Research On System Identification And Control Method Based On Neural Network

The artificial neural network has many functions, including the parallel processing capabilities, self-learning ability, adaptive ability and the ability that can approximate any nonlinear function with arbitrary precision. Therefore, the artificial neural network has been widely applied in many fields such as pattern recognition, system identification, system control and so on. In this paper the application of neural networks in system identification and system control is developed. Several novel methods are given for system identification and model parameter identification. And the algorithm of parameter adjustment of the neural control system is improved.In this paper, considering the slowly training speed of BP neural network, easy to fall into local minima, and weak generalization ability, the conjugate gradient method is proposed to adjust the neural network parameters. This method improves the quality of identification of feed forward neural networks in some extent. The neural network is used to identify Linear discrete time-varying systems defined in the Hilbert space. And the Approximation capability to the original system is discussed. Further, the impact that the orthonormal basis of Hilbert space to neural network’s training is analyzed.Considering System control, the basic structure of the neurons in closed-loop control system, parameter adjustment, the relationship of training and system stability are researched on the basis of training methods of artificial neural PID regulator. A previous experience and momentum tem are used in adjustment algorithm of system parameters, so that the adjustment of network parameters not only depends on the nature of the neural network. And there is a contact with previous training error. Thus the sharp oscillation can be avoided in the training process.