How do you evaluate the performance and cost-effectiveness of MPC for industrial control?

1 Define your objectives and metrics

Before you apply MPC to your industrial control problem, it is essential to define your objectives and metrics. What are you aiming to optimize? How do you measure the quality of the control and process? How do you compare different MPC strategies and alternatives? Common objectives and metrics for MPC include setpoint tracking, which evaluates how well the controlled variables follow the desired values, disturbance rejection which measures how well the system responds to external or internal disturbances, constraint satisfaction which monitors how often and how much the system violates physical or operational constraints, energy efficiency which calculates how much energy the system consumes or saves, and economic performance that evaluates how much profit or cost the system generates or reduces.

2 Choose an appropriate model and solver

MPC relies on a mathematical model of the system to predict its future behavior and optimize the control actions, so it is essential to choose an appropriate model and solver. Factors to consider include the model accuracy, complexity, structure, solver speed, and robustness. The model needs to capture the dynamics and uncertainties of the system accurately, while the structure can be linear, nonlinear, discrete, continuous, or hybrid. The solver must be able to find the optimal solution for the MPC problem quickly and handle numerical issues, feasibility, and stability effectively.

3 Test and validate your MPC design

Once you have selected your model and solver, you need to test and validate your MPC design on different scenarios and conditions. Simulation tools, experimental data, or real-time platforms can be used to evaluate the transient response, steady-state performance, robustness and reliability, and scalability and flexibility of your MPC. The evaluation will help determine how the system reacts to changes in setpoints, disturbances, or parameters; how it reaches and maintains the optimal operating point; how it copes with model errors, measurement noise, or control failures; and how it adapts to changes in system size, configuration, or requirements.

4 Compare and benchmark your MPC results

Finally, when comparing and benchmarking your MPC results with other methods or standards, you can use criteria and indicators such as a performance index, a Pareto front, and key performance indicators. These can help you evaluate the performance and cost-effectiveness of MPC for industrial control more systematically and comprehensively. Performance indices may include scalar values summarizing the quality of the control and the process, such as integral of squared errors or integral of economic costs. Pareto fronts are sets of optimal solutions that trade off different objectives, such as energy efficiency and economic performance. Key performance indicators consist of metrics reflecting the specific goals and targets of the system, such as production rate, product quality, or safety margin.

5 Here’s what else to consider

This is a space to share examples, stories, or insights that don’t fit into any of the previous sections. What else would you like to add?