Five lessons from the February freeze in Texas’s power market

Exceptionally cold weather conditions in mid-February throughout Texas triggered a large number of outages across the thermal and wind fleets, forcing ERCOT, the operator of the electrical grid, to resort to rolling blackouts throughout the state. Data that has emerged since then points to few measures that, in hindsight, could have reduced the severity of the consequences. Here are some of the main learnings from the extreme weather event.

1. 1. Wind availability was not adequately accounted for in scenarios or real-time view

Substantial wind outages already present on Feb. 10 should have given a few days' warning and shifted the focus on the availability of thermal plants. More than 12 GW of wind capacity, over 40% of the installed base, were offline from Feb. 11 through Feb. 15—when thermal outages eventually surged.

And the surge in unavailable capacity—at plants that included coal, gas and nuclear—went as high as 25.3 GW in the morning of Feb. 16, a level much higher than the 14.0 GW that ERCOT assumed in their most extreme scenario for the winter.

Interestingly, in that scenario, ERCOT kept its wind generation assumptions unchanged from normal (while de-rating all thermal capacity), even if in very cold weather potential outages might be expected. A scenario with extremely low wind generation, while identified, was not part of this view.

2. 2. Gas generation resource was overestimated even under normal conditions

When comparing aggregated data published by ERCOT right after the events and the latest information on plant-by-plant outages another element also stands out.

In contrast to wind (and solar, coal and nuclear, not shown in the chart), aggregated data for gas outages (blue solid line in the chart) comes with a gap of about 7 aGW over plant-by-plant data (dashed line).

Note that, following a Public Utility Commission Order, this data is based on the disclosure by all plants affected by the outages, also those that initially did not agree to make their information public before the standard 60-day confidential period. Any difference between the solid and dashed profiles is therefore very likely due to long-term outages.

While quite large, this difference is very close to the thermal capacity technically offline that ERCOT assumed for the winter in their Seasonal Assessment of Resource Adequacy. Note however that if the shortfall applied also to earlier periods (the data in this case only starts on Feb. 14) it meant that ERCOT had overestimated their available resources even under (closer to) normal weather conditions, as any unexpected outage would have come on top of such baseline.

3. 3. Some gas supply issues were avoidable with existing safeguards

Regardless, the sharp increase in gas plant outages on Feb. 15 is probably the most striking feature of the chart above. Initial data didn't specify the nature of the outages, but information published by ERCOT on April 6 confirmed that fuel limitations played a large role.

In fact, in the early morning of Feb. 16 at the time of maximum generators' unavailability, fuel limitations caused outages and derates affecting 6.1 GW of capacity. Being about fuel limitations, this necessarily applies to thermal plants, and most likely to gas ones given that nuclear fuel is not affected by weather conditions and coal should also be largely shielded unless it gets wet or frozen. On this point, there have been no reports of widespread disruptions of coal supply.

As a result, assuming the above fuel outages were mostly at gas plants, their share of total outages affecting the gas fleet comes to around 30%, which is substantial. This result excludes the around 7 aGW previously identified as long-term outages, but when including those the share is still high at 25%.

The comparison between the timeline of gas plant capacity going offline and the percentage of Texas gas production going offline also supports this view. Using S&P Global Platts Analytics' interstate production sample as a proxy, the chart below shows how gas production posted the largest day-on-day drop on Feb. 15, continued to decline in the following days and took longer to recover than power capacity.

While some of this behavior is certainly a feedback loop on gas production linked to users remaining offline or operating at reduced capacity, the forced shutdown of gas wells due to lack of power supply was also a contributing factor.

This could have been easily avoided. Several gas facilities reportedly failed to apply for outage exemptions before the storm, a once-a-year process where the owners fill out a two-page form and submit it to the local utility company. By doing so, facilities are classified as critical infrastructure and prevented from load shedding. The measure was already present among the recommendations contained in the FERC/NERC report on the February 2011 blackout, which included a number of winterization options around power and gas infrastructure. In contrast to many others, this measure would have come at no cost.

