In 2020, if all goes according to plan, the state of California will get 33 percent of its electricity from renewable power, including solar and wind, as required by the state's Renewable Portfolio Standard. But wind doesn't blow all the time, and the sun doesn't provide as much power on cloudy days—renewable power is intermittent. This poses a problem for the electric grid's operators, who need to be able to exactly match the generation of electrical power with the demand for it at any given moment.
Because swings in the amount of available renewable power don't match the total demand for power from moment to moment, the balance has to come from somewhere, while the total power from clean energy sources must average out at 33 percent.
Another solution is to store renewably generated electricity in large, stationary grid-connected batteries, and supply power from them when renewable sources aren't providing enough in the moment. With current technology, this would be expensive, although research efforts are under way to develop less expensive energy storage technologies for the grid.
Scientists at the Environmental Energy Technologies Division (EETD) of Lawrence Berkeley National Laboratory (Berkeley Lab) have developed another solution that can help deal with intermittency: adjusting the power demand of large buildings automatically to more closely match the conditions on the electric grid. Their technological approach is called automated demand response (AutoDR), and their research was supported by the California Energy Commission's Public Interest Energy Research program, California utilities, the Bonneville Power Administration, and the New York State Energy Research and Development Authority.
Demand response is a set of activities usually carried out in commercial, industrial, and sometimes residential buildings that change, shed, or shift electricity use with the goal of improving electric grid reliability and managing costs. When demand for and the cost of electricity is high, building managers in large commercial and industrial facilities can, for example, dim lights, or turn them off in unused areas, temporarily raise a building's temperature setpoint by a degree or two to reduce air conditioning use without impacting the building's occupants, or defer the use of certain industrial equipment until later, when power is cheaper. This reduces electric demand, or "load," and the possibility of grid failure.
"Our study suggests that fast-acting automated demand response in the commercial and industrial is more cost-effective than grid-scale battery storage," says EETD scientist Sila Kiliccote, principal investigator of the research. "It offers grid operators a less expensive tool for managing the grid than battery storage. The infrastructure for demand response already exists and is growing in California in elsewhere."
Since the 1990s, EETD scientists have been working with California utilities, the state's Public Utilities Commission, commercial and industrial power consumers, and utilities around the world to test and deploy AutoDR technologies. AutoDR is a significant element of the "Smart Grid," an expression meaning that the operators, utilities, power generators, and customers on an electrical grid can respond in real-time to changes in the cost and availability of power with the help of software and hardware that monitors the state of the grid continuously.
The benefits of AutoDR and the Smart Grid include reduced chances of the grid failing during periods of high demand (a more reliable grid), lower power bills for customers who can reduce demand during these periods, and in the long run, greater energy efficiency and reduced greenhouse gas emissions.
Research by EETD scientists in cooperation with California electric utilities and large industrial and commercial customers has demonstrated that AutoDR reduces peak power use during periods of high demand. In response, the California Public Utilities Commission mandated the use of AutoDR by California's investor-owned electric utilities as a tool for managing the grid. Currently, there is more than 250 MW of AutoDR in California. Electric power authorities globally are also beginning to add AutoDR to their grid management toolkits.
EETD scientists developed OpenADR, an Internet-based communications specification used by utilities, their customers, and electric grid authorities to implement automated demand response in practice. It is one of the early Smart Grid standards. An organization of private sector companies, utilities, and research institutions called OpenADR Alliance is supporting OpenADR's adoption as a Smart Grid standard by providing certification for the devices that use OpenADR standard.
To address the intermittency of renewable power, a group of EETD researchers set out to determine whether automated demand response could substitute for batteries or other forms of energy storage to balance the grid. Today, grid operators balance the demand for electricity exactly by purchasing power one day beforehand from sources such as hydroelectric plants, and gas-powered combustion turbines.
These sources provide the peak load needed by the grid to meet the extra demand that shows up on hot summer afternoons (for air conditioning) or cold winter evenings (for heat). Grid operators make power purchases to meet peak load based on forecasts of how much demand is expected the following day.
Balancing the load on a grid supplied by a high percentage of intermittent renewables is a different problem from managing peak load, which can be estimated in advance. Power availability from renewable generation can ramp up or down rapidly, and requires grid operators to respond much faster than what the day-ahead market can provide—usually within minutes. Gas-fired power plants can provide fast response now, but because they emit air pollution, they do not meet the 33 percent renewable power requirement. Can demand response help?
The answer is yes, according to a scoping study just published by EETD researchers David Watson, Nance Matson, Janie Page, Sila Kiliccote, and Mary Ann Piette, and colleagues at KEMA.
