Robust and Stochastic Control Methods for Sustainable Engineering:
For a variety of social and economic reasons, designing engineered systems to ensure their operation in a sustainable manner is becoming increasingly important. The focus of the workshop will be on the development and application of control methods to address societal problems related to energy efficiency and pollution reduction in both small- and large-scale systems. One key challenge to achieving improvements in the sustainability of these systems is being able to handle their fundamentally inherent uncertainties. This necessitates the use of robust, stochastic, and adaptive control methods.
The workshop will broadly cover new control methods that have been developed in order to improve the operation of specific systems. Within buildings, variations in occupancy negatively impact the energy characteristics of automation systems; coupling control methods with scheduling and statistical adaptation can reduce average and peak energy usage. Wind turbines can also benefit from such techniques. At the larger scale, the integration of renewable energy into the smart grid may benefit from the use of stochastic and distributed control methods. Similarly, reductions in the pollution and energy consumption of transportation networks are possible by designing infrastructure to better handle the congestion characteristics found in these systems.
Energy-efficient building HVAC using statistical control methods - Anil Aswani (UC Berkeley)
Achieving high efficiency and performance in engineered systems can be challenging because of the difficulty in identifying accurate models. For instance, heating, ventilation, and air-conditioning (HVAC) modulates building environment to ensure occupant comfort. And though HVAC can be described by simple physics-based processes, the impact of building occupants makes it difficult to create models for energy-efficient HVAC control.
This talk describes two new techniques that rigorously combine statistical methods with control engineering, for the purpose of identification, analysis, and control of systems for which models are not well known. The first is a control method called learning-based model predictive control (LBMPC) that allows statistics to improve performance through identification of better system models, while providing theoretical guarantees on the robustness and stability of the closed-loop control. The second is a regression technique that leverages differential geometry to better reduce noise when identifying linear or local-linear models of systems in which measurements have either manifold or collinear structure. The improvements possible with these techniques are illustrated with applications to energy-efficient HVAC systems and high performance control of semi-autonomous systems.
Nash-Stackelberg games in transportation networks: leveraging the power of smartphones for traffic monitoring and management - Alexandre Bayen (UC Berkeley)
The first part of this work investigates the problem of real-time estimation and control of distributed parameters systems in the context of monitoring traffic with smartphones. The recent explosion of smartphones with internet connectivity, GPS and accelerometers is rapidly increasing sensing capabilities for numerous infrastructure systems. The talk will present theoretical results, algorithms and implementations designed to integrate mobile measurements obtained from smartphones into distributed parameter models of traffic. The models considered include Hamilton-Jacobi equations, first order conservation laws and systems of conservation laws. Other techniques developed will be briefly presented as well, relying on ensemble Kalman filtering.
In the second part of the work, we develop a game theoretic framework for studying Stackelberg routing games on parallel networks with horizontal queues, applicable to transportation networks. We ﬁrst introduce a new class of latency functions to model congestion with horizontal queues, then we study, for this class of latency, the Stackelberg routing game: assuming a central authority can incentivize the routes of a subset of the players on a network, and that the remaining players choose their routes selﬁshly, can we compute an optimal route assignment (optimal Stackelberg strategy) that minimizes the total cost? We propose a simple strategy, the Non-Compliant First (NCF) strategy, that can be computed in polynomial time, and we show that it is optimal. We also show that it is robust, in the sense that some perturbations of the NCF strategy are still optimal strategies.
The talk will illustrate the results using a traffic monitoring system launched jointly by UC Berkeley and Nokia, called /Mobile Millennium/, which is operational in Northern California and streams more than 60 million data points a day into traffic models. The talk will also present a new program recently launched in California, called the Connected Corridor program.
Environment-friendly airport operations - Hamsa Balakrishnan (MIT)
Increased surface congestion at major airports results in increased taxi times, as well as the associated fuel consumption and emissions. In this talk, we discuss how more environment-friendly airport operations can be enabled through congestion control techniques, while meeting the practical requirements of the real system. In particular, we describe Pushback Rate Control, a strategy to reduce taxi times and fuel burn by controlling the rate at which flights push back from their gates. The proposed approach addresses two main challenges: the random delay between actuation (at the gate) and the server being controlled (the runway), and the need to develop control strategies that can be implemented in practice by human air traffic controllers.
We model the runway system as a semi-Markov process using surface surveillance data, and use approximate dynamic programming to derive optimal pushback policies that control congestion. Finally, we present the results of the real-world implementation and field-testing of this congestion control protocol at Boston Logan International Airport.
