The complexity of operating a transmission system has increased over the years. The failure of a single critical element, if not managed correctly, can result in cascading outages and distribution of the entire transmission system. Intervention by system operators is often not possible because voltage collapse can take place in fractions of a second. Maintaining reliability sometimes requires automatic load-shedding schemes. The Public Service Company of New Mexico's (PNM) undervoltage load-shedding (UVLS) scheme is one design that prevents either a fast or slow voltage collapse for low probability events involving the loss of multiple transmission elements.

PNM (Albuquerque, New Mexico, U.S.) is an investor-owned utility that serves native retail load customers and provides network integration and transmission service to third parties in the Northern New Mexico portion of the Western Electricity Coordinating Council (WECC) grid. The majority of PNM's generation assets are located at the San Juan Generating Station. In essence, PNM's bulk transmission system transmits power from Northwestern New Mexico to load centers in the northern half of New Mexico through WECC Path 48, as shown in Fig. 1.

Fast voltage collapse in Northern New Mexico can occur with the simultaneous loss of both 345-kV lines from the Four Corners/San Juan area to the Albuquerque area (FW and WW 345-kV lines) during heavy import levels on Path 48. Prior to implementing an automatic UVLS scheme, the only protection against this scenario was manual load shedding. UVLS also provides the benefit of added flexibility for maintenance and repair activities that require scheduled outages when the system load levels may not be sufficiently low to ensure that transmission limits are not exceeded for the next contingency.

UVLS schemes implemented by other utilities generally fall in one of the following categories:

• Those that require the use of individual undervoltage relays at specific locations to disconnect load within a specific time when local voltage drops below a predetermined threshold.

• Those that use measurements from the SCADA system to determine when to trip load.

The PNM system is particularly susceptible to a fast voltage collapse scenario. The UVLS implementations described above would not be effective for the PNM system, because they rely on the assumption that voltage collapse will occur within a relatively long time frame (up to several seconds). A relay-based UVLS scheme would require short delay times, which could result in excessive or insufficient load shedding as well as misoperation for system faults. On the other hand, the time delays associated with the SCADA UVLS systems would be too long to be effective in this application.

PNM designed and implemented a robust Import Contingency Load-Shedding Scheme (ICLSS) to provide reliable system protection against voltage collapse under a wide range of system operating conditions.

Simulation Studies

The most limiting N-1 contingency is the outage of either the WW or FW line. For study purposes, the case with the FW line initially out of service (FW-ios case) was used as the initial condition to determine the requirements for UVLS. Beginning with the FW-ios case, the loss of the remaining 345-kV transmission line into the Albuquerque area (WW line) could lead to voltage instability in two distinct time frames. In the first time frame (0 to 20 seconds), voltage collapse can occur as a result of induction motor characteristics, loss of the Blackwater high-voltage direct current (HVDC) converter or the effects of generator over-excitation limiters (OEL). In the second time frame (tens of seconds to several minutes), voltage collapse could occur as a result of load restoration by transformer load tap changers (LTC).

Preliminary analyses were conducted with the transmission system represented down to individual distribution substation transformers. Loads were represented at the low-voltage side of the substation transformers. Transient stability simulations were performed for the outage of the WW line in the FW-ios case. All loads were initially modeled as constant current for real power and constant admittance for reactive power. The dynamics of induction motor loads were not modeled in the initial study. Figure 2 shows a simulation of the WW line outage without a fault. Note that voltages were initially above 1.0 pu and quickly dropped below 0.6 pu after the outage. Although this case is “stable,” load bus voltages in the Albuquerque area dipped below 60% for long intervals, indicating that if induction motors had been modeled, many of them would have tripped or stalled.

The transient response of generators close to load centers is critical to voltage stability. For the simulations, it was assumed that the Pruitt-Escalante Generating Station (PEGS) in Western New Mexico was operating near its maximum output. Figure 3 shows that the WW line outage forced the PEGS field voltage to its maximum output. Field current reached approximately two times its rated value. Reactive output reached 450 MVAR. Under this condition, the machine would trip off line within an estimated 20 to 30 seconds.

