Tilting at windmills?
Inside the data from the Winter Storm Uri power crisis in Texas
The blame game started almost immediately. With approximately four million Texas households out of power on Monday, February 15, 2021 (mine included), everyone was looking for an explanation. National and state politicians blamed windmills; others pointed fingers at renewable energy sources more broadly. Market analysts and media were quick to point out problems with fossil fuel-based energy production and the Texas power grid.
Sitting home that week with my family, eating pancakes by candlelight (we still had gas on our stovetop), I wondered what could have caused such a massive failure of infrastructure. Chatter online was hard to follow — it seemed as though everyone had a cherry-picked data point to support their point of view. It was clear that the situation was complex and nuanced, but I wanted to see inside the data for myself. The main question I sought to answer was:
How did our sources of electricity perform compared to historical tendencies and each other during the crisis?
Using publicly available data from the U.S. Energy Information Administration (EIA) and the National Oceanic and Atmospheric Administration (NOAA), I assembled a data set to explore Texas electricity generation in relation to the weather over nine days from February 11 to February 19, 2021. In that period, the statewide daily average temperature fell from 36 degrees on February 11 to a low of 14 degrees on February 15. At the same time, statewide electricity production fell from 1,450 GW on February 14 to 1,064 GW on February 17, even as electric demand reached unprecedented heights.
Data Sources
Daily electricity generation and demand data for 967 days from July 2, 2018 to February 22, 2021 were obtained from the U.S. EIA’s Open Data API U.S. Electric System Operating Data and Hourly Electric Grid Monitor datasets. U.S. Electric System Operating Data included hourly electricity generation figures in Texas from seven sources: natural gas, coal, wind, nuclear, solar, hydro, and other; “other” sources may include electricity generated from biomass, fuel cells, geothermal heat, waste, or wood. Hourly Electric Grid Monitor data included electric demand forecasts for days of the crisis in which demand exceeded production, which were used to determine what electricity consumption would have been if demand could have been met during the crisis.
Daily weather and atmospheric conditions data were obtained for the same 967 days from the NOAA’s Local Climatological Data subset of the Climate Data Online Data Tools. Sixteen Texas weather stations in population centers were selected for geographic balance, including stations in: Abilene, Amarillo, Austin, Beaumont, Brownsville, Corpus Christi, Dallas, El Paso, Houston, Laredo, Lubbock, Midland, Paris, San Antonio, Tyler, and Wichita Falls. Four data points were obtained from each station: average daily temperature, average daily wind speed, heating degree days, and cooling degree days. Heating and cooling degree days defined here.
Non-crisis operational benchmarks
Data from EIA and NOAA were combined into a single data set and categorized three ways for benchmarking: the nine days of the February 2021 power crisis, statewide cold weather days (days in which average daily temperature did not exceed 65 degrees Fahrenheit at any of the 16 weather stations in the dataset), and high electricity demand days (days with the highest 12% of electricity production).
The larger data set contained four subsets: crisis days (n = 9), non-crisis days (n = 955), cold weather days (n = 172), and high demand days (n = 113). For each subset I calculated mean net daily electricity generation from all sources, mean net daily electricity generation from each source individually, mean percentage that each source contributed towards daily electricity generation from all sources, the maximum observed single-day value for net daily electricity generation from all sources, the maximum observed single-day value for net daily electricity generation from each source individually, and the day-to-day variance for each source on non-crisis days. Table 1 summarizes these results.
The All Days, High Demand, and Cold Weather means indicate a range of expected production for each energy source given a variety of relevant circumstances. The maximum observed value of each source represents maximal production capacity of that source under optimal circumstances. All Days Variance is a rough indicator of the consistency and reliability of each source, with lower percentages denoting sources that vary little in terms of day-to-day production (more consistent and reliable) and higher percentages denoting sources with greater day-to-day variance (less consistent and reliable). Thus, nuclear power production varies the least over time while hydro and solar vary the most.
Crisis electricity production
Table 2 displays electricity generation figures during the nine days of the power crisis. It shows that there was an electrical generation deficit in seven out of those days, with a greater than 10% deficit on five days and a greater than 25% deficit on three (2/15, 2/16, and 2/17, the three coldest days of the period). The five high-deficit days from February 15 to February 19 will be examined most closely in this study.
Figure 1 below shows electricity demand, electricity demand forecast, and net electricity generation for the nine days of the power crisis. Note that net electricity generation fell precipitously late in the evening on February 14 and continued to fall until around midday February 15.
Figure 2 shows electricity generation by source over a longer time period. A closer look shows that wind power began falling late in the day on February 14, followed a few hours later by sharp declines in all other sources besides solar (because it was nighttime). Wind power fell the most dramatically and dropped to extremely low levels on February 15 and 16 before beginning its recovery on February 17–18. Statewide average temperatures rose to a much more typical 40 degrees on February 20 and 51 degrees on February 21, essentially ending the crisis for most Texans.