Universal laws and architectures: Foundations for sustainable infrastructures - John Doyle (Caltech)
This talk will focus on progress towards a more “unified” theory for complex networks involving several elements: hard limits on achievable robust performance, the organizing principles that succeed or fail in achieving them (architecture), the resulting high variability data and “robust yet fragile” behavior observed in real systems and case studies, and the processes by which systems evolve. Insights can be drawn from three converging research themes. First, detailed description of components and a growing attention to systems in biology and neuroscience, the organizational principles of organisms and evolution are becoming increasingly apparent. Second, while the components differ, advanced technology’s complexity is now approaching biology’s and there are striking convergences at the level of organization and architecture, and the role of layering, protocols, and feedback control in structuring complex multiscale modularity. Third, new mathematical frameworks for the study of complex networks suggests that this apparent network-level evolutionary convergence within/between biology/technology is not accidental, but follows necessarily from their universal system requirements to be efficient, evolvable, and robust to perturbations in their environment and component parts. Case studies in classical problems in complexity will be drawn from statistical mechanics, turbulence, cell biology, human physiology and medicine, neuroscience, wildfire, earthquakes, economics, the Internet, and smartgrid.
Optimization over graph with application to power systems - Javad Lavaei (Columbia)
The operation of next generation electric grids will likely rely on solving large-scale, dynamic optimization problems involving hundreds of thousands of devices jointly optimizing millions of variables, due in part to the presence of distributed generators, batteries, deferrable loads and curtailable loads. These problems are not only large scale but also non-convex. The non-convexity is imposed by nonlinear physical laws and can introduce inferior local solutions. To address this issue, we derive various theories for both transmission and distribution networks, proving that the non-convexity may be eliminated by exploiting the physics of transmission lines and transformers. We discuss the implementation of our algorithm in a custom solver, which is able to solve a 10,000-bus optimal power flow problem in less than 1 second on a single core machine.
In the second part of the talk, we study to what extent energy-related optimizations can be simplified. To this end, we consider an arbitrary real or complex-valued optimization problem whose variable is a positive semi-definite (PSD) matrix. The objective is to investigate under what conditions this optimization has a low-rank (e.g., rank 1 or 2) solution. The motivation is that a broad class of optimization problems, including polynomial optimization, can be cast as a rank-constrained matrix optimization. To solve the problem, the structure of the optimization is mapped into a generalized weighted graph, where each edge is associated with a real/complex weight set. The notion of "sign definite real/complex set" is introduced, based on which we show that the existence of a low-rank solution can be related to the topology of the graph characterizing the optimization as well as the sign definiteness of its weight sets. As a by-product of this result, several classes of optimizations are shown to be polynomial-time solvable. In particular, when the underlying graph of the optimization is acyclic, bipartite or weakly cyclic, the optimization will not be NP-hard under mild conditions. As an application, we demonstrate that optimization over a power circuit can be mapped into a generalized weighted graph, where each weight set is guaranteed to be sign definite due to the physics of power systems. This result improves upon our previous result on zero duality gap for the optimal power flow problem, and can be readily applied to abstract optimizations (e.g. max cut) as well as optimizations over other physical systems.
On the robustness of cyber-physical systems to security attacks - Bruno Sinopoli (CMU)
Cyber Physical Systems (CPS) refer to the embedding of widespread sensing, computation, communication, and control into physical spaces. Application areas are as diverse as aerospace, chemical processes, civil infrastructure, energy, manufacturing and transportation, most of which are safety-critical. The availability of cheap communication technologies such as the internet makes such infrastructures susceptible to cyber security threats, which may affect national security as some of them, such as the power grid, are vital to the normal operation of our society. Any successful attack may significantly hamper the economy, the environment or may even lead to loss of human life. As a result, security is of primary importance to guarantee safe operation of CPS. We will study the effects of false data injection attacks on control systems. We assume that the control system is monitoring and controlling a linear time-invariant system. The attacker's goal is to destabilize the system by compromising a subset of sensors and sending altered readings to the state estimator. The attacker also wants to guarantee that its action can occur undetected. Under these assumptions, we will give robustness conditions. We will provide several illustrative examples in the context of power systems.
Coordinated stochastic dispatch of the grid - Ram Rajagopal (Stanford)
Dispatching resources to mitigate uncertainties due to increased penetration of renewable generation is a difficult problem as it has to account for physical power flow constraints and information constraints due to market structure. In this talk we describe a dynamic procedure for approximate control that achieves provably good performance in various important grid dispatch problems.
