Shortly after midnight on the 15th, 8000 MW of gas power plants shut down because of fuel shortages or freezing equipment and 2000 MW of wind went offline due to low winds, frozen equipment such as substations, or precipitation that caused ice formation on turbine blades . With projected demand significantly exceeding supply, ERCOT ordered firm load shed (blackouts) on Monday the 15th at 1:25am to prevent the collapse of the grid (see Fig. 2 below)
[...] While wind power provides more than 20% of the state’s electricity needs over the course of the year, it is a resource whose availability changes by location, time of day and time of year. As such, grid operators use seasonal expectations for wind’s contribution to grid generating capacity for their reliability planning. The forecast for this period suggested wind production would only provide about 7% of the state’s power demand , with an expected hourly output of 6.1 GW for wind compared to the installed capacity of 28 GW. Knowing this, ERCOT did not plan for much generation from wind during this period.
All major fuel sources underperformed against expectations save for solar. Natural gas was responsible for nearly ⅔ of the total deficit. Gas underperformed by 37% compared to its expected output – more than 18 GW below expectations – and was 21% below the extreme scenario. Coal was 43% below expectations and 28% below the extreme scenario. Wind was 46% below expected but better than the extreme weather scenario. One of Texas’ four nuclear reactors was offline for 2.5 days during the freeze due to a feedwater pump issue , , , which meant that nuclear underperformed by about 20% (see Table 1 below).
|Expected capacity (GW)||Extreme Scenario Capacity (GW)||Actual Average Generation (GW)|
% Deficit [From Expected Capacity]
Deficit Extreme Scenario (GW)
% Deficit [From Extreme Scenario Capacity]
Source: ERCOT data compiled by Blake Shaffer.
The primary culprit for the electricity system failure was problems in electricity production from natural gas.
About 40% of natural gas production was not available during the crisis. Texas’ gas, electricity, and water systems are inter-linked so failures in one of them can lead to cascading effects on the others 
. The natural gas system relies on electricity, and the electrical system relies on gas. Thus, constrained gas limits the ability to generate electricity and constrained electricity limits the ability to supply gas which in turn further limits the ability to generate power in a vicious circle. Power outages in turn can lead to failures in the water supply 
Ravikumar summarized four failures in supply chains that ultimately led to a sharp drop in electricity from gas at the height of the crisis, (1) freezing
at natural gas wells, (2) freezing
of gathering lines, (3) power outages at compressor stations, and (4) equipment malfunctions at power plants 
. Power plants themselves faced their own challenges operating in cold weather, as many power stations in the southern United States are not contained in buildings; doing so reduces the risk of overheating during warm periods but leaves them exposed during the cold 
One reason gas producers lost power is that many of them intentionally sign up for interruptible power contracts as a way to reduce their electricity bills. Furthermore, a significant number of them did not fill out a short form requesting that they be identified as critical infrastructure, compounding the problem. Electricity providers unwittingly shut power off to some gas production and processing facilities, which reduced gas pressure further, thereby forcing more generating capacity offline when fuel was unavailable for gas-fired electricity plants. According to ERCOT, more than 9 GW of outages, about 20% of the total and enough for 1.8 million homes, resulted from insufficient gas supplies reaching power plants 
. A similar sequence of events had occurred in 2011, and post-event regulatory reviews had warned of this problem’s ongoing risk, but little had been done by the gas industry to rectify it.
A number of producers entered the market since 2011 and were unaware of the form they needed to fill out to request this designation. For example, in the midst of the freeze, Oncor, a utility, turned on the power to 150 gas facilities in the Permian Basin after receiving frantic calls that they had been turned off 
. In addition, the shale boom in the Permian mostly occurred since 2011, which exacerbated risks: Permian gas production is highly electrified (and therefore affected by power outages) and liquids-rich (and therefore at risk of freeze-offs).
Another reason less capacity was available is that a number of thermal plants were down for scheduled maintenance to take advantage of Texas’ normally mild winters when demand is low. This schedule is typically interpreted as good planning because it was intended to ensure sufficient reserve in summer time. ERCOT in its seasonal winter maintenance schedule had forecast 4 GW of scheduled maintenance
. Analysts from Wood Mackenzie wrote that the actual amount of plants down for outages in the week leading up to the freeze was more like 14 GW
, consistent with what happens annually in so-called “shoulder season” when demand is light. ERCOT in the week prior to the freeze tried to induce some plants to come back by issuing a Operating Condition Notice, but they noted plants were unlikely to be able to secure gas on short notice and could and did disregard the request, with 14 GW still offline going in to Monday morning 
Demand for power has to be tightly matched to supply. Grids based on alternating current need a frequency within a very narrow margin of 60 Hz in the U.S. to maintain stability. Below a frequency of 59 grids face cascading blackouts that can cause the entire grid to shut down. ERCOT thus avoided a complete shutdown which would have led to power outages for 26 million Texans covered by the Texas grid. On the morning of February 15, the Texas grid’s frequency declined to below 59.4 Hz for four minutes and 23 seconds. Had it remained there for a full 9 min, ERCOT said the grid would have collapsed, leaving much of the state in darkness (see Fig. 3) , , . At its peak on February 17th, nearly 49% of the grid capacity (52.2 GW out of 107.5 GW was offline) .
With this close call, the entire system was within minutes of collapse, which would have required a “black start” that could have taken days if not weeks (or even months) to implement. To restart would require a slow process of starting individual plants and then building the grid back up gradually.