An improved real-coded genetic algorithm for parameters estimation of nonlinear systems

This paper presents a searching method for parameters estimation of nonlinear system by using a modified real-coded genetic algorithm (GA). It is well known that GA method is an optimal or near-optimal search technique borrowing the concepts from biological evolutionary theory. The ordinary form of GA used for solving a given optimization problem is a binary encoding during operating procedures. However, in the real applications a real-valued encoding is usually used and is easy to directly implement the programming operations. Thus, in this paper we develop a multi-crossover real-coded GA and utilize it to estimate the parameters of nonlinear process systems, even though those have the term of the time delay or are not linear in the parameters. The effectiveness of the proposed algorithms is compared with different evolutionary algorithms. Simulation results of two kinds of process systems will be illustrated to show that the more accurate estimations can be achieved by using our proposed method.