The model-based control strategy that has been most widely applied in the process industries is model predictive control (MPC). It is a general method that is especially well suited for difficult multi-input, multi-output (MIMO) control problems where there are significant interactions between the manipulated inputs and the controlled outputs. 
However, in engineering practice today, the term predictive control does not designate anymore to only a specific control strategy but a wide range of control algorithms which make explicit use of a process model and a general control approach that has integrated a cost function minimization to continuously obtain the optimization strategy.

Model Predictive Control – more than a control approach

There are many analogies to present MPC as a control approach. The name of “model predictive control” is referring to controlling something based on and with the help of prediction. The prediction is stored and based in the form of a mathematical model. Having the prediction as a possibility can be a precious support from many different aspects, in the terms of the process this would be: safety, stability and optimization. In the terms of the everyday life, having a GPS in your car is a good example – we feel much more in control of a situation and have a solid ground to optimize our itinerary. Similar is when playing a chess game: the more future moves we can predict based on the knowledge of playing that game, the more success we can achieve and easier beat our opponent. 

The same is with the process: the more we know about the process behavior, the more we are able to predict how it will behave in future and the better is our potential to control it to our best benefit.

The key step is to “store” the process knowledge. The more knowledge we store in the forms of the mathematical models – the more we will gain out of it in future. 

There are numerous of vendors who own licenses for MPC programs widely used in industry. The architecture of those tools is mainly based on the controller characteristics:

Predictive – the controller uses internal dynamic models to predict the process outputs response based on past input history. The controller outputs are calculated so as to minimize the difference between the predicted process response and the desired response.
Multivariable – the controller is able to handle the dynamic interactions between variables found in processes with multiple inputs and outputs.
Multi-objective – controllers are mainly able to handle multiple objectives which are a part of optimization strategy, but are differing by the way the hierarchy of the multiple objectives is defined.
Optimizing – the controllers have either linear or non-linear algorithms for solving optimization problems to find the most economical operating point that satisfies the limits of all variables. Those are applied at every controller execution.
Constraints handling – one of the most important problems in process control applications is handling of constraints, so all major MPC controllers on the market have their algorithms of handling the constraint and adjusting the optimization strategy accordingly.

The current widespread interest in MPC techniques was initiated by pioneering research performed by two industrial groups in the 1970s. Shell Oil (Houston, Tex.) reported its Dynamic Matrix Control (DMC) approach in 1979, while a similar technique, marketed as IDCOM was presented by Richalet et al. in 1976 conference and owned by Setpoint.

More details about the history of APC, can be found in the article: Advanced process control: a history overview.

Since then, there have been thousands of applications of these and related MPC techniques in oil refineries and petrochemical plants around the world. Thus, MPC has had a substantial impact and is currently the method of choice for difficult multivariable control problems in these industries. However, relatively few applications have been reported in other process industries, even though MPC is a very general approach that is not limited to a particular industry.

What are MPC strengths and what are still challenges?

Advantages and Disadvantages of MPC Model predictive control offers a number of important advantages in comparison with conventional multiloop PID control:

1. It is a general control strategy for multivariable processes with inequality constraints on input and output variables.
2. It can easily accommodate difficult or unusual dynamic behavior such as large time delays and inverse responses.
3. Because the control calculations are based on optimizing control system performance, MPC can be readily integrated with online optimization strategies to optimize plant performance.
4. The control strategy can be easily updated online to compensate for changes in process conditions, constraints, or performance criteria.

Over the decades, a traditional way of controlling the units has changed significantly. Operators' behavior and focus have changed because units are run more efficiently and in more narrow working area. Also, there are less frequent unit start-ups and shut-downs which don't allow the operators to learn about process behavior outside the normal operating area. Disturbances and emergency situations are also occurring far less frequently and the number of their interventions have also decreased. To be able to run the unit as close to optimal operating conditions, operators get help from a number of tools they must learn how to use. These are all the reasons that their process knowledge had vanished a little over time. MPC addresses this problem successfully because during the analysis and testing of the unit for the purposes of APC development, the process knowledge is transferred and stored into Model Predictive Controller and constantly employed to run the unit optimally.
You can read more about a practical application of MPC and dynamic models which are storing the process knowledge of a typical column in the article Distillation column: minimizing energy requirements.

However, in comparison with conventional multiloop control, MPC applications have some challenges which are important to address during the project execution, such as:

1. The MPC strategy is very different from conventional multiloop control strategies and thus initially unfamiliar to plant personnel. However, software versions are usually very user-friendly and operators can easily learn to use it effectively.
2. The MPC calculations can be relatively complicated e.g., solving a linear programming (LP) or quadratic programming (QP) problem at each sampling instant and some algorithms can be in some occasions a subject of mathematical constraints and show the lack of robustness. 
3. Because empirical models are generally used, they are valid only over the range of conditions considered during the plant tests.

MPC has been widely used and has had considerable impact, there is a broad consensus that its advantages far outweigh its disadvantages.

A key reason why MPC has become a major commercial and technical success is that there are numerous vendors who are licensed to market MPC products and install them on a turnkey basis. Consequently, even medium-sized companies are able to take advantage of this technology with payout times of 3 to 12 months have been widely reported.

©SimulateLive.com