在解釋多變量控制前，我們首先需要了解過程單元所需要解決的主要問題。簡單地說，一個典型的過程單元（如圖1所示）是運用原料和公用工程生產一種或多種產品的過程。

Figure1: Typical process unit

? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? 圖1：典型過程單元

當然，過程單元致力于利潤最大化方式生產產品。在許多情況下，利潤最大化對應于單位生產量最大化。而有時候，因市場條件原因限制了產品銷售量，這一情況下，利潤最大化是在保證產品質量的前提下，以原料及公用工程最小化，生產效率最大化的方式生產定量產品。

還有一些情況，需要在增加生產量與其帶來原材料及公用工程成本的增加之間做出權衡。鑒于生產量的增加會使過程單元效率降低，這種情況下需要平衡產品的價值與制造成本。

對任一過程單元，可接受的操作區(qū)域被限定在下列類別的各種約束及限制之內：

·執(zhí)行器限制（例：一個閥門是開啟還是關閉）

·設備限制（例：容器最大工作壓力或溫度）

·操作約束（例：壓縮機喘振限制，塔壓差）

·產品質量約束（例：產品雜質上限）

最簡單的操作方式是操作點始終保持在可接受操作區(qū)域的中心，遠離任何約束。這將允許足夠的時間裕度來響應某些將單元操作點推向不可接受操作區(qū)的干擾。然而，從經濟角度出發(fā)的最佳操作點將總是同時處于幾個約束條件下（見下圖2）。

Figure2: Optimum operating point versus the operator's "Comfort Zone"

? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? 圖2：與操作員“舒適區(qū)”相對的最佳操作點

無論過程單元是否運行在最大生產量，這都是正確的。造成多約束過程單元操作困難有許多原因。其中一個原因是，過程單元控制的一些最佳處理參數將因為各種限制而無法控制。

另一個原因是過程單元操作點接近不可接受操作區(qū)域時，當干擾變量如進料品質變化，暴風雨等出現時，必須迅速進行補償使單元操作保持在可接受操作區(qū)域。

事實上比上述圖2更復雜的是，大多數過程單元是高度耦合的，一個變量（執(zhí)行器）的變化會影響到系統(tǒng)中許多其他變量。

最后，鑒于環(huán)境條件、進料組成及公用工程品質的變化，一天內最佳操作點可能在過程中改變多次。

鑒于以上所述原因，傳統(tǒng)的PID反饋控制是難以勝任多約束條件下的控制過程的。當操作員識別過程中的多變量、強耦合性質時，不能期望他們在監(jiān)控幾十個過程變量的同時每分鐘都對變量進行合理的調整。

這造成的結果是必須給予操作員一定量的“裕度”；也就是說，過程操作點必須離約束一定的距離，以提供給操作員識別及響應干擾進入過程的時間。

另一個問題是需要找到最佳操作點。最佳操作點在一天里會發(fā)生好幾次變化，而且通常這個最佳操作點并不明顯。

事實上，“操作裕度”存在于所有過程單元操作，以及技術原因導致缺乏實時辨識經濟最佳操作點，意味著經濟效益空間的存在。

如果可以創(chuàng)建一個控制方案辨識各控制器執(zhí)行時的實時經濟最佳操作點，將過程單元推向此最佳操作點，并且在存在干擾的情況下使過程單元在此操作點穩(wěn)定，這個經濟效益就可以實現。

這就是DMCplus多變量控制軟件所解決的問題。

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