Aspen DMCplus多变量控制软件包基于流程强耦合、多变量约束控制技术设计。
这一技术操作于流程管理监控级。典型的执行周期大约在1~4min，DMCplus控制器通常是通过写入调节控制器设定点控制。
以下是DMCplus控制器的特点，并论证了MVC的功能。
DMCplus是多变量控制器
在普通过程单元中，每个独立变量会影响若干个非独立变量。任何需要组合独立变量与非独立变量的控制技术都会遇到困难，因为一个回路偏离设定点需要对独立变量改变时，独立变量的改变会干扰其它回路。
控制器将互相对抗。而DMCplus计算操作变量MV动作时，将会同时考虑PV偏离SP值，以及MV造成的所有相关变量的偏离值。这一动作将考虑系统中所有相互作用的影响，并将所有的影响变量保持在设定点附近。DMCplus控制器“知道”系统是多变量的。
DMCplus是模型预测控制器
工厂数据将被用于建立包含重要相关变量的过程线性动态模型。在开环回路中，模型通过过去一段时间操作变量（MV）和干扰变量（DV）的动作变化预测未来一段时间被控变量（CV）的变化直至稳定。这个稳定的时间被称为过程的稳态时间。
这将允许DMCplus控制器预测未来一段时间CV是否会违反约束，并且控制器可以提前动作以防止CV违反约束。每一控制周期中，预测值将与实际测量值校正以消除模型失配带来的影响。预测模型将允许一些特殊的动态建模（如长死时间及响应方向相反等）。
DMCplus是约束控制器
MV和CV的约束处理都将得到显式表达。迄今为止所有的讨论都提到CV的设定点。然而，DMCplus可以将CV控制在上下限范围之间。事实上，大多数CV并没有一个固定的设定值，但却有一个可接受值的上下限范围。
同时，DMCplus还对MV进行了约束处理。每个MV都有上下限和变化率限制。重要的是未来约束也得到处理。
当DMCplus计划了对一扰动进行补偿，它计算出未来MV需要移动的准确值。需要注意的是，在执行MV动作计算时，计算出的MV动作将不会违反MV上下限值。否则控制器将不会使用计算出的MV动作。DMCplus对这些未来约束做出明确处理，以确保计算动作是可以被执行的。
DMCplus是优化控制器
DMCplus采用稳态求解器寻找控制器执行每步的最经济操作点的稳态优化解决方案。该优化解决方案综合考虑了CV预测稳态值、MV当前值以及原料、产品、公用工程的成本信息。这些值被用来计算满足所有MV和CV约束条件下的最佳稳态工作点。
将这一稳态工作点问题施加于控制过程中，就是一个动态优化方案。这一动态优化方案将最大限度地减少CV偏离所计算出的稳态工作点的误差值，同时防止MV违反约束规则。
DMCplus是严谨的控制器
DMCplus假设所控制系统可以被描述或近似成线性微分方程组。某些其它软件对模型假设处理时将其限制成一阶或二阶加纯滞后过程。但DMCplus没有假设模型的形式；任何形式都是允许的。这将尽可能准确地预测CV未来值。
DMCplus根据提供的成本与约束值，基于稳态优化函数求解出经济最优工作点。动态控制过程将最大限度地减少每个CV当前及未来值的偏差，一直到稳态。DMCplus也为MV未来值提供约束处理。
附原文：
The Aspen DMCplus Multivariable Control software package is based on technology designed for controlling highly interactive, multivariable processes at several constraints simultaneously.
Thistechnology operates at the supervisory control level. Typical executionintervals are on the order of one to four minutes, and the DMCplus controllernormally manipulates the set points of regulatory controllers.
Followingare characteristics of the DMCplus controller, which demonstrate the power ofMVC.
DMCplus is a multivariable controller
Inthe normal process unit, each independent variable affects several dependentvariables. Any control technique that requires pairing independent anddependent variables will have difficulties, since the independent variablemoves required to correct one loop's error will cause a disturbance to otherloops.
Thecontrollers will fight against one another. DMCplus formulates the controlproblem to simultaneously consider all dependent variable errors, or deviationsfrom set point, when solving for the manipulated variable moves. This move planconsiders all the interactions in the system and is consistent with holding alldependent variables at their set points. The DMCplus controller"knows" the system is multivariable.
DMCplus is a model predictive controller
Plantdata is used to build a linear dynamic model of the process that contains allsignificant interactions between variables. The model is then used to predictthe open loop behavior of the controlled variables for a period of time intothe future which is long enough to allow the effects of all past changes in themanipulated and disturbance variables to settle out. This settling time iscalled the steady-state time of the process.
Thisallows DMCplus to anticipate, or predict, future constraint violations, so thatcontrol action can be taken well in advance of the actual violation. Thisfuture prediction is reconciled with actual controlled variable measurements ateach control cycle in order to eliminate model mismatch. This model predictivecapability allows for the modeling of processes with unusual dynamics such aslong deadtimes or inverse responses.
DMCplus is a constrained controller
Constraintson both controlled and manipulated variables are dealt with explicitly. Alldiscussions so far have referred to controlled variables set points. DMCplus,however, can also control those variables between upper and lower limits. Infact, most controlled variables do not have a fixed value set point, but haveupper and lower limits on their range of acceptable values.
DMCplusalso deals with manipulated variable constraints. Each manipulated variable hasupper and lower limits and also has rate of change limits as well. An importantpoint is that future constraints are handled as well.
WhenDMCplus plans how a disturbance is to be compensated for, it calculates currentand future moves in the manipulated variables. When calculating this move plan,care must be taken that the moves do not violate the upper and lower limits. Otherwise,the controller might plan moves that could not be implemented. DMCplusexplicitly handles these future constraints, ensuring that the calculated planis one that can be implemented.
DMCplus is an optimizing controller
DMCplusincorporates a steady-state solver to determine a steady-state optimizationsolution for the most economic operating point at every execution of thecontroller. This optimization solution uses the predicted steady-state valuesof the controlled variables and the current values of the manipulatedvariables, along with cost information on raw materials, products, andutilities.These values are used to calculatethe optimum steady-state operating point that satisfies the limits on allmanipulated and controlled variables.
Thissteady-state operating point is imposed on the control problem, in which adynamic optimization problem is solved. This dynamic optimization problemminimizes controlled variable error away from the calculated steady-stateoperating point, while preventing manipulated variable limit violations.
DMCplus is a rigorous controller
DMCplusis based on the assumption that the system to be controlled can be described orapproximated by a system of linear differential equations. Other technologiesmake assumptions about the form of the model, limiting it to first or secondorder plus deadtime.In DMCplus, no assumptionsare made about the form of the model; any form is allowed. This permits themost accurate prediction of future controlled variable values.
DMCplus provides the steady-state optimization function to solve for an economic optimum operating point, based on the costs and constraint values provided. The dynamic control problem minimizes the current and future error in each control variable all the way to steady state. DMCplus also provides constraint handling for future values of the manipulated variables.
2015.9.8