Gautam Gowrisankaran, Stanley S. Reynolds, and Mario Samano have a new JPE paper on this very important question, here is their abstract:
A key problem with solar energy is intermittency: solar generators produce only when the sun is shining, adding to social costs and requiring electricity system operators to reoptimize key decisions. We develop a method to quantify the economic value of large-scale renewable energy. We estimate the model for southeastern Arizona. Not accounting for offset carbon dioxide, we find social costs of $138.40 per megawatt hour for 20 percent solar generation, of which unforecastable intermittency accounts for $6.10 and intermittency overall for $46.00. With solar installation costs of $1.52 per watt and carbon dioxide social costs of $39.00 per ton, 20 percent solar would be welfare neutral.
20 percent solar for Arizona as the social welfare break-even point does not strike me as especially impressive, but of course the infrastructure integration technologies may yet advance. In gross terms, intermittency costs exceed carbon costs. Note also that forecastable intermittency accounts for most of the costs, and so with perfect storage solar would be a much more efficient technology.
Here are ungated versions of the paper.