A First Course in Predictive Control, Second Edition [Bookshelf]

Abstract

s the demand for individualized products increases, there is a need for advanced control at the supervisory level, where model predictive control (MPC) becomes an established solution in industry. Recent surveys [1], [2] indicate that industry is ready for and open to MPC, given the technology’s maturity and high deployment ability. During sessions with academia and industry at the Third IFAC Conference on Advances in Proportional-Integral-Derivative (PID) Control [3] held in Ghent, Belgium, May 9–11, 2018, several bottleneck problems were noted for research translation and control education, among which two are relevant here: 1) the lack of off-the-shelf tools for MPC and 2) the lack of trained control engineers for the industrial environment. A First Course in Predictive Control delivers solutions to both problems. This is an excellent text that provides numerous MPC algorithms supported by highly relevant, illustrative industry examples. The book opens with a comprehensive chapter about industry’s need for predictive control, reminding readers that MPC originated in business and expanded into academia, where it captured the attention of several seminal researchers. To establish a foundation for the MPC algorithms presented later in the book, Chapter 1 proceeds through numerous, essential concepts and discusses the limitations of PID. Prediction is an essential feature of MPC, and Chapter 2 discusses the properties of model-based control and the requirements for MPC’s use. An overview of several model structures and their advantages and disadvantages guides the reader through the decision-making process behind MPC design. Predictive function control (PFC) has been widely applied in industry across a range of sectors, mainly for single-input, single-output loops. Developed specifically to address industry-related problems, it is an effective solution with minimal implementation demands. Chapter 3 introduces PFC as a simple approach to MPC and contains a practical guide for tuning PFC parameters. Chapter 4 presents the basic predictive-control algorithm, with an emphasis on what is known as generalized predictive control (GPC), and an insightful view of closedloop performance and stability. GPC tuning is tackled in Chapter 5, which contains practical questions, such as when we would use large prediction horizons and when to use a higher control horizon. Those choices can help to strike a balance between cost control and system performance, which is the essence of an industrial-control engineer’s practical vision. Chapter 6 offers an overview of the principal components of dual-mode MPC. To support a better understanding of the theoretical concepts that are presented, examples and illustrations are used to discuss the various advantages and limitations of dual-mode MPC. The ever-ambiguous question of how to tune the weight matrices for performance and control effort is discussed from a practical perspective. Next, Chapters 7 and 8 address the constrained case. Each chapter tackles the GPC and dual-mode MPC formulation with constraint-handling mechanisms. Implementation issues and pitfalls to be avoided are discussed, and effective solutions are given. The book has a set of Matlab code (from MathWorks) and sources to help readers reproduce the illustrations in each chapter. A First Course in Predictive Control can be considered as a textbook, since it includes introductory subchapters to guide lecturers and help filter the sections for teaching from areas outside the course. In summary, this book provides an excellent course in MPC, presented in the form of a textbook, and great support for lecturers. It also offers a substantial starting point for MPC-related research for industrial applications.