Wind production exceeded all forecast metrics, solar production exceeded short-term forecasts but not operational forecasts.
TL;DR: On 2/15 both solar and wind exceeded the short term forecasts which do not account for local operating conditions. The operational wind forecast was about 60% low for the whole day, seemingly due to a conservative application of the icing derate built into the forecast. The operational solar forecast was about 20% high, because ERCOT does not forecast solar for when the grid is under extreme/contingency conditions.
What does "forecast" mean in this context? Obviously solar and wind are weather-dependent resources, which is the most significant factor in the forecast amounts. However, the operational forecast also accounts for technical problems and unexpected shutdowns -- things beyond just wind and sunshine availability (in the short term forecast). Due to it's relatively low capacity statewide, solar forecasts are not included in contingency planning, so was not updated during the event. For wind, since the storm was already factored into the forecast, the fact that it exceeded this forecast indicates that it experienced less technical challenges than expected. Ultimately this says at least as much about the forecast methodology as it does about the capabilities of wind turbines in Texas.
What caused the blackouts? This is still being investigated in detail, but the most significant factors appear to be that load and outages were both under-estimated. Wind out-performed contingency expectations. Outage estimates do not include solar.
ERCOT, the grid operator covering most of Texas, provides a number of detailed generation reports for all resources on the Texas grid, including hourly values for forecast and actual production from wind and solar resources. The data is updated every hour with data for the previous 48 hours. The files give raw data, which I've plotted and explained below:
Wind

Metric |
Total value (MWh) |
Actual vs forecast |
Actual |
73,396 |
- |
COP HSL |
46,408 |
158% |
STWPF |
48,434 |
152% |
Solar

Metric |
Total value (MWh) |
Actual vs forecast |
Actual |
20,134 |
- |
COP HSL |
25,385 |
79% |
STWPF |
19,781 |
102% |
Forecasting definitions and methodology
From the ERCOT glossary with some other sources noted.
- COP HSL: current operating plan high sustained limit. The "maximum sustained energy production capability of the resource." The COP HSL is a seven-day forecast, updated by 2:30pm every day. By definition, the HSL is equal to or lower than the short-term forecasts. It accounts for factors specific to each resource, so would exclude any resource that is offline for repair, maintenance, etc. This is what ERCOT plans for because it accounts for things about each resource that ERCOT would rely on the operator to inform them of.
- STWPF: short-term wind power forecast. This training presentation from ERCOT provides more detail: "Statistically 'most probable' forecast of production potential for each Wind Generation Resource." By definition, there is a 50% chance that the actual production will exceed this value.
- STPPF: short-term photovoltaic power forecast. Like the short-term wind forecast, the amount of power that is 'likely' to be produced by PV resources on the grid -- by definition there is a 50% chance that actual production exceeds this value.
The forecasting process documentation indicates that the short term forecasts are generated internally by ERCOT and provided to the QSEs (qualified scheduling entities -- whoever operates and gets paid for each generation resource on the grid) who use it to generate the "operational forecast" -- the COP HSL.
The wind forecast process includes two different derating factors for extreme weather, which specifically accounts for the probability that turbines will be off-line due to icing. The solar forecast accounts for weather, but does not appear to account for snow accumulation on panels.
Commentary
Wind. Without additional commentary from ERCOT, it would appear that ERCOT forecasters and wind operators were somewhat conservative in their application of the extreme weather derates when they updated their output forecasts, and the actual production ended up greatly exceeding both forecast metrics. There is no comprehensive comparison of actual to forecast values that I could find, but spot-checking a few other days indicates that actual values tend to be within +15% or so of the forecast, which is what we could expect given the definition of the forecast.
Solar. In this case, it looks like ERCOT accounted for snow accumulation in derating their forecast, which came in slightly less than the actual production. However (and I'm speculating here) since the forecast procedure does not appear to account for snow accumulation, the PV operators did not adequately derate their forecasts. This would appear to indicate a shortcoming in the process, as the operating plans, by definition, should not exceed the short term forecast.
So what caused the rolling blackouts?
Obviously there will be a detailed postmortem of this, but we can say that wind and solar were certainly not the most significant factor. The blue line on this real time plot shows available standard capacity (excluding the operating reserves which are dispatched to respond to acute failures) minus actual load -- any time that blue line goes below zero, there are problems. Here's the plot for February 15, showing the value at the low point at 5:13pm:

ERCOT forecasts load and non-dispatchable generation (wind and solar) on a number of timescales to predict how much energy will be needed on a day-ahead and hour-ahead basis, and then sets prices using the operating reserve demand curve. This curve dictates the price per MWh that is paid on the market, and increases as capacity becomes constrained. The maximum is fixed at $9,000 per MWh, which is considered the VOLL -- value of lost load. At some point generating power is too expensive, and blackouts are cheaper. ERCOT has decided that $9,000 is the price at which this occurs. The intent of the system is to incentivize more capacity to become available as that price goes higher, to ensure that blue line stays at a sufficient height above zero.
On February 15, the prices hit $9,000 throughout the day, and yet rolling blackouts were required to keep the system from collapsing. This means that one of three things went wrong:
1. The load forecast was too low, meaning more capacity was needed than expected?
Energy consulting firm ICF did an analysis of ERCOT's various long term forecasts for the winter and produced this table:
Capacity, GW |
Expected Forecast |
Extreme/Contingency Forecast |
Actual Conditions (8am 2/15) |
Peak Load |
57.7 |
67.2 |
74.5 |
Resource Outages |
8.6 |
14.0 |
26.6 |
Wind Output |
7.1 |
1.8 |
4.5 |
Solar Output |
0.3 |
[not forecast] |
0.0 |
Total Generating Capacity |
73.1 |
68.6 |
53.4 |
Remaining Reserve Capacity |
16.2 |
1.4 |
-21.1 |
Operational Conclusion |
Normal operations |
Emergency measures |
Widespread outages |
This shows that the load under extreme conditions was underestimated by 7.3 GW, or about 13% of normal load.
Answer: Yes
2. The solar and wind forecast was too low, meaning less capacity was available than expected?
As discussed above, the wind forecast accounted for the storm, and the solar forecast appeared to correct itself partway through the day. These forecasts ended up being overly conservative -- the system expected to have less wind and solar available than were actually present.
Answer: No
3. The VOLL is too low, and generators didn't have sufficient incentive to stay on-line?
Obviously, generators wouldn't have waited until February 14 to see the price, and then make infrastructure changes so they can produce power the following day despite the cold. The market design is supposed to entice them with the allure of high prices to be prepared well in advance, spending some money up-front so they can be prepared to take advantage of price spikes. Given that there were no outages in far colder parts of the country, it's clear that the technology needed to stay online during such storms is available and cost-effective. So if the VOLL had been higher, perhaps they would have planned ahead and done this? Especially since something similar had happened in 2011?
Answer: It's complicated