Wide-Area Monitoring and Control of Power Systems using Real-Time Hardware-in-the-Loop Simulations

Wide-Area Monitoring and Control of Power Systems using Real-Time Hardware-in-the-Loop Simulations Matthew Weiss Thesis advisor: Dr. Aranya Chakrabortty 7/28/2016 1

Power grids are envisioned to be come green and smart in the coming decades. PMU measurements and PMU technology are becoming much more common. Relatively little effort has been made to explore how synchrophasors can be used for automatic feedback control over a wide geographic area. Local PSS and AVR control commonly in use today. In this research, Introduction A breakdown of methodology used to create a functional and accurate reduced-order model is presented. The model is validated and contingencies regarding renewable energy are explored and a major problem identified. Two wide-area controller designs are presented using PMU measurements. A hardware-in-the-loop test-bed is presented for implementation of these control schema in a real-world setting for performance validation. It is concluded that wide-area control schema presented here are successful in 2 stabilizing otherwise unstable power-system conditions!

Thesis Outline Identification of Power System Models using Synchrophasors Impacts of Wind Penetration Wide-Area SVC Control Design Wide-Area PSS Design using LQR Hardware-in-the-Loop Implementation Wide-Area Control using Cloud Computing 3

Identification of Power System How can synchrophasor measurements be used to construct a reliable dynamicequivalent model? How does the implemented model react to different types of contingencies? 4

The Western Electricity Coordinating Council (WECC) is a large power system on the west coast of United States. The WECC 500 kv power system is divided into five coherent generation areas interconnected by long transmission lines. This leads to the emergence of slow inter-area power oscillations in the range of 0.1 Hz to 1 Hz. WECC and its Geography 5

Reduced-Order Topology A reduced-order equivalent of the WECC 500 kv system can be constructed. A pilot bus is selected from each area based on the following criteria: The bus must have a PMU installed All generators within that area must lie electrically behind this bus. The area behind the pilot bus is represented by an aggregated synchronous generator (ASG) Five-machine equivalent of WECC 6

Aggregate Machines The ASG is modeled as a second-order damped oscillator described by the swing equations: Each pilot bus is connected to adjacent pilot buses through long transmission lines. In the steady state, each ASG is represented in the network by its Thevenin equivalent. Estimation of all system parameters will be done using PMU data. 7

8 PMU Data Modes

Model Parameter Calculations From this data it is possible to derive : Tie line impedances Inter-area impedances Inertias Damping values 9

Model Validation RSCAD software was used to realize a model of WECC. Real-Time-Digital-Simulators (RTDS) was used to run RSCAD models in real-time with a 50 micro-second time-step. Real-time application will be of utmost importance later in this research 10

Calculated Values Intra-area Impedance Inter-area Impedance Machine Inertia Machine Damping Easily Derived Data Values Pre-fault Inter-area Voltage Phase Angle Post-fault Inter-area Voltage Phase Angle Pre-fault Voltage Post-fault Voltage Needed but not Provided Values 11 Machine Power Rating Machine Power Generation Load Placement in System Effective Shunt Capacitance Model Parameters

Station 3 Ended up Negative in Least Squares Calculations. Real bus contains significant capacitance Real bus represents a very heavy load Without capacitance, voltage sag occurs In the model, it wasn t possible to use this value. Capacitor added at Station 3 Value tuned such that bus voltage was correct Intra-area impedance substituted with another stations value Line Reactance 12

Bus Voltage Tuning Voltages only of secondary interest System voltage may vary locally Voltage has little effect on overall power flow Just needs to be close, defined as within 2% error Voltages tuned by varying the PV bus voltage inside the machines. 13

Angle (degrees) Fault Power Flow Matching Most pilot buses exhibit a large, instantaneous phase angle and power flow change. Station 1 - Station 2 is the exception Station 4 - Station 5 is very large. Almost 25 degrees. Station 5 is a loss of generation fault location in the real world power system This can t be recreated by adjusting the governor load reference setting due to slow dynamics. Resistive loads added or dropped to recreate instantaneous power changes, and thus phase angle changes. Resistance value calculated to appropriately add 14 or subtract net power injection at each pilot bus -4-6 -8-10 -12-14 Angle Between Area 4 and Area 3 Real Transient Response 5 10 15 20 25 30 Time (seconds)

Angle (degrees) Phase Angle Tuning Resistive Loads insufficient to match steadystate phase angles Steady-state phase angle values matched by adjusting machine Pm reference points. These points change when the fault occurs in the model. Allows slower, non instantaneous phase angle adjustment Exact steady-state phase angle matching possible 10 5 0-5 -10-15 Angle Between Area 5 and Area 4 Real Transient Response Model Transient Response 5 10 15 20 25 30 Time (seconds) 15

