No abstract

This paper addresses the selection of controlled variables, that is, "what should we control". The concept of self-optimizing control provides a systematic tool for this, and we show how it can be applied to the Tennessee Eastman process, which has a very large number of candidate variables. In this paper, we present a systematic procedure for reducing the number of alternatives. One step is to eliminate variables that, if they had constant setpoints, would result in large losses or infeasibility when there were disturbances (with the remaining degrees of freedom reoptimized). The following controlled variables are recommended for this process: optimally constrained variables, including reactor level (minimum), reactor pressure (maximum), compressor recycle valve (closed), stripper steam valve (closed), and agitator speed (maximum); and unconstrained variables with good self-optimizing properties, including reactor temperature, composition of C in purge, and recycle flow or compressor work. The feasibility of this choice is confirmed by simulations. A common suggestion is to control the composition of inerts. However, this seems to be a poor choice for this process because disturbances or implementation error can cause infeasibility.

Most (if not all) available control theories assume that a control structure is given at the outset. They therefore fail to answer some basic questions that a control engineer regularly meets in practice (Foss 1973): 'Which variables should be controlled, which variables should be measured, which inputs should be manipulated, and which links should be made between them?' These are the questions that plantwide control tries to answer.There are two main approaches to the problem, a mathematically oriented approach (control structure design) and a process oriented approach. Both approaches are reviewed in the paper.We also provide some definitions of terms used within the area of plantwide control.

We consider control structure selection, with emphasis on “what to control”, for a simple plant with a liquid-phase reactor, a distillation column, and recycle of unreacted reactants. Plants of this kind have been studied extensively in the plantwide control literature. Our starting point is a clear definition of the operational objectives, constraints, and degrees of freedom. Active constraints should be controlled to optimize the economic performance. This implies for this case study that the reactor level should be kept at its maximum, which rules out many of the control structures proposed in the literature from being economically attractive. Maximizing the reactor holdup also minimizes the “snowball effect”. The main focus is on the selection of a suitable controlled variable for the remaining unconstrained degree of freedom, where we use the concept of self-optimizing control, which is to search for a constant setpoint strategy with an acceptable economic loss. Both for the case with a given feed rate where the energy costs should be minimized and for the case where the production rate should be maximized, we find that a good controlled variable is the reflux ratio L/F. This applies to single-loop control as well as multivariable model predictive control.

No abstract