Predictable isn't good enough.

Predictable is a reason for a government regulation.

Without that, you need quantifiable.

What is the probability of a polar vortex this year? (P) How many hours of high prices would that cause? (H) What will the price be during those hours ? (Pr) What is the probability that our grid uplink will go down and cut us off? (Pd)

So, on the profit side: P x H x Pr x (1-Pd)

Liability:

What is the probability that our failure to winterize will cause a blackout? (Pb) How much damage will that cause ? (D) How much of it is our legal responsibility? (L)

Pb x D x L

Now: how much will winterizing cost? C

The big question:

C ?>? Pb x D x L + P x H x Pr x (1-Pd)

Complicating factors: in Texas, L is 0. For everyone. And P,H,Pr, and Pd are all non-zero, non-negligible, but not quantifiable.

How are you going to get investment money for a project that may, or with some probability may not, hit the jackpot sometime during the next century but otherwise lose money all the time?

Why do once-in-a-century weather events seem to happen so much more than once in a century?