I would model the main wing and 'up' winglet with a blended wing component.
The inboard part of the wing with the straight trailing edge would be the first wing section.
The main wing portion would be the second wing section.
The blend transition would be the third.
The main winglet would be the fourth.
I might add a fifth section to model the scimitar blended tip -- but OpenVSP might struggle with this a bit.
The 'down' winglet would be a second wing -- possibly attached in U,V coordinates -- but with a slight negative delta to 'bury' the root section. It would get the same tip treatment as the main winglet.
Depending on the analysis you want to perform, the winglet junction may not work very well. For example, you don't want a gap for VLM analysis.
Something like this...

Of course this wing is un-twisted and uses NACA 0010 airfoils. It is just meant as a demonstration. To properly design a wing, you need to select appropriate foils and then twist the wing to achieve the desired lift distribution. Adding winglets does not guarantee a drag reduction -- the winglets must also be properly twisted and designed to have a beneficial effect. It would not surprise me if this naive wing performed worse with the winglets than without.
Rob