Improve Model Predictive Controller (MPC) Quickly, Without Intrusive And Time-Consuming Plant Perturbations

Model Predictive Control (MPC) technology, such as: DMC (Dynamic Matrix Control), RMPCT (Robust Model Predictive Control Technology), PredictPro, Connoisseur and other Model Predictive Control (MPC) systems are widely used in chemical, petrochemical, paper, power plant and oil refining industries. A Model Predictive Control (MPC) system can have anywhere from 3 to 50 MVs (Manipulated Variables) and 5 to 100 CVs (Controlled Variables). Therefore, any Model Predictive Controller (MPC) can have anywhere from 10 to 100 dynamic models.

The shapes of the MPC dynamic models are the heart and soul of any Model Predictive Control (MPC) system. With time, Model Predictive Controllers (MPCs) can deteriorate and produce ugly oscillatory behavior, and this is due to:

• Aging equipment, hardware, process, operating and economic changing conditions and process nonlinearities.
• Incorrect design of certain MPC dynamic models: During MPC dynamic model identification stage, only one MV is moved at a time in order to avoid correlation problems. If more than one MV (slave PID setpoint) is changed at the same time, then there is ambiguity and uncertainty regarding the impact of the multiple variables and subsequently dynamic models could be wrong due to correlations. However, when the Model Predictive Control (MPC) is running, the Model Predictive Control (MPC) moves various MVs (slave PID setpoints) simultaneously in order to keep the CVs at their targets. Therefore, SISO (Single-Input-Singe-Output) MPC dynamic model identification can lead to uncertain Model Predictive Controller (MPC) performance.
• Identification of MPC dynamic models based only on the small plant step-tests which can differ from the actual models seen when the Model Predictive Controller (MPC) is ON and makes bigger moves: Most MPC dynamic models are based on making small plant step-tests of about 1-3 % of the current MV values. If these step changes are too large, this may upset the process and the steps cannot be held for too long. This leads to uncertain or even a wrong determination of model gains. When the Model Predictive Controller (MPC) is running, it uses these models that were identified using the small steps to calculate the trajectory for various MVs in closed-loop mode. When a measured disturbance changes, the Model Predictive Control (MPC) calculates compensating moves based on the models that were identified using the small steps. The difference now is that a measured disturbance signal (feedforward in Model Predictive Controller) may have changed by 20 % (not by 1-3 % like during step-tests) and these larger changes can lead to nonlinearities that can cause wrong Model Predictive Controller (MPC) behavior.
• Judicious selection and deletion of certain dynamic MPC models.
• Poor selection or wrong location of certain MV and CV variables inside of Model Predictive Controller (MPC).
• Bad performance of Model Predictive Controllers (MPCs) due to presence of unmeasured disturbances which comprises of a mixture of fast random noise, medium frequency drifts and slow unmeasured disturbances.
• Poor PID tuning of slave PID control loops connected to the Model Predictive Controller (MPC).

Consequently, the root cause of Model Predictive Controller (MPC) problems and deterioration in quality are most definitely due to bad design or due to changes in the dynamic models behavior inside the Model Predictive Controller (MPC). Unfortunately, when MPC models change, or were wrong to begin with, it is a very difficult job to identify which model(s) is/are wrong and to fix them by identifying the correct models. Due to above reasons, Model Predictive Controllers (MPCs) are often turned off with subsequent loss of benefits and profits.

PiControl Solutions LLC has developed COLUMBO, the new technology for Model Predictive Control (MPC) maintenance and improvement, which:

• Improves any Model Predictive Control (MPC) technology, such as: Dynamic Matrix Controller (DMC) from Aspen Tech, Robust Model Predictive Control Technology (RMPCT) from Honeywell, Predict Pro from Emerson.
• Uses open-loop or completely closed-loop data, even data when Model Predictive Controllers (MPCs) were running and making moves to slave PID controllers in cascade mode.
• Works well without any need for intrusive and time-consuming plant step-tests and perturbations, which are required by other Model Predictive Control (MPC) competitors.
• Works well admits fast random noise, medium frequency drifts and slow unmeasured disturbances.
• Detects and isolates the pattern of unmeasured disturbance while identifying true process models.
• Allows fixing all known parameters like time to steady state, dead time or time constant or even process gain based on operator experience/knowledge, engineering calculations, vessel dimensions and vendor data.
• Identifies as many as ten Model Predictive Control (MPC) dynamic models simultaneously.