The electric power system is unique in that it must match aggregate production and consumption instantaneously and continuously. Several types of controllable reserves are maintained to help the system operator achieve this required generation/load balance. The continuous random minute-to-minute fluctuations in load and uncontrolled generation are compensated for with regulating reserves. Frequency deviations are compensated for with frequency-responsive
reserves, and the daily cycling of load through load following and generator dispatch. Sudden failures of generation and transmission are compensated for with three additional reserves: 10-minute spinning reserve, 10-minute non-synchronized reserve, and 30-minute operating reserve.
Conceptually the generation/load balance can be maintained by controlling generation, load, or both. Historically, system operators have tended to control generation almost exclusively. Generators are typically in the business of providing their services to the power system, so their business model (whether they are owned by an integrated utility or are independent) accommodates following system operator directives. Communication and control technology also made it easier to monitor and control a few large resources than numerous smaller resources. Consequently, the rules governing how the power system is operated were developed at a time when large generators were essentially the only resources available to support system reliability. Rules were prescriptive as to the actions to be taken and the technologies to be used, rather than being results oriented (i.e., performance based).
While responsive load can theoretically provide almost any service the power system requires (black start may be the only exception), most loads are best suited to provide contingency reserves. Contingency reserves restore the generation/load balance after the sudden unexpected loss of a major generator or transmission line. Power system frequency drops suddenly when generation trips, as shown in Fig. 1. There is no time for markets to react. In the case illustrated in Fig. 1, frequency-sensitive generator governors responded immediately to stop the frequency drop. Spinning and supplemental reserves successfully returned frequency to 60 Hz within 10 minutes. Power systems typically keep enough contingency reserves available to compensate for the worst credible event (contingency). This is typically the loss of the largest generator or the largest importing transmission facility. In Texas, the simultaneous loss of two nuclear plants is credible (as shown by the event recorded in Fig. 1), so the Electric Reliability Council of Texas requires over 2600 MW of contingency reserves. Frequency response, 10-minute spinning, 10-minute non-synchronized, and 30-minute operating reserves operate in a coordinated fashion, as shown in Fig. 2.
Restructuring has changed the business relationships between generators and the system operator. Technology has advanced to allow loads to be responsive. Energy costs have risen (in some regions) and become more volatile (almost everywhere) from hour to hour, providing incentives for loads to respond. Rules established by regulators and technical organizations are being changed to accommodate this new set of circumstances.
REGULATIONS AND POLICIES
While the general concepts of system operations and reliability are well established, implementation details continue to evolve as the industry is restructured. FERC, NERC, NPCC, NYSRC, NYISO all have rules and procedures that govern contingency reserve requirements. These rules tend to become more specific organizationally closer to the system operator. These rules are not yet consistent among organizations, but the trend toward open, technology-neutral, market-based solutions is clear.
CONCLUSIONS AND RECOMMENDATIONS
Responsive load has the potential to be a more reliable and lower-cost supplier of contingency reserves, especially spinning reserve. This report provides detailed results from one example technology, the Carrier ComfortChoice responsive thermostats deployed in the LIPAedge program to provide peak demand reduction through central control of residential and small commercial air-conditioners. Preliminary analysis indicates that these responsive loads could be excellent providers of spinning reserve. Roughly three times the load reduction capacity is available for contingency events (75 MW) as is available for peak reduction (25 MW); two-thirds of the capacity (50 MW) is still available to supply spinning reserve when the loads are already curtailed for peak reduction. This added spinning reserve capability is currently available and should be utilized (tested) during the summer of 2003. Large numbers of these responsive thermostats have also been deployed by ConEd, SCE, and SDG&E and are similarly good candidates for immediate use and testing.
The preliminary analysis shows that load response is likely to be faster and more effective than generation response for providing spinning reserve. Commands are received and response typically completed in 90 seconds as compared with 10 minutes for generation. Extensive testing and monitoring should be conducted before and throughout the summer.
There are two areas where this technology does not currently meet strict interpretations of spinning reserve requirements: monitoring speed and frequency response. Frequency response capability could be added relatively cheaply, for perhaps $1–$10/device for equipment not yet installed. All of the signals and expensive control equipment are already there. But this capability will not be added unless manufacturers see that there is a real demand and that response specifications are established and stable. NERC, NPCC, NYSRC, and the NYISO need to decide if frequency response is a spinning reserve requirement or an AGC requirement.
Real-time supervisory control and data acquisition monitoring (2–8 seconds) is the only requirement that is fundamentally difficult to meet. To avoid overwhelming the paging network, responses from individual units must be staggered when the entire system is being polled. It can take 90 minutes for 20,000 units to respond. But unlike large generators that can completely fail to respond because of an equipment problem at the generator, it is unlikely that 20,000 individual air-conditioners will fail simultaneously. The communication backbone could fail, but that can be monitored separately at any rate desired. Monitoring requirements that are appropriate for an aggregation of small resources need to be established.
LIPA should begin discussions with NYISO immediately to determine what additional research and analysis is needed to allow this responsive load to supply 10-minute spinning reserve and obtain ICAP and hourly spinning reserve payments for the service. If necessary, an exception to NYSRC rules should be requested to allow LIPAedge response to be tested as spinning reserve during the summer of 2003.