Based On Wavelet Neural Network Model Predictive Control Design And Research In More Water Tank

With the advent of economic times, in order to meet people's demand for highquality products, improved production technology becomes more and more urgent,however, complex industrial site, so that changing any minor error experiments will not only cause damage to the device, more time will be wasted in the face of these problems, can simulate most of the actual object becomes a key industry, the emergence of four tank water level control system largely solves this problem,particularly non linear and coupling characteristics such that when the control nonlinear predictive control system also can delve into a subject in order to achieve adequate control and meet customer needs. Therefore, this paper nonlinear coupling characteristics and how the predictive control strategy, carried out research work.Wavelet neural network nonlinear approximation ability is good or bad, is often a critical part of the model accurately or not, this article will advance the parameters of wavelet neural network optimization, process optimization, linear reduction factor proposed strategy enables variance showing a decreasing trend, from the role of balancing algorithm, followed by the establishment of the exact model and dynamic matrix control strategy combining the nonlinear multi- tank were studied and controlled. Faced with multi- tank coupling characteristics, we use neural network decoupling techniques to deal with this problem, the main principle is to identify and feed-forward neural network decoupling theory, decoupling channel, the internal structure of the neural network, the relevant parameters have been designed to give the neural network decoupling specific strategies.Optimized wavelet neural network dynamic matrix control first two- tank controlled experiment, and the results of PID simulation results for comparison,then in the middle of the controller with a four- tank add neural network decoupling controller, thus completing the accused objects, namely decoupling c ontrol four- tank system. From the simulation and experimental results that,based on the rise time of wavelet neural network model predictive control fast,no overshoot, shorter settling time and a strong anti-jamming capability, a control strategy to control the effect of this design made relatively good and after the addition of a fixed source of interference is detected stability and ro-b ustness is relatively good, in order to solve the nonlinear coupling of a feasibility issues.