Abstract: In order to solve an MRT (mass rapid transportation) train operation simulation model and obtain the optimum operation curve,an improved multi-objective GA (genetic algorithm) was proposed.In this improved GA,gene encoding is based on train control shift position,and a chromosome composed of genes represents a train operation plan to produce the original population.The fitness function includes the objectives of train control,such as stop deviation,time deviation and energy consumption.Every individual is checked with some rule before accepted as new population in course of selection,crossover and mutation,and each algorithm contains elitist reservation in the improved GA.In addition,the proposed GA was tested through a comparison with the multi-particle model by using a case. The result shows that the improved GA can save energy above 10% and give many alternative train control schemes.