Genetic algorithm is a kind of random searching method directed by fitness function. It simulates the biology evolution in nature. GA is simple and easy to implement. Especially, it does not need the special field knowledge, so it has been using in very broad fields such as Numerical Function Optimization, Combination Optimization, Artificial Intelligence, Intelligent Control, Image Processing.Genetic Algorithm is still a new method of optimization algorithm, and its theoretical groundwork is weak. There are still lots of problems to study and develop. GA has some drawbacks when it is used in engineering practice. Especially, premature convergence sometimes occurs.Firstly, this thesis briefly summarizes the basic technologies and theories of Genetic Algorithm and introduces the basic flow of Simple Genetic Algorithm.Secondly, I analyze the cause of premature phenomena in the article, and give a hybrid improved genetic algorithm with the criterion of premature convergence based on the study of other scholars. The improved algorithm can find the occurrence of premature phenomena using the wave of population entropy. Breaking out of the trap of the part best value by means of the mutation directed by the monotony coefficient of population when the premature phenomena is occurred, the algorithm restore the ability of evolution.Finally, the algorithm is tested by three test functions of numerical optimization .The emulational experiment results show that this improved algorithm has greater probability of convergence.
