Many engineering applications, including fault detection, fault diagnosis, automatic control, and simulation, require mathematical models of dynamic systems. There are two basic approaches to the construction of mathematical models; one is analytic based on the laws of physics, and the other is experimental based on fitting a model to recorded data by assigning numerical vahes to its parameters. The latter approach, known as system identification may involve the following steps: (1) select a model structure, based on physical knowledge; (2) parameterize the model; (3) design experiments and collect data; (4) estimate parameters; (5) verify the estimated model. This paper describes the application of genetic algorithms to a numerical search problem in the identification of dynamic systems. The prediction error identification method is used, allowing nonlinear-in-the-parameters models[3].