The electric power industry recently has undergone dramatic institutional changes. Participants are being faulted for failing to invest adequately in the electric power grid and allowing reliability to slip. Although there may be some truth to these claims, a look at historical planning practices and the evolution of the electric power grid over the last 30 years reveals something more insidious is at work.

At the heart of today's reliability problem is a natural evolution of the grid that resulted from traditional N-1 planning practices and the ever-present goal to produce new transfer capability with the least investment. For many years, lines and transformers have been added to the grid to fill gaps that limit transfer capability. Each addition has increased the loading on many lines and transformers upstream and downstream of the new line or transformer, as well as on parallel paths.

As this activity has moved the grid “mesh” from quite sparse to relatively dense, even lower investments are needed to increase the loading on ever-larger areas. While this activity occurred in accordance with planning standards, the reliability consequences have been unwitting and unfortunate. Traditional thinking suggests that filling gaps in the mesh improves reliability or at least does not degrade it. However, the resulting widespread heavier grid loading has decreased grid reliability.

In addition, phase-angle regulators and series capacitors have been added to control line flows in congested areas, thereby further increasing line loadings over large areas. The flexibility in placement of gas-fired generating plants makes it possible to place them where existing grid capacity is available, increasing the loading on lines that previously provided some margin.

Consequences of Higher Grid Loading

Thirty to 40 years ago, in most systems, one most-severe N-1 contingency would limit transfers. That N-1 event would push just one other line or transformer to its loading limit. With today's more densely meshed system, higher line loadings cause more contingencies to have significant impacts and push more lines to or close to thermal limits. Today, a system may have several relatively severe N-1 events, and each of those events will cause several lines or transformers to approach thermal limits. In effect, each contingency “tests” the system more rigorously and the tests come more often. The risk of triggering hidden or latent failures, overload cascading, voltage collapse or operator error is much higher than it was in the days of a relatively sparse network.

The higher loading puts lines higher on their reactive loss curves. The result is much higher reactive losses under normal conditions and a greater increase in reactive losses following contingencies. Coupled with the difficulty in identifying troublesome voltage contingencies and their risk, higher loading has led to a growing list of voltage-collapse events and close calls. In addition, low voltage greatly exacerbates overload cascading.

Latent failures, such as miss-set protective relays, or ones with the proverbial bent contact, can initiate cascading. When loadings are high, latent failures are more likely to be triggered with the resulting “nuisance trips,” compounding otherwise routine N-1 events. Likewise, more lines are sitting closer to sag limits, thereby increasing the risk of tree contact and overload cascading. Tree contact also occurs more quickly when the pre-contingency conductor temperatures are high.

With more lines operating near continuous ratings and many close to or at emergency ratings following a contingency, operators can more easily fail to address a contingency in a timely manner or make mistakes that cause or contribute to cascading. With today's complex and unpredictable flows, system adjustments, such as load shedding to address a contingency, are more difficult to define. Moreover, contingencies affect larger areas, increasing risk and potentially leaving neighboring operators unaware that events have put their system in jeopardy.

Increasingly onerous load characteristics are compounding the risk of voltage collapse and cascading. An industry effort to identify load characteristics in the 1970s and 1980s showed changes for the worse. Since the early 1990s, little effort has been made to stay on top of evolving load characteristics, while the increasing use of personal computers and compact fluorescent lighting and other changes continue to affect load characteristics in ways that degrade grid reliability.

The combined evolution of load characteristics and increasing grid meshing has made the traditional out-of-step block-and-trip protection, which is intended to break up unstable systems, largely ineffective on most of the U.S. grid.

Out-of-step block-and-trip protection has never been widely used, although this matters little today because it is no longer effective. The problem is that the cascading behavior involving well-distributed generators and motor loads is complex and bears little resemblance to the two-machine models on which the protection technology and device settings are based. For instance, with motors nearly everywhere, angular instability leads to motor stalling and vice versa — a much more complex failure process than was the case 30 to 40 years ago. Thus, we are without a means to deal with cascading once it reaches the stage of angular instability and motor stalling.

There are other side effects of increased meshing of the grid and the attendant higher average line loadings. A state estimator may fail to converge on a heavily stressed system after one or more contingencies have occurred. The practice of using reduced models of the “outside world” in on-line security analysis is no longer effective because remote contingencies unknown to the model have widespread impacts.

The above are some key effects of the evolution or “maturation” that the electric power grid has experienced. There are others, and all portend a lower reliability despite a more highly meshed grid.

Solutions That May Not Help

Solutions to these problems are not simple. Some will be very costly and others may not work as well as expected, or may actually degrade reliability. We need to understand our evolving reliability and the risks we are taking and get a better handle on just what the solutions are or are not doing for us.

