A Penalty-Based Evolutionary Algorithm for Constrained Optimization

Evolutionary algorithm based on penalty function is a new kind of efficient method for constrained optimization problems, however, the penalty parameters are usually difficult to control, and thus the constraints can not be handled effectively. To handle the constraints effectively, a new constraint handling scheme based on a continuous penalty function with only one control parameter is proposed. Moreover, it does not add extra local optimal solutions to the primary problem. In order to decrease the amount of computation and enhance the speed of the algorithm, the uniform design is combined into the proposed algorithm to design a new crossover operator. This operator can generate a diversity of points and exploit the search space efficiently. Based on these, a novel evolutionary algorithm is proposed and the simulation on 5 benchmark problems is made. The results show the efficiency of the proposed algorithm with less computation, higher convergent speed for all test problems.