The research team looked at the hour-by-hour electric demand in California in the commercial and industrial sectors, estimating what percentage of their electric load could be shed through automatic demand response programs, and how long these loads could be shed. The study looked at loads that could be shed for two-hour durations and for twenty-minute durations. These intervals give grid operators a range of flexible resources to match fluctuations from renewable sources. They used data from the California End Use Survey 2004 to determine what commercial and industrial loads were suitable for demand response, and currently or potentially controllable through energy management and control systems (EMCS) or system control and data acquisition systems (SCADA).
The study shows that these "fast AutoDR" sheds currently could provide between 0.18 and 0.90 gigawatts of load shedding. "With modest investments to upgrade and expand use of automated control systems in commercial and industrial facilities," says Kiliccote, "the estimated shed potential could approximately double to between 0.42 and 1.8 gigawatts." One gigawatt is one billion watts of power.
The load shedding they identified is substantially less than what would be required to balance out the load fluctuations from intermittent renewable sources providing 33 percent of the state's power needs.
However, the study also found that the cost of using AutoDR to shed load is roughly one-half to one-quarter of the deployed cost of grid-scale battery storage using current battery technologies.
"Automated demand response has the potential to balance renewable intermittency in a cost-effective way," says Kiliccote. "Combined with grid-scale energy storage and other methods, it could be an important element of a suite of tools to help operators manage the grid."
Thinking of buildings as energy storage devices is a key to understanding how demand response can be an active player in a Smart Grid system. Just as batteries store energy chemically, buildings (including refrigerated warehouses) store heat (or retain coolness) in their thermal mass.
The building operator can reduce a building's HVAC load—the energy required to heat, ventilate, or air-condition the building—temporarily, because a building will (within limits) retain its temperature for some period of time that depends on its mass, outside temperature, and other factors. The EETD researchers estimate that some buildings can shed 60 percent of HVAC-related electric demand for two-hour load shed events and 80 percent for 20-minute events for facilities with rooftop chiller units that can turn off compressors. These buildings usually include small offices, office areas within warehouse facilities, schools, lodging, and other facilities. Large office and college buildings that can make setpoint adjustments to reduce demand can shed 50 percent for both two-hour events and 20-minute events.
From previous studies, researchers know that buildings can dim or turn off lighting to reduce lighting electricity demand averaging 33 percent for a demand response event of up to four hours. Studies of retail facilities have shown a shed of 25 percent of their lighting electricity demand—display lighting requirements reduce their ability to shed load slightly compared to other commercial facilities.
Refrigerated warehouses are known from prior research to be able to reduce their loads by at least 25 percent or more for two hours without a serious change in temperature. Data centers can temporarily reduce their HVAC and lighting use as well, and water pumping for agriculture can be shifted to off-peak hours to respond to automated demand response events.
Together, these load sheds are a resource that can give grid operators one tool they need to manage an electric grid with intermittent supply resulting from a high percentage of renewable power. The grid will still need other tools for storing energy, such as grid-connected batteries and compressed air or pumped water storage. But the low cost of AutoDR makes it an attractive option for supplying some of the slack the grid will need as the state's and the world's renewable power generation capacity grows.
Follow-up research is planned. Says Kiliccote, "We need to better understand what percentage of each of these types of load sheds is available to address intermittency throughout the year. Also needed is a quantitative economic analysis of the scale up of AutoDR as a grid resource integrated with renewable and energy storage."
David S. Watson, Nance Matson, Janie Page, Sila Kiliccote, Mary Ann Piette (Lawrence Berkeley National Laboratory); Karin Corfee, Betty Seto, Ralph Masiello, John Masiello, Lorin Molander, Samuel Golding, Kevin Sullivan, Walt Johnson, David Hawkins (KEMA). Fast Automated Demand Response to Enable the Integration of Renewable Resources. California Energy Commission. Publication number: CEC-XXX-2012-XXX.
S. Kiliccote, M.A. Piette, G. Ghatikar, E. Koch, D. Hennage J. Hernandez, A. Chiu, O. Sezgen J. Goodin. Open Automated Demand Response Communications in Demand Response for Wholesale Ancillary Services November 2009 Presented at the Grid-Interop Forum 2009, Denver, CO, November 17-19, 2009. LBNL-2945e.
Sila Kiliccote, Pamela Sporborg, Imran Sheikh, Erich Huffaker, Mary Ann Piette. Integrating Renewable Resources in California and the Role of Automated Demand Response. November 2012. LBNL-4189e.
This research was supported by the California Energy Commission's Public Interest Energy Research program, and the U.S. Department of Energy.