Optimal Operation of a
Process Industry Plant
by Means of Override Control

Rainer Scheuring
TH Koeln

The optimal operation point of a plant in the process industry refers to the set of conditions under which the plant operates most efficiently while meeting production goals, safety standards, and environmental regulations. Identifying and maintaining the optimal operation point of a process industry plant, while considering process constraints, involves finding a balance between efficiency, productivity, and compliance with operational limitations. This requires a holistic approach that integrates process engineering, control strategies, and continuous optimization. Here’s an overview of how this can be accomplished:

1. Process Constraints

The optimal operation point of a process industry plant is often defined by process constraints because these constraints impose essential limits within which the plant must operate to achieve maximum efficiency, reliability, safety, and profitability. Understanding and managing these constraints is crucial because they directly impact the plant’s performance and ability to meet its objectives. Here’s why process constraints play such a defining:

• Equipment and Technological Limits:
  Physical and technical constraints of equipment, including capacity limits and technology capabilities, dictate what is feasible in the short and long term.
• Safety Considerations:
  Process constraints are critical for maintaining safe operating conditions. Constraints related to pressure, temperature, chemical concentrations, and equipment capacity ensure that the plant operates within safe limits, preventing accidents and equipment failures.
• Regulatory Compliance:
  Environmental, health, and safety regulations impose constraints that plants must adhere to avoid legal penalties and ensure the well-being of workers and the community. These include limits on emissions, waste discharge, and hazardous material handling.
• Quality Assurance:
  Product quality constraints ensure that the final output meets customer specifications and industry standards. Operating outside these constraints can lead to substandard products, leading to customer dissatisfaction and potential financial losses.
• Resource Optimization:
  Constraints on resource availability, such as raw materials, energy, and utilities, necessitate efficient use of available resources to minimize costs and waste, driving the need for optimal operations within these limits.
• Economic Factors:
  Economic constraints, such as budget limits and cost-effectiveness of operations, define the financial feasibility of different operating conditions. The plant's optimal operating point must align with economic goals like maximizing profit margins or minimizing operating costs.
• Interdependent Processes:
  In complex plants, multiple processes are interdependent. Constraints in one process can affect others, requiring a balanced operation to ensure overall plant efficiency and effectiveness.
• Dynamic Market Demands:
  Constraints also reflect market dynamics, where fluctuating demand and supply chain variability can define operational boundaries, requiring flexible and agile adjustment of the optimal operation point.
• Sustainability Goals:
  Environmental sustainability goals impose constraints, such as reducing carbon footprints or increasing resource recycling, influencing how the plant must operate to meet corporate responsibility and sustainability targets.
• Operational Reliability:
  Constraints ensure operational reliability by maintaining conditions that avoid stressing systems beyond their designed capabilities. Ensuring operations within these physical constraints avoids costly downtimes and extends equipment life.

In essence, process constraints are critical because they shape the boundaries within which a plant must operate to achieve its goals while ensuring safety, compliance, efficiency, and reliability. The optimal operation point is therefore a strategic balance that maximizes performance subject to these real-world limitations.

2. Process Modelling and Simulation

• Process Models:
  Mathematical and stationary process simulation models can be used to understand process behaviour under constraint conditions. These models can help predict system behavior in various scenarios.
• Dynamic Simulation:
  Dynamic simulation can be used to develop appropriate control concepts. Furthermore, different operating scenarios and process constraints can be tested using dynamic simulations to observe how changes to one part of the process affect others, providing a more comprehensive understanding of overall system behavior.

3. Optimization Techniques

For large and complex plants, it is sensible to use optimization methods such as linear, nonlinear, quadratic, sequential quadratic programming, or other optimization techniques to find the optimal operating point and optimal operating strategies. The optimization techniques typically use objective functions, such as maximizing yield or minimizing energy use, and are capable of taking process constraints into account.

However, in the case of less complex plants or units, the optimum or required operating point is often determined by external specifications, boundary conditions or the overall system in which the plant or unit is operated. The optimal operating point is often obvious or can be determined using relatively elementary mathematical or engineering calculations.

