过程块分散控制系统结构设计

Control configuration design of block decentralized control system

化工过程一般为多变量系统,即输入输出变量之间往往存在相互关联,但是在大多数的工业过程中,输入变量和输出变量之间的耦合作用范围是有限的,一般不包括全部范围,因此这些过程可以根据耦合强弱以分块的方式进行控制形成块分散控制系统。块分散控制综合了多变量集中控制高性能和完全分散PID控制易实现性和易维护性的特点,既能降低计算量,又能保证控制质量,因此块分散控制给复杂工业过程的控制提供了极大的便利。块分散控制系统设计的首要工作是进行输入变量和输出变量的块划分和块配对。块相对增益的概念来源于相对增益阵,可以在不设置控制器的条件下对块分散控制结构的闭环特性进行分析并进行块分散控制系统的结构设计。本文引入了块相对增益的定义和性质,介绍了利用块相对增益进行块分散控制系统结构筛选的方法。最后通过实例分析说明了块分散控制系统结构设计的过程,通过与预测控制的对比验证了块分散控制的控制性能不亚于多变量集中控制。

Chemical processes are usually multivariable systems, and there are always interactions between input variables and output variables. But in most industrial processes, the interactions are limited in scope and do not include the full scope of the plant. Thus, these processes are most suitably controlled in a block-wise fashion based on the interaction degree. Block decentralized controllers balance the high performance given by full multivariable centralized controllers and the easier implementation and maintenance associated with fully decentralized PID controllers, which not only reduce computational complexity but also ensure the high control performance, thus block decentralized controllers provide enormous convenience for complex process control. The first step of block decentralized control design is to find the suitable block partitions and block pairings of inputs and outputs. From the concept of relative gain array, block relative gain is an effective tool to estimate closed-loop properties without actual block decentralized controllers and to find the suitable pairings for block decentralized control. In this paper, we introduced the definition and important properties of block relative gain, as well as the block pairing rules of inputs and outputs using block relative gain. Finally, a case study showed the processes of finding the suitable alternative of block decentralized control, and the comparison with model predictive control showed that the control performance of block decentralized control is not less than multivariable centralized control.