Most new transmission lines will allow grid loadings to increase with all the attendant problems listed above. As such, new transmission lines are generally not a solution to low reliability. In fact, new regional transmission lines may simply increase the size of the area impacted by contingencies and lead to larger blackouts. Even “local” transmission line additions that increase loadings over a large area will increase the risk of overload cascading and voltage collapse, and make it more difficult to halt cascading.

Sag limits are not a new problem, but they become increasingly problematic as grid loading increases. Improved line monitoring may help with the sag problem, but also may allow higher loadings under today's planning standards, thereby exacerbating other seemingly unrelated problems such as voltage collapse. This is an example of a promising technology that may involve more issues than meets the eye, and thus have unforeseen reliability impacts.

Latent failures frequently compound otherwise routine N-1 events and thus demand unprecedented attention to detail during maintenance. However, while there seems to be potential for improvement, this is a long-standing problem and little progress has been made in the past. An aging and increasingly complex system seems to just stay one step ahead of efforts to solve this problem.

New technologies such as FACTS seem promising, but if only applied to increase transfers, they will reduce already thin margins. Such devices further complicate an already too complex system and exacerbate many of the side effects of higher grid loading. FACTS will give dispatchers additional controls, but the increasingly complex and fast-moving contingencies will make dispatching decisions that much more difficult and prone to error.

Composite conductors that promise to double or triple existing line thermal ratings will have dire reliability consequences. These conductors will be used to increase thermal capability of lines that constrain grid usage. Doing so will involve all of the side effects outlined above for new line additions. While it may be many years before a reconductored line sees a 100% increase in loading, just the reactive power implications are astounding. A 50-mile (80-km) 345-kV line loaded to 1300 MW during a contingency consumes over 500 MVAR. At 2600 MW this same line will require more than 2000 MVAR.

Wide area measurement systems (WAMS) may lead to wide-area protection that can replace out-of-step block and trip protection, but this solution is years away. Also, it will be fully effective only if dispatcher training and tools that address wide-area effects are greatly improved. Workable area protection, whether line-by-line or in a central EMS-based system, does not exist and is not on the horizon.

Some Things That May Help

Enforcement of current planning standards may be helpful but will not address the risks of an evermore heavily loaded grid. Current planning standards are based on experience that predates the characteristics of today's highly meshed grid. A major overhaul is needed that addresses recent experience and foresees impacts of the continuing evolution of the grid. Both deterministic and probabilistic criteria should be considered. Tools and computing power to implement probabilistic criteria have long been available but are not being applied.

Power-flow contingency analysis, especially for maximum credible disturbance (MCD) or possible but improbable (PBI) event analysis, should include constraints imposed by protection. This is a long-overdue improvement to contingency analysis, both for planning and on-line security analysis.

Segmentation of ac grids by HVDC (back-to-back and conversion of ac tie lines to dc) has been discussed since the early 1980s and may improve reliability, albeit at a high cost. Segmentation with HVDC can improve reliability while increasing transfer capability by limiting the propagation of disturbances. Seeing the benefits to segmentation requires thinking outside the box.

Because contingencies can affect large areas in a tightly meshed grid, dispatchers need information from a larger area and they need it more quickly. They also need better training to deal with things like decaying voltage. Dispatcher training simulators are limited and should at least include the “slow dynamics” of today's systems (automatic LTCs, load dynamics and excitation limiters). Additionally, dispatchers must have responsibility for coherent sections of the grid.

Undervoltage load shedding (UVLS), developed in the 1970s, is woefully underutilized. Today, every system is at some risk of voltage collapse and can benefit from UVLS. While guides and standards for voltage collapse and UVLS are evolving, much work remains to help utilities recognize the benefits.

Evaluating the Present

The situation we are in today simply happened. All industry participants — from government regulators to the universities preparing engineers — have been unwitting participants, even consultants. But placing blame will get us nowhere. What we need is a thorough and honest assessment of where we are, how we got here and how we can do things differently in the future.

If we find that things should be done differently, we must understand how the grid has evolved, and why standards based on the experience of years past are no longer adequate. We need engineering studies that tell us how to change planning standards, operating practices, protection and dispatcher training that, over time, will reverse recent trends. This is a complex task, given the vagaries of a competitive market. It is not only essential, but must be done in the near term and not become an R&D agenda that might bear fruit in 20 years.

If the grid evolution of the last 30 or 40 years continues, we will see normal-condition average line and transformer loadings edge ever closer to thermal limits with ever thinner thermal margins to handle contingencies. Likewise, we will face dramatically higher reactive demands, which have already risen from a minor problem in the 1960s and 1970s to a major cause of outages today. It's time for a hard look and some changes.

Harrison Clark has developed planning and operating standards, managed grid planning projects and investigated seven blackouts, including the 1977 New York City blackout. He served on the Blue Ribbon Panel assembled to examine the 1996 Western Interconnection black-outs. He also helped define the voltage collapse problem and developed tools and solutions, including VQ analysis and undervoltage load shedding. He spent 26 years with Power Technologies Inc.

hkc@hkclark.com