How do heat waves, droughts, wildfires and cold snaps break energy systems, and how do planners prepare?

Hit the Plains Grid with four hazards, alone and two at once, then choose defences under a budget

Analyst lesson A9 of Energy Systems, Step by Step: sort what each hazard hits, read Winter Storm Uri from the FERC–NERC report, stress the Plains Grid with hazards, and choose defences under a budget.
  • Analyst · L2
  • 25 min guided · about 30 min practice
  • After F6 and A5

Lesson A5's Plains Grid in a year when each hazard strikes (500 resampled years, seed 2025; illustrative)

  1. A heat wave raises demand; a drought warms or starves cooling water; a cold snap raises demand, freezes plants and cuts gas; wildfire shutoffs switch lines off.
  2. In Winter Storm Uri (February 2021) units of every kind failed: 58% of the 1,045 affected units were gas, 27% wind and 6% coal.
  3. Unserved energy rises from 108 MWh a year with no hazard to 5,802 MWh in a cold-snap year (table below).
  4. A heat wave and a drought together add 1,541 MWh a year, 2.6 times the 591 MWh of the two alone.
  5. Within $60M a year, winterizing, hardening lines and demand response together leave the least unserved energy across the four hazard years: 686 MWh.
Year with When LOLE (days a year) Unserved energy (MWh a year) With winterizing, hardened lines and demand response Switched off directly (MWh a year), without → with those defences
No hazard – 0.228 108 – 0
Heat wave 15 July to 14 August 0.738 451 210 0
Drought July to September 0.818 357 137 0
Cold snap 14 to 20 January 2.792 5,802 288 0
Wildfire shutoffs 1 to 21 August, 12:00 to 20:00 0.556 230 51 28,871 → 11,548

With JavaScript on, you sort what each hazard hits, run each hazard and pairs through the simulator, and choose defences yourself.

  • Stress tests, not forecasts: each hazard strikes in every one of A5's 500 resampled years, so a result reads "in a year when it strikes".
  • The hazards' dates and sizes, and the defences' effects and yearly costs, are illustrative; the real events quoted come from the primary reports named with them.
  • Supply measures (LOLE, unserved energy) count shortfalls for the customers still connected. Customers on switched-off lines lose power whatever the supply; that energy is counted on its own.
  • Left out: the network inside the grid, imports, repair crews, and how often each hazard strikes.

Need a refresher? F6 (the grid and outages) and A5 (adequacy and accreditation).

What you just did, as diagrams

Correlated failure
Many units failing together from one cause, such as cold, which independent outages miss.
Compound hazard
Two hazards at once; together they can do more harm than each alone added up.
Stress test
A grid run through a hazard on purpose, to see what breaks and what a defence buys.
Defence
A measure that eases one hazard or several: winterizing, cooling upgrades, hardened lines, storage, flexible demand.
Go deeper · AnalystWhy a stress test is not a risk estimate
  • Conditional, not expected. Each result assumes the hazard strikes every year. A risk estimate also needs how often it strikes, and that is changing with the climate.
  • One cause, many failures. Uri’s outages shared causes: freezing (44.2%) and fuel problems (31.4%). Elliott’s were 31% and 24%.
  • Cost per MWh avoided. Compare defences by MWh avoided per dollar, hazard by hazard; a defence for one hazard may do nothing for another.
Go deeper · ModelerHow the hazards enter the model
  • Multipliers by the hour. In each hazard’s window, demand, wind and each unit type’s output are multiplied, in every simulated year.
  • A defence keeps a share \(k\) of a loss. A multiplier \(m\) becomes \(1 - k(1 - m)\); demand is never eased.
  • Shutoffs. Demand switched off directly leaves the supply balance: supply need not serve it. Its energy is counted on its own.

Check yourself

Myth or fact? “Winter Storm Uri was caused by frozen wind turbines.”

  • Myth
  • Fact
Answer

Myth. Every kind failed: by count of affected units 58% were gas, 27% wind and 6% coal, and freezing and fuel problems caused 75.6% of the outages.

Myth or fact? “Two hazards at once do the harm of each one added up.”

  • Myth
  • Fact
Answer

Myth, here. A heat wave and a drought together added 2.6 times the unserved energy of the two alone: both squeeze the same summer evenings.

Myth or fact? “Wildfire shutoffs can cut far more customers' energy than the supply shortfall they cause.”

  • Myth
  • Fact
Answer

Fact, in this model: 28,871 MWh a year switched off directly, against 230 MWh of supply shortfall in that year.

Myth or fact? “A big battery is the best defence against every hazard.”

  • Myth
  • Fact
Answer

Myth. A 4-hour battery alone leaves 3,365 MWh unserved in the cold-snap year, against 528 MWh with winterizing at less than half its cost.

With no hazard, 108 MWh a year go unserved; with a heat wave 451, with a drought 357, with both 1,649. Subtract the no-hazard 108 from each: the increase with both is how many times the sum of the two separate increases?

  • 1.0 times
  • 2.6 times
  • 1.9 times
  • 0.5 times
Answer

(1,649 − 108) ÷ ((451 − 108) + (357 − 108)) = 1,541 ÷ 592 ≈ 2.6 times (from the rounded numbers).

Numbers, sources and what’s next

Numbers to remember

Number What it means Source
75.6% Share of Uri's unit outages, derates and failures to start caused by freezing or fuel problems (February 2021). FERC–NERC report, printed p. 15 · November 2021
20,000 MW Firm load ERCOT shed at the worst point of Uri. FERC–NERC report, printed p. 9 · November 2021
90,500 MW Unplanned outages, derates and failures to start at the same time at Winter Storm Elliott's worst point (December 2022), not counting units already out. FERC–NERC Elliott report, p. 5 · October 2023
1 October 2025 NERC's cold-weather standard EOP-012-3 takes effect: generator owners must plan for extreme cold. NERC EOP-012-3, purpose A.3

Keep going

Back to the course map

Sources and further reading

Spotted an error, or stuck? Write to hlli@ksu.edu. Corrections are made in place and dated here.

Text licensed CC BY 4.0, code MIT: teachers and students may reuse and adapt it, with credit to Honglin Li.