Industrial Model Predictive Control Research Articles

A practically implementable reinforcement learning control approach by leveraging offset-free model predictive control

This work addresses the problem of designing an offset-free implementable reinforcement learning (RL) controller for nonlinear processes. RL-based controllers can update the control policy using observed data obtained online based on the controller-process interactions. This allows to alleviate the regular model maintenance step that is essential in advanced control techniques such as model predictive control (MPC). However, random explorations are required for an RL agent to find optimal state–action regions so as to finally reach the optimal policy. This is not implementable in practical situations due to safety concerns and economic objectives. To address this issue, a pre-training strategy is proposed to provide a secure platform for online implementations of the RL controllers. To this end, an offset-free MPC (representative industrial MPC) optimization problem is leveraged to train the RL agent offline. Having obtained similar performance to the offset-free MPC, the RL agent is utilized for online control to interact with the actual process. The efficacy of the proposed approach to handle nonlinearity and changes in plant operating conditions (due to unmeasured disturbances) are demonstrated through simulations on a chemical reactor example for a pH neutralization process. The results show that the proposed RL controller can significantly improve the oscillatory closed-loop responses, obtained by running the offset-free MPC due to the plant-model mismatch and unmeasured disturbances.

Closed-loop robust steady-state target calculation for model predictive control

Industrial model predictive control (MPC) typically includes a steady-state target calculation (SSTC) layer that provides targets for the dynamic MPC controller to track. This paper develops the first closed-loop robust SSTC method, which explicitly addresses the closed-loop uncertainty, for stable linear systems with time-invariant parametric uncertainty. The problem is initially formulated as a stochastic bi-level optimization problem, where the lower-level optimization problem describes the nominal MPC law. The problem is then transformed into a deterministic mixed-integer second-order cone program for ellipsoidal uncertainty. Case study results show that the proposed method is more robust than the classical nominal SSTC in that it helps reduce cycling of the steady-state targets. In addition, the proposed method is less conservative than a typical open-loop robust SSTC method, in that it drives the system to the steady state more quickly and can push the steady state closer to the operating boundary when desired.

MPC model monitoring and diagnosis for non-square systems

Many industrial model predictive control applications, called non-square systems, have more variables to be controlled than manipulated variables available. At these cases, the control objectives are related to keep the controlled ones within a range, instead in reference value (setpoint). Assessing the model quality of these controllers is fundamental, however, most available MPC assessment methods are setpoint dependent, providing misleading results when these are unavailable. The methods proposed by Botelho et al. (2015, 2016a, 2016b) allow the model assessment of linear MPCs independently from the setpoints. Even though such methodologies have this advantage, there is not a procedure to assess this kind of controller. Therefore, this paper presents a guideline to apply Botelho et al. (2015, 2016a, 2016b) methods in non-square MPCs. The Shell Heavy Oil Process is used as a case study. The results demonstrate that the procedure is capable of estimate the effect of modeling problems and indicate the associated controlled variable, as well as whether the problem is due to a model-plant mismatch or unmeasured disturbance in non-square controllers.

Dynamic Matrix Control and Tuning Parameters Analysis for a DC Motor System Control

Model predictive control (MPC) in system control industry overrides the challenges of conventional controllers in controlling complex systems. However, for efficient control, it is essential to find the best combination of parameter values. In this paper, we present the implementation of a multivariable dynamic matrix control (DMC) algorithm. An industrial system consisting of a DC motor, coupled to a mechanical load, the assembly associated with an electronic speed variator was considered to test the implemented DMC controller. DMC’s tuning parameter analysis on the manipulated inputs and their variations on the controlled outputs was performed. Results guarantee that efficient control was presented.

