在解释多变量控制前，我们首先需要了解过程单元所需要解决的主要问题。简单地说，一个典型的过程单元（如图1所示）是运用原料和公用工程生产一种或多种产品的过程。
图1：典型过程单元
当然，过程单元致力于利润最大化方式生产产品。在许多情况下，利润最大化对应于单位生产量最大化。而有时候，因市场条件原因限制了产品销售量，这一情况下，利润最大化是在保证产品质量的前提下，以原料及公用工程最小化，生产效率最大化的方式生产定量产品。
还有一些情况，需要在增加生产量与其带来原材料及公用工程成本的增加之间做出权衡。鉴于生产量的增加会使过程单元效率降低，这种情况下需要平衡产品的价值与制造成本。
对任一过程单元，可接受的操作区域被限定在下列类别的各种约束及限制之内：
·执行器限制（例：一个阀门是开启还是关闭）
·设备限制（例：容器最大工作压力或温度）
·操作约束（例：压缩机喘振限制，塔压差）
·产品质量约束（例：产品杂质上限）
最简单的操作方式是操作点始终保持在可接受操作区域的中心，远离任何约束。这将允许足够的时间裕度来响应某些将单元操作点推向不可接受操作区的干扰。然而，从经济角度出发的最佳操作点将总是同时处于几个约束条件下（见下图2）。
图2：与操作员“舒适区”相对的最佳操作点
无论过程单元是否运行在最大生产量，这都是正确的。造成多约束过程单元操作困难有许多原因。其中一个原因是，过程单元控制的一些最佳处理参数将因为各种限制而无法控制。
另一个原因是过程单元操作点接近不可接受操作区域时，当干扰变量如进料品质变化，暴风雨等出现时，必须迅速进行补偿使单元操作保持在可接受操作区域。
事实上比上述图2更复杂的是，大多数过程单元是高度耦合的，一个变量（执行器）的变化会影响到系统中许多其他变量。
最后，鉴于环境条件、进料组成及公用工程品质的变化，一天内最佳操作点可能在过程中改变多次。
鉴于以上所述原因，传统的PID反馈控制是难以胜任多约束条件下的控制过程的。当操作员识别过程中的多变量、强耦合性质时，不能期望他们在监控几十个过程变量的同时每分钟都对变量进行合理的调整。
这造成的结果是必须给予操作员一定量的“裕度”；也就是说，过程操作点必须离约束一定的距离，以提供给操作员识别及响应干扰进入过程的时间。
另一个问题是需要找到最佳操作点。最佳操作点在一天里会发生好几次变化，而且通常这个最佳操作点并不明显。
事实上，“操作裕度”存在于所有过程单元操作，以及技术原因导致缺乏实时辨识经济最佳操作点，意味着经济效益空间的存在。
如果可以创建一个控制方案辨识各控制器执行时的实时经济最佳操作点，将过程单元推向此最佳操作点，并且在存在干扰的情况下使过程单元在此操作点稳定，这个经济效益就可以实现。
这就是DMCplus多变量控制软件所解决的问题。
附：原文
A description of Multivariable Control first requires a description of the major issue in a process unit that it solves. Basically, a typical process unit(shown below in Figure 1), takes in raw materials and utilities, and produces one or more products.
The purpose of the process unit is, of course, to produce the products in a manner that maximizes profits.
In many cases, maximum profitability is achieved at maximum unit throughput. In other cases, market conditions dictate that only a given amount of product can be sold. In such cases, maximum profitability is achieved by producing the specified amount of products in the most efficient manner possible, by minimizing use of raw materials and utilities while still maintaining product quality.
Instill other cases, there is a trade off between increased throughput and the cost of raw materials and utilities required to achieve this throughput. This case requires balancing the value of the products with the cost of making them,as the efficiency of the process unit decreases at increased throughputs.
For any process unit, the acceptable operating region is defined by various constraints or limits, which fall into one of the following categories:
·Actuator limits (e.g.,a valve is either open or closed)
·Equipment limits (e.g.,the maximum vessel working pressure or temperature)
·Operational constraints(e.g., a compressor surge limit, a tower differential pressure)
·Product quality constraints (e.g., upper limit on product impurities)
The simplest point from which to operate the unit is in the center of this acceptable operating region, far from any constraints. This allows the maximum amount of time to respond to disturbances that would drive the unit to an unacceptable operating point. However, the optimum operating point from an economic standpoint will always be at several constraints simultaneously (seethe figure below).
This is true whether the unit is run for maximum throughput or not. Operation of the process at multiple constraints is difficult for several reasons. One reason is that some of the best handles for controlling the process will be at their limits, and will not be available for control.
Another reason is that since the unit is operating near the unacceptable region,disturbances such as feed quality changes, rainstorms, etc., must be compensated for promptly in order to keep the unit in the acceptable operating region.
Further complicating the picture is the fact that most process units are highly interactive; a change in one variable (actuator) will affect many other variables in the system.
Finally,the optimum operating point can change several times over the course of a day,as ambient conditions, feed compositions, and utility qualities change.
Traditional PID feedback control is not adequate for controlling a process at several constraints for the reasons described above. And while an operator recognizes the multivariable, interactive nature of the process, that operator cannot be reasonably expected to monitor dozens of process variables and make adjustments every minute.
This results in a certain amount of "wrap"; that is, the process must be operated a certain distance away from the constraints in order to give the operator time to recognize and respond to disturbances entering the process.
Another issue is finding the optimum operating point. The optimum operating point will change several times over the course of the day, and often it is not obvious where this optimum operating point is.
The fact that some amount of "wrap" exists in the operation of all processes, and the lack of knowledge on a minute-by-minute basis about where the actual economic optimum operating point lies, implies that an economic opportunity exists.
If a control scheme could be created to detect this economically optimum operating point at each controller execution, drive the process operation to this point,and operate the process stable at this point in the face of disturbances, this economic potential could be realized.
This is the issue that the DMCplus Multivariable Control Software has been created to solve.
2015.9.7
