Research And Simulation Of Adaptive Model Algorithmic Control

In order to solve the problem of imprecise model, predictive control appeared and developed. Due to its three basic features — predictive model , rolling optimization and feedback compensation , predictive control has reduced the influence brought by variable structure and parameters of industry object and uncertainty of circumstance effectively. Thus predictive control has been applied to many industries successfully.Predictive control based on nonparametric model which can be easily obtained on the spot, has been put into practice widely. It has many model parameters and needs much calculation. With the advance of computer's speed of calculation and the advent of fast control algorithms, predictive control based on nonparametric model will be applied to more domains.This paper mainly discusses adaptive model algorithmic control based on impulse response model. Present algorithm treats the length of impulse response sequence constant when identifying the model of object. But in fact, the dynamic process of impulse response may become shorter or longer when the parameters of controlled object change. So it is possible that more accurate model of controlled object and better effect of adaptive model algorithmic control are obtained if the length of impulse response sequence changes accordingly. Thus this paper changes adaptive model algorithmic control. To acquire the model of controlled object, both the length of impulse response sequence and its components are identified. Simulation demonstrates improved adaptive model algorithmic control can promote the speed of control system's response.Adaptive model algorithmic control needs much calculation. Especially the calculation of convolution is complicated. So this paper studies the fast algorithms of convolution. Research indicates the FFT algorithm is simple and effective if sequences have a great many components whereas it is more complicated than the direct calculation of convolution if sequences have not so many components. On the other hand, multidimensional fast algorithm of convolution can reduce calculation effectively whether sequences have a great many components or not. So the latter can be applied to adaptive model algorithmic control to reduce calculation and response time.