由于计算中变量工程单位可以有不同的时间单位，所有稳态成本必须在同一时间基础上（例如，$/hr，K$/day）仔细计算。对于稳态优化而言，只有稳态成本的相对大小是非常重要的;在不影响稳态解的情况下它们都可以用相同的数相乘。
正的稳态成本意味着减少操作量可以降低成本，从而增加利润。负稳态成本则相反：为增加利润MV必须尽可能大。
对于只有少数几个MV的简单控制器，通常是其中1~2个MV支配经济性能。在这种情况下，系统有可能为所需成果“弥补”稳态成本。然而，在更复杂的（也更典型）应用中并没有任何捷径。先前描述的技术必须用于获得稳态成本。
LP标准
到目前为止，我们已经讨论了如何从经济角度计算稳态成本。通常情况下，1个MV的DMCplus应用对系统经济性能影响甚少。在这种情况下，无论动作向上或向下，通过惩罚该变量动作都是有利的。这要求将MV作为一个最小动作变量而非成本变量，正如我们迄今所讨论的一样。LP Criterion是用于指定变量被当做最小动作变量还是成本变量类型的一个整定参数。
稳态成本最小动作变量实际上是一个“动作惩罚”，必须是正值。稳态成本（动作惩罚）的大小可以影响其他MV的优化过程能力。如果稳态成本设得较大，移动这一变量会非常“昂贵”，因此它不会移动，即使该动作将使其他MV往一个有利可图的方向移动。
如果稳态成本（动作惩罚）被设成一个小数值，其它MV可以自由地往一个有利可图的方向移动，即使这可能需要最小动作MV移动。
在某些情况下，无论哪种方式设置稳态成本（动作惩罚）都可以应用。获取所希望性能的最好方法是使用DMCplus Simulate测试一系列真实场景，然后将这些模拟结果用于建立稳态参数。
附原文：
All steady-state costs must be carefully calculated on the same time basis (for example, $/hr, K$/day) since the engineering units of the variables in the calculation can have dissimilar time units.As far as the steady-state optimization is concerned, only the relative magnitude of the steady-state costs is important; they can all be multiplied by the same number, without affecting the steady-state solution.
A positive steady-state cost implies that reducing the manipulated variable reduces cost, and therefore increases profit. A negative steady-state cost implies the opposite: the manipulated variable should be increased in order to increase profit.
For simple controllers with only a few manipulated variables, it is often the case that one or two manipulated variables dominate the economics.In this case,it may be possible to "make up" steady-state costs that have the desired effect.However, in more complex (andmore typical) applications there is no shortcut. The technique described previously must be used for obtaining the steady-state costs.
LP criterion
So far, we have discussed how the steady-state costs are calculated from an economic standpoint.Often, a manipulated variable in a DMCplus application has little or no economic impact on the system. In this case, it is advantageous to penalize movement in this variable, whether the movement is up or down.This requires that the manipulated variable be treated as aminimum movement variable rather than a cost variable, as we have discussed so far. The LP Criterion for a manipulated variable is a tuning parameter that specifies whether the variable is to be treated as a minimum movement variable or a cost variable.
The steady-state cost for aminimum movement variable is actually a "move penalty" and must havea positive value.The magnitude of the steady-state cost (move penalty) can affect the ability of the other manipulated variables to optimize the process. If the steady-state cost is set large, moving this variable may be so "costly" that it will not move,even though moving it would allow the other manipulated variables to move in a profitable direction.
If this steady-state cost (move penalty) is set to a small number, the other manipulated variables are free to move in a profitable direction, even though this may require that the Minimum Movement manipulated variable be moved.
In certain situations, either way of setting the steady-state cost (move penalty) can apply. The best way to assure the desired performance is to use DMCplus Simulate to test a series of real world scenarios, then use the results of these simulations to establish to steady-state parameters.
2015.10.10
