cheap protocol for 'global' minimum search in liquid phase

[Rafał Roszak]

hello all,

I want to find reasonable candidate(s) for global minimum of
molecular complexes in organic solvent. My first idea was force-field
simulated annealing but my systems contain exotic species (e.g.
radicals, carboanions) which are not supported by most (all?) of
existing force field. The systems are to big for simulated annealing at
DFT level, so I thinking about semi-empirical simulated annealing. Here
is my question: 
1. Is this reasonable idea? Could you suggest any better method for this purpose?
 
There are several semi-empirical methods implemented in cp2k. At least
some of them have problem with periodic system (e.g.
https://groups.google.com/forum/#!msg/cp2k/eN54pEEbQ3Y/cOLsAYeQyWsJ),
so:
2. What you recommend? Which method and what about other settings?

Regards,

Rafał