Texas Blackouts–Critical New Data Revealed
FEBRUARY 24, 2021tags: Texas
By Paul Homewood
This is the best analysis I have seen yet regarding the Texas blackouts. Yes Energy are a professional outfit of energy analysts, so have no axe to grind whatsoever.
By now most of the world knows about the cold snap in ERCOT the week of 2/15/21 (if not, check out our blog post here), but do you know the generation side of the story? You may know some of the story from watching the news, like we have – including claims that it was a wind issue or a natural gas supply issue.
History tells us that it will take months to determine the root causes of this event, but what was the data telling us in real-time during the crucial hours and minutes leading up to the event? Traders with access to market data in Yes Energy alongside real-time generation data from Live Power had critical insight into the generation side of this story. What happened and what role did generation play in this historic week in ERCOT?
First, prices in ERCOT were at their market cap of $9,000 per MWh for days on end the week of 2/15/21. This was not price node or load zone specific, this was system-wide in ERCOT for many hours and many days on end. The chart below shows the real-time LMP (with scarcity adders) at Hub North in ERCOT from 2/9/21-2/18/21, but this could have been any price node in ERCOT that week because this cold snap had system-wide implications.
The map below is a snapshot from 2/15/21 at 2 am from Yes Energy’s PowerSignals product that provides a visual representation of the high prices throughout ERCOT and the pervasive transmission constraints. In this visual, all of the red dots are price nodes in ERCOT (all are at or around the market cap of $9,000 per MWh) and the yellow circles & lines are transmission constraints binding in the real-time market. It is worth noting that during this time, as well as throughout the week of 2/15/21, ERCOT was in an Energy Emergency Alert Level 3 (EEA3), which is due to an imbalance of supply & demand, thus rolling blackouts are required.
What happened right before this in the wee hours of 2/15/21? At 1:10 am CT Live Power data gave an early indicator of generation decay. This gave Live Power customers 45 minutes to react prior to 1:55 am CT when ERCOT frequency, which is the heartbeat of the grid (60 Hz is a normal beat) suddenly dipped to 59.38, which is significant. Had it dipped lower than that there could have been cascading blackouts across the entire state, which would have taken days or weeks to bounce back from. What led to this dip in frequency and what critical insight did Live Power customers have?
At the same time as ERCOT frequency dipped, there was a 6 GW loss of generation across all of Live Power’s monitored plants in ERCOT. What does that mean?
It means that ERCOT needed to shed load, and shed it fast because balancing load and supply on the grid is absolutely critical for the security and reliability of the grid. As seen below, shortly after the loss of generation, ERCOT quickly shed 10 GW of load to regain balance on the grid. This meant far reaching blackouts across the state.
What did Live Power customers know about this loss in generation? Within 60 seconds of this loss in generation, Live Power customers not only knew the relative magnitude of the lost generation, but they also knew the part of the supply stack that was affected, as well as which plants tripped offline. Circled in yellow below are the gas and coal facilities that Live Power monitors that had a sudden loss in generation at 1:55 am, which lead to the dip in frequency and the subsequent shedding of load. This sudden loss of generation was partially due to the widespread natural gas issues that were caused by the historic low temperatures across the state. For example, critical parts of the gas supply chain were freezing (valves, pipes), electric motors could not run due to the power outages, and there were compression losses throughout the system due to the cold temperatures.
Drilling into the Live Power monitored plants that suddenly tripped offline at 1:55 am, the chart below hones in on three plants in particular that contributed to this decay in generation, which in turn led to a decay in frequency. The green line is Live Power generation data for Colorado Bend II, the blue line is the Live Power generation data for Sandy Creek, and the pink line is Live Power generation data for Wolf Hollow. All of this generation data comes from Live Power and is piped into the Yes Energy platform every 60 seconds. Yes Energy and Live Power customers were able to layer this generation data in context with market data like frequency, which gave them critical insight during this historic week in ERCOT.
There’s quite a lot to take in, but these seem to be the salient points:
1) Market prices had already been flagging up danger signals days before the crash.
