Generate synthetic high-frequency irradiance data from hourly data

[simon-d-b]

Is your feature request related to a problem? Please describe.
Some of the pvlib.iotools functions such as iotools.get_pvgis_tmy generate hourly irradiance data. It would be interesting to interpolate this data to a more fine time-step, say 5 minutes.
This would be interesting for, in my eyes, two reasons:

• Generate nicer, more smooth results and figures
• Generate data which is more realistic

Describe the solution you'd like
I think there are multiple ways of solving this, depending on desired fidelity:

• Interpolate the hourly data with, for example a cubic spline after setting the values at night to 0. This solution would solve the first problem but very physically accurate as it would not preserve the irradiance's variability and statistical properties. It would be very easy to select any desired time-step
• This paper by Grantham et al. proposes an algorithm to generate synthetic irradiance data for 5 minute time-steps. This model could be implemented to generate more physically representative data.

Describe alternatives you've considered
The introduction of the above-mentioned paper discusses some alternative algorithms, but none of them preserves the relation between GHI and DNI.
The paper mentions that there is little difference between 5 minute and instantaneous data, the implementation could be extended to use an arbitrary time step by using a simple interpolation on the 5-minute data, but this does not seem really necessary to me.

[kandersolar]

Thanks @simon-d-b for the thoughtful write up. Can you repost it in #788? Better to have all the information in one place.