Model Predictive Control Algorithm Of Distribution Process

With the rising demand for performance to industrial products, the traditional control methods can not often meet the design requirements of the control system. For example, the polyolefin products are in need of good rigidity and high impact strength in machinery and automobile industries, while in other fields require the properties of lightweight, wear-resistant, and corrosion-resistant to replace metal, and the performance index are transform from the average molecular weight to the molecular weight distribution, so the problem of controlling the distribution process has become a research hotspot. Model predictive control algorithm develops a control strategy of prediction model, rolling optimization and feedback correction, and is an effective control algorithm to deal with multi-variable constraints of complex processes at present for its easy modeling, good performance, strong robustness and ability to deal with constraints. Therefore, this paper launches research on model predict control algorithm of distribution process.The main work and contributions of the thesis are listed as follows:(1)According to the characteristics of distribution process, this paper extend from conventional predictive control algorithm to controlling the distribution process, and propose a model predictive control algorithm of controlling the distribution process based on the integral square error. First, the method using state-space model base on B-splines to predict the shape of the curve in future. Second, using composite trapezoidal rule for discretizing proposition of rolling optimization. Finally, realizing feedback correction with cubic spline interpolation.(2)Because the distribution process is a micro distribution curve controlled by macro control variables, and not arbitrarily shaped curve can realize unbiased trace, this paper propose a offset-free model predict control algorithm of distribution process based on feasibility analysis of steady. The method adds feasibility analysis of steady on existing predictive control algorithm of controlling the distribution process, which using the quadratic programming method to calculate the offset-free curve based on steady state model of distribution process. It proved well performed at simulation experiment.(3)As integral square error just consider partial geometry properties like area, and ignore inner structure of distribution curve, this paper propose a multiresolution predictive control algorithm of distribution process based on the Haar wavelet decomposition. This algorithm makes distribution curve decomposed to a infrastructure framework and serveral layers of deatil by Haar wavelet, and describes the shape of the feature information from whole to part. Then designing predictive controller of distribution process based on multiresolution details, and achieving the natural transition of shape features.