Angle (degrees) Inertias and Damping Values Calculated values produced one strange result Angle between Station 1 and Station 2 had a very high frequency component. Original Station 1 Inertia was skeptically insufficient This value was changed such that the transient responses shared slow-mode frequencies with PMU data. 19 18.5 18 17.5 17 16.5 16 Angle Between Area 1 and Area 2 Real Transient Response Model Transient Response 5 10 15 20 25 30 Time (seconds) 16

Transient Response Validation Plots of PMU data compared against transient response of WECC model in RSCAD Reduced-order model then used for various contingencies. 17

Contingencies Increase in intra-area line impedance Line trips Change in generation location Renewable energy sources located further from Station Lines intra-area Impedance increased until marginal stability was observed. Some buses more sensitive to increases than others Model excited with impulse fault on all 5 pilot buses and phase angles recorded Decrease in aggregate inertia Inertia-less renewable sources displacing synchronous machines 90% of inertia eliminated from each machine, one at a time. Model excited with loss of generation fault and phase angles recorded 18

Line Loss Contingency Effects of line loss are demonstrated here in these four plots along with data regarding underlying frequencies and damping values. 19

Inertia Loss Contingency Effects of inertia loss are demonstrated here in these four plots along with data regarding underlying frequencies and damping values. 20

Wind Power in the US Wind power generation increasing in the US Oil and coal prices rising Great for the environment For this trend to continue, prevalent issues regarding renewable energy integration need to be solved PAST PRESENT FUTURE

Wind Availability vs. Use Wind located far from US population centers Population lies along East and West coasts of US Midwest lightly populated and wind abundant Most population in low wind areas Poses issues for US power grid Stability issues Local area control insufficient in some cases Long distance transmission challenges present Wide-area control a solution for this problem!

Wind in WECC Expected to increase in Southern California Area 4 represents this area 500MW of wind added to RSCAD model DFIG turbine model 24

25 Wind and Bus Inertia

Wind Simulation Results WECC model faulted with a four-cycle line-toground fault on all five pilot buses simultaneously Phase angles between pilot buses recorded and plotted for two cases WECC with 500MW wind penetration on area 4 WECC with no wind penetration 26

27 Wind Simulation Plots

Wind Bus Sweep Model transient response was observed when 700MW of wind was placed on different pilot buses in WECC 28

Summary of Wind Simulations Wind increases severity of power swing on model Increase in intra-area line impedance Decrease in bus aggregate inertia RSCAD simulation shows this phenomenon Poorer damping, higher residue values Next question. How to counteract this? 29

Need a way to combat system destabilization due to wind penetration Must exist in real-world Must increase damping SVC offers a solution Typically regulates immediate bus voltage Device regulates via reactive power control, allowing control of phase angles SVC in the WECC 31

Real-world SVC exists geographically and topologically halfway between Areas 4 and 5. Geographically local to planned increases in wind penetration. SVC Location 32

SVC Parameters One 117 MVAR variable inductive element Two 91 MVAR switchable capacitive elements Droop typically 1-10% in Industry. 4% was used. 33

SVC Controls Basics Input is a per-unit local bus voltage measurement Filtered Relative to reference Droop of 4% Output controls inductive reactor elements and capacitive reactor switches Passes through PI controller PI controller will be tuned 34

SVC PI Tuning System very complex, unknown, even in model Even more unknown in real-world or with wind/svc added to model Exact system identification based tuning methods impractical Zeigler-Nichol s method used Great for unknown/complex systems RSCAD allows necessary tests/data collection 35

Zeigler-Nichol s RSCAD Process: Controller input, VPU, disconnected Step input applied to controller Resulting change in per-unit voltage collected in RSCAD Collection data shows SVC Process Reaction Curve 36

SVC Step Response RSCAD Data shown below for Process Reaction Curve Lag time, L measured Rise time, T, measured Change in amplitude, A, measured 37

SVC Local Control Results Data collected from RSCAD when a fourcycle line-toground fault was applied to area three. Just a comparison of SVC, no WAC implementation. 38

Input to Controller Must be a measure of some quantity in power system Rotor angles of generators Speed deviations of generators Machine output power Branch power flows/phase angles Chose inter-area phase angles Generators in model are fictitious Model tuned around phase angle recreation, not voltage recreation Pilot buses exist in real-world, with PMUs installed 39

Selection of an Input Signal Must be robust to changes in steady-state operating point Need to devise a test to search for robustness Test devised to randomly vary steady-state point Test shall randomly vary power injection on each bus Vary governor load-reference at each bus Vary wind penetration at area 4 40

