Hi Salaki,
The Savonius style of rotor is (i think) inherently inefficient.
The only way to improve the wind action that i can see,
is to mask or shield the wind from the half that is detracting.
This could concievably be done by using a tail which would place
such a shield across the half that detracts.
***************************
Since my explanation of this concept is poor,
i will try to explain it another way.
The savonious style of rotor has one side moving into the wind.
This side could be shielded by a movable part with a tail.
The movable part with a tail would be positioned by the wind,
and would sheild the section of the savonius rotor which
moves into the wind.
***************************
That explanation is not much better.
But it will have to do.
**************************
I can see no way to improve the efficiency of a savonius style
rotor very far, this suggestion of adding a separate part to act
as a sort of wind-guide or shield should make a big difference,
even though it does not address the efficiency of the rotor
but changes the way the wind blows upon it.
Best of luck with your project,
and let us know how you get on.
John