Abstract: Based on conventional genetic algorithms, a beneficial modification is done in some aspects to enable them to avoid inherent prematurity and getting bogged down in local optimization. Hybrid encoding method is introduced to make the algorithm more practical. Recombination and screening operators used in forming virgin population will make the distribution of initial solutions more reasonably so as to benefit the improvement of computation efficiency and convergence of the algorithm. In the implementation of the algorithm the mixed genetic operators are selected suitable for both binary coding and real value coding. The modified genetic algorithm has been used in parametric optimization of a sliding-vane compressor, by which the result shows that the modified genetic algorithms is reliable and efficient, as proved by the reasonable parameters.