Real-Time Load-Side Control of Electric Power Systems Thesis by Changhong Zhao In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy CALIFORNIA INSTITUTE OF TECHNOLOGY Pasadena, California 2016 Defended May 6, 2016 ii © 2016 Changhong Zhao ORCID: 0000-0003-0539-8591 All rights reserved iii ACKNOWLEDGEMENTS I owe my deepest gratitude to Professor Steven Low, for his generous support and helpful guidance during my PhD studies. Steven is a great scholar. He always keeps learning, and is enthusiastic to tackle challenging research problems that make real impacts. He is also the best advisor I could ever imagine. He understands what the students do from big picture to technical details, and is ready to provide advice whenever the students need. His professionalism and personality have set up a role model for me. I feel fortune to have worked with him. I am grateful to my colleagues around the world, including Scott Backhaus, Janusz Bialek, Lijun Chen, Misha Chertkov, Florian Dörfler, Na Li, Feng Liu, Enrique Mallada, Kiyoshi Nakayama, Ufuk Topcu, and Yingjun Zhang, for our fruitful discussions and collaborations. Some of them indeed worked as my mentors at different stages of my research. I also want to thank the members of my candidacy and thesis committees: Professors K. Mani Chandy, John Doyle, Victoria Kostina, Richard Murray, P. P. Vaidyanathan, and Adam Wierman. Their comments and suggestions have greatly helped improve my work. Among them, I would like to give special thanks to Adam for his help on scientific writing and presentation. The EE and CMS Departments and RSRG at Caltech provide an open, creative, and friendly atmosphere that is ideal for a student and a researcher like myself. I have spent an enjoyable time with all my friends here: Subhonmesh Bose, Desmond Cai, Niangjun Chen, Krishnamurthy Dvijotham, Lingwen Gan, Minghong Lin, Zhenhua Liu, Qiuyu Peng, Xiaoqi Ren, and more. Moreover, the administrative staff, including Sydney Garstang, Lisa Knox, Maria Lopez, Christine Ortega, and Tanya Owen, have helped us a lot with their hard work. Last but not least, I want to thank my parents, Ms. Cuisong Song and Mr. Junqing Zhao, for their endless love and support from China. It must be a hard time for them to have their only child in another continent that is thousands of miles away. I hope my efforts and achievements would make them proud. iv ABSTRACT Two trends are emerging from modern electric power systems: the growth of re- newable (e.g., solar and wind) generation, and the integration of information tech- nologies and advanced power electronics. The former introduces large, rapid, and random fluctuations in power supply, demand, frequency, and voltage, which be- come a major challenge for real-time operation of power systems. The latter creates a tremendous number of controllable intelligent endpoints such as smart buildings and appliances, electric vehicles, energy storage devices, and power electronic de- vices that can sense, compute, communicate, and actuate. Most of these endpoints are distributed on the load side of power systems, in contrast to traditional control resources such as centralized bulk generators. This thesis focuses on controlling power systems in real time, using these load-side resources. Specifically, it studies two problems. Distributed load-side frequency control: We establish a mathematical framework to design distributed frequency control algorithms for flexible electric loads. In this framework, we formulate a category of optimization problems, called optimal load control (OLC), to incorporate the goals of frequency control, such as balancing power supply and demand, restoring frequency to its nominal value, restoring inter-area power flows, etc., in a way that minimizes total disutility for the loads to participate in frequency control by deviating from their nominal power usage. By exploiting distributed algorithms to solve OLC and analyzing convergence of these algorithms, we design distributed load-side controllers and prove stability of closed-loop power systems governed by these controllers. The general framework above is adapted and applied to different types of power systems described by different models, or to achieve different levels of control goals under different operation scenarios. We first consider a dynamically coherent power system which can be equivalently modeled with a single synchronous machine. Decentralized controllers are derived through a dual gradient algorithm for the OLC problem on this single-machine system, where the dual gradient, which is indeed the total mismatch between load and generation, is calculated at each load by an input estimator using local frequency measurements. Adding neighborhood communication between the loads can improve robustness of the controllers to the noise in local frequency measurements. Through analysis and simulations, we also investigate various practical issues that affect the performance of the proposed v control scheme, such as asynchronous