No. If you hit start, the first generation (generation 0) is always a random generation. If you wish to improve the result of a simulation, one way is to let it run longer. Of course, it also relates back to your design problem and the way you set it up.
On a different note, if you wish to run sequential simulations, one way is to treat the results of the first simulation as constants for the second simulation. For example, you run a simulation for a tower design, mainly for massing studies. You select one individual and then use that individual to run another simulation for facade studies.