A while back I claimed that an all-dispatchable grid would require 75 hours of storage. I have been definitively proven wrong. Nothing like that is required.
In fact, tiny amounts of storage may do the trick if the grid is large enough -- and making the grid large enough is not that expensive. We may already have in place what we need for a completely renewable grid.
The 75-hour figure came from studies of single, isolated wind farms. But as you add wind farms, the odds of two wind farms being down -- or low producers at the same time -- drops.
I came across confirmation by accident: a Vehicle to Grid study (PDF) that evaluated data from eight sites showed that storing just 36 minutes of nameplate capacity would allow the widely dispersed farms to meet a firm power commitment ~90% of the time.
There are tricky steps involved. I am not necessarily advocating a 95%+ wind grid, as there are many ways to generate renewable electricity. But just as coal dominates power production in our current system, I suspect wind will dominate in any future grid. So consider the following a limiting case for wind.