Merging Optimization and Control in Power Systems
Description
A novel exploration of distributed control in power systems with insightful discussions of physical and cyber restrictions
In Merging Optimization and Control in Power Systems an accomplished team of engineers deliver a comprehensive introduction to distributed optimal control in power systems. The book re-imagines control design within the framework of cyber-physical systems with restrictions in both the physical and cyber spaces, addressing operational constraints, non-smooth objective functions, rapid power fluctuations caused by renewable generations, partial control coverage, communication delays, and non-identical sampling rates.
This book bridges the gap between optimization and control in two ways. First, optimization-based feedback control is explored. The authors describe feedback controllers which automatically drive system states asymptotically to specific, desired optimal working points. Second, the book discusses feedback-based optimization. Leveraging the philosophy of feedback control, the authors envision the online solving of complicated optimization and control problems of power systems to adapt to time-varying environments.
Readers will also find:
* A thorough argument against the traditional and centralized hierarchy of power system control in favor of the merged approach described in the book
* Comprehensive explorations of the fundamental changes gripping the power system today, including the increasing penetration of renewable and distributed generation, the proliferation of electric vehicles, and increases in load demand
* Data, tables, illustrations, and case studies covering realistic power systems and experiments
* In-depth examinations of physical and cyber restrictions, as well as the robustness and adaptability of the proposed model
Perfect for postgraduate students and researchers with the prerequisite knowledge of power system analysis, operation, and dynamics, convex optimization theory, and control theory, Merging Optimization and Control in Power Systems is an advanced and timely treatment of distributed optimal controller design.
More details
Other editions
Additional editions
Persons
Feng Liu, PhD, is Associate Professor in the Department of Electrical Engineering at Tsinghua University in Beijing, China.
Zhaojian Wang, PhD, is Assistant Professor in the Department of Automation at Shanghai Jiao Tong University in Shanghai, China.
Changhong Zhao, PhD, is Assistant Professor in the Department of Information Engineering at the Chinese University of Hong Kong, Hong Kong SAR, China.
Peng Yang is a PhD Candidate in the Department of Electrical Engineering at Tsinghua University in Beijing, China.
Content
Foreword xv
Preface xvii
Acknowledgments xix
1 Introduction 1
1.1 Traditional Hierarchical Control Structure 2
1.1.1 Hierarchical Frequency Control 2
1.1.1.1 Primary Frequency Control 4
1.1.1.2 Secondary Frequency Control 5
1.1.1.3 Tertiary Frequency Control 5
1.1.2 Hierarchical Voltage Control 5
1.1.2.1 Primary Voltage Control 6
1.1.2.2 Secondary Voltage Control 7
1.1.2.3 Tertiary Voltage Control 7
1.2 Transitions and Challenges 7
1.3 Removing Central Coordinators: Distributed Coordination 8
1.3.1 Distributed Control 11
1.3.2 Distributed Optimization 12
1.4 Merging Optimization and Control 13
1.4.1 Optimization-Guided Control 14
1.4.2 Feedback-Based Optimization 16
1.5 Overview of the Book 17
Bibliography 19
2 Preliminaries 23
2.1 Norm 23
2.1.1 Vector Norm 23
2.1.2 Matrix Norm 24
2.2 Graph Theory 26
2.2.1 Basic Concepts 26
2.2.2 Laplacian Matrix 26
2.3 Convex Optimization 28
2.3.1 Convex Set 28
2.3.1.1 Basic Concepts 28
2.3.1.2 Cone 30
2.3.2 Convex Function 31
2.3.2.1 Basic Concepts 31
2.3.2.2 Jensen's Inequality 35
2.3.3 Convex Programming 35
2.3.4 Duality 36
2.3.5 Saddle Point 39
2.3.6 KKT Conditions 39
2.4 Projection Operator 41
2.4.1 Basic Concepts 41
2.4.2 Projection Operator 42
2.5 Stability Theory 44
2.5.1 Lyapunov Stability 44
2.5.2 Invariance Principle 46
2.5.3 Input-Output Stability 47