Predictive control for embedded systems

Recent years have seen a drive to gain the theoretical guarantees and proven practical benefits of model predictive control (MPC) in a broad range of new application areas. The underlying theory of MPC has become well-established over the last decade, and it is now a simple and powerful paradigm for implementing complex control laws. MPC is currently the only technique available for synthesizing controllers that can explicitly ensure constraint satisfaction by design and easily allows for the incorporation of nonlinear dynamics. Contrary to the classic application areas of MPC involving expensive products, slow sampling rates, and massively complex plants, these new systems are characterized by their aggressive demand for fast sample rates, high reliability, low cost, and/or efficient use of energy. This special issue contains seven papers addressing new control and optimization methods, as well as computational software, for implementing predictive control on embedded platforms for high-speed dynamical systems. Hartley et al. 1 presents a field programmable gate array (FPGA)-based model predictive controller (MPC) for two phases of spacecraft rendezvous. The developed interior-point solver is accelerated by moving the main computational component, solving a system of linear equations, onto FPGA hardware. The measured timing indicates a substantial acceleration in computational time. Herceg et al. 2 presents a review of common active set algorithms that have been deployed for the implementation of fast MPC. The extensive theoretical and simulation-based analysis suggests directions for improvements specific to MPC computation. Liniger et al. 3 develops a nonlinear receding horizon-based controller for the racing of small (1:43 scale) RC cars. The nonlinear optimization problems are linearized at each time step and the resulting quadratic programs (QPs) are solved in embedded hardware in milliseconds, resulting in a 50 Hz sampling rate with cars traveling more than 3 m/s. Necoara et al. 4 presents a novel linear model predictive control scheme for use with quadratic programing software that may return a suboptimal or even infeasible solution. The scheme uses an estimate of infeasibility and suboptimality to adaptively tighten constraints and, thereby, ensures recursive satisfaction of constraints and asymptotic stability of the closed-loop system. Holaza et al. 5 develops a novel technique for synthesizing reduced complexity explicit MPC laws, which provide closed-loop stability and recursive satisfaction of constraints. The process involves the automatic generation of a simpler control law over a reduced partition via the solution of a convex optimization problem. Quirynen et al. 6 presents a tutorial of online optimization algorithms and efficient code implementations for embedded nonlinear model predictive control. The ACADO Toolkit for MATLAB is used to explain and showcase how these methods can solve practically relevant problems in only a few tens of microseconds. Kufoalor et al. 7 demonstrates how an existing PC-based industrial MPC solution can be migrated to an embedded platform. The real-time considerations necessary to achieve a functional high-performance predictive controller on a low-cost embedded hardware, with limited computational resources are detailed. We thank the editors of Optimal Control Applications and Methods for hosting this special issue, especially Martin Wells for his invaluable help and guidance.

Aircraft Control System Using Model Predictive Controller

This paper concerns the application of model-based predictive control to the longitudinal and lateral mode of an aircraft in a terrain following task. The predictive control approach was based on a quadratic cost function and a linear state space prediction model with input and state constraints. The optimal control was obtained as the solution of a quadratic programming problem defined over a receding horizon. Closed-loop simulations were carried out by using the linear aircraft model. This project thesis provides a brief overview of Model Predictive Control (MPC).A brief history of industrial model predictive control technology has been presented first followed by a some concepts like the receding horizon, moves etc. which form the basis of the MPC. It follows the Optimization problem which ultimately leads to the description of the Dynamic Matrix Control (DMC).The MPC presented in this report is based on DMC. After this the application summary and the limitations of the existing technology has been discussed and the next generation MPC, with an emphasis on potential business and research opportunities has been reviewed. Finally in the last part we generate Matlab code to implement basic model predictive controller and introduce noise into the model.

Aircraft Control System Using Model Predictive Controller

This paper concerns the application of model-based predictive control to the longitudinal and lateral mode of an aircraft in a terrain following task. The predictive control approach was based on a quadratic cost function and a linear state space prediction model with input and state constraints. The optimal control was obtained as the solution of a quadratic programming problem defined over a receding horizon. Closed-loop simulations were carried out by using the linear aircraft model. This project thesis provides a brief overview of Model Predictive Control (MPC).A brief history of industrial model predictive control technology has been presented first followed by a some concepts like the receding horizon, moves etc. which form the basis of the MPC. It follows the Optimization problem which ultimately leads to the description of the Dynamic Matrix Control (DMC).The MPC presented in this report is based on DMC. After this the application summary and the limitations of the existing technology has been discussed and the next generation MPC, with an emphasis on potential business and research opportunities has been reviewed. Finally in the last part we generate Matlab code to implement basic model predictive controller and introduce noise into the model.

Fast Online Computation of a Model Predictive Controller and Its Application to Fuel Economy–Oriented Adaptive Cruise Control

The recent progress of advanced vehicle control systems presents a great opportunity for the application of model predictive control (MPC) in the automotive industry. However, high computational complexity inherently associated with the receding horizon optimization must be addressed to achieve real-time implementation. This paper presents a generic scale reduction framework to reduce the online computational burden of MPC controllers. A lower dimensional MPC algorithm is formulated by combining an existing “move blocking ” strategy with a “constraint-set compression” strategy, which is proposed to further reduce the problem scale by partially relaxing inequality constraints in the prediction horizon. The closed-loop stability is guaranteed by adding terminal zero-state constraint. The tradeoff between control optimality and computational intensity is achieved by proper design of the blocking and compression matrices. The fast algorithm has been applied on intelligent vehicular longitudinal automation, implemented as a fuel economy-oriented adaptive cruise controller and experimentally evaluated by a series of real-time simulations and field tests. These results indicate that the proposed method significantly improves the computational speed while maintaining satisfactory control optimality without sacrificing the desired performance.