measurements and actuations, robustness to model inaccuracies, and scalability of performance. We then extend our framework to a multi-machine power network. We prove that the dynamics of these machines and the linearized power flows between them can be identified as part of a real-time primal-dual algorithm to solve the OLC problem. Then we implement the other part of the primal-dual algorithm as distributed con- trollers. With this method, we design completely decentralized primary frequency control where every controllable load only measures its local frequency without communicating with any other, and prove global asymptotic stability of the closed- loop system. The controller design and stability analysis are then extended to the scenarios with a more accurate nonlinear power flow model, generator dynamics and control, and secondary frequency control, et cetera. For the case with non- linear power flows, the controllers designed from the OLC framework guarantee local stability. We also design a different version of secondary frequency control through local integral of frequency deviation, which ensures global convergence of the closed-loop system. The OLC problem can be solved by adding distributed averaging filters to the local integral controllers. Two-timescale voltage control: The voltage of a power distribution system must be maintained closely around its nominal value in real time, even in the presence of highly volatile power supply or demand. For this purpose, we jointly control two types of reactive power sources: a capacitor operating at a slow timescale, and a power electronic device, such as a smart inverter or a D-STATCOM, operating at a fast timescale. Their control actions are solved from optimal power flow problems at two timescales. Specifically, the slow-timescale problem is a chance- constrained optimization, which minimizes power loss and regulates the voltage at the current time instant while limiting the probability of future voltage violations due to stochastic changes in power supply or demand. This control framework forms the basis of an optimal sizing problem, which determines the installation capacities of the control devices by minimizing the sum of power loss and capital cost. We develop computationally efficient heuristics to solve the optimal sizing problem and implement real-time control. Numerical experiments show that the proposed sizing and control schemes significantly improve the reliability of voltage control with a moderate increase in cost. vi TABLE OF CONTENTS Acknowledgements . iii Abstract . iv Table of Contents . vi List of Illustrations . viii List of Tables . xii Chapter I: Introduction . 1 1.1 Distributed load-side frequency control . 3 1.2 Two-timescale voltage control . 8 1.3 Thesis outline . 10 Chapter II: Load-Side Frequency Control in Single-Machine Systems . 11 2.1 System model and problem formulation . 11 2.2 Estimating load-generation mismatch . 16 2.3 Decentralized load control: algorithm and convergence . 18 2.4 Asynchronous measurements and actuations . 23 2.5 Simulations . 25 2.6 Conclusion . 32 Appendices . 34 2.A Proof of Proposition 2.1 . 34 2.B Proof of Theorem 2.1 . 34 2.C Proof of Theorem 2.2 . 38 2.D Proof of Theorem 2.3 . 38 2.E Proof of Theorem 2.4 . 38 Chapter III: Load-Side Frequency Control in Multi-Machine Networks . 40 3.1 System model and problem formulation . 41 3.2 Load control and system dynamics as primal-dual algorithm . 47 3.3 Convergence analysis . 51 3.4 Generator and load-side primary control with nonlinear power flow . 58 3.5 Generator and load-side secondary control with nonlinear power flow 66 3.6 Decentralized frequency integral control . 72 3.7 Distributed averaging-based proportional integral control . 77 3.8 Simulations . 79 3.9 Conclusion . 89 Appendices . 92 3.A Frequently used notations . 92 3.B Frequency behavior of the test system . 92 3.C Proof of Lemma 3.1 . 93 3.D Proof of Lemma 3.2 . 94 3.E Proof of Lemma 3.3 . 94 3.F Proof of Lemma 3.4 . 95 vii 3.G Proof of Lemma 3.5 . 96 Chapter IV: Two-Timescale Voltage Control . 97 4.1 Distribution circuit and load models . 97 4.2 Optimal voltage control and device sizing . 100 4.3 Heuristic solution and implementation . 104 4.4 Numerical results . 111 4.5 Conclusion . 116 Bibliography . 117 viii LIST OF ILLUSTRATIONS Number Page 1.1 A four-day real power consumption profile (sampled every five sec- onds) of the high performance computing load studied in this thesis. 9 2.1 The single-machine power system model where ∆g denotes a gener- ation drop and ∆! denotes the frequency deviation. Load i obtains a measured value of the frequency deviation which may differ from ∆! by a stochastic noise ξi. Based on the measured frequency deviation, load i is reduced by ∆di......................... 12 2.2 The home area network (HAN) supports the communication be- tween appliances and smart meters. The neighborhood area network (NAN), which is used in our load control, aids the communication between utilities and smart meters.