by Planning Engineer
The costs of major grid outages are staggering and recovery from such outages is challenging; therefore the North American grids are planned and operated to ensure high levels of reliability.
Despite changing conditions and various threats, it is widely expected that that current levels of reliability will be maintained or improved upon. The grid is impacted by multiple electro-mechanical effects that planners have learned to model and plan for over time and through experience. The rapid deployment of any new technology will present both modelling and operational challenges to maintaining high levels of grid reliability.
With the increased focus on reducing fossil fuel generation the question frequently comes up as to, “How much solar and wind can be integrated with the grid without unduly impacting system reliability?” The increase in renewables relative to conventional generation is often referred to as “penetration”. The US grids have sufficient robustness such that small penetration levels do not pose excessive risk, however high levels of penetration raise serious reliability concerns. This post will argue that there is not a single answer and that the answers are not easy, therefore estimates will involve considerable uncertainty. Casual readers may want to read the “Key Points” and then skip to the “Conclusions” or specific topics of interest. Those seeking a more optimistic assessment may want to read Volume 4 of the National Renewable Energy Laboratory’s Renewables Electricity Futures Report.
- There has been a high value placed on having an extremely reliable bulk grid as the costs and consequences of bulk grid outages are severe
- The bulk grid supports and is supported by conventional rotating generators (Coal, natural gas, hydro, nuclear, biomass) which provide “Essential Reliability Services” (ERSs)
- Wind and solar provide increased reliability risks because they are new changing technologies, they are intermittent and they do not as readily provide ERSs
- Current high levels of reliability depend upon experience gained over time through the gradual adoption of new technologies
- Wind and solar can be made to provide approximations of ERSs, but that requires significant increased costs and reduced generation output
- Because of the complexity of impacting factors and the high level of reliability maintained for the US grids, systemic degradation of the reliability of the grid is hard to detect and measure
- The amount of renewable penetration that can be accommodated will vary from area to area and power system to power system – There is not a single answer
- Because conventional resources produce an abundance of ERSs, accommodation of low levels of renewables may be accomplished with negligible risks
- Because current renewables do not provide adequate ERSs high penetration levels provide significant risks
- Between the above two levels there is a gap of (wicked?) uncertainty.
- For assessing grid reliability, the maximum penetration of wind and solar during times of stress is the key number not the “average” contribution of wind and solar
- Increased penetration of such asynchronous resources, all else equal, will likely adversely impact bulk grid reliability
- As the penetration level of asynchronous generation increases this will either increase cost, limit operational flexibility, degrade reliability or most likely result in a combination of all three factors
The above statements have the following important caveats
- In some situations renewable resources may have some practical benefits and better support reliability in some limited applications For example:
- Air quality standards often prohibit the location of new generation resources in congested areas. If renewable resources are allowed to be located close to load centers –the system may see benefits
- Electronic emulation of ERSs in some cases will not be as good as actual synchronous machines, but with proper controls it may also be better in some cases
- Given time the reliability risk associated with new technology can be reduced as more experience is gained so that penetration levels can be increased
What is Meant by Bulk Grid Reliability and Why is it So Hard to Measure?
Bulk Grid Reliability applies to the high voltage backbone system that supports bulk generation and the load serving distribution systems. Bulk reliability is concerned with preventing voltage collapse, instability, cascading outages and uncontrolled separation. Events of this sort make national and international news when they occur on modern grids and are called blackouts. They are widespread unplanned, unintentional loss of load. There is a fundamental difference between these events and brownouts. The term brownouts sometimes refer to periods of low voltage and these can damage equipment and disrupt service. When load is deliberately shed (outages occurs by design in a controlled manner) this is also sometimes referred to as a brownout, blackout or as rotating blackouts. This situation can result from poor reliability, but is actually employed as a reliability measure to protect the bulk grid from a major outage. Avoiding needed load curtailments (in order to avoid adverse public relations and the economic burdens imposed by such curtailments) is a reliability risk itself. The impacts of uncontrolled outages are far more severe than controlled or contained area loss of load.
The Northeast Blackout of 2003 is an example of a bulk system blackout. It impacted 55 Million people in the US and Canada, causing an estimated $6 billion in damage, shutting down major cities, interrupting industrial processes, leaving many businesses, residences and industries without power for days, some for nearly a week and contributing to at least 11 deaths.
