Since I'm plugging for nuclear power, I feel the need to address nuclear accidents.
To date, there have been three, though the anti-nuke folks will tell you with a straight face that there have been "tens of thousands" of accidents, including the recent "accident" in Taiwan where Taipower replaced two broken bolts during a routine maintenance shutdown (by this method of calculating, there must have been thousands of accidents involving wind power too, but I digress).
Of the three accidents, only the first, Three Mile Island, is a "true accident." I say that because it was simply defective design. A valve in the cooling system failed, the reactor overheated and was destroyed. Although only a small amount of radiation was released and no one was killed, there is no denying that the reactor was inherently unsafe by design. The only upside is that the nuclear industry was forced to do a lot of safety redesigning as a result.
The second accident was Chernobyl, which was a generation one reactor. It was also a bad design, built to produce as much plutonium as possible (for nuclear weapons) - producing electricity was secondary. Despite the poor design, ironically the disaster wasn't a true accident in the sense that it wasn't caused by a mechanical failure or even an operator error. What happened is that a team of "experts" was sent from Moscow to conduct a dangerous experiment on the reactor, which involved shutting down the reactor with all the safety systems disconnected. The operators of the Chernobyl reactor objected to this, but they were overruled. They gritted their teeth and obeyed orders, unhooked backup safety equipment, knowing it could easily turn into a disaster. The experiment was conducted near midnight (because the demand for electricity is low then), so when the evening shift departed, the graveyard shift came on duty not even knowing what was being done. The reactor overheated and caught fire, an entirely predictable outcome. My understanding is that several people later went to jail because of their role in creating the disaster. Of course, whether or not one wishes to call it an "accident" or "sabotage," there is no denying the fact that it was indeed a disaster, the most deadly one in the history of nuclear reactors.
Last accident was, of course, Fukushima Daiichi, a generation-two power plant composed of six reactors. First off, the reactors all came through the earthquake OK. The emergency earthquake detection system did its job and shut down all six reactors almost instantly, the emergency cooling system (which is diesel powered) started automatically. But this points to a major weakness in generation-II design - you need an active cooling system running even when the reactor is turned off. The tsunami arrived about 20 minutes after the earthquake, and carried away the diesel fuel tanks (which were stored outside the building housing the reactors). With no fuel, the diesel generators shut down. There was a backup cooling system which ran off of batteries, and this kicked-in as it was designed to do, but the batteries could only power the pumps for eight hours. Once the batteries were discharged, the cooling system went down, the hot reactor started boiling the water and it evaporated, exposing the core to the open air and causing a meltdown.
There are a few lessons that should be learned from Fukushima. First would be to locate reactors at least 20 meters above sea level (Fukushima was 10 meters, the tsunami was 15 meters). Ironically, a 45-meter high hill was bulldozed when the power plant was constructed, lowering the site to 10 meters - the landfill was used to construct the harbor for delivering uranium. I used Google Earth to look up the elevations of Taiwan's reactors - sadly, the fourth nuclear powerplant (currently under construction) is only 10 meters high - like Fukushima, the site is actually being lowered to make it easier to deliver fuel by ship. That was dumb, and makes me wonder if the project shouldn't be abandoned - it might actually be cheaper to do so, since the design being used is already obsolete and a generation three-plus plant could be up and running in three years.
Another lesson that should be learned is that we need to use generation-four reactors, which eliminate the whole problem of needing a cooling system running for weeks after shutdown. As an added bonus, generation-four plants burn their own plutonium waste (the most troublesome kind of waste, by far) - the Gen-IV plants can burn the waste of Gen-II and Gen-III plants, using it as fuel.
As Finley said, nuclear power is complicated, and the details matter. Unfortunately, the public and politicians want energy policy to be so simple that everything we need to know can be fit on a bumper sticker.
Looks like the USA will soon begin building two new gen-III-plus reactors, the first in decades:
http://en.wikipedia.org/wiki/AP1000