Model Predictive Control in Industry: Challenges and Opportunities

With decades of successful application of model predictive control (MPC) to industrial processes, practitioners are now focused on ease of commissioning, monitoring, and automation of maintenance. Many industries do not necessarily need better algorithms, but rather improved usability of existing technologies to allow a limited workforce of varying expertise to easily commission, use, and maintain these valued applications. Continuous performance monitoring, and automated model reidentification are being used as vendors work to deliver automated adaptive MPC. This paper examines industrial practice and emerging research trends towards providing sustained MPC performance.

Drill-down diagnosis of deficient models in MPC

Model maintenance is the most time-consuming and cost-intensive in industrial model predictive control. In this paper, a drill-down diagnosis algorithm for deficient models of industrial MPC via a model quality index (MQI) is proposed. The CVs with poor models can be detected first by MQI values with all controlled variables. Then, a leave-one-out algorithm is proposed to further diagnose which sub-models are deficient for the CVs with poor model performance. Thus, the effort and cost of model maintenance can be reduced. The application result to the Wood-Berry distillation column process indicates the effectiveness of the proposed assessment method.

A CRUDE UNIT DATASET FOR INDUSTRIAL MPC IDENTIFICATION

A crude unit dataset from an industrial identification test is provided as a benchmark for studying system identification. The identified model was used in an industrial MPC (model predictive control) project. Requirements of industrial identification are outlined. Experiences with several identification methods are mentioned.

ILC strategy for progress improvement of economic performance in industrial model predictive control systems

A novel approach to progress improvement of the economic performance in model predictive control (MPC) systems is developed. The conventional LQG based economic performance design provides an estimation which cannot be done by the controller while the proposed approach can develop the design performance achievable by the controller. Its optimal performance is achieved by solving economic performance design (EPD) problem and optimizing the MPC performance iteratively in contrast to the original EPD which has nonlinear LQG curve relationship. Based on the current operating data from MPC, EPD is transformed into a linear programming problem. With the iterative learning control (ILC) strategy, EPD is solved at each trial to update the tuning parameter and the designed condition; then MPC is conducted in the condition guided by EPD. The ILC strategy is proposed to adjust the tuning parameter based on the sensitivity analysis. The convergence of EPD by the proposed ILC has also been proved. The strategy can be applied to industry processes to keep enhancing the performance and to obtain the achievable optimal EPD. The performance of the proposed method is illustrated via an SISO numerical system as well as an MIMO industry process.

Benefits of optimisation and model predictive control on a fully autogenous mill with variable speed

Autogenous (AG) milling is utilised around the world for particle size reduction. The system exhibits highly non-linear behaviour in addition to being subject to unmeasured variability associated with most ore bodies. Anglo American Platinum aimed at improving online optimisation of the circuit by implementing industrial model predictive control (MPC) to reduce system variability and continuously drive towards the optimal operating point within system constraints.The industrial dynamic matrix controller commissioned on the AG mill with a variable speed drive resulted in a 66% reduction in power and a 40% reduction in load standard deviation. These are the main controlled variables of the mill. The controller also improved the objective function, effective power utilisation, by 11%. This reduction in operated variable variability enabled a test campaign where the mill was controlled at various operating regions in order to establish the conditions conducive to the finest product size at a given mill feed rate.Moving the mill operating region from the benchmarked plant to the optimal grind environment and stabilising the mill at this point with the model predictive controller provided an estimated potential recovery increase of 0.32% (absolute) due to better liberation.

A model predictive control package for undergraduate education

Model predictive control (MPC) techniques are extremely profitable control strategies and are well accepted in the chemical processing industry so it is important that chemical engineering graduates have a fundamental understanding of MPC. This understanding will help them make contributions in industry where these control strategies abound. Without such knowledge, graduates would not understand a major part of the control structure present in modern manufacturing systems and would have difficulty understanding how to modify and improve those chemical manufacturing systems to take advantage of new technology.In this paper we describe a new software package developed and tested by the authors for teaching undergraduates the fundamentals of MPC including its suggested application in the classroom. The package is similar to existing industrial model predictive control packages in that the same steps are required to implement a model predictive controller as follows: model identification, controller configuration, controller simulation and tuning. We describe our experience using the package to introduce MPC to an advanced undergraduate process control II class. The package has also been used to provide an MPC laboratory experience for a graduate class on industrial process control.