The PEGS OEL was modeled as a simplified controller intended to mimic both the actual automatic OEL protection and the expected operator intervention. If the field current exceeds its rated value, the field voltage is ramped back to its rated value after a short delay. This action limits the steady-state reactive output of the generator to a value near the upper limit of the generator reactive capability. Figure 3 shows that 3.3 seconds after the line trips, the OEL acts to bring the field voltage back to its rated value.

Effect of Load Models on UVLS

For transient stability simulations, the real and reactive power portions of the load are typically modeled as constant current and constant impedance, respectively. At low voltages, actual load response can vary significantly from these typical assumptions, which can affect the requirements for a UVLS scheme.

As power systems are operated closer to their voltage stability limits, it is important to represent load characteristics more accurately. For most utilities, obtaining good modeling data for power system simulations is difficult. Load characteristics in an area of interest may differ significantly from generic load models and guidelines. In particular, the amount and type of induction motors can have a significant impact on system dynamics. As the motors slow down (slip increases), they draw increasing amounts of current. Many motors, particularly those driving compressor loads, will stall at sustained low voltages, drawing four to six times normal current. Stalled motors are typically tripped by thermal protection within several seconds. For voltages below 70%, larger motors, such as commercial air conditioners, are likely to trip within five cycles via undervoltage relays.

In recognition of the importance of load modeling, PNM and EPRI sponsored a research effort with the primary objective of developing dynamic load models to assist in the design of PNM's UVLS scheme. Based on the results of this research, 30% of the load was converted to induction motors for transient simulations. The remaining load was modeled as constant impedance. Undervoltage relays were modeled on each of the induction motors.

Figure 4 shows a simulation of the WW outage with the modeling refinements described above. It should be noted the back EMF of the induction motors helped support the voltage for a short period of time before the voltage dips to 0.7 pu after the outage. The results should be considered optimistic since a large portion of the motors tripped off line involuntarily. Because of many unknowns regarding motor control and protection, disconnection of load should not be relied upon to prevent voltage instability.

Development of Load-Shedding Scheme

Figure 5 shows an overview of the ICLSS diagram. ICLSS will shed load by tripping radial and looped subtransmission lines. If it is necessary to relieve underlying system overloads, the ICLSS will instruct the distribution SCADA computer to shed load by tripping individual distribution substation feeders. Note that permissive signals from undervoltage relays at the subtransmission stations are used to improve the security of ICLSS.

Implementation of High-Speed Load Shedding

Figure 6 shows bus voltages within the Albuquerque area after the WW outage, assuming that the ICLSS is in service. A total of 13 load-shedding stages were initiated by ICLSS, tripping 341 MW of load as noted in the figure. This action prevented a fast voltage collapse scenario. An extended simulation shows that for this particular case, ICLSS would shed an additional 279 MW of load to mitigate the Ambrosia — West Mesa 230-kV line (WA line) overload. Therefore, ICLSS would have shed a total of 620 MW to stabilize the system for the FW-ios case and WW line outage.

Implementation of Slow-Speed Load Shedding

A slow voltage collapse scenario could take place several minutes after an outage, as the LTCs act to restore load to pre-disturbance levels. Even in this time frame, it may not be possible to prevent voltage collapse by operator intervention.

Long-term simulations with LTC dynamics included were conducted to demonstrate the effectiveness of ICLSS in preventing slow voltage collapse scenario. Typical PNM LTCs have ±10% regulation capability with 32 tap steps. The tap relay bandwidth is 3 V on a 120-V base. For voltages outside of the control bandwidth, the LTCs act to change taps after a predetermined time delay.

Long-term simulations show that for the loss of WA and FW lines, the transmission system is subjected to a slow voltage collapse condition as load is restored by LTC action (Fig. 7). This figure shows the transformer primary and secondary voltages (115 kV and 4.16 kV, respectively) at a unit substation in the Albuquerque area. As the transformer increased taps, the secondary voltage increased while the primary voltage declined steadily. After a 10-second delay, with the West Mesa and Sandia 115-kV bus voltages below 0.90 pu, ICLSS started to trip distribution feeder loads to arrest the voltage decline. ICLSS dropped a total 16.7 MW of load during this simulation.

