ERCOT February 2021

It has been a month and our industry is still coming to terms with the impacts of the events in Texas this February. As more date becomes available we are able to understand better what occurred. ERCOT released a list of outages during the events. https://www.ercot.com/content/wcm/lists/226521/ERCOT_Letter_Re_Feb_2021_Generator_Outages.pdf.

With a little effort it is possible to convert this data to an hour by hour outage rate by fuel capacity.

It is clear that Wind and Gas represented a majority of the outage in the ERCOT dataset. We can use EIA Form 860 data to understand the asset mix on the ground in Texas at the end of 2020. https://www.eia.gov/electricity/data/eia860m/

The November 2020 EIA data was used to generate this figure of Texas capacity by technology by install date.

It is possible then to look at these two different datasets to try and understand a little better what occurred in ERCOT. In the following graph the gas generation capacity is normalized as a percent of the 2020 total by year and the total amount of gas capacity outage (from the ERCOT data) is also normalized by year of install of the units. Orange dots are the normalized outage data, blue are normalized capacity data.

If age of the unit was not a factor in terms of capacity loss these two trends should track each other pretty closely. We see in the graph a bit of a difference, but it doesn't necessarily look too significant. This is where human brains can fool us. So on the next graph this is just a straight up ratio of the outage normalized data to the install normalized data. If age was not a factor we should expect a random distribution around the 100% or 1.00 number.

Which is exactly what we see for the newer plants. But for the older plants it is clear they disproportionately had more outages versus installed capacity. So older plants failed during the ERCOT event at a noticeable higher rate than newer gas plants.

Now for the purists out there they would note that there is a bit of a cheat in the above graph due to using cumulative numbers, and they are right!

So for legitimacy this is a straight ratio of the percent outage per install year over percent capacity per install year (not cumulative). The same conclusion (older assets failed at higher rates) is reached in this graph - it was normalized around a 0% value by subtracting 1.

It is also possible through data mining to figure out previous year capacity factors on the gas plants that resulted in outages. Not surprisingly the older gas plants that failed at a higher rate had lower capacity factors.

So rarely run, older gas units failed at appreciably higher rates on average than newer assets.

As we look to build resilience and improve our planning going forward this type of analysis can be relevant.

When the question of improving resilience is brought forward, how is it justified for a rarely run power plant that is 40+ years old? How is resilience improvements paid for?

As our grid transforms and dispatchable assets are increasingly called upon only on extreme days, how can we be sure that these plants can be ready and available?

These are not simple questions. There are not simple answers.

At EPRI we are working collaboratively to understand these issues and develop solutions.