Cyber-secure and resilient control for sustainable critical infrastructures - Karl H. Johansson (KTH)
Safe and reliable operation of cyber-physical systems and critical infrastructures is of major societal importance. These systems need to be engineered in such a way so that they can be continuously monitored, coordinated, and controlled despite a variety of possible cyber-attacks and system disturbances. Unlike other IT systems where cyber-security mainly involves encryption and protection of data, attacks on cyber-physical systems may influence the physical processes through the digital controllers or the communication infrastructure. Therefore focusing on encryption of data alone may not be enough to guarantee the security of the overall system, especially not for its physical component. In order to increase the resilience of these systems, one needs appropriate tools to first understand and then to protect against such attacks. In this talk, we will present some recently developed methods to analyze and design cyber-secure control systems. Motivating applications from the power grid, process industry, and building automation will be discussed in some detail. It will be shown that the state estimator used in the control of the transmission grid can be vulnerable to malicious deception attacks on the measurements resulting in biased estimates, which can have severe consequences for the efficient control of the system. How to formalize such sensor attacks and how to protect against them will be discussed. Vulnerabilities in wireless sensor networks used in control systems in the process industry will also be presented. The presented work has been done in collaboration with several colleagues at KTH.
A distributed control framework for smart grid development - Jinfeng Liu (U. of Alberta) and Panagiotis Christofides (UCLA)
The traditional electrical grid involves large, centralized power plants that feed power over an electro-mechanical grid to end users using one-directional power flows. While the traditional electrical grid has been successful, in recent years there have been numerous calls for the development of "smart" electrical grid by expanding the traditional electrical grid with distributed, medium-scale renewables-based energy generation systems and digital technologies, for example, communications, computing, sensing and automation, to better meet the increasing energy demand and environmental regulations. Incorporated with two-way communication networks, digital devices and distributed optimization and control systems, the so-called "smart grid" is expected to be more reliable, more secure, more energy efficient and more environmentally friendly. One important feature of the smart grid is its capability of integrating distributed energy resources and generation, for example, renewable energy resources, into the electrical grid. Renewable energy resources, like wind and solar-based energy generation systems, are receiving national and worldwide attention owing to the rising rate of consumption of fossil fuels. In addition to the environmental benefits, solar and wind renewable energy generation systems also have a reduced investment risk and an increased energy efficiency. However, integrating renewable energy generation systems with the electrical grid requires addressing key fundamental challenges, for example, variable output, in the operation of intermittent renewable resources like solar- and wind-based energy generation systems. One approach to deal with the variable output of wind and solar energy generation systems is through the use of energy generation systems using both wind and solar energy integrated with loads, the electrical grid and distributed energy storage systems.
In this presentation, we propose a conceptual distributed control framework for electrical grid integrated with distributed renewable energy generation systems in order to enable the development of the so-called "smart electrical grid". First, we introduce the key elements and their interactions in the proposed control architecture and discuss the design of the distributed control systems which are able to coordinate their actions to account for optimization considerations on the system operation. Subsequently, we discuss our results on 1) short-term supervisory control of wind-solar energy generation systems, 2) long-term optimal maintenance and operation of wind-solar generation systems, and 3) distributed supervisory predictive control of distributed wind and solar energy generation systems. In the discussion, detailed simulation models of the wind and solar systems as well as a reverse osmosis water desalination process will be developed. The approach of handling the interactions between these subsystems and the electrical grid will be discussed and extensive simulation results will be shown.
Green scheduling for peak power demand reduction in buidling systems - Rahul Mangharam (UPenn)
Commercial electricity customers are often subject to peak-demand based pricing. In this pricing policy, a customer is charged not only for the total electricity consumption but also for the maximum demand over the billing cycle. The unit price of the demand charge is usually very high to discourage the use of electricity under peak load conditions since they can cause issues such as low quality of service and service disruptions, which affect the reliability of the grid. High peak loads also lead to a higher cost of production and distribution of electricity. Therefore, peaks in electricity demand are inefficient and expensive for both suppliers and customers.
Building systems such as heating, ventilating and air conditioning (HVAC) systems, chillers and boilers usually operate independently of each other and frequently trigger concurrently, resulting in temporally correlated high power demand surges. While there exist several different approaches to balance power consumption by load shifting and load shedding, they operate on coarse grained time scales and do not help in de-correlating energy sinks. In this talk, we will present an approach, named Green Scheduling, for fine-grained scheduling of building systems within an aggregate peak demand envelop while ensuring the custom climate conditions are maintained within the desired ranges. We will provide theoretical results on robust schedulability of affine dynamical systems under constrained disturbances. Scalable periodic schedule design will be presented, as well as scheduling policies based on computation of the maximal robust control invariant set. Finally, several applications in building energy systems will be described, including radiant heating systems and chiller scheduling with limited thermal storage.