返回
类型 基础研究 预答辩日期 2017-12-04
开始(开题)日期 2012-09-20 论文结束日期 2017-11-01
地点 礼西2楼会议室 论文选题来源 国家社科规划、基金项目     论文字数 6.3 (万字)
题目 多目标非线性预测控制及其在机炉协调优化控制中的应用研究
主题词 机炉协调系统,非线性预测控制,多模型预测控制,多目标预测控制,干扰主动控制
摘要 在国家节能减排政策的引导下,火电机组朝着大容量、高参数的超超临界机组发展。而由于其非线性、多变量耦合、大惯性、控制受约束、易受燃料变动等不可测扰动影响等特点,使得常规控制方法难以取得良好的控制效果。为此,本文利用超超临界机组协调系统研究先进控制策略,以改进火电机组控制品质。 本文主要包括: (1) 针对超超临界机组协调系统分别设计了基于微分代数方程的模型预测控制器和基于多模型的预测控制器。其中,基于微分代数方程的模型预测控制为了消除扰动的影响实现无偏跟踪,输出方程中被加入累积跟踪误差引入积分作用。通过采用Lagrange插值多项式和Radau配置法对过程变量进行离散化,采用Gaussian求积公式将目标函数离散化,并构造对应的配置方程和连续性方程。经此数学变换将原非线性预测控制问题转化为非线性规划问题进行求解。对于多模型预测控制,在对协调系统模型进行非线性度分析的基础上用多模型对其描述,并设计增量形式的预测控制实现负荷无偏跟踪。通过协调系统大范围变工况及模型失配工况时的控制性能和计算时间比较,为后续章节选择多模型预测控制提供依据。 (2) 为了在协调系统负荷调节动态过程中实现节约燃料量、减少节流损失等经济目标,本文提出两种改进的Utopia点跟踪多目标预测控制算法。算法采用双层结构,改进方案分别在上层求解关于稳态折衷点的拟无穷时域预测控制和模糊控制,利用求得值函数的次优性构造下层Utopia点跟踪多目标预测控制的稳定性约束。改进算法可以保证多目标预测控制算法的可行性,同时提高下层多目标优化的裕度。在此基础上,本文探讨了协调系统的多目标运行模式,仿真表明该算法能够优化协调系统负荷调节的经济性。 (3) 为克服协调系统运行中存在的如燃料变化等不可测因素的影响,本文提出一种基于改进扩张观测器的稳定预测控制。主控制器采用稳定预测控制算法,可以在求解控制作用时显式地考虑约束。改进的扩张状态观测器可以处理超超临界机组协调系统的“直馈”特性,将外部不可测扰动和内部模型不确定性作为集总扰动估计出来,并通过合适的前馈增益补偿掉,以增强预测控制器的扰动抑制能力。通过对协调系统的仿真,验证了算法具有同时处理其强非线性和扰动抑制问题的能力。 (4) 结合协调系统多目标预测控制和干扰抑制控制相关内容,本文提出一种可行的多目标预测控制抗扰动设计。利用扩张状态观测器对集总扰动进行估计,通过稳态折衷点修正计算消除扰动对稳态点的影响。通过对协调系统的进行仿真实验验证所提算法的有效性。
英文题目 Multi-objective nonlinear predictive control and its application in the control of boiler-turbine unit
英文主题词 Boiler-turbine unit,Nonlinear predictive control,Multi-model predictive control,Multi-objective control,Active anti-disturbance control
英文摘要 Following the national policy of energy saving and emission reduction, ultra-supercritical (USC) boiler-turbine units, which operate at higher temperature and pressure level, have been greatly promoted and gradually become the main devices in the power industry. Due to the severe nonlinearity, coupling among variables, large time-delay, tight operating constraints, vulnerability to unmeasurable disturbance such as the variation of fuel, the control performance is limited under the conventional control methods. Given these reasons, advanced control strategy is investigated based on the USC boiler-turbine unit to improve the control performance of the power plants. The main contributions of this thesis are as follows: (1) The differential-algebraic equations (DAEs) based predictive controller and the multi-model based predictive controller are developed for the USC boiler-turbine unit. In the DAEs based predictive control, Lagrange interpolation polynomials and Radau collocation methods are used to discretize the process variables, and the objective function is discretized by Gaussian quadrature. Collocation equations and continuity equations are formulated to convert the optimization problem into a nonlinear programming problem. To eliminate the steady-state deviation, integral action is introduced by adding the accumulated tracking error in the output equations. In the multi-model based predictive control, multi-model for the boiler-turbine unit is established on the basis of nonlinearity analysis. To achieve the offset-free tracking of the set-points, multi-model based predictive control in incremental form is used. By comparison of the control performance and the computing time for these two controllers, multi-model based predictive control is chosen as the basis for the advanced algorithm design in the following chapters. (2) To optimize the economic objectives of the boiler-turbine unit in transient process,such as