>>>> PENALTY: penaltypar(1) penaltypar(2)
all third-level commands in block > OBSERVATION
This command reads parameters of the penalty function used to impose non-equality constraints on model output. The lower and upper bounds of the model output (zbound) are provided through command >>>> CONSTRAINT. Model outputs beyond that range (i.e., if z(p) < zbound- or z(p) > zbound+) are penalized by a power law function:
zâ€™(p,t) = z(p) + a*tb
t = |z(p) – zbound|
a = penaltypar(1)
b = penaltypar(2)
It should be noted that the penalty is not applied to the objective function directly, but to the calculated system response z(p). The values of a and b as well as the target value z* need to be selected accordingly.
This option is useful for determining control parameters that lead to an optimized system behavior under constraints.
>>> ELEMENT: BH__0
>>>> CONSTRAINT: 0.0 0.01
>>>> PENALTY: 1.0D6 2.0
>>>> ZERO DATA