Periodic optimal control, dissipativity and MPC

Lars Grüne, Sébastien GrosEPub 2014

TL;DR

This paper extends the concept of "dissipativity" to periodic systems, a crucial step for proving the stability of economic Model Predictive Control (MPC) when optimal operation involves cycles rather than a fixed equilibrium. By developing new definitions of periodic dissipativity, the authors provide a theoretical framework to guarantee that economic MPC schemes will converge to these optimal periodic behaviors, leading to better system performance.

What Problem Does This Paper Address?

Economic MPC aims to optimize a system's performance directly, rather than just tracking a reference. A key challenge in proving the stability of these schemes is that the optimal operating point might not be a static equilibrium but rather a periodic trajectory. Traditional stability proofs for economic MPC often rely on the concept of "dissipativity," which ensures that the system's energy doesn't grow uncontrollably. However, this concept was primarily developed for steady-state operations. This paper addresses the gap of extending dissipativity to periodic systems, enabling stability guarantees for economic MPC when optimal operation is cyclical.

Methods

The core contribution of this paper is the development of three novel definitions for "periodic dissipativity." These definitions extend the existing notion of dissipativity to handle systems that exhibit periodic behavior.

  1. Periodic Dissipativity with Periodic Storage Function: This definition directly generalizes the concept of dissipativity to a periodic orbit, using a set of storage functions that cycle with the orbit's period.
  2. P-step Strict Dissipativity: This approach considers a "P-step" system, which effectively "unrolls" the periodic dynamics over P time steps. Dissipativity is then checked on this larger, but now static, P-step system.
  3. P-step Strict Dissipativity with Respect to a Set: This is a more generalized version of the P-step dissipativity, allowing for a set of optimal periodic trajectories rather than a single one. It uses a distance metric to the set of periodic orbits.

These definitions are illustrated with examples, showing how they can be used to prove stability for economic MPC schemes operating with periodic optimal trajectories.

Datasets

Not discussed in the paper. The paper focuses on theoretical development and uses illustrative examples with defined dynamics and cost functions rather than real-world datasets.

Key Insights and Ideas

  • Periodic operation can be more optimal: The paper highlights that for many economic systems, steady-state operation might not be the most efficient. Periodic trajectories can yield better overall performance, making the study of periodic optimal control crucial.
  • Dissipativity is key for stability: Just as dissipativity is vital for proving stability in steady-state economic MPC, extending this concept to periodic systems is essential for guaranteeing that MPC controllers will find and converge to these optimal periodic behaviors.
  • Multiple ways to define periodic dissipativity: The authors propose three distinct but related definitions of periodic dissipativity, offering flexibility in analyzing different types of periodic optimal behaviors. This suggests that the "best" way to define dissipativity might depend on the specific characteristics of the periodic orbit.
  • Generalization is necessary: The examples demonstrate that some periodic optimal behaviors might not be captured by simpler definitions of dissipativity, underscoring the need for more generalized approaches like the P-step dissipativity with respect to a set.

Key Results

The paper presents theoretical results that, under the proposed periodic dissipativity conditions, imply:

  • The system is optimally operated at a periodic orbit.
  • Economic MPC schemes with these properties will find the optimal periodic orbit and achieve (approximately) optimal average performance.

The paper illustrates these results with numerical examples (Figures 1, 2) showing the trajectories of MPC schemes applied to systems with periodic optimal behaviors. For instance, Figure 1 shows a trajectory converging to a periodic orbit for Example 3.3, and Figure 2 illustrates the behavior for Example 3.8 where multiple optimal trajectories (periodic and steady-state) exist.

Limitations and Future Work

  • Computational Complexity: While the theoretical framework is established, the practical implementation and computation of these periodic dissipativity conditions, especially for high-dimensional systems or long periods, might be computationally intensive.
  • Generalization to Non-linear Systems: The paper focuses on nonlinear systems, but the complexity of analyzing dissipativity for general nonlinear periodic systems remains a significant challenge.
  • Future Work: The authors suggest that their work provides a foundation for further research into economic MPC for periodic systems. Future directions could include developing more efficient algorithms for checking periodic dissipativity, exploring its application to specific classes of systems (e.g., hybrid systems), and investigating its implications for robustness and performance guarantees in the presence of uncertainties.

Summary updated

Abstract

Recent research has established the importance of dissipativity for proving stability of economic MPC in the case of a steady state. In many cases, though, steady state operation is not economically optimal and periodic operation of the system yields a better performance. In this paper, we propose three different ways of extending the notion of dissipativity for periodic systems and illustrate them with three examples.
BibTeX
@inproceedings{grune2014periodic,
  author = {Lars Grüne and Sébastien Gros},
  title = {{Periodic optimal control, dissipativity and MPC}},
  booktitle = {EPub},
  year = {2014},
}
Figure