Hi Zirui,
to get different results you should probably feed different trials with different seeds. For example, if you use different "stimSourceParams" for a NetStim stimulation, you can attach the dict-entry "seed" with the trial number. Also, every trial may instantiate other connectivity (probabilistic) patterns via "simConfig.seeds = {'conn': i}. If you effectively want to reset the network, after the simulation you can use "sim.clearAll()". Hope it helps,
Eugenio