4. Advanced Process Control

Once the optimal operating point has been determined or is known, it is important that the plant is operated at this point during its operation (Fig. 1).

Figure 1: Optimal operation point.

From an automation and control point of view, it is crucial to assess whether the optimum operating point is located at a process constraint or not. If not, it is often sufficient to automate the plant using standard PID controllers.
However, if the optimal operating point is at a process constraint, this is frequently not sufficient. With standard PID controllers, the system often cannot be reliably operated at the process constraint.

In complex plants where maintaining optimal performance within specified limits is essential model predictive control (MPC) is useful. Model Predictive Control (MPC) is an advanced control strategy used for managing multivariable systems. It employs a dynamic model to predict future system outputs over a given horizon. At each control step, MPC solves an optimization problem to determine control actions that minimize a cost function while considering constraints on inputs and outputs. This method operates on a receding horizon principle, implementing only the first control action and recalculating at each step with updated data.
MPC is a vast field of study that has been the subject of intensive research for several decades. It is not possible to provide a comprehensive overview of MPC here. Generally speaking, however, it can be noted that implementing MPC solutions requires relatively high expertise and frequently involves complex and costly technical implementation and maintenance.

As an alternative, override control offers a much simpler yet similarly effective solution for many applications. A key advantage of override control is that it can be easily implemented on any automation system, even on simple programmable logic controllers (PLC) with limited PLC programming libraries.
In the following sections, override control is thoroughly explained. A comparison of the advantages and disadvantages will subsequently be presented.

5. Override Control

The basic principle of override control is that multiple controllers, typically PI or PID controllers, act on a single actuator. Especially when process constraints need to be taken into account in the control of a plant, override control presents itself as an ideal solution by having one or more controllers ensure compliance with each process constraint. In a typical application, one controller attempts to maximize a process variable, such as flow rate, while one or more other controllers ensure adherence to specified process limits, such as pressures or temperatures.

If several controllers act on one actuator, the question arises as to how the active controller is selected?
One conceivable way would be to select the active controller using discrete or Boolean logic.
However, a different approach is typically chosen: instead of discrete controller switching, the active controller is selected using either a low or high selector (Fig 2). In this way, at any given time, the controller with the greatest need for action is chosen. With suitable low or high selection, the process constraint controller whose limit is currently exceeded is selected. As long as no limit is exceeded, the other controller(s) can adjust the process value toward their setpoints.

Figure 2. Override controller with low selector.

Another important aspect is the coordination of the operating points respectively integral components of the various PI or PID controllers.
Just as a note: If the control deviation respectively the error of a PID controller is equal to zero, i.e. if the setpoint is equal to the process value, then the output variable of the controller is equal to the value that the integral component has at this point in time. This value is the value or operating point that the control loop needs to ensure that the controlled variable is equal to the setpoint.
If the operating points or integral components of two PID controllers do not match at the time of switching, this potentially leads to undesirable jumps in the manipulated variable of the override controller.
This problem can be elegantly solved by using the series or interactive form for implementation instead of the parallel form of the PID controllers (Fig. 3).
In the series implementation, the lag element takes on the function of the integral component.

Figure 3. Override PID controller in series form.

By giving all the lag elements of the various PID controllers of the overide controller the same input value or the same manipulated variable, it is ensured that the PID controllers operate synchronously.
If the various lag elements have similar time constants, which can usually be easily achieved by adjusting the controller gain factors, then the basic structure of the override controller can be further simplified:
Instead of providing each PID sub-controller of the override controller with its own lag element, it is sufficient to use a common lag element for the entire override controller.
Another option is that the proportional or derivative functions of the series implementation can be replaced by other functions. Nonlinear functions and switching functions are particularly suitable.

Example Reciprocating Compressor

Reciprocating compressors are widely used in various industries for gas compression applications. They operate using a piston within a cylinder that moves up and down to compress gas.

Reciprocating compressors can achieve high compression ratios and deliver gas at high pressures, making them suitable for demanding applications. They are capable of compressing a wide range of gases, including toxic, flammable, and non-flammable varieties, with variations in molecular weight and specific gravity. These compressors can be designed to handle a broad array of flow rates and pressures, enhancing their adaptability to different application needs. Moreover, they tend to be more energy-efficient than other types of compressors when operating at full capacity.

