Research On Solving Optimization Problem By Real-coded Genetic Algorithm

Real-coded genetic algorithm(RCGA) is one of the important methods in genetic algorithm and it is a global optimization algorithm based on the principle of natural selection and genetics. it has become one of the most important methods for solving optimization problems. Although real-coded genetic algorithm has achieved great success in the theoretical study and practical application, but there are still some deficiencies which need to be further developed and improved.From the existing literature on real-coded genetic algorithm,I have found real-coded genetic algorithm has the following shortcomings:(1)In real-coded genetic algorithm, the outstanding individuals are easily destroyed during crossover and mutation procedure. The elitist strategy is often used to prevent this phenomenon.But the diversity of the population is destroyed. It becomes an urgent problem to assure the diversity and protect the excellent individuals of the population in the same time.(2) Real-coded genetic algorithms generates the population in random during crossing. But individuals can’t step out of the constraints on the range in this way. The scalability of genetic algorithm is limited.(3)The penalty function is adopted by real-coded genetic algorithm to solve the constrained optimization problems, and the value of punish factors is different in different problems. What’s more, it is very difficult to determine the best punish factors.Aiming at the defects of real coded genetic algorithm, the improved genetic algorithm is studied in this paper. Then, the improved real-coded genetic algorithm based on the idea of adaptive step has been proposed. And the study is comprehensively applied in practical optimization problems. What’s more, the method of solving the constrained optimization problems based on the repair strategy has been proposed after the study has been introduced before.This paper adopts the research method that the theory analysis first and then uses the computer simulation to verify the correctness of theoretical analysis. This paper can promote the development and application of real-coded genetic algorithm.(1) A new method of generating the initial population is proposed.In solving the constrained optimization problems, the initial population is often generated in random. Much more time has been cost in solving the complex problem. The method of generating the initial population in this paper synthesizes the strategies of penalty function, gradient method and repair strategy. The initial interior point is generated by gradient method in the first. Then the other individuals of the population are generated by repairing strategy. The speed of generating theinitial population has been improved more in this method.(2) The strategy of adaptive step has been used in crossing.The strategy of gradient method has been used in crossing. The value of crossing factor is determined by fitness function. By this way, we can ensure that more new individuals will be created near the excellent solutions. At the same time, progenies can step out the constraint by using the adaptive step. The speed of the algorithm’s convergence has been improved much more.(3)A strategy of repairing infeasible solutions is proposed in solving constrained optimization problems.In order to ensure the feasible of every individual in solving the constrained optimization problems, the strategy of repairing infeasible solutions is put forward. It is able to transform the solution from infeasible individual to feasible individual in finite steps.Finally, on the basic of summarizing the paper, the prospect of RCGA is made.