2.6 Passivity and Dissipativity Theory 49
2.6.1 Passivity 49
2.6.2 Dissipativity 51
2.7 Power Flow Model 52
2.7.1 Nonlinear Power Flow 53
2.7.1.1 Bus Injection Model (BIM) 53
2.7.1.2 Branch Flow Model (BFM) 54
2.7.2 Linear Power Flow 55
2.7.2.1 DC Power Flow 55
2.7.2.2 Linearized Branch Flow 56
2.8 Power System Dynamics 56
2.8.1 Synchronous Generator Model 57
2.8.2 Inverter Model 58
Bibliography 60
3 Bridging Control and Optimization in Distributed Optimal Frequency Control 63
3.1 Background 64
3.1.1 Motivation 64
3.1.2 Summary 66
3.1.3 Organization 67
3.2 Power System Model 67
3.2.1 Generator Buses 68
3.2.2 Load Buses 69
3.2.3 Branch Flows 70
3.2.4 Dynamic Network Model 72
3.3 Design and Stability of Primary Frequency Control 74
3.3.1 Optimal Load Control 74
3.3.2 Main Results 75
3.3.3 Implications 79
3.4 Convergence Analysis 79
3.5 Case Studies 88
3.5.1 Test System 88
3.5.2 Simulation Results 89
3.6 Conclusion and Notes 92
Bibliography 93
4 Physical Restrictions: Input Saturation in Secondary Frequency Control 97
4.1 Background 98
4.2 Power System Model 100
4.3 Control Design for Per-Node Power Balance 101
4.3.1 Control Goals 102
4.3.2 Decentralized Optimal Controller 103
4.3.3 Design Rationale 105
4.3.3.1 Primal-Dual Algorithms 105
4.3.3.2 Design of Controller (4.6) 105
4.4 Optimality and Uniqueness of Equilibrium 108
4.5 Stability Analysis 112
4.6 Case Studies 120
4.6.1 Test System 120
4.6.2 Simulation Results 122
4.6.2.1 Stability and Optimality 122
4.6.2.2 Dynamic Performance 123
4.6.2.3 Comparison with AGC 124
4.6.2.4 Digital Implementation 124
4.7 Conclusion and Notes 128
Bibliography 131
5 Physical Restrictions: Line Flow Limits in Secondary Frequency Control 135
5.1 Background 136
5.2 Power System Model 137
5.3 Control Design for Network Power Balance 138
5.3.1 Control Goals 139
5.3.2 Distributed Optimal Controller 141
5.3.3 Design Rationale 142
5.3.3.1 Primal-Dual Gradient Algorithms 142
5.3.3.2 Controller Design 143
5.4 Optimality of Equilibrium 144
5.5 Asymptotic Stability 148
5.6 Case Studies 155
5.6.1 Test System 155
5.6.2 Simulation Results 156
5.6.2.1 Stability and Optimality 156
5.6.2.2 Dynamic Performance 158
5.6.2.3 Comparison with AGC 158
5.6.2.4 Congestion Analysis 158
5.6.2.5 Time Delay Analysis 161
5.7 Conclusion and Notes 165
Bibliography 165
6 Physical Restrictions: Nonsmoothness of Objective Functions in Load-Frequency Control 167
6.1 Background 167
6.2 Notations and Preliminaries 169
6.3 Power System Model 170
6.4 Control Design 171
6.4.1 Optimal Load Frequency Control Problem 172
6.4.2 Distributed Controller Design 173
6.5 Optimality and Convergence 176
6.5.1 Optimality 176
6.5.2 Convergence 178
6.6 Case Studies 183
6.6.1 Test System 183
6.6.2 Simulation Results 184
6.7 Conclusion and Notes 187
Bibliography 188
7 Cyber Restrictions: Imperfect Communication in Power Control of Microgrids 191
7.1 Background 192
7.2 Preliminaries and Model 193
7.2.1 Notations and Preliminaries 193
7.2.2 Economic Dispatch Model 194
7.3 Distributed Control Algorithms 195
7.3.1 Synchronous Algorithm 195
7.3.2 Asynchronous Algorithm 196
7.4 Optimality and Convergence Analysis 198
7.4.1 Virtual Global Clock 199
7.4.2 Algorithm Reformulation 200
7.4.3 Optimality of Equilibrium 203
7.4.4 Convergence Analysis 204
7.5 Real-Time Implementation 206
7.5.1 Motivation and Main Idea 206
7.5.2 Real-Time ASDPD 208
7.5.2.1 AC MGs 208
7.5.2.2 DC Microgrids 208
7.5.3 Control Configuration 210
7.5.4 Optimality of the Implementation 211
7.6 Numerical Results 213
7.6.1 Test System 213
7.6.2 Non-identical Sampling Rates 214
7.6.3 Random Time Delays 217
7.6.4 Comparison with the Synchronous Algorithm 217
7.7 Experimental Results 219
7.8 Conclusion and Notes 222
Bibliography 224
8 Cyber Restrictions: Imperfect Communication in Voltage Control of Active Distribution Networks 229
8.1 Background 230
8.2 Preliminaries and System Model 232
8.2.1 Note and Preliminaries 232
8.2.2 System Modeling 233
8.3 Problem Formulation 234
8.4 Asynchronous Voltage Control 235
8.5 Optimality and Convergence 237
8.5.1 Algorithm Reformulation 238
8.5.2 Optimality of Equilibrium 242