Angle (degrees) Angle (degrees) Angle (degrees) Angle (degrees) RSCAD Impulse Responses Each case was tested for an impulse located on additional control input of the controller Resulting phase angle responses were recorded 0.15 Angle Between Area 1 and Area 2 0.1 Angle Between Area 2 and Area 3 0.1 0.05 0.05 0 0-0.05-0.05-0.1 0 10 20 30 Time (seconds) Angle Between Area 4 and Area 3 0.5-0.1 0 10 20 30 Time (seconds) Angle Between Area 5 and Area 4 0.6 0.4 0 0.2 0-0.2 41-0.5 0 10 20 30 Time (seconds) -0.4 0 10 20 30 Time (seconds)

42 Modal Data

43 Modal Variance

Variance Results Controller input must be robust to changes in system steady-state Finding this by observing variance in mode residue Mode Phase 1 Phase 2 Phase 3 Phase 4 1 2.9410 1.8491 0.0648 0.0487 2 0.8818 0.4235 0.2351 0.1941 Phase angles three and four show much less variation than one and two Also in closer proximity to controller, more realizable in real-world scenario 44

Use of both phase angles three and four as inputs Requires three data sources from three buses Output is reactive power injection between buses four and five Limited by SVC operational limits Goal is to reduce inter-area oscillation intensity and increase damping Tests will be conducted to observe and quantify 45 controller performance Wide-Area Control Structure

Controller Overview Supplementary controller composed of several parts Input to be phase angle measurements Output to lead to PI controller input, which leads to reactive SVC elements Composed of a Low-Pass filter, Lead-Lag filter, and Washout Filter in series for each slowmode. Two branches for each phase angle input 46

47 Controller Transfer Function

Data Required Frequency, Residue, and Damping for both modes across both angular inputs. Gained from RSCAD and ERA Impulse applied to power system Data collected Data analyzed with ERA Different than base paper s method Used to calculate controller parameters 48 Angle Mode Frequency Residue Damping 3-4 1 1.2478 45.5570 0.2719 3-4 2 1.9131 42.2355 0.2099 4-5 1 1.2784 41.0623 0.2763 4-5 2 1.7526 64.0806 0.2813

49 Low Pass Filter

50 Lead-lag Filter

51 Washout Filter

Total Transfer Function Controller consists of two parallel transfer functions for each phase angle input Two phase angle inputs One for each modal frequency Four total transfer functions in parallel Displayed on next slide. 52

53 Controller Transfer Function

54 Recap of Wide-Area Controller

Controller Tests Compare performance of power system both with and without supplementary wide-area controller Default SVC case One angle input Two angle inputs Fault applied on area 3 Wind on area 4 55

Controller Test Plots Controller performance with two, and then three pilot buses transmitting data were compared against the baseline case with no widearea controller. 56

Power System Transience ERA used to find damping reduction of primary slow-mode frequencies More sufficiently damped Phase Angle Mode 1 Baseline Mode 1 Controller 1 Mode 1 Controller 2 1 0.3653 0.3311 0.3426 2 0.6006 0.7816 1.0000 3 0.2166 0.3747 0.2856 4 0.2774 0.3001 0.4143 Phase Angle Mode 2 Baseline Mode 2 Controller 1 Mode 2 Controller 2 1 0.1923 0.1616 0.1674 2 0.2213 0.2762 0.2840 3 0.2763 0.3589 0.2856 4 0.1472 0.2151 0.4029 57

SVC WAC on Other Operating Points 58

Controller Conclusion Improvements in power system stability Improvements in three of four phase angles, with large improvements in geographically close angles Real-world applicable SVC exists in real-world Pilot buses exist in real-world PMUs exist in real-world Further tests desired to include hardware contingencies and implementation Up to now, model and controller are both in RSCAD It is desired to create PMUs, and controller using real-world equipment 59

LQR Wide-Area Control Base model used was reduced-order fivemachine WECC equivalent model. We next design a linear quadratic regulator state feedback controller u(t) = Kx(t) Gain matrix K computed offline Excitation system voltage used as control input Bus angle, frequency, and voltage used for estimating machine states. 61

State-Space Equations Each machine i represented by three orders of differential equations 62

Arranged in matrix form: State-Space Equations Partials taken, not shown here State Space formed Initial conditions taken from WECC in RSCAD 63

Wide-Area Control Transform into via use of the PSS stabilizer input on each machine. Choice of matrix K will minimize: For our design, R was I 3n Selection of Q was more challenging. 64

Relative Angles Product X T QX is a problem because X contains absolute angles. 65

Selection of Q Matrix Q was chosen such that the product X T QX contained only relative angles. 66

K Matrix Issue The feedback matrix K will have the same issue as Q. Define: Rearranged as: Controller output equates to: Only if: We force this: 67

State-Space Impulse Comparison Alteration of K Matrix could influence controller performance. A brief test done simply in MATLAB verifies alteration of matrix is negligible. 68

