抄録
In this paper, a new identification method is proposed which can obtain a good accuracy of identification of nonlinear time-lag systems under noisy measurements, on the basis of combination of genetic algorithm and sequence method. First, the identified system is described as a discrete model of a polynomial type with unknown parameters using Kolmogorov-Gabor's method. The task of system identification is to determine these parameters. Though the parameters of the identified system can be obtained through the search of GA, there is a potential risk in using a simple GA that a solution is usually stuck at a local minimum. Second, to solve this problem, a new GA search method is proposed by adding a sequence search which is carried out nearby the value of each estimated parameter coming from a simple GA. By this method, individuals whose fitness are larger are found. As a result, the solution escapes from the local minimum and converges to the optimum one. Finally, through simulation, some examples are given to demonstrate the validity of the proposed identification method.
