The UV LEDs are nice, but can be somewhat expensive and have a short lifetime (although they are getting much much better and that may not be as much of a concern theses days). The problem you have is that UV-B LEDs may be too weak for for your application. The UV LEDs may be too narrow bandwidth for your application.
I think the problem is how close you want to simulate the sun's spectrum.
If you need high intensity, the answer is probably different than if you just want a spectral source.
In the past I have used a monochromator and a combination of deuterium and Xeon lamps then I could calibrate against a standard photodiode.
If cost is really a driving factor, You might want to consider using a Xeon lamp with a quartz or fused silica envelope, and a deuterium lamp with appropriate filters.