In the following example, the control task is to regulate the pressure in the discharge tank to a specified pressure setpoint value despite fluctuating delivery quantities. In doing so, various process limits must be taken into account:

• The suction pressure must be kept within a permissible pressure range. There is a minimum lower limit and a maximum upper limit.
  
• The compressor power consumption is limited.
  
• The outlet temperature of the compressor must not exceed an upper limit.
  

There is also a permissible pressure range for the discharge pressure:

• If the discharge pressure becomes too high, gas must be recirculated.
  
• Too low a discharge pressure is not relevant here, as the capacity of the compressor is sufficient to cover the maximum consumption.
  

In this example, there are three actuators:

• Suction pressure control valve,
  
• Compressor speed, and
  
• Recirculation valve.
  

The recirculation valve can be controlled by a standard PID pressure controller that opens when an upper pressure limit is reached.

The compressor speed can be set using a linear function that maps the suction pressure range to the compressor speed range.


However, the control of the suction pressure control valve can and should be managed by an override controller, which is capable of executing the task of regulating the discharge pressure while taking into account the various process limits (Fig. 4).

Figure 4. Override controller reciprocating compressor.

6. Override Control versus MPC

Choosing between override control and model predictive control (MPC) depends on the specific requirements, complexity, and objectives of the process in question. Each control strategy has its own advantages and is suited for different applications. Here's a comparison to help decide which might be more appropriate for a given scenario:

Override Control - Advantages

Simplicity and Reliability:
Override control is typically simpler to implement and understand. It relies on straightforward logic and mathematical operations to maintain process variables within given limits.
Robustness:
PID control, as basis of override control, has proven in countless applications to be a robust technology that can effectively handle process uncertainties and parameter variations.
Low Computational Demand:
Override control requires minimal computational resources and can be implemented with existing control hardware and software, like PLCs (programmable logic controllers) or DCSs (distributed control systems), making it cost-effective.

Override Control - Limitations

Limited Predictive Capability:
Override control does not predict future states of the process, which might lead to less efficient operation compared to predictive control methods.

Model Predictive Control (MPC) - Advantages

Predictive Capability:
MPC uses a model of the process to predict future behavior, allowing it to minimize the impact of measurable disturbances on the system, resulting in highly optimized control performance as long as the model accurately represents reality.
Handling of Multi-variable Systems:
MPC excels in processes where multiple interrelated process variables need to be controlled simultaneously, making it well-suited for complex industrial applications.
Optimization:
MPC continuously optimizes control signals to keep the process operating at desired levels, considering future constraints and performance objectives.
Flexibility:
MPC can handle various constraints and changes in process conditions dynamically.

Model Predictive Control (MPC) - Limitations

Complexity and Cost:
More complex to implement, requiring detailed process modeling and significant computational resources. Initial setup and tuning can be time-consuming and costly.
Model Dependence:
The performance heavily depends on the accuracy of the process model used. Mis-modeling can lead to degraded performance or instability.

Choosing Between Override Control and MPC

Use Override Control when:

• The primary goal is to maintain process variables at desired setpoints and prevent process limit violations.
  
• The system is relatively simple and doesn’t justify the cost and complexity of implementing MPC.
  
• There are clear, discrete thresholds and process limitations.
  

Use Model Predictive Control when:

• The process involves multiple interacting variables and demands continuous optimization.
  
• Predictive capabilities can significantly enhance process performance and efficiency.
  
• There is sufficient budget and expertise to develop, implement, and maintain a robust model.
  
• The additional financial returns achievable through MPC, minus the costs associated with it, are higher than the financial returns achievable through override control.
  

Ultimately, the choice may not always be between one or the other; in advanced applications, override control can function alongside MPC. This hybrid approach ensures both predictive optimization and the simplicity and robustness of override control, leveraging the strengths of both techniques.

Biography:

Prof. Dr. Rainer Scheuring is a distinguished professor at the Institute of Automation & Industrial IT at TH Koeln, specializing in the modeling, simulation, and control of process industry systems.

Contact Information: rainer.scheuring@th-koeln.de