SIA OpenIR  > 工业控制网络与系统研究室
预测控制稳态目标计算与动态跟踪控制策略研究
Alternative TitleResearch on Steady-State Target Calculation and Dynamic Tracking Control Strategy for Predictive Control
刘博文1,2
Department工业控制网络与系统研究室
Thesis Advisor邹涛
Keyword预测控制 增量分辨率 LP或QP问题 稳态目标 动态控制
Pages67页
Degree Discipline控制工程
Degree Name硕士
2019-05-17
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Abstract在过程控制领域,预测控制能够有效实现工业过程安全、经济与控制目标优化问题,因此自问世以来便受到了广泛的应用,并成为一种高级的过程控制方法。双层结构预测控制主要包括稳态优化层与动态控制层两个部分,在稳态优化层中通过可行性判定以及软约束调整,结合稳态目标计算策略得到优化的目标点,在动态控制层中,通过滚动优化实现对目标的跟踪控制。然而实际应用中,双层结构预测控制理论还面临着许多问题,本文主要以解决工程问题为导向,旨在缩小双层结构预测控制理论与实际应用的差距。主要研究内容包括:(1)针对工业现场中存在的噪声和不可测扰动造成控制器波动以及大动作阀门无法跟踪控制小动作这两种问题,结合目前工业现场的控制模式,提出了具有输入增量分辨率的多变量预测控制策略,将增量分辨率分成了动态阈值跟静态阈值两种形式。仿真验证动态阈值策略有效保证了控制变量的稳定,避免了频繁的调节执行机构,适用于现场工业控制。静态阈值策略在一定程度上保证控制变量平稳,主要解决现场大动作阀门无法跟踪小动作的问题。(2)针对稳态优化层存在的通用标准化以及计算机求解方面的问题,针对工程应用中参数的配置问题,首先研究了稳态优化层中的两阶段求解策略,将稳态问题转化为了基于优先级的一系列标准LP、QP规划问题,并通过引入辅助变量解决目标函数带绝对值的问题,通过边界条件加极小值解决数值截断误差造成的优化不可行问题,最终形成可计算机求解的通用描述形式。之后从目标函数与约束空间之间的几何关系及代数关系两方面讨论了两阶段稳态优化求解策略中LP及QP解的特性,并结合时间复杂度给出二者在工程应用中的选择原则,进一步讨论了模型失配对LP与QP解的特性的影响,为实际应用中预测控制器稳态解的稳定性分析提供一定依据。(3)针对模型失配情况下双层结构预测控制稳态目标计算存在偏差以及稳态目标计算没有参考实际测量值的问题,基于稳态优化层中稳态目标计算策略的分析,将动态控制层中的预测误差通过数字滤波后作为稳态误差,实时校正稳态目标来确保经济最优性。进一步通过添加稳态步幅约束保证稳态目标的计算参考实际测量值,防止模型失配下稳态目标严重偏离实际值。仿真说明该策略保证被控对象在模型失配的情况下有较好控制效果和性能指标,验证了本策略的有效性。
Other AbstractIn the field of process control, predictive control can effectively realize industrial process safety, economic and control target optimization problems. Therefore, it has been widely used since its inception and has become an advanced process control method. The two-layer structure predictive control mainly includes two parts: the steady-state optimization layer and the dynamic control layer. In the steady-state optimization layer, the feasibility target and the soft constraint adjustment are combined with the steady-state target calculation strategy to obtain the optimized target point in the dynamic control layer. In the tracking control of the target through rolling optimization. However, in practical applications, the two-layer structure predictive control theory still faces many problems. Therefore, this paper is mainly oriented to solving engineering problems. It aims to narrow the gap between the two-layer structure predictive control theory and practical application. The main research contents include: (1) Aiming at the two problems of controller fluctuations caused by noise and unmeasurable disturbance in industrial scene and large control valve unable to track control small action, combined with the current control mode of industrial site, a multivariable predictive control strategy with input incremental resolution is proposed. The incremental resolution is divided into dynamic thresholds and static thresholds. The simulation verification dynamic threshold strategy effectively ensures the stability of the control variables, avoids the frequent adjustment of the actuator, and is suitable for on-site industrial control. The static threshold strategy ensures that the control variables are stable to a certain extent, mainly solving the problem that the large-motion valve on the spot cannot track small movements. (2) Aiming at the general standardization and computer solving problems in the steady-state optimization layer, combined with the parameter configuration problem in engineering application, this paper first studies the two-stage solving strategy in the steady-state optimization layer, and transforms the steady-state problem into priority-based problem. A series of standard LP and QP planning problems, and by introducing auxiliary variables to solve the problem of absolute value of the objective function, solving the optimization infeasibility caused by the numerical truncation error by adding the minimum value of the boundary condition, and finally forming a general description form that can be solved by computer. . Then, the characteristics of LP and QP solutions in the two-stage steady-state optimization strategy are discussed from the geometric relationship between the objective function and the constrained space and the algebraic relationship. The combination of the time complexity is given in engineering application. The selection principle is used to further discuss the influence of the model mismatch on the characteristics of LP and QP solutions, which provides a basis for the stability analysis of the steady state solution of predictive controllers in practical applications. (3) For the model mismatch, the two-layer structure predictive control has a bias in the steady-state target calculation and the steady-state target calculation does not refer to the actual measured value. Based on the analysis of the steady-state target calculation strategy in the steady-state optimization layer, the dynamic control layer will be the prediction error is digitally filtered and used as a steady-state error to correct the steady-state target in real time to ensure economic optimality. Further, by adding a steady-state stride constraint, the calculation of the steady-state target is made to refer to the actual measured value, and the steady-state target under the model mismatch is prevented from seriously deviating from the actual value. The simulation shows that the strategy ensures that the controlled object has better control effect and performance index in the case of model mismatch, which verifies the effectiveness of this strategy.
Language中文
Contribution Rank1
Document Type学位论文
Identifierhttp://ir.sia.cn/handle/173321/25204
Collection工业控制网络与系统研究室
Affiliation1.中国科学院沈阳自动化研究所
2.中国科学院大学
Recommended Citation
GB/T 7714
刘博文. 预测控制稳态目标计算与动态跟踪控制策略研究[D]. 沈阳. 中国科学院沈阳自动化研究所,2019.
Files in This Item:
File Name/Size DocType Version Access License
预测控制稳态目标计算与动态跟踪控制策略研(1257KB)学位论文 开放获取CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[刘博文]'s Articles
Baidu academic
Similar articles in Baidu academic
[刘博文]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[刘博文]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.