Guest Op-Ed: Grid-Scale Gaslighting

Over the last few years I have noticed a disturbing trend in discussions of tight grid conditions, generation shortfall near-misses, and general grid reliability. The discussions often take many forms and the target of the discussion may vary, but in general the nature of the phenomena at hand is: whenever a grid nears disaster due to resource adequacy, or is forecasted to approach disaster, a plethora of “experts” emerge declaring one of two conclusions. If variable renewable energy sources (VREs), which are wind and solar, are strongly producing, then these VRE sources “saved” the grid. If VRE output is low, it is concluded that thermal sources, such as natural gas, coal, and nuclear, are not dependable. These hot takes are deeply intellectually dishonest, they are in fact, grid-scale gaslighting.

The purveyors very well know, or at least should know given their credentials, how grid planners assess whether or not there will be enough electricity generation to meet demand under various scenarios.

I will use the Electric Reliability Council of Texas as an example because ERCOT has had numerous close calls this summer with a Level 2 Energy Emergency declared in the evening hours of Sep 6th, 2023.

Several times a year ERCOT releases a “Seasonal Assessment of Resource Adequacy (SARA). These are fairly simple models wherein ERCOT adds up the expected output for all generators during that season, calculates a forecasted peak load, and subtracts the forecasted peak load from the expected generator output. The result is the “reserve margin” and this value should be greater than 0, and really it should be closer to 10% of peak load. Now grid planners recognize that things do not always go as planned, equipment under heavy stress breaks down, freak weather events can spike demand far above what would normally be expected, and sometimes those same events have significant impacts on VRE generation and can even impact thermal generation. These models and the reports derived from them are available from ERCOT as excel spreadsheets and pdf documents respectively here.

These documents detail exactly what the “expected” output of different generation sources are under various scenarios. However, I rarely see these materials referenced during these discussions.

For example on Aug 29th, 2023 at 7:45pm, gridstatus.io and ERCOT were reporting >11,000MW of non-VRE outages while tight grid conditions existed. At this time total thermal generator output—gas, coal, nuclear—was 60.4GW, with 4.7GW of reserves. That means roughly 65.1GW of thermal generation was either generating or available to call upon. If we add up what ERCOT expects from the SARA we find that thermal resources + available mothballed capacity + capacity from private networks + planned thermal resources works out to 69.4GW.

No piece of machinery is perfect; things break down. ERCOT knows this and so they expect roughly 5GW of thermal capacity to be offline due to planned or unplanned outages. Subtracting the available generation on Aug 29th from the ERCOT expected capacity of 69.4GW we can conclude that there was roughly 4.3GW less thermal generation than ERCOT’s perfect scenario, and about 0.7GW more than the expected contribution of thermal generation to peak load.

Like what you’re reading? Click the button to get Grid Brief right in your inbox!

So, where does the 11GW come from? Well, that number is based off of the installed nameplate capacity of the plants, which for a variety of reasons, is typically larger than the capacity expected from those same generators during the summer. For thermal generation, this number is 73.2GW in ERCOT.

Disclaimer: the numbers do not perfectly align as we do not have knowledge of what the breakdown of the 4.7GW of reserves is nor what the breakdown of the 11,000MW of non-VRE outages (note the outages are not thermal generation, all VRE outages get lumped together and all non-VRE outages get lumped together).

One might also ask, “Well, the non-VRE outages were 11GW, but what about the VRE outages?”

Well at that same time, the non-weather related VRE outages in ERCOT totaled nearly 7.5GW… Total VRE output was 6.4GW - 7.4GW out of a SARA expected 22.7GW / 53.4GW installed. Note: in the baseline scenario ERCOT assumes typical wind and solar output. When they compute their scenario for low wind and solar, which still expects a combined 11GW of output, the result is a 2.4GW shortfall. So we have VREs producing roughly 63% of their expected output, while thermal generators are producing 101% of their SARA expected output.

Bottom line: the problem is not that thermal generation is not dependable, it is extra-dependable, but rather the simple fact that ERCOT does not have enough thermal generation on their grid. This problem has been building for years now and it is largely being swept under the rug in the SARAs by assigning expected capacities for wind and solar that are far too high. This monster reared its ugly head back in 2021 during Winter Storm Uri when Wind was delivering <1GW at a time that ERCOT was counting on at least 7GW to keep the lights on under those conditions. At the time ERCOT did not include a low wind adjustment for the extreme weather scenario based on the 2011 storm because it would show that there was a potentially significant generation shortfall should such a storm occur.

Well, it did and there was a significant generation shortfall. ERCOT made an attempt to rectify this deadly accounting error by including low wind and low solar adjustments in subsequent SARAs. As expected, they show shortfalls. ERCOT still maintains that the low wind and low solar conditions are rare and therefore grid resource adequacy is satisfied. I’m not so sure that a condition that occurs multiple times a week, has resulted in 16 weather watches and conservation calls, plus one energy emergency alert level 2 during the summer of 2023 can be considered rare.

In order to drive home this point I present the horrifying frequency chart from the evening of Sep 6, 2023.

Here we see the grid frequency steadily decrease for nearly 30 minutes from 60Hz to just shy of 59.8Hz. At 59.3Hz ERCOT initiates a 5% “Load Relief” as in they turn off 5% of the load on the grid. At 58.9Hz ERCOT drops an additional 10% of the system load, and finally at 58.5Hz it drops a final 10% for a total of 25% load shed. If conditions stay below 59.4Hz but above 58.4Hz for more than 9 minutes, then roughly 25% of the power plants on the grid start tripping offline in order to protect themselves. This period drops to 30 seconds below 58.4Hz, 2 seconds below 58Hz, and is instant below 57.5Hz.

It is not entirely uncommon for the grid frequency to rapidly drop and then recover when a large (~1GW) power plant trips offline for whatever reason. These are common enough that common practice is to ensure that the grid always has at least 2x the power output of the largest single power plant on the grid in spinning reserves. If/when that plant trips offline the frequency will drop, the mechanical or digital control systems on the spinning reserves (and really all other thermal generation) will sense the drop, and in concert increase power to bring the system back to 60Hz. This occurs automatically, across the entire grid, in a fraction of a second.

What happened on Sep 6th was VRE generation dropping from 11.9GW to 6.9GW over the course of 30m, the equivalent of 5x of those large power plants tripping offline in quick succession. The frequency response show us something truly terrifying: for that entire 30 minute period, demand exceeded supply, and the rotational energy of hundreds of tons of metal slowing down from 3600rpm to 3588rpm was making up the difference. The grid was off balance for half an hour. During that time supply continuously fell short of demand. Had the difference been greater, had it taken longer to get additional supply up, had the bitcoin miners kept mining, load shedding would have been inevitable.

As we saw in Feb, 2021, the consequences of a grid with inadequate supply are deadly serious. The grid is not something that we can fudge the numbers on in an attempt to meet some arbitrary climate goals. The grid is our primary defense against a climate that, even before climate change, can easily snuff out human lives. Let us treat it with the respect that it deserves.

Joshua Payne is an MIT trained nuclear engineer, pilot, high altitude mountaineer, and amateur machinist. After spending 8 years doing computer science and complex systems modeling at Los Alamos National Laboratory, Joshua's primary focus has shifted toward energy system resiliency and reliability.