You probably already know about the Morris Worm, but in case you don't, here's a short summary:
In 1988, Mr Morris, a grad student, wanted to highlight the inherit risks of weak email passwords. He decided to raise awareness by creating a short code that would be sent by email to someone, take a hold of that person's computer, then send copies of itself to each of that person's email contacts.
This was the first ever software that you could consider to be a computer virus. Although Morris never intended it to be very serious, his worm caused a denial of service typeattack on the infected computers, due to the overload of messages being sent by computers with multiple copies of the worm on it.
It could be fun to bring back this sort of miscalculation to create an accidental global blackout. Imagine an eco-terrorist group wanting to highlight our society's dependance on electricity. One of their hackers develops a code worm that infects a city's power grid to shut it off for one hour (except vital services, like hospitals etc), then turn it back on, and then the group can send a message out for big impact.
But something goes wrong in the deployment. Maybe the hacker didn't configure the "turn back on" sequence properly, so the power grid misfires and melts down. Maybe someone higher up in the organisation (less versed in IT) wants to spread the virus to other cities, all across the globe, for higher impact. Maybe while the power is out, something goes wrong due to a lack of power, like someone trying to force it back on (the hospital still has power, but does the prison? Does the military base?), and it damages the system. Maybe some of all this at the same time!
Just this might not hit the whole world, but as others have mentioned, once a few things go wrong, it can be quite easy for everything to cascade out of control. This could go quite well with @Terraformer8's scenario of power grids trying to replace their technicians with an AI.