Instead, now the focus is mainly on the other recommended winterization measures, for wind facilities as much as thermal ones.

The options for wind turbines range from ultra-sonic de-icing to hot air injection, to the addition of heaters on the surface of blades. For gas plants, viable options include firm gas supply contracts and dual fuel capability.

From the profile of outage durations for gas plants we can see that the most frequent period was less than half a day, and retrofitting them with oil tanks holding fuel for only a day or two might have already made the difference between a tight system and rolling blackouts.

Cost-wise, all the above measures reportedly fall between 5-15% of overnight costs of new wind and gas plants.

4. 4. Future winterization measures will be driven by regulation and economics

While decisions on winterization measures might be driven by the outcome of a number of bills introduced in the Texas Legislature and potentially by an overarching decision by Congress, the near term consequences of the mid-February events are more concrete and include bankruptcies, large financial losses and rating downgrades impacting utilities and individual operators.

At the same time, the recovery of losses could also be hampered by the Peaker Net Margin (PNM) rule, which currently prevents the maximum hourly price from reaching the $9,000/MWh (commonly known as HCAP) seen in previous summers, and instead limits it to $2,000/MWh or 50 times the natural gas price index (commonly known as LCAP). A proposal by the Public Utility Commission further aims to remove the 50-times-natural gas price index element from the LCAP, leaving only the $2,000/MWh one.

As the reduction to $2,000/MWh or, depending on the outcome of the proposal, 50 times the natural gas price index, directly impacts the maximum marginal prices to be expected during the hottest days of the summer, winterization decisions might be encouraged by the operators' need to capture the high margins of the summer. In fact, as the situation of this year demonstrates, equipment failure at times of low temperatures and lower margins at times of high ones can mean a major double blow to the balance sheets of operators and utilities.

The prices for the summer are still to be determined, but the comparison between market prices – whose premium implies a level for the cap above $2,000/MWh – and our base case – which assumes the price cap at $2,000/MWh—can provide a guide for the margins being lost in the latter case. As Platts Analytics' latest Short-Term Forecast report shows, the difference for the reference hub of ERCOT-North is more than $10/MWh for the on-peak prices of July and approaches $40/MWh for those in August.

5. 5. Climate change impact is likely but wind still well placed to provide winter generation

Longer term, the trade-off between costs and benefits of winterization measures should be considered in the context of the record-low temperatures recorded on Feb. 15.

If we take the measurements at Abilene airport as a reference (it's conveniently located in Central Texas), the chart below shows the distribution of daily average temperature since Jan. 1, 2005 over the 3-month period Dec-Feb. This choice is justified by the fact that the coldest day of the winter has historically occurred in one of these three months.

The approximation of the history with a normal distribution shows that the temperature of Feb. 15 was an extremely rare event. In fact, it corresponds to a probability of about 1 in 20,000 days, or more than 200 years given that our analysis is limited to a 3-month period. Using a separate dataset with incomplete daily temperatures since 1975 confirms these results. We must however note that, given the rarity of the event, fitting the data with a normal distribution might not yet capture the shifting frequency of extreme conditions resulting from climate change. Therefore, what on paper might look like a 1 in 200 years event, in practice it might be closer to a 1 in a century or less. Indeed, similar temperatures already occurred in 1983 and 1989.

Still, wind has much to gain from these trends if suitably winterized. In fact, if we take again the measurements at Abilene airport as a reference (not only is it located in central Texas, it is also surrounded by several windfarm clusters), wind speeds were rather strong leading to the coldest days of February. The fact that output dipped below 1 aGW during the period (and daily output was also much lower than expected) was therefore largely due to equipment failure rather than low wind speeds. By contrast, assuming typical plant availability, progressively higher Texas loads in the winter driven by climate change, and higher wind resources in the presence of very cold spells, can offer added bonuses to the business case for wind.