8.5.3 Convergence Analysis 243
8.6 Implementation 245
8.6.1 Communication Graph 245
8.6.2 Online Implementation 246
8.7 Case Studies 246
8.7.1 8-Bus Feeder System 247
8.7.2 IEEE 123-Bus Feeder System 250
8.8 Conclusion and Notes 253
Bibliography 254
9 Robustness and Adaptability: Unknown Disturbances in Load-Side Frequency Control 257
9.1 Background 258
9.2 Problem Formulation 259
9.2.1 Power Network 259
9.2.2 Power Imbalance 260
9.2.3 Equivalent Transformation of Power Imbalance 261
9.3 Controller Design 263
9.3.1 Controller for Known P _in j 263
9.3.2 Controller for Time-Varying Power Imbalance 264
9.3.3 Closed-Loop Dynamics 265
9.4 Equilibrium and Stability Analysis 266
9.4.1 Equilibrium 266
9.4.2 Asymptotic Stability 269
9.5 Robustness Analysis 274
9.5.1 Robustness Against Uncertain Parameters 274
9.5.2 Robustness Against Unknown Disturbances 275
9.6 Case Studies 277
9.6.1 System Configuration 277
9.6.2 Self-Generated Data 279
9.6.3 Performance Under Unknown Disturbances 282
9.6.4 Simulation with Real Data 282
9.6.5 Comparison with Existing Control Methods 284
9.7 Conclusion and Notes 286
Bibliography 287
10 Robustness and Adaptability: Partial Control Coverage in Transient Frequency Control 289
10.1 Background 289
10.2 Structure-Preserving Model of Nonlinear Power System Dynamics 291
10.2.1 Power Network 291
10.2.2 Synchronous Generators 292
10.2.3 Dynamics of Voltage Phase Angles 293
10.2.4 Communication Network 294
10.3 Formulation of Optimal Frequency Control 294
10.3.1 Optimal Power-Sharing Among Controllable Generators 294
10.3.2 Equivalent Model With Virtual Load 295
10.4 Control Design 296
10.4.1 Controller for Controllable Generators 296
10.4.2 Active Power Dynamics of Uncontrollable Generators 297
10.4.3 Excitation Voltage Dynamics of Generators 298
10.5 Optimality and Stability 298
10.5.1 Optimality 298
10.5.2 Stability 300
10.6 Implementation With Frequency Measurement 306
10.6.1 Estimating ¿ I Using Frequency Feedback 306
10.6.2 Stability Analysis 307
10.7 Case Studies 310
10.7.1 Test System and Data 310
10.7.2 Performance Under Small Disturbances 312
10.7.2.1 Equilibrium and its Optimality 312
10.7.2.2 Performance of Frequency Dynamics 313
10.7.3 Performance Under Large Disturbances 316
10.7.3.1 Generator Tripping 317
10.7.3.2 Short-Circuit Fault 318
10.8 Conclusion and Notes 321
Bibliography 322
11 Robustness and Adaptability: Heterogeneity in Power Controls of DC Microgrids 325
11.1 Background 325
11.2 Network Model 328
11.3 Optimal Power Flow of DC Networks 329
11.3.1 OPF Model 329
11.3.2 Uniqueness of Optimal Solution 331
11.4 Control Design 334
11.4.1 Distributed Optimization Algorithm 334
11.4.2 Optimality of Equilibrium 335
11.4.3 Convergence Analysis 338
11.5 Implementation 344
11.6 Case Studies 346
11.6.1 Test System and Data 346
11.6.2 Accuracy Analysis 348
11.6.3 Dynamic Performance Verification 348
11.6.4 Performance in Plug-n-play Operations 352
11.7 Conclusion and Notes 353
Bibliography 354
Appendix A Typical Distributed Optimization Algorithms 357
A.1 Consensus-Based Algorithms 357
A.1.1 Consensus Algorithms 358
A.1.2 Cutting-Plane Consensus Algorithm 359
A.2 First-Order Gradient-Based Algorithms 362
A.2.1 Dual Decomposition 363
A.2.2 Alternating Direction Method of Multipliers 366
A.2.3 Primal-Dual Gradient Algorithm 368
A.2.4 Proximal Gradient Method 371
A.3 Second-Order Newton-Based Algorithms 374
A.3.1 Barrier Method 374
A.3.2 Primal-Dual Interior-Point Method 375
A.4 Zeroth-Order Online Algorithms 377
Bibliography 379
Appendix B Optimal Power Flow of Direct Current Networks 385
B. 1 Mathematical Model 385
B.. 1 Formulation 385
B.1. 2 Equivalent Transformation 387
B. 2 Exactness of SOC Relaxation 388
B.2. 1 SOC Relaxation of OPF in DC Networks 388
B.. 2 Assumptions 388
B.2. 3 Exactness of the SOC Relaxation 389
B.2. 4 Topological Independence 396
B.2. 5 Uniqueness of the Optimal Solution 396
B.2. 6 Branch Flow Model 397
B. 3 Case Studies 399
B.3. 1 16-Bus System 399
B.3. 2 Larger-Scale Systems 401
B. 4 Discussion on Line Constraints 402
B.4. 1 OPF with Line Constraints 402
