动态求解器
为了触碰到MOVRES约束，需要对操作解决变量的动态动作规划进行检查，看动作是否应该被推迟或加速。
考虑到对于接近稳态的目标，未来的动作并不需要违反MOVRES约束。换句话说，为了确保目标可以实现可能会有一些动作超过MOVRES。
若动作被推迟，其将被添加到动作累积（MOVACC）项。该项用来保留算法中永久推迟的小步伐。作为动作积累，当MOVACC超过MOVRES时将被执行。
一些后果：
•如果用户修改了限制，使得VIND超出界限，小的动作也将被执行。
例如，考虑一个MV控制离散风扇数目的情况。若变量运行时触碰到20风扇的上限，然后你将上限（ULINDM）更改至19.9，则控制器将给出-0.1的动作。应该在培训操作员使用动作解决变量时对其进行处理。
动态动作规划选项（DYNOPT）
有几种DYNOPT选项可用于改变计算出的动态动作规划路径。
不使用随时间变化的ECEs
使用随时间变化的ECEs
不使用随时间变化的ECEs，但使用先进动作规划
使用随时间变化的ECEs和先进动作规划
随时间变化的ECEs（同等关注误差）
当前的DMCplus算法是计算出一个平均动态CV权重。其将被基于CV值的沿时域变化的CV权重所代替。
注意大型控制器要支持该运算可能会比较昂贵。
若要使用此功能，可将DYNOPT值改成1或3。
除了处理随时间变化的ECEs，该算法的修改也解决了不可行的稳态目标问题。
当前的DMCplus动作规划并不具有对不可行稳态目标进行特殊考虑的能力。这样做有几个不良后果—若系统有大的前馈干扰会使动作延迟，并且快速驱使CV进入不可行解决方案。
鉴于其可能只是瞬时现象，我们期望的动作是尽可能长地延迟通往不可行方案。因此，该算法也被设计成能够尽可能长地延迟CV通往不可行方案的动作，通过调整随时间变化的ECEs。
先进动作规划
当前的DMCplus动作规划使用了快速但可能是次优的策略处理MV约束。这已经被更严格的程序所替代。
该选项算法使用二次规划来解决约束动态动作规划，而不是使用具有限幅的最小二乘法，以满足MV约束。
要激活此功能，将DYNOPT值修改成2或3。
附原文：
Dynamic solver
The dynamic move plan for a move resolution variable is inspected to see if moves should be postponed or accelerated, in order to meet the MOVRES constraint.
Consideration is given, regarding the proximity to the steady state target, so that future moves need not violate the MOVRES constraint. In other words, some moves may exceed MOVRES to ensure that the target can be achieved.
When moves are postponed, they are added to the move accumulation (MOVACC) entry.This entry is used to keep the algorithm for perpetually postponing small moves.As moves accumulate, they will get implemented when MOVACC exceeds MOVRES.
Some ramifications:
·If the user has moved the limits, such that the VIND is out of bounds, then a small move can get implemented.
For example, consider an MV that controls a discrete number of fans. If the variable is running against an upper limit of 20 fans, and you change the upper limit (ULINDM) to 19.9, then the controller will make a -0.1 move. This should be handled through training operators on the use of move resolution variables.
Dynamic move plan options (DYNOPT)
There are several DYNOPT options available to change the way the dynamic move plan is calculated:
Do not use time-varying ECEs
Use time-varying ECEs
Do not use time-varying ECEs but use advanced move plan
Use time-varying ECEs and advanced move plan
Time varying ECEs (Equal Concern Errors)
The current DMCplus algorithm calculates an average dynamic CV weighting. This has been replaced by a CV weighting that varies along the horizon, based on the value of the CV.
Note that the calculations required to support this can be expensive for large controllers.
To use this feature, change the value of DYNOPT to 1 or 3.
In addition to addressing time varying ECEs, this algorithm change also addresses a problem with infeasible steady state targets.
The current DMCplus move plan does not have any special consideration for infeasible steady state targets. This has several undesirable effects—including delaying moves if there are large feedforward disturbances, and generally forcing the CV to the infeasible solution as fast as possible.
The desired behavior is to delay the movement to the infeasible solution as long as possible, because it may be due to a transient phenomena. Therefore, this algorithm has also been designed to delay the CV from moving to the infeasible solution, for as long as possible, by making adjustments to the time varying ECEs.
Advanced move plan
The current DMCplus move plan uses a quick, possibly sub-optimal, strategy in handling MV constraints. This has been replaced by a more rigorous procedure.
The option selects an algorithm that uses quadratic programming to solve the constrained dynamic move plan, rather than using a least square algorithm with clipping, to meet MV constraints.
To activate this, change the value of DYNOPT to 2 or 3.
2015.9.25