Grid Reliability is very different from distribution reliability, which is more concerned with individual consumer outages (or even city wide blackouts in some cases) that occur due to localized distribution networks. Distribution outage data is not directly related to grid reliability and cannot serve as a measure of grid reliability, although it is not uncommon to see articles touting the strength of the German grid based on measures of distribution reliability. Assessing grid strength based on measures of distribution performance is as inappropriate as would be assessing the foundational strength of a bridge based upon the crash performance of its guardrails.
Major power outages (blackouts) are costly infrequent events and are hopefully are becoming rarer and more infrequent on modern grids. Unless caused by major natural disasters they are typically associated with multiple things going wrong. In a complex, 24 hours a day system over the course of time it is inevitable however that at times multiple things might go wrong, however power systems should be sufficiently robust that such confluences of events do not lead to major disturbances. There is not a clear easy answer as to what is “robust enough”.
As noted major grid outages are rare. If the reliability of the system is significantly reduced, we may not see additional major outages for years anyway. Moderate sized utilities may spend tens of millions of dollars to reduce risks related to major outages that might have less than a 50/50 chance of occurring in a 50 year time span. The results of such valuations of reliability are not readily observed in any available performance measures. The industry is struggling to develop such measures. The best thing we have going for us beyond our models is the experience with current technologies. If the current level of reliability were to be greatly reduced it could be a while before the impacts become obvious. But over time consequences will emerge. If bulk system reliability is allowed to degrade it will be challenging to turn it around and return to today’s high levels.
Planners have confidence that at low levels of penetration the existing conventional synchronous rotating machines will likely provide sufficient robustness in most cases. At very high levels of penetration from intermittent renewables the system will disproportionately lack contributions from synchronous machines which support the grids stable and reliable operation. The “uncertainty range” is large. Generally the more synchronous rotating machines with inertial mass the better. Planners desire a large amount of margin (as it can be eaten up by unusual and unanticipated events).
Conventional Rotating Generators Support the Grid in Ways That Solar and Wind Do Not
Conventional generation has characteristics that support the stability and operation of the grid. They have inertial mass and spin in synchronism with the wave forms powering the system while readily providing voltage and frequency support. The grid was built upon and depends on the characteristics associated with large rotating machines. For additional, more detailed information on this topic see the following posts: “Transmission Planning: wind and solar”, “More renewables? Watch out for the Duck Curve”, and “All megawatts are not equal”.
The North American Electric Reliability Corporation (NERC) recently issued an announcement stating that “New generation resources must provide adequate levels of frequency support, ramping capability and voltage control to maintain the reliability of the bulk power system during its ongoing transformation”. This NERC Report provides more detail as do these basic introductory videos on load ramping, voltage and frequency. They describe the collective desirable characteristics of generation known as essential reliability services (ERSs).
Modern wind resources do not economically spin in synchronism with the grid so they are electrically decoupled from the system. Solar generation does not involve rotating machinery. Solar and wind do not inherently provide ERSs. In addition wind and solar are intermittent resources. This dramatically increases the number of potential operating scenarios to be studies and increases the chances that unanticipated scenarios might cause problems.
An additional emerging area of concern centers on protecting the system from faults. When faults occur on the system it is essential that that the impacted parts of the system be quickly disconnected and isolated from the grid. Extended faults can cause significant damage and lead to system collapse. When a fault occurs, conventional generators contribute an inrush of “fault current” (short circuit current). The protective devices (relays) sense parameters associated with fault current and know how and when to operate to properly isolate the fault. Closer relays respond more quickly and more distant relays operate more slowly in order to serve as backup in case of failure. This approach minimizes outraged elements while providing redundancy for relay failures. When inverter-based generation (wind and solar) replaces traditional synchronous generation the fault current contribution is approximately cut in half. Fault currents will vary significantly depending on whether there is low or high renewable penetration at any given time. With higher than expected fault currents, distant relays may trip too soon. With lower than expected fault current all relays may not operate quickly enough. Accommodating higher and varying penetration levels will require comprehensive studies, system-wide relay coordination efforts, large capital improvement projects and labor intensive testing and engineering in order to implement the new relay schemes.