B.4. 2 Exactness Conditions with Line Constraints 403
B.4. 3 Constructing Approximate Optimal Solutions 406
B.4.3. 1 Direct Construction Method 407
B.4.3. 2 Slack Variable Method 408
Bibliography 409
Index 411
1
Introduction
Modern power grids rely on hierarchical control architectures to achieve stable, secure, and economic operation, which involves various kinds of advanced measurement, communication, and control techniques [1, 2]. Under the pressures of global climate change and energy shortage, power systems have been undergoing fundamental changes. The past decade has witnessed the leaping penetration of renewable energy resources and distributed generations, the proliferation of electric vehicles, and the active participation of customers, all of which are devoted to recreating a more reliable, flexible, sustainable, and affordable power grid.
On the generation side, fossil fuels are gradually giving place to environment-friendly renewable generations. By the end of 2020, the total installed capacity of global renewable energies reached 2802.004 GW, including 1332.885 GW of hydropower, 732.41 GW of wind energy, and 716.152 GW of solar energy [3]. The rapid growth of renewable energy shows signs of speeding up in the near future. On the consumption side, many new forms of loads have emerged and started participating in system operation with unprecedented enthusiasm. These include, just to name a few, electric vehicles, active participation of load demand [4, 5], energy storages [6, 7], and microgrids [8].
Despite the tremendous environmental and economical benefits, the ongoing changing trends on both generation and consumption sides are gravely challenging the traditional power system control technologies. Renewable energies such as photovoltaics (PVs) and wind power generations (WTGs) are intrinsically uncertain [9], leading to volatile operating conditions. Besides, most PVs and WTGs are integrated via power electronic devices with low/zero inertia and may operate in various control modes. The load-side diversity and participation also complicate the system control problem, which requires a careful design of the interaction protocol to achieve the smart grid vision.
We have to thoroughly and carefully address all these issues to facilitate the grand ambition of the ongoing system revolution. This extremely challenging task calls for advanced future power system control technologies to handle volatile operating conditions and massive controllable participants. Unfortunately, the existing power system control paradigm that features a centralized hierarchical structure may fail to achieve this goal. Here, we envision a new paradigm that reshapes the hierarchy and merges optimization with control, which may provide a new opportunity to tackle the task. This chapter will first introduce the traditional control methods in power systems and then introduce some state-of-the-art methods.
1.1 Traditional Hierarchical Control Structure
The functional operation of a power system depends on its control systems. As one of the largest and most complicated man-made systems, the modern power system must keep the frequency strictly synchronized and the voltages around their nominal values among thousands of generators and loads spanning over continents and interconnected through tens of thousands of miles of electric wires and cables. Therefore, an appropriate control architecture appears to be highly crucial to a reliable and efficient operation of power systems, if not the most. As a matter of fact, during the past one hundred years, power system control has evolved to be a layered/hierarchical structure encompassing diverse time scales and control objectives, ranging from millisecond to years. Figure 1.1 illustrates the time scales of power system controls with different control objectives. Generally, a slow time-scale layer is more concerned with the economy of operation during a long-time period, while a fast time-scale layer focuses on stability and security during a short-time dynamic process.