Operating Experience with ICLSS

On March 18, 2000, prior to the commissioning of ICLSS, PNM experienced a multiple contingency disturbance that lead to the blackout of the Northern New Mexico area. Prior to the disturbance, the San Juan-Ojo 345-kV line had been taken out of service for maintenance (Fig. 1). The FW, WW and WA tripped within 20 minutes of each other due to a grass and brush fire in the Four Corners area. System voltages in the Albuquerque area fell to nearly 60%. The remaining 345-kV line interconnection to the south was severed as voltage collapsed in the Northern New Mexico system. PNM lost approximately 815 MW of load (355,000 customers). It is estimated that other utilities in New Mexico lost more than 500 MW of load. This event was simulated in sufficient detail to determine that ICLSS would have been effective in preventing voltage collapse in the Northern New Mexico system.

Recently, the PNM system has experienced two events where ICLSS, as designed, did not shed load, even though the events were severe. In both cases, the system relied on ICLSS to survive the next contingency. The sequence of events for each these two disturbances are explained below:

March 7, 2003

PNM lost the FW and WA transmission lines due to an aviation accident north of Albuquerque. PNM subsequently lost a third 345-kV transmission line when the Blackwater HVDC converter station tripped. Generation in metro Albuquerque allowed the system to serve the load until all lines had been returned to service. Load shedding by ICLSS was not required and did not take place, and the system remained within acceptable voltages and equipment loadings.

Oct. 3-4, 2003

On Oct. 3, 2003, the FW line faulted and tripped due to an insulator string failure. On Oct. 4, 2003, while the FW line was being repaired, the WW line faulted and tripped. Under this condition, ICLSS would send a transfer trip signal to shed load, with no intentional delay. However, since the voltage at West Mesa did not fall below 95%, ICLSS responded correctly and did not shed load. System voltages did not fall significantly due to the presence of on-line generation in metro Albuquerque and low load levels.

Conclusions

UVLS is an effective (technically and economically) control-based measure to prevent voltage instability and collapse for operating conditions and disturbances that have a low probability of occurring. UVLS provides fast, controlled, reliable, and predictable load shedding. It eliminates the delays and unpredictability of conventional UVLS and removes the burden from the system operators who would otherwise be forced with the impossible task of shedding load to avoid voltage instability.

The voltage dependency of the load has a tremendous influence in voltage security analyses. In particular, induction motor modeling is critical due to the increased reactive power demand from these devices at reduced voltage. The effect of load modeling assumptions needs to be taken into account in the design of UVLS.

In the examples of March 7 and October 3-4, PNM was able to continue to serve load without implementing load shedding as a preventive measure in preparation for the next contingency.

As a method to increase system security, UVLS is less effective than adding transmission lines. However, UVLS allows utilities to manage the risk of voltage collapse for contingencies that have a relatively low likelihood of occurring. Controlled load shedding reduces a potentially catastrophic, widespread disruption to a series of short-term localized outages.

Jeff Mechenbier is the manager for the Transmission Analysis Department with Public Service Company of New Mexico, where he is responsible for the planning of the transmission system. He has a BSEE and MSEE degrees from New Mexico State University. He is a senior member of IEEE and a registered professional engineer in the state of New Mexico and Oregon.
jmechen@pnm.com

Abraham Ellis is senior engineer with Public Service Company of New Mexico, where he is responsible for transmission planning studies. He has BSEE, MSEE and Ph.D. degrees from New Mexico State University. He is a senior member of IEEE and a registered professional engineer in the state of New Mexico.
aellis@pnm.com

Richard Curtner is the principal engineer with Public Service Company of New Mexico, where he is responsible for protection and control of the transmission system. He has a BSEE degree from Ohio State University and an MSEE degree from New Mexico State University. He is a member of IEEE and a registered professional engineer in New Mexico.
rcurtne@pnm.com