模型预测控制算法研究及其在水泥回转窑中的应用

发布时间：2018-05-26 01:06

本文选题：区间特性 + 模型预测控制　；参考：《燕山大学》2015年硕士论文

【摘要】：水泥回转窑是水泥生产的核心设备,我们通过深入研究其机理,选取其有关变量。本课题正是以水泥回转窑系统为研究对象,基于模型预测控制(Model Predictive Control,MPC)理论,通过结合线性区间MPC和非线性MPC,设计集成了MPC高级算法设计控制器。该控制器对实现复杂多变量系统的良好控制具有重要的理论和实际意义。具体研究工作如下:首先,从预热分解、窑内煅烧、窑尾废气处理和熟料冷却四个方面研究了水泥回转窑控制的煅烧工艺,并通过操作员经验和理论研究选取了操作变量以及被控变量。其次,针对线性MPC算法,建立了一种基于区间特性和变量软约束的MPC算法。利用预测输出与区间的位置关系,模仿设定值偏差项的设计原理,构造了区间跟踪偏差项;另外变量软约束项是由经典MPC算法的MV软约束和CV的软约束项构成,这二者共同构成了新的变量软约束项。然后,为保证控制器在约束条件下始终有可行域,同时使CV尽可能的在输出软约束空间内,在每一步滚动优化求解之前,根据输入空间和输出软约束空间的交集情况进行区间约束的可行性判定,当交集解域为空时,引入松弛变量,并给出松弛变量的求解方案,使优化重新可行。最后,研究了一种非线性过程控制策略方法。当非线性较弱时,采用区间MPC算法,以线性代替非线性,简化了算法。当被控对象处于非线性较强的情况下,控制器切换为基于LSSVM模型的非线性MPC算法,通过多步迭代的方法代替了多步预测,省略了建立非线性预测模型的这一过程,只需要训练出被控对象的单步预测模型即可。
[Abstract]:Cement rotary kiln is the core equipment of cement production. Taking cement rotary kiln system as the research object, based on the model predictive control (MPC) theory, a MPC advanced algorithm design controller is designed and integrated by combining linear interval MPC and nonlinear MPCs. The controller has important theoretical and practical significance for the good control of complex multivariable systems. The specific research work is as follows: firstly, the calcination process controlled by cement rotary kiln is studied from four aspects: preheating decomposition, kiln calcinations, waste gas treatment and clinker cooling. Through operator experience and theoretical research, the operating variables and the controlled variables are selected. Secondly, for linear MPC algorithm, a MPC algorithm based on interval characteristic and variable soft constraint is established. Based on the position relation between the predicted output and the interval, the interval tracking deviation term is constructed by imitating the design principle of the set value deviation term, and the variable soft constraint term is composed of the MV soft constraint of the classical MPC algorithm and the soft constraint term of the CV. Both of them constitute a new variable soft constraint. Then, in order to ensure that the controller always has a feasible region under the constraint conditions, and make CV as far as possible in the output soft constraint space, before each step of rolling optimization solution, According to the intersection of the input space and the output soft constraint space, the feasibility of interval constraint is determined. When the solution field of the intersection is space, the relaxation variable is introduced, and the solution scheme of the relaxation variable is given to make the optimization feasible again. Finally, a nonlinear process control strategy is studied. When the nonlinearity is weak, the interval MPC algorithm is used to simplify the algorithm. When the controlled object is in strong nonlinearity, the controller is switched to a nonlinear MPC algorithm based on LSSVM model, and the multistep prediction is replaced by the multi-step iterative method, which omits the process of establishing the nonlinear prediction model. Only a single step prediction model of the controlled object needs to be trained.
【学位授予单位】：燕山大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TQ172.622;TP273

【参考文献】