Can Wind and Solar Provide ERSs?
Wind and solar through electronic emulation can be made to operate more like conventional generation, but it generally comes at significant additional costs. For example a wind or solar facility can have the capability of providing an extra boost of energy to the bulk system when it is needed. However to do that, for all other hours you have to “waste” a portion of the output to serve as backup reserve. The economics of building a facility and only using 90% of its output so that 10% can be kept to support the grid during limited times of system stress may be challenging, but that will become more feasible if costs drop relative to other resources.
Wind and solar may have special advantages in some cases. For example, if wind and solar can be located near load centers where regulations do not allow fossil fuel generators. Also electronic emulation may allow the artificial response of generation resources to be more supportive in some cases. Synchronous generators sometimes over-react to system disturbances. So a simulated response may be better in some cases. It will be challenging determining what those responses should be so that they work well across a myriad of scenarios. At this time the limited benefits and emulation capabilities associated with asynchronous generation have potential but these resources should not be viewed as equivalent to synchronous resources or fully worthy competitors just yet.
New Technology imposes risks – We don’t know what we don’t know
Conventional technology also benefits from considerable experience gained over the years. Power systems are the largest most complex machines in the history of the world. There are many electro-mechanical interactions that can adversely impact the system. To ensure their reliable operation various models on components of their performance that span differing time frames and differing conditions are employed. The models and modelling techniques have been developed and refined over long time periods as new technology is employed, sporadically at first, and then expanded as the technology proves itself and becomes more widely adopted. Policies to increase wind and solar may lead to unprecedented changes for the bulk power system. No one can say with any certainty that our existing models and study approaches are sufficient to guarantee that new problems associated with new technology (and the interaction of such technology with conventional technology) will not emerge.
For example series compensators were put into long transmission lines to enable long distance transfers of power without having excessive voltage drops. As modelled they worked well and no problems were anticipated. In practice it was discovered that they could produce catastrophic results from sub synchronous resonance (SSR) between the interconnected generators. (Rough translation – noise from generators is shared and compounded below the normal 60 Hz frequency, such that the generators start to oscillate, shake and possibly break.) This phenomenon was outside the normal study arena at the time. Solutions were found and today series compensators are widely used for their benefits.
Another risk that was observed before it was detected by models was fault induced delayed voltage recovery (FIDVR). The problem emerges when there is a large amount of induction motors located far from var producing resources. Unlike a lightbulb which helpfully reduces its demand for current when voltage is low, induction motors demand more current and vars when voltages drop. (Rough translation- a local area sees stress from air conditioners which suck extra current and vars during a voltage drop, and the problem is compounded because there are not enough local generators to support the system during the disturbance.) Load models were not accurate enough to pick up this problem as it emerged. (It’s easier to model large generators than a host of different load elements.) Luckily FIDVR problems did not hit the entire US at once. Some systems (hot urban areas with restrictions on new generation within the air quality management zones) discovered the problem as they studied real world performance issues and tried to “reconcile that with modelled performance.
Rapid changes in generation resources employing a variety of new technologies will create uncertainty. When new technology is introduced across a wide area at a fast pace at levels never seen before, it definitely will heighten the risk that major outages will occur before the problems are identified and fixed. With new technology problems may emerge outside of our existing study areas. In addition our existing modeling methods will be challenged. A model of a 1000 MW coal plant will likely be more accurate than the aggregate of the many models simulating a vast array of solar panels from different production runs, with different technological tweaks and possibly different user settings employed. It’s naïve at best to presume that rapid sweeping changes do not impose special risks.
Experience is Crucial
The 2013 Super Bowl held in the New Orleans Superdome was delayed due to a 22 minute partial power outage. No power supplier wants that kind of publicity. Watching the game, I felt sure that the problem was not the result of scrimping on costs or the use of older equipment or technology. Turned out the system had been recently upgraded and a few months before the game there were still concerns that there was a “chance of failure”. So they spent an extra $1 Million on upgrades. New relay equipment was installed to “ensure” continuous power supply in case one supply line failed. They switched from a system that worked in practice to one that was theoretically better on paper. But for the uprates, I believe there would not have been any outage related impacts upon the game. There has been some dispute over how those relays were set and why they did not operate correctly, but it is clear that before the event on paper the new system looked great. Mathematical models can only go so far with ensuring that any technology will work. Real world experience cannot easily be replaced with modeling.
