Research On System Modeling Theory And Optimization Based On Dynamic Neural Networks

The technique of system modeling has been an important tool now in analyzing and researching all kinds of systems, especially complex systems. Because it's economic, trusty, easy to be implemented and can be used repeatedly, it has been an efficient means in analyzing, designing, experimenting and evaluating actual systems, and has extensive applications in the fields of science and engineering. With the continually widening of the application fields of scientific and engineering, the problems people encountered become more and more complex. Most objects to be modeled exist many kinds of indetermination and the non-linearity character which is hard to quantitatively describe, and the requirements to improve the describing ability of system models and the flexibility of modeling method, the generality and the intelligence level of modeling methods. Traditional methods of system modeling no longer meet current needs in many ways, exist difficulties in model selection and model, low precision and difficulty solving problems.In recent years, Intelligent information processing theory and technology are valued and developed, provides an effective way to solve the modeling of complex and nonlinear dynamic systems. As dynamic neural network has highly nonlinear mapping capabilities, large-scale parallel distributed processing, and good adaptive learning mechanism, it shows a greater advantage in modeling complex systems. Therefore, the study of dynamic neural network applied to the system modeling has a good prospect.The paper presents and establishes a class of feedback dynamic neural network model, process neural network model, continuous dynamic neural network model, dynamic loop network model and other models. The models have the features of strong adaptability, high learning efficiency and easy modeling of specific practical problems and so on. The paper has constructed three dynamic neural network learning methods which are based on the training algorithm of Gradient-Newton Combined, the training algorithm of quadratic spline fit and the training algorithm of numerical integration. Optimization principles and implementation techniques of the models and algorithms have been given. Finally, combined with the practical problems in the development and production of oilfield, the paper proposes the specific applications of the dynamic neural network modeling.