Demand response, aka load shifting, or virtual power plants, keep power grids in balance by curtailing non-essential loads, during times that would otherwise require ramping up peaker plants, or when grids simply do not have enough capacity overall. The costs of demand response are estimated at c$60/kW/year, across 25 case studies in this data-file, for a net present 20-year cost of $500/kW. This is less than batteries or peaker plants and supports the potential for smart grids.
Voltus is the largest platform in North America for coordinating demand responses, aka load shifting or virtual power plants. By 2025, it has 7.5GW of resources under its control, across 70+ programs at 20,000+ sites across 50+ industries, in each of North America’s ISOs; and over the past decade has paid its customers $200M+ by allowing them to flex their loads and participate in wholesale markets.
Voltus remains privately owned, after cancelling plans to go public via a $1.3bn SPAC transaction in 2021-22.
What are the costs of demand response programs? We have aggregated data from 25 case studies on the Voltus website. In each case, a customer installed Voltus’s 30-second monitoring device and gained access to the Voltus demand response platform, which allows real-time consumption to be visualized, and to be reduced during times of grid strain.
In the average case study, a typical customer has 6MW of capacity enrolled in demand response, generating revenues of $60/kW per year. The amount of MWH shifted per year per MW of load is also back-calculated in the data-file. This is an annual number, which translates into a net present cost (to the utility) of around $500/kW over a 20-year basis (chart below).
CO2 savings are also achieved. For example, if we think that the average CO2 intensity of the US grid is 0.4 kg/kWh, and falling; while a typical peaker plant might have a CO2 intensity of 0.57 kg/kWh, then demand shifting is going to yield a saving of 0.17 kg/kWh; or more if the loads can specifically be moved to times of elevated wind/solar generation.
The breadth of potential customers stood out in the data-file, spanning residential, industrial, commercial, utility and data-center customers. As did the ease of onboarding customers to the Voltus platform. This reinforces our conclusions that demand response could ramp up to smooth c20% of peak grid loads.
In one example, a 24MW midstream company slows down all pumping equipment via their SCADA systems during demand response events. In another example, a commercial real estate firm in New York pre-chills select spaces ahead of anticipated demand response events, and also shuts down select elevator banks. In other examples, a mine shuts down water pumps, a food processing plant reduces on-site CO2 production or a crypto-miner shuts down entirely.
We see great potential to smooth out volatility in power grids via AI-coordinated load-shifting. Average costs of demand response, as quantified in this data-file are clearly much lower than grid-scale batteries and peaker plants. Further details are in the data-file. Underlying details come from the Voltus website.