复杂大化工过程的分布式广义预测控制研究

项目名称： 复杂大化工过程的分布式广义预测控制研究

项目编号： No.61203072

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 自动化学科

项目作者： 李丽娟

作者单位： 南京工业大学

项目金额： 26万元

中文摘要： 预测控制是化工企业稳定生产、提高效益的关键技术，但化工过程的大型化、复杂化使传统的集中式预测控制遇到了计算上的困难，针对复杂大化工过程的分布式预测控制(MPC)研究具有非常重要的应用价值。鉴于目前的分布式MPC理论假定状态空间模型已知，预测控制器不能及时根据对象特性变化作调整，且协调策略会导致优化解偏离Pareto最优解的问题，本课题提出研究大化工过程的分布式广义预测控制(GPC)算法。首先从大化工过程的子系统分解方法入手，基于相关性分析、离散PSO算法和两步辨识法研究分布式CARIMA模型表达形式及在线辨识算法；提出研究具有竞争机制的局部目标加权协调策略，构建包含主控制变量及邻域子系统控制变量的优化问题，采用最优性条件和快速障碍内点优化算法分别求解相应的无约束和有约束优化问题；分析控制系统收敛性和稳定性条件并进行实验验证，初步构建大化工过程分布式GPC基本理论，提高分布式MPC的实用性。

中文关键词： 模型预测控制；大系统；分解；建模；控制策略

英文摘要： Predictive control strategy is a key technology to stabilize processes and increase economic benefit in chemical enterprises. But now it is inefficient for the computational difficulty with the chemical processes becoming large scale and complicated. The research of distributed model predictive control(MPC), aimed at large scale chemical processes, is of great value in application. The fashionable distributed MPC theory is based on known state space models, and the predictive controllers are incapable of regulating with the object fluctuation, and moreover, the optimal solution sometimes deviates from Pareto optimal solution for nowadays cooperative strategy. Hence，the research of distributed generalized predictive control (GPC) algorithm for large scale chemical processes is proposed in the project. Firstly, the decomposition method of large scale chemical processes is explored. Then the CARIMA model formulation and online identification algorithm are studied based on relativity analysis, discrete PSO algorithm and two step identification method. And then the proposed competitive cooperative strategy by weighted local objectives is researched, and optimization question is formulated including master operative variables and these from neighbor sub-systems. Then，corresponding unconstrained and constrained optimiz

英文关键词： Model predictive control；Large scale system；Decomposition；Modeling；Control strategy