| |
MPC Performance Monitoring
Improves Model Predictive Control Performance
ExperTune's PlantTriage®
includes a comprehensive MPC monitoring facility. This
facility not only monitors all the important and well known
statistics but also includes new innovative performance
assessments (high value assessments) to actually detect
and track when a controller is making moves which result in
cycling between constraints. All of the following MPC
assessments are performed automatically for every MPC controller
during every assessment interval. Basic MPC
Assessments
- User specified MVs CVs for each MPC controller
- Model Analysis:
- Model stability analysis in terms of singular value decomposition
- Model interaction analysis in terms of Relative Gain Arrays
- Model analysis in terms of mismatch between the actual
values and predicted values (Prediction Error)
- Constraint analysis
- Percentage of time when MVs are saturated at low or high
limits
- Percentage of time when CV violates its low or high limits
- Percent time that the MV moves are at ROC limits
- Each MPC CV has the following loop assessments performed:
- Variability
- Noise Band
- Target Crossings
- Target Changes
- Average error
- Integrated Absolute Error
- Oscillating
- Oscillating from Field Equipment
- Oscillating from Tuning
- Oscillation from Disturbance
- Variance
- First Period
- Second Period
- Third Period
- Energy from First Period
- Energy from Second Period
- Energy from Third Period
- Each MPC MV has the following loop assessments:
- MV Standard Deviation
- MV Noise Band
High Value MPC Assessments
These innovative performance assessments actually detect
and track when a controller is making moves which result
in cycling between constraints. PlantTriage detects
unnecessary moves to the process, unnecessary alerts to
the operator and reduced profit from the controller, even
one with a high service factor. MPC monitoring from
ExperTune has highly innovative metrics available with our
new monitor and are only available from ExperTune.
- Degrees of Freedom assessment—A unique ExperTune
metric —A
controller that is well designed and configured should have a DOF
Assessment value close to zero, indicating that MVs are free to
control the process. Large values may indicate MV limits that are too
conservative or restricting.
- Controller Entropy—A unique ExperTune metric —
A low gradient in the polytropic LP tableau can very often cause high
entropy. In layman's terms: The controller can be moving around lot,
but no improvements or benefits result. This innovative performance
assessment actually detects and tracks when a controller is making
moves which result in cycling between constraints.
ExperTune has developed this metric which gives a composite indication of
controller movement and resulting benefit.
- Controller optimizer Objective function costs—Monitors the
economic value of each MV.
- Controller optimizer Objective function value—Overall cost function for the
entire MPC controller.
|