Planning and Operating the future system with increased renewables is problematic especially assuming rapid changes
There are multiple electromechanical forces operating in vastly differing time frames and interactions among these can adversely impact the grid. As noted above in the example of sub synchronous resonance, studying the system at one level might not inform you of problems at other focal points. The reliability of the grid depends upon both modelling and experience.
Current load models of the system are lacking compared to models for generators. For large generators it is fairly easy to get good models of their behavior because the units are large, are made by a limited number of manufacturers and the units can be tested. Load elements are much more diverse and vary more hour to hour and seasonally, contain too many elements to be modelled individually and test can’t be run on system wide loads. The greater integration of renewables will introduce more uncertainty in to the generator side of the equation as it will be harder to get good modelling information for dispersed resources and for resource blocks containing multiple varying components. Today simulations can be done using “good” generation data and sensitivities changing load models. Performing simulations and making system assessments that account for unknowns in both load and generation may exponentially increase study challenges.
Planners perform scenario outages of the system under stressed conditions with various outages modelled. When the system response is inadequate the planning standards require that system improvements (reinforcements) be made. I will share a planning secret. We don’t really think that the specific outage and the specific conditions which were identified in the study will actually occur and the system will be “saved” by that particular fix. We have learned over time that planning that way results in a system that is sufficiently robust so that system operators can sufficiently recover when unanticipated events happen across variety of circumstances. Planners will model the new technology as best they can, but if adoption of new technology is rapid, they will not have the needed experience behind them to justify confidence in the models. Because of the larger potential for unanticipated actual configurations, system operators will be functioning under conditions of greater risk and uncertainty trying to ensure reliability across a widening set of operational parameters.
What options are available to maintain high levels of reliability?
Given that new technology introduces greater uncertainty as to modelling, scenario analysis and performance, what can be done to maintain system reliability? One answer is to invest more heavily in the system to better support system reliability. This can be accomplished through requiring extra generation to be on-line, adding stronger transmission ties, employing more sophisticated electronic protection devices and adding a variety of specialized power system equipment such as synchronous condensers and static and dynamic VAR devices. Beefing up the bulk system could easily double transmission costs.
Another answer is to limit the operational flexibility of the system. For example when the operation of intermittent resources creates a dispatch scenario that appears problematic or goes beyond what can comfortably be compared to modelled scenarios, the system operators may choose to curtail (or force the operation of) certain generation. The generation likely to be curtailed in such situations will likely be the lower incremental cost intermittent renewable generation. The generation forced to run will likely be the higher incremental cost conventional generation. The operators may also cut transfers from one area to another to limit risks. For example regional wind transfers may have to be cut when they would otherwise provide the most value. The cost impacts of maintaining reliability in this manner will be large as well. Additionally operators may find the system will react better to small disturbances if they are willing to cut consumer load to help keep the system strong. This approach can make the bulk backbone system more reliable but will have serious consequences. The US grid will function more as a third world grid with increased small scale outages, costly resource inefficiencies and the inability for renewables to operate as needed achieve their planned economics and production goals.
Problems with Picking a Target Penetration Level for Wind and Solar
Often renewable targets are set without any regard for bulk reliability concerns. When such concerns are recognized they often not well understood. Targets typically refer to average values of generation provided by different resource types. For system reliability the average does not matter. It’s the penetration level at specific operating states. If a target is set at 20% of energy needs to be met by renewables, at times renewables will make up less of the system load and so at other times must make up a larger percentage. Depending on the renewable resource mix system operation may require very high penetration levels at some times in order to achieve much lower average targets.
Occasionally you will read about some “system” where renewables served a large portion of the load. This typically occurs when load demand is low and conditions are good for wind and/or solar generation. Renewable enthusiasts often gush over such reports, but this is likely not a good situation from the perspective of grid reliability. Such periods create a tension between maximizing the economics of renewable resources and the reliability needs of the transmission system.
