Traditional genetic algorithms put all the individuals in one population to cross and adopt the same set of evolutionary parameters and genetic operators to guide the evolution, which will easily lead to local convergence and poor searching efficiency. A multi evolutionary system co-exist genetic algorithm is developed to overcome the fluctuations of the whole evolution process through dividing individuals into several sub-populations according to the fitness value. Moreover, the improved algorithm prevents the early convergent and increases the diversity of individuals by supplying these sub-populations different evolutionary systems. The effectiveness and feasibility of the algorithm are verified by typical genetic algorithm test functions and an engineering case. The results show that the genetic algorithm has a good versatility, high convergence rate and solution precision.