EIA data includes an overall electricity demand forecast for each day, but does not include a forecast for each source. Having an expected production figure for each source would enable us to see more clearly how each source performed compared to its observed historical production capacity and actual production during the crisis. To calculate Expected Production for each source, then, I multiplied the overall EIA Electricity Demand Forecast for each day by the mean percentage that source supplied during historical high demand days. Table 3 includes the resultant figures.
Actual — Expected Production in Table 3 presents the absolute disparity between what each source should have produced compared to what it actually produced. However, since absolute figures in this case are not a fair method of comparison across source types, I have included % of Expected Production to better evaluate each.
Grading Natural Gas: B-
On the plus side, natural gas provided about two-thirds of Texas’ electricity at the height of the winter storm power crisis. No single day or three-day period during the 2.6-year range examined in this study saw a higher percentage of electricity come from any single source. Only hydro and solar power produced higher percentages of their expected output and no sources came close to natural gas’s absolute contribution to the grid during the crisis period. It was indispensable for maintaining what remained of Texas’ electric production at the height of the crisis.
On the other hand, natural gas electricity production was well below maximum capacity (855 GW on August 23, 2019) in the five worst days of the crisis and below average production levels for high demand days (719 GW) on three out of five of those days. Electricity from natural gas peaked at 893 GW on 2/13 and 890 GW on 2/14 (the two highest totals observed for the entire period examined in this study) before falling to 773 GW on 2/15 and 694 GW on 2/16, a decline of 22% in four days. Given the state’s dependence on and observed capacity for generating electricity from natural gas, this decline was likely decisive in triggering the crisis.
Grading Coal: D
Good news is hard to come by for coal in this case. From a high of 262 GW on February 14, coal electricity production fell 33% in three days and only slightly recovered thereafter. Where Texas counts on 261 GW of power from coal on high demand days, it never came close to that figure during the five worst days of the crisis. Neither did it come close to its dataset maximum of 349 GW/day. Thus, coal did not meet high demand day expectations and stayed far below its observed historical capacity. Hard to see this as anything but a failure.
Grading Wind: F
There’s really no sugarcoating wind’s dismal performance during the crisis. Wind typically provides about 21% of the state’s electricity needs, but only provided around 8–11% during the worst days of the crisis. Production plummeted 65% from a high of 174 GW on February 14 to only 61 GW on February 17. Overall, wind generated only 42% of its expected production and had the largest absolute five-day deficit between expected and actual production (719 GW) of all sources. To be fair, the statewide average wind speed on most of those five days was below historical averages, but that explanation is insufficient to explain such a sizable drop. Wind is a high variance electricity source under normal circumstances, but this event was outside wind’s normal range of variance.
Grading Nuclear: C+
Nuclear power is easily the most consistent energy source in Texas, but even it fell off in the recent crisis. From a high of 123 GW on February 14, it fell 26% to 91 GW on February 16. It was back to full capacity by February 19, however, which made it the first major power source to fully recover. It only provided 77% of its expected production in the five worst days of the crisis, but its contribution to overall production was roughly consistent with historical high demand day averages. A mixed performance.
Grading Solar: A
The other side of the renewables coin has to be viewed as a success story in this case. Solar averaged 21 GW/day during the five worst days of the power crisis despite having fewer daylight hours to produce them (solar has historically averaged only 13 GW/day during statewide cold weather days). Some of that production can be chalked up to solar’s enormous recent growth in Texas, but the resilience of solar is difficult to deny. (Anecdotally, I talked to one of my neighbors with solar panels recently and he told me his panels continued producing normally even when they were half covered in ice and snow.)
Solar electricity production during the height of the crisis was above average for high demand days, its contribution to overall electricity production was higher than normal, and it met 95% of expected production (second highest of any source). Solar remains a highly variant electricity source, but it definitely helped ease the crisis in this case. The only tragedy is that we didn’t have more of it to make up for shortfalls almost everywhere else.
Grading Hydro: A+
I don’t know how hydro electricity produced more power during the crisis considering the state was an ice rink, but it did. Hydro generated 155% of its expected electrical production — the only source to exceed expected production — by staying strong and consistent throughout the worst five days of the crisis. Production peaked at 3.9 GW on February 15 and fell 39% over the next two days before stabilizing around 2.4 GW. It just wasn’t enough, however. The state’s hydroelectric capacity is very small and really only reaches upper limits during periods of flood.
Final Thoughts
There are no scapegoats and few heroes here. It’s clear after reviewing these data that there’s plenty of blame to go around for the recent winter storm power crisis. Fossil fuels and renewables both come off pretty mixed, and I haven’t even touched the issues of weatherization, grid management, or Texas’s grid being separate from the rest of the country. I will leave those questions to the energy experts, but I hope my analysis provides an objective viewpoint that can inform smarter energy policy. As much as I enjoyed it, I would rather not eat pancakes in the dark again.
