A very naive way of modelling this is to use the solar radiation values on the facade (measured at the points of your grid) as the value with which you scale the panels (radius of circles). This is actually using n attraction points, one for each of the n faces of your mesh (facade) grid. It is still a simple way of approaching this however, meant more to make you work yourself through the native GH definitions required for this.
When you run this for yearly average values you will notice the results are (most probably) uninspiring. This is because the yearly average variation of solar radiation on a facade isn’t usually 'interesting. You can make it a bit more responsive if you track values through seasons, weeks, days, or time of the year, or normalizing results as well.
Just play around with that. Let us know what you find!