Back calculating exposure using variable ND filters for long exposure times

I want to use my variable ND filter to take a photograph of a busy street corner and want the exposure to be at least 15 to 20 min long so that it appears to be a completely vacant of people. But the problem I have is back calculating the length of exposure to ND without having to do several 20 min tests. I've found sites what give simple back calculations but nothing that would cover a lengthy exposure.

I wonder if I could do some tests and then on a piece of graph paper extrapolate back what exposure to ND would get me to 20 mins? Are there any former mathematicians in the audience? Or better yet, computer programmers that can write a simple program for such calculations?

Personally I have never tried it but my reference book prescribes an exposure of 1/2 second at f11 to begin to blur people and an exposure of 2 seconds at f22 to eliminate them (base normal exposure @ 1/30 f2.8) . 15 to 20 minutes seems like an awful long time to me - I am thinking maybe several minutes in poor light/night time???

The examples I'm going off of show anywhere from 15 to 30 min to completely wipe people out. I agree it does seem like a long time doesn't it? In the mean time I did find this little tidbit

For each stop of light that the filter blocks, you need to double the exposure time, so for a 10-stop ND filter, multiply the exposure by approximately 1,000.

If you want to do it in daylight, my guess is that 20-30 seconds would be good enough. Just take a few shots at different interval and then layer them together carefully blending out the unwanted.