fuel consumption, throttle loss and so on, two improved Utopia tracking based multi-objective predictive control are proposed, in which the hierarchical structure is utilized. In the upper layer, quasi infinite horizon model predictive control and fuzzy control are designed for the steady-state compromise solution, respectively. The stability constraint for the multi-objective predictive control in the lower layer is established via the suboptimality condition of the value function in the compromise solution tracking problem. In addition, the multi-objective operation mode of the boiler-turbine unit is proposed, in which the economic objectives are optimized while the output power is regulated. Simulation results on a 1000-MW USC boiler-turbine unit demonstrate the e?ectiveness of the proposed approach. (3) To overcome the influence of internal nonlinearity and unknown disturbances simultaneously, an extended state observer (ESO) based stable model predictive control is proposed for the USC boiler-turbine unit. The stable model predictive controller is devised on the multi-model using output cost function for the purpose of wide range load tracking. Input constraints are taken into consideration when the control action is solved. The improved ESO which can estimate plant behavior variations and unknown disturbances regardless of the direct feedthrough characteristic of the unit, is synthesized with the predictive controller to enhance its disturbance rejection property. Closed-loop stability of the overall control system is guaranteed. The proposed method is validated through simulations on a 1000-MW USC boiler-turbine unit. (4) Based on the multi-objective predictive control and anti-disturbance control of the boiler-turbine unit, a feasible disturbance rejection structure for the multi-objective predictive control is proposed. The lumped disturbance is estimated by the extended state observer, and then used for the compromise solution correction calculation. The influence of disturbance can be attenuated from the output channel in steady state. The simulations results on a 1000-MW USC boiler-turbine unit verify the merits of the proposed algorithm.
学术讨论
主办单位时间地点报告人报告主题
能源与环境学院 2014.10.08 礼西2楼会议室 张帆 Review of economic model predictive control methods
能源与环境学院 2015.05.15 礼西2楼会议室 张帆 配置法在动态优化问题中的应用
能源与环境学院 2015.09.22 礼西2楼会议室 张帆 空冷凝气系统特性及研究现状
能源与环境学院 2015.12.16 礼西2楼会议室 张帆 中国自动化大会参会总结
能源与环境学院 2016.03.25 礼西2楼会议室 张帆 白城电厂直接空冷机组调研报告
能源与环境学院 2016.05.09 南高院4楼会议室 张帆 扩展状态观测器在扰动抑制中的应用
能源与环境学院 2016.10.14 南高院4楼会议室 张帆 基于扩增状态观测器的模型预测控制
能源与环境学院 2017.04.14 南高院4楼会议室 张帆 基于多模型的多目标预测控制
     
学术会议
会议名称时间地点本人报告本人报告题目
IEEE Conference on Control Technology and Applications 2017.8.29 Hawaii Extended State Observer Based Fuzzy Model Predictive Control for Ultra-Supercritical Boiler-Turbine Unit
Lehigh University Doctoral Candidates‘ Seminar 2013.10.9 Lehigh University Model Predictive Control of Repetitive Processes using linear Matrix Inequalities
     
代表作
论文名称
Nonlinear model predictive control of ultra-supercritical once through boiler-Turbine unit
Extended State Observer Based Fuzzy Model Predictive Control for Ultra-Supercritical Boiler-Turbine
Fuzzy disturbance rejection predictive control of ultra-supercritical once-through boiler-turbine un
 
答辩委员会组成信息
姓名职称导师类别工作单位是否主席备注
吕剑虹 正高 教授 博导 东南大学
王培红 正高 教授 博导 东南大学
李益国 正高 教授 博导 东南大学
李东海 副高 副教授 博导 清华大学
刘志远 正高 教授 硕导 南京工程学院
      
答辩秘书信息
姓名职称工作单位备注
孙立 其他 讲师 东南大学