You can find a lot of varying estimates, based on other varying estimates, and they're almost always from a source with a vested interest in skewing the statistics one way or another.
And numbers for accidents caused by windmills?
Here's my attempt at nuclear vs windmills (except I'm only mildly biased in the pro-nuclear direction and trying to be as fair as possible):
Nuclear
The vast majority of nuclear power plant deaths occurred from Chernobyl. This has been estimated anywhere from 57 deaths to 985,000 deaths. Obviously these numbers are both BS. A realistic estimate appears to be about 5,000. (See Have several hundreds of thousands of people died because of the Chernobyl disaster? )
Later, Fukushima happened, which released radiation that may result in 130-640 cancer deaths.
From 1965-2015, nuclear power plants worldwide have generated 82,337 TWh, if you add up all the totals from BP Statistical Review of World Energy
Together, this is 0.068 deaths / TW⋅h = 600 deaths / TW⋅yr
Wind
CWIF lists 165 windfarm-related fatalities up to to September 2016. Some of these are dubious (car accidents caused by distraction from shadows), but there are certainly others unreported, so we'll just use that number.
It's really hard to find a single statistic for the total amount of energy produced by wind. Wind advocates always quote "installed capacity" because it sounds bigger, but that's just the maximum theoretical power a wind farm could produce under constant high wind. The actual amount of energy produced in reality is less than 1/4 of this.
Combining several incomplete sources (EIA, Paul Gipe, IEA, WWEA) that seem to match each other more or less, I get a total of 4799 TWh total from 1980 to 2015. (data) (Actually, the BP source says 4778 TWh total now.)
Together, this is 0.034 deaths / TW⋅h = 301 deaths / TW⋅yr
This matches Paul Gipe's estimate of 0.032.
So based on (pretty imprecise) death estimates, nuclear still has to catch up with wind.
As I said in a comment, the Next Big Future source is being dishonest by extrapolating the nuclear numbers 25 years into the future and assuming no accidents during that time, which they don't do for other power sources.