The time-scale decomposition and the hierarchical control structure in traditional power systems greatly simplify the control synthesis problem. For example, in most cases, detailed fast time-scale dynamics can be neglected in slow time-scale control problems and vice versa. Such decomposition works well when the time scales of the two layers are significantly different. Even if the difference is less noticeable, e.g. between seconds and minutes, it is still acceptable. Nonetheless, the recent transition of our power system shows that it might be more suitable to combine layers in different time scales, say, merging slow optimization and fast control. This idea sets the first motivation for us to write the book.
In the rest of this section, we shortly introduce the hierarchies of traditional frequency and voltage controls.
1.1.1 Hierarchical Frequency Control
In an alternating current (AC) power system, frequency reflects the active power balance across the overall system. The frequency goes down when generation is less than load and vice versa. Therefore, a power system must adopt frequency control to maintain its frequency within a small neighborhood of the nominal value, such as 50 or 60 Hz.
Figure 1.1 Time-scale decomposition of controls in a traditional power system.
In traditional power systems, most electric power is supplied by large-capacity synchronous generators. The huge rotating inertia of generators can serve as a buffer of kinetic energy to mitigate moderate power imbalance, limiting frequency changes instantaneously. For example, a sudden load demand increase will be naturally supported by extracting the kinetic energy from synchronous generators. Consequently, the frequency will drop. However, the kinetic energy stored in the inertia is quite limited, which is inadequate to cope with large or long-term frequency deviation. Therefore, intentional frequency control becomes a must to maintain system frequency more effectively and flexibly.
In accord with the control hierarchy mentioned above, frequency control includes three layers with respect to three different time scales, i.e. the primary control with a typical time scale in tens of seconds, the secondary control in several minutes, and the tertiary control in several minutes to tens of minutes, as shown in Figure 1.2. The first two layers act in a fast time scale that involves physical dynamics such as excitation control and governor control, while the last layer in a slow time scale involves operational or market dynamics such as economic dispatch (ED) and electricity market clearing. It is obvious from Figure 1.2 that the hierarchy of frequency control heavily relies on a control center.
Figure 1.2 Diagram of hierarchical frequency control.
1.1.1.1 Primary Frequency Control
Primary frequency control is designed to limit frequency deviation within an acceptable range. It is usually fulfilled via automatic governor regulation of generators. Denote by the mechanical power of generator and the frequency deviation from the nominal value at bus . Let the power compensation be with , and send it to change the valve opening of the prime mover, such that the system frequency regains a state of operating equilibrium.
Obviously, the primary frequency control is indeed a proportional feedback control, or droop control. It responds fairly fast to frequency deviation since only local frequency measurement is required. However, it may not restore system frequency to the nominal value due to the proportional control strategy.
In practice, not all generators in the system need to be equipped with a governor control. Those generators, however, usually are competent to provide fast power support. Hence one can still categorize them into primary frequency control when needed.
1.1.1.2 Secondary Frequency Control
Secondary frequency control is designed to eliminate frequency deviation. It is usually fulfilled via automatic generation control (AGC). In a multi-area power system, the area control error (ACE) is a linear combination of the deviations of system frequency and the tie-line powers delivered to or received from its neighboring areas [10]. For the th area, the ACE is defined as
where is a constant that stands for the responsibility of this area in response to the frequency deviation, which is referred to as area frequency response coefficient (AFRC). is the power deviation of the tie-line connecting areas and . As a matter of fact, when the ACEs of all control areas converge to zero, the system frequency restores to the nominal value. Therefore, traditional AGC uses ACEs as the feedback signals to compute the control command that reflects the total required power compensation in the area. The obtained control command is then distributed to individual generators in proportion to their participation factors.
AGC typically adopts a proportional-integral (PI) control to drive ACEs to zero. However, as a power system always works in a time-varying environment, the ACEs do not converge to zero but rather fluctuate around zero. So does the system frequency.
1.1.1.3 Tertiary Frequency Control
Although the secondary frequency control can eliminate the...
