Penetration of Solar Power without Storage

I'll preface this post by saying that it's a plug for my own research, but it's also published so (at least some people think) that I'm not making it up. Since the work is published in a journal, normally I wouldn't be able to just link to it, due to access limitations, as Fat Knowledge has lamented. However, some professor at a university in Texas has made it part of his course, so I'm going to direct you there for the whole article:

Solar Power without Storage
The idea is that while solar panels are expensive on their own, if you need to build a storage system to use their energy at night, they become even less affordable. Clearly, solar power can't provide all of our electricity needs if we skip the storage, but they can still help out in the daytime. In fact, in most parts of the country, the electrical demand is already highest when the sun is out, so solar can help with these peaks. The higher daytime demand is currently met by turning on "dispatchable" power sources that are (relatively) inexpensive to build, cost a lot to run, and can be turned on and off in very little time. The ones that run less frequently cost less to build (by design) but even more to run, since they won't be generating very much. In fact, the plants that run to fill the demand during the highest 30 or so hours of the year can cost 5-10 times as much (/kWh) as your typical coal or nuclear plant. (The reason that coal and nukes don't produce all the electricity is that these "baseload" plants can take half a week to start up and shut down, and could never respond to the daily variations of demand.)

My goal is to see how much solar capacity can be installed so that the panels mainly replace the dispatchable plants. Specifically, what is the maximum deployment that permits 95% of the annual output from PV to be utilized without reducing the output of the baseload plants? I used hourly solar intensity and electrical demand data from 32 regions across the country to accomplish this.

This map shows the regions and their possible penetration. Note that the share is in effect a measure of the correlation between electrical use and sunlight, and not of the amount of sunlight. Locations with large amounts of sunlight in times of low demand, such as noon in winter or anytime in the spring or fall, will have a lower possible deployment. In this sense, New England’s grey days in October and November help improve the matching because the clouds
reduce sunlight when the electricity use is lowest.

My total installed possible capacity in the 32 regions is 59 GW. These regions cumulatively consume 30% of America's electricity, so if panels are as effective in the rest of the country, the whole US could use 196 GW. This would reduce the energy currently provided by dispatchable power plants by 23% and would represent over 7% of the present total annual electrical load in the US. Although that share may not seem like a lot, it requires nearly 8 doublings from the 865MW of PV which was the installed base in the US in 2007, showing that in the near future, bringing down the cost of the panels is more important than worrying about the dark.