2) “At 1:10 am CT Live Power data gave an early indicator of generation decay. This gave Live Power customers 45 minutes to react prior to 1:55 am CT when ERCOT frequency, which is the heartbeat of the grid (60 Hz is a normal beat) suddenly dipped to 59.38, which is significant.”
In other words, it was clear at 1.10 am that something was happening.
This is a critical piece of information, that the drop in frequency and associated plant shutdowns occurred precisely at 1.55 am. Previously we have only been told that it happened between 1 am and 2 am.
4) Frequencies had already to started to decline before 1.10 am, which had obviously flagged up the warning.
5) Up until that time however, the three highlighted gas power plants were still operating normally. Indeed all three were varying their outputs, up and down, presumably reacting to demand.
6) Circled in yellow below are the gas and coal facilities that Live Power monitors that had a sudden loss in generation at 1:55 am, which lead to the dip in frequency and the subsequent shedding of load (6GW)
The analysis then goes on to claim:
This sudden loss of generation was partially due to the widespread natural gas issues that were caused by the historic low temperatures across the state.
It is absurd to claim that all of these plants packed up at 1.55 am, because they had all suddenly frozen up at precisely the same time.
Indeed it is obvious from the map that they are scattered across the state. While Wolf Hollow, for instance, is up north near Ft Worth, Colorado Bend is near the coast near Houston, where weather conditions would have been much different.
The media has been full of reports of how gas plants shut down because of freezing pipes, valves, shortage of gas and so on. But, and I might be wrong, I have yet to see any hard evidence of this at this particular time. If it had happened, surely it would have been easy to prove that XYZ power plant shut at such and such a time, because of frozen pipes or whatever.
From this new information, we can now infer:
1) The sudden drop in wind power of 4 GW on Sunday evening left the already highly stressed grid in a critical state. It was only the ramping up of gas power that saved a calamitous total collapse there and then – one that would have been much worse.
2) Something happened around 1.00 am to push the whole thing over the edge, as the 1.10 am warning confirms.
3) Whatever this “something” was (and it may have been supply or demand orientated), it caused the tripping out of 6 GW of generating capacity (mainly gas) at precisely 1.55 am.
4) At that stage there was no easy way back, and the widespread blackouts were the inevitable result.
To sum up, the loss of gas generation resulted from tripping out, and not because of freezing up.
Gas v Wind
There has been much debate about whether wind or gas power were most to blame, but it has been a meaningless, not to say misleading, one, based on a misinterpretation of the data.
To put the various roles of the two in perspective, let’s first consider this graph:
When wind power largely disappeared on Sunday evening, it was gas power that stepped up to keep the grid working. Clearly the opposite could not have happened.
Secondly, for the rest of the week, when demand was still high because of the cold weather and when supplies were still being restricted, gas generation remained flexible and stable, responding to rises and dips in demand, as well as the ups and downs of wind power.
During the days after the shutdown, gas power fluctuated between 25 and 35 GW. Indeed, it was running at 33 GW immediately after the crash. The total CCGT capacity in Texas is 41 GW, and about 4 GW was offline due to annual maintenance. There is some extra gas engine and single cycle capacity, but this is designed for peaking, not 24/7 running. It appears that it was the latter which helped gas power to peak at 43 GW at 11.00 pm. It may even have been problems at some of this peaking capacity which triggered the subsequent tripping out.
So, clearly the gas power grid was actually working pretty well, all things considered. To consistently run at over 30 GW, when available capacity is 37 GW, does not sound like a failure to me.
In comparison, wind power was effectively absent without leave for the whole of last week.
Finally, let us consider generating capacity in Texas:
Bearing in mind that demand peaked at 68 GW on Sunday evening, Texas only has 64 GW of 24/7 dispatchable power – CCGT, coal and nuclear.
Peakers are fine for operating for an hour or two at times of peak demand, but should not be relied on for continuous operation. (In summer, for instance, demand cycles from around 45 GW at night to 65 GW late afternoon. Peakers are ideal for meeting this predictable surge in demand for an hour or two each day).
ERCOT got itself into this mess because of a naive belief that wind power would always be available in large quantities. This was an accident waiting to happen. And it will happen again if Texas does not get at least another 10 GW of proper dispatchable capacity.