Distributed dissipativity and graph theoretic properties in distributed economic MPC

通过局部控制来控制大型动力系统网络的能力是众多现代工程系统功能的必要前提。这类系统的例子包括智能电网，它使用分布式控制，将可再生能源整合到智能“发电机和负载”网络中；智能工厂，它通过控制运输和制造设备来提高生产能力；或自主移动系统，其中多组移动代理必须协作执行复杂的任务。尽管这些设想的系统存在固有的自然差异，但从事其中任何系统的工程师都面临着一个根本的挑战：动力系统(例如，发电机、生产单元、机器人等)。这个项目提出了一种建设性的方法来解决分布式经济模型预测控制框架内的上述挑战。为此，将开发一种新的模型预测控制方案来解决涉及多个分布式决策者的控制问题，这些决策者都配备了单独的成本函数。我们的目标是开发具有保证系统理论性质的算法方法，以确保在这种分布式和多目标设置中的公平和效率。该项目的主要贡献将是将数值分布式优化算法和分布式模型预测控制方案联系起来。这些结果将对分布式模型预测控制对底层数值优化算法的要求有一个新的理解。本项目由拜罗伊特大学的Lars Grüne教授提出的一个合作项目作为补充。这两个项目将从广泛的基础上解决分布式经济模型预测控制的挑战，并研究从性能分析到数字实现的概念和实际问题。

The ability to control large networks of dynamical systems by local controlactions is an imperative prerequisite for the functionality of numerous modern engineering systems. Examples of such systems are smart grids, which use distributed control to integrate renewable energy sources into a network of intelligent" generators and loads; intelligent factories, which improve the production capabilities by controlling the transportation and manufacturing devices; or autonomous mobile systems, where groups of mobile agents have to perform complex tasks cooperatively. Despite the natural differences inherent to these envisioned systems, engineers working on any of them are all faced with one fundamental challenge: The dynamical systems (e.g., generators, production units, robots, etc.) have to be controlled in a way such that the control actions applied to a single system are consistent with the control actions applied to the other systems.This project proposes a constructive approach to address the above challenges within the distributed economic model predictive control framework. For this a novel model predictive control scheme will bedeveloped to solve control problems involving several distributed decision makers, all equipped with individual cost functions. We aim to develop algorithmic methods with guaranteed systems theoretic properties to ensure fairness and efficiency in such distributed and multi-objective set-ups. The key contribution of this project will be a connection between numerical distributed optimization algorithms and distributed model predictive control schemes. These results will provide a novel understanding of the requirements that distributed model predictive control imposes on the underlying numerical optimization algorithms.The present project is complemented by a partner project, proposed by Prof. Lars Grüne (University of Bayreuth). Together, the two projects will address the challenges of distributed economic model predictive control from a broad basis, and investigate conceptual as well as practical issues, ranging from a performance analysis up to numerical realizations.