Model analysis and performance analysis of two industrial MPCs

This paper presents two case studies on the performance evaluation and model validation of two industrial multivariate model predictive control (MPC) based controllers: (1) a 7-output, 3-input MPC with three measured disturbance variables for controlling a part of kerosene hydrotreating unit (KHU) and (2) a 8-output, 4-input MPC with five measured disturbances for controlling a part of naphtha hydrotreating unit (NHU). The first case study focuses on potential limits to control performance due to constraints and limits set at the time of controller commissioning. The root causes of sub-optimal performance of KHU are successfully isolated. Data from the NHU unit with MPC ‘on’ and with MPC ‘off’ are analyzed to obtain and compare several different measures of multivariate controller performance. Model quality assessment for the two MPCs are performed. A new model index is proposed to have a measure of simulation ability and prediction ability of a model. Closed-loop identification of KHU and closed-loop identification of NHU are conducted using the asymptotic method (ASYM) proposed by Zhu (1998).

Multi-step prediction error approach for controller performance monitoring

Performance monitoring of model predictive control (MPC) systems has received a great interest from both academia and industry. In recent years some novel approaches for multivariate control performance monitoring have been developed without the requirement of process models or interactor matrices. Among them the prediction error approach has been shown promising, but it is based on single-step prediction and may not be compatible with the MPC objective that is based on multi-step prediction. This paper develops a multi-step prediction error approach for performance monitoring of model predictive control systems, and demonstrates its application in a real industrial MPC performance monitoring and diagnosis problem.

A Pragmatic Approach for Assessing the Economic Performance of Model Predictive Control Systems and Its Industrial Application

In this article, an approach for economic performance assessment of model predictive control (MPC) system is presented. The method builds on steady-state economic optimization techniques and uses the linear quadratic Gaussian (LQG) benchmark other than conventional minimum variance control (MVC) to estimate the potential of reduction in variance. The LQG control is a more practical performance benchmark compared to MVC for performance assessment since it considers input variance and output variance, and it thus provides a desired basis for determining the theoretical maximum economic benefit potential arising from variability reduction. Combining the LQG benchmark directly with benefit potential of MPC control system, both the economic benefit and the optimal operation condition can be obtained by solving the economic optimization problem. The proposed algorithm is illustrated by simulated example as well as application to economic performance assessment of an industrial model predictive control system.

Sensitivity analysis for selective constraint and variability tuning in performance assessment of industrial MPC

This paper is concerned with economic performance assessment of industrial model predictive control (MPC) applications for processes with constraints. The interest of this paper is to find sensitive process variables, which are the most contributive to the economic performance of MPC. Optimization algorithms for sensitivity analysis are presented for both constraint and variability tuning of industrial MPC applications. Several new problems related to sensitivity properties of process variables, which arise in the actual MPC economic performance assessment, are addressed. Industrial case studies are included to demonstrate how the proposed sensitivity analysis can be used to provide practical and selective tuning guidelines in industry.

Performance Assessment and Model Validation of Two Industrial MPC Controllers

This paper presents two case studies on the performance evaluation and model validation of two industrial multivariate model predictive control (MPC) based controllers at Suncor Energy Inc., Fort McMurray, Canada: (1) a 7 controlled variable (CV), 3 manipulated variables (MV) kerosene hydrotreating unit (KHU) with three measured disturbance variables that are used for feedforward control; and (2) an 8 CV, 4 MV naphtha hydrotreating unit (NHU) with 5 measured disturbances. The NHU and KHU controllers are implemented on the product stripping distillation towers. The first case study focuses on potential limits to control performance due to constraints and limits set at the time of controller commissioning. The root causes of sub-optimal performance of KHU are successfully isolated. Data from the NHU unit with MPC on and with MPC off are analyzed to obtain and compare several different measures of multivariate controller performance. Model quality assessment for the two MPCs are performed. A new model index is proposed to have a measure of simulation ability and prediction ability of a model. Open-loop identification of KHU and closed-loop identification of NHU are conducted using the asymptotic method (ASYM).

IMPROVED TARGET CALCULATION FOR MODEL PREDICTIVE CONTROL

In industrial Model Predictive Control (MPC) applications, it is common to perform target calculation at each sample instant. The purpose of the target calculation is to translate operational targets supplied by higher level optimization functions into control targets that are feasible in the face of current disturbances. This paper shows that the commonly used target calculation formulation is flawed, and that this can lead to significant economic loss. A method for dealing with the identified problem is proposed.