Adaptive Control of Electricity Demand

The legacy electrical grid upper-bounds a customer’s energy demand using a circuit breaker. The breaker is conservatively sized so that it rarely ‘trips,’ interrupting supply and inconveniencing the customer. This results in a system whose peak load can be much higher than its average load. Such a system can be made reliable only by being provisioned, at great cost, for infrequent peaks. The goal of demand management is to reduce the peak load and the resulting infrastructure provisioning cost. The most widely used such approach is to price a unit of energy higher during peak times, giving an incentive to consumers to reduce their demand during these times. Although this does reduce the peak load by 2-10%, this ‘human in the loop’ approach is onerous and frequently ineffective. Instead, we propose a radically different approach that draws upon the Internet traffic management approaches of proactive and reactive congestion control for fine-grained management of customer demand without human intervention. We show that this direct adaptive control of electricity demand can significantly reduce the peak load and therefore infrastructure costs. Moreover, it allows individual users to place a greater load than in the existing system, as long as it does not hurt other customers or the grid itself, thus increasing revenues. We believe that our approach has the potential to avert grid congestion, reduce capital costs, and eliminate a portion of the carbon footprint of the grid.