This means you need more communications between all the various parts running the grid. So it's not just the switch that opens the valve at the gas turbine power plant that you have to worry about (which I agree is easy to be air-gapped and interlocked). You also have to worry about manipulation of the communication ecosystem that let all the various grid actors coordinate so everything stays in sync.
EDIT: Also, remember that power engineers are extremely good at running the grid and keeping the lights on. They are NOT experienced at protecting it from cyber attacks. Unfortunately, neither the intelligence community or tech sector that is actually good at cyber defense (Google, Apple, Facebook, etc.) has not offered to help add cyber defenses and these physical infrastructures. So, the only parties who have been doing the grid cyberdefense work have been security theater companies ("enterprise" security).
It’s hard enough with non-safety-critical, pure-software systems. Industrial systems with life-safety implications and release cycles measured in decades cannot be made secure against nation-state adversaries at the drop of a hat.
The United States has never tried to bomb-proof every building. Our military strategy is our capability to bomb you right back a thousand times over. Attribution and offensive capabilities will be a lot more important to military strategists accustomed to that mindset.
One reason we have to do this work instead of just warning the world "if our power plants get hacked we're going to war" is that the world in general has no way to judge our claims about hacks we may or may not have suffered. In contrast, if our buildings get bombed that's kind of obvious. We don't have a great deal of credibility when it comes to war justifications, just based on recent form.
I agree we need to fix it, but I'm not surprised that it's slow going or ineffective. Likely we're going to see some very expensive projects to merely install antivirus packages, where what we really needed was to engineer completely different protocols.
Edit: all of the above is factual, apart from the evaluation of them being negligent of their responsibilities solely due to being overly busy. And if you reasonably hold that the NSA are failing to perform their duties to help protect US infrastructure from cyber attacks for other reasons than being too busy spying on the American people -- would you like to suggest what such reasons are?
Fact: the NSA records all electronic communications within the USA.
Fact: the NSA collects backdoors, and has bribed US companies to create them.
Fact: the NSA seeks and collects major OS bugs for exploitation.
Fact: the NSA has systematically avoided revealing the above bugs to US manufacturers. If they had done so, as expected according to their charter, the bugs would have been fixed -- and thereby would have been protecting US infrastructure, government systems, companies, and last but not the least, the people.
That's why you are downvoted I think. Unless you have some proof on this activity being a source of NSA'a problems in achieving it's other goals.
So, since you think they're not too busy with spying on us, then what do you think accounts for their dereliction of duty?
Which really sad and depressing for the American tech sector. Not only do they have foreign state actors trying to hack into their infrastructure, they also know that they're own government won't help by disclosing the holes they hoard (or worse, joining the foreign attackers in exploiting the holes they have found).
America has so much to lose and very little gain from exploits. It's so discouraging that the intelligence community is still valuing offense over defense.