Studies On Improvement Of Genetic Algorithm And Its Applications In Optimization

As an important branch of science, optimization has been rapidly popularized and applied in many domains such as economic, production, engineering technique, transportation logistics and so on, and it has become an indispensable tool in many practical works. How to solve the optimization problem fast and effectively has become an important and meaningful object that demands to investigate on. Traditional optimization algorithm theory has been extremely maturity, but when encountering some large-scale and complicated problems, it shows its shortages.Genetic algorithm (GA), as a kind of intelligent algorithms, is a global auto-adapted optimization technology which simulate是the natural evolution process. During the process of solving the continuous state space optimization and the combination optimization, it shows robustness, global, interoperability implicit parallelism, which makes it quickly become a widespread intelligent optimization algorithms. In this paper, on the basis of summary and analysis of optimization and genetic algorithms, to solve the deficiencies of genetic algorithm when solving the continuous state space optimization and traveling salesman problem, the corresponding improved methods are proposed. The main research works are presented as follows:In function optimization, some methods are proposed to solve the problems of premature convergence, poor local search, slow convergence speed, low accuracy in the genetic algorithm. In this paper, based on real coding, through introducing an orientation factor called compound form method which can converge to the local optimal solution rapidly, I proposed a new hybrid genetic algorithm.On the combination optimization—traveling salesman problem(TSP) which is NP-hard problem, I analyze the characteristic of TSP itself: Every individual forms a chromosome ring; approximate optimal route and the best individual have many identical gene. I propose a kind of genetic operator which can not only inherit most of genes fragments but also obtain new excellent short mode genes. In addition, I improve the production method of the initial colony and propose an improved genetic algorithm to solve TSP.I test the improved algorithm in MATLAB environment, on the problem of continuous state space optimization, I use Banana function and Rastrigin's function as examination functions, Banana function is difficult to optimize with traditional optimal method such as the steepest descent method, and Rastrigin's is a multimodal function and is very easy to converge to local optimization. It shows that this new algorithm is high efficiency and flexibility compared with standard genetic algorithms through numerical experiment. On TSP, by compared with Ref. 14, and applying to less than 100 cities TSP of ShanXi province, the optimal route obtained in experiments can demonstrate the high efficiency of the algorithm.