The achievable penetration level will vary from system to system depending on factors such as the characteristics of available synchronous generation, the characteristics of the load, specialized transmission elements and the distance between existing generators and load centers. It will also vary from time to time within a given system in response to the differing generation dispatch mixes and load levels. Historically the greatest reliability concerns have occurred during peak periods. Large base load coal plants were dependably on line during off-peak periods and supported the system with ERSs, such that the system was usually very robust at those times. The unfolding future will spread the risk across more hours and scenarios and it may be that the greatest risks will occur at lower load levels. As suggested above it will be challenging to model and assess low lower load levels with questionable models and uncertainty across a host of potential dispatch options. The question from these studies is not really how much of the generation can be asynchronous (wind and solar), but rather the flip side of that question – how much synchronous generation with inertial capability and other ERS parameters, needs to be on line.
When discussing achievable penetration levels – keep in mind that the overall power system is the major frame of reference. We often hear about the high penetration levels that occurred in Germany – but Germany is not an independent bulk power system. Germany is part of an international interconnected grid that includes vast hydro resources as well as considerable coal. These conventional units provide ERSs for the grid that would be lacking if the entire grid had the same penetration level of renewables as Germany.
Similarly renewable enthusiasts often crow about some “load” being served reliably with 100% renewables. One thing you can count on is that they are not referring to an independent reliable system serving a high proportion of its load by wind and solar power. In some of these cases rotating hydro may make up a large portion of the generation. Often it’s just “accounting” to credit some sub-load with all the renewables from a system. It may be that they are referring only to residential load being served by renewables when that load is dwarfed by the commercial and industrial. Or as noted above it may be that the system is just a small part of a larger system such that the overall penetration level is low. It is often all of the above but it could also be an accident waiting to happen. Despite headlines, no large regional “system” today can reliably operate with extreme levels of asynchronous renewable penetration.
What is Happening Today?
The performance characteristics of major grids are decreasing. As noted it’s hard to measure and quantify the results but there are indicators. Frequency response is an indicator of how well a system can recover from a disturbance. The chart below, available in this NERC report, shows how the frequency response to a 2,750 MW generation trip in Texas has decreased and is projected to decrease as ERCOT continues to increase the penetration of renewable resources.
The figure below from the NERC Essential Reliability Services Task Force concept paper illustrates future potential gaps associated with reliability.
Are there alternatives to the bulk grid?
The bulk grid provides significant benefits. Bulk grids connect sub-systems making them stronger and more stable. They allow backup service which enables electric service during maintenance periods and emergencies. Additionally bulk grids support markets, arbitrage and the efficient use of resources.
Micro-grids can have high levels of reliability but replacing the bulk grid system with micro-grids would be extremely costly and thus infeasible for most applications. Microgrids may provide acceptable benefits for dense areas with a high availability of resources, but much of the population could not be reliability and economically served from microgrids at this time. As noted in this posting microgrids likely will not facilitate the expansion of solar and wind.
Solar and wind raise reliability concerns because: they employ significant new technologies, they do not as readily provide ERS’s, they impose modelling challenges, the transition is expected to be rapid and they operate intermittently. Each factor by itself would present a challenge; however in combination the interaction of all the factors greatly complicates the task of ensuring high levels of reliability.
Maybe it’s worth giving up our existing high levels of reliability to achieve other societal goals. But if so the action should be undertaken with the understanding that reliability risks will be increasing, not in denial of such risks. As we work towards accommodating greater penetration of renewables, there is not a single answer for appropriate penetration levels. Risks will vary based on the characteristics of the individual integrated power systems. Until we have more experience with the newer technologies as they develop and reach maturity, estimating the “safe” level of penetration presents challenges and any definitive statements should be suspect.
The bulk grid has been able to adapt to new technology and will continue to do so. The major concern is with the rapid forced expansion of solar and wind technology. The proposed changes are unprecedented in terms of both scope and speed. If the growth of these resources were driven by the economics and their demonstrated performance characteristics, the bulk system would better adapt and maintain traditional reliability levels while these resources were gradually integrated over time within the system. But with unprecedented change the options are to increase transmission system costs greatly, limit operational flexibility or see reliability degrade. The most likely scenario is that we will see a combination of increased transmission costs, limited operational flexibility and degraded system reliability.
JC note: As with all guest posts, keep your comments civil and relevant.