Creating an Unstable Test Case To test the controller s performance, a case was created in WECC that was unstable. K matrix was recomputed around new operational point: 69

70 Test of Unstable Case

Controller Performance Test Controller performance in damping a transience was compared against baseline model with lack of control. 71

Feedback of Just Local Voltage State Variables All terms in the feedback matrix K were zeroed out except for terms responsible for local voltage feedback. This was compared against the full K matrix controller implementation. 72

Feedback of Just Local Frequency State Variables All terms in the feedback matrix K were zeroed out except for terms responsible for local frequency feedback. This was compared against the baseline model with no control. 73

Standard PSS on WECC A PSS was applied to each aggregate machine and compared against performance of the Wide-Area LQR Controller. Note the issue of applying a double PSS here. 74

Unstable Baseline Case for LQR and SVC To further test performance, another case was created in WECC that was unstable even with PSS. K matrix was recomputed around new operational point: 75

Controller Performance Conclusion Case unstable with PSS was tested with both SVC-based WAC and LQR-based WAC. Both widearea control methods were capable of damping this unstable system! 76

Hardware architecture RTDS: Software component GTAO: low-level analog signals PMUs: analog data collection PDC: PMU/computer interface RTAC: hardware controller GPS: universal timestamp Design Capabilities Integration of hardware controllers Includes hardware measurement devices Built-in software simulators such as RSCAD Real-time operation PMU SEL421 Hardware-in-the-loop Implementation RTDS GTAO Card RTAC PMU1 PMU2 GPS PDC PMU3 To Computer RTDS & GTAO Card 78 GPS RTAC PDC SEL3373 PMU SEL487

79 Controller in Hardware

Controller Differences Model tested for two cases WAC in software in RSCAD exclusively WAC brought into hardware, and created in RTAC Fault applied to bus 3 of duration eight cycles from line to ground Controller output recorded for both cases 80

Controller Differences Differences come from many sources Measurement errors Analog noise Discretization errors System delays Fourth order discrete transfer function Time constant of 16.67ms Theoretical transfer function delay of 67ms 81

Controller Differences Delay found to be 75 milliseconds 67 ms from discrete transfer function 8 ms from network delays or other delays in system 82

Controller Performance Performance of RSCAD-based SVC-WAC compared against a hardware implementation using the RTAC. 83

Controller Performance 2 Mode damping compared using ERA Phase Angle Mode 1 in Software Mode 1 in Hardware Mode 1 No Controller 1 0.3426 0.3408 0.3653 2 1.0000 1.0000 0.6006 3 0.2856 0.3162 0.2166 4 0.4143 0.3193 0.2774 Phase Angle Mode 2 in Software Mode 2 in Hardware Mode 2 No Controller 1 0.1674 0.1721 0.1923 2 0.2840 0.2872 0.2213 3 0.2856 0.3162 0.2763 4 0.4029 0.3498 0.1472 84

Summary of HIL Set-up Controller functions in a hardware laboratory setting 67ms of delay present innately Graphically, controller performance shows no degradation Damping values comparable and show little degradation HIL controller can still provide adequate damping to the WECC model 85

Wide-Area Monitoring and Control Validate the distributed applications for wide-area monitoring and control through the cyber-physical distributed cloud computing test-bed 87

Ribbon cables Ethernet cables PMUs and Cloud Computing ExoGENI Oakland, CA Rack Site GTNET 10.0.0.9 Internal Fiber Optic Cable GTAO RTDS 152.14.125.32/33/232 Lab Computers 152.14.125.109/ 88 10.0.0.X VM5 VM1 VM3 Control Signals PMU1 10.0.0.3 PMU2 10.0.0.4 PMU3 10.0.0.5 PMU4 10.0.0.6 PMU5 10.0.0.7 PMU6 10.0.0.8 VM2 VM4 VM6 Control Signals Netgear Switch VLAN904 BEN BEN port PMU based WAMS at NCSU PMU data

units units units units units 0.3 0.2 1sec Control Signals from ExoGENI P1 0.3 0.2 Comparison of Control Signals Control Signals from RSCAD G1Input 0.1 0-0.1 t1 0.1 0-0.1 t2-0.2-0.2-0.3-0.3 0.3 P2 0.3 G2Input 0.2 0.2 0.1 0.1 0 0-0.1-0.1-0.2-0.3 0.3 0.2 P3-0.2 t1-t2 = -0.3 0.2sec 0.3 0.2 G3Input 0.1 0.1 0 0-0.1-0.1-0.2-0.2-0.3-0.3 0.3 P4 0.3 G4Input 0.2 0.2 0.1 0.1 0 0-0.1-0.1-0.2-0.2-0.3-0.3 P5 G5Input 0.3 0.3 0.2 0.2 0.1 0.1 0 0-0.1-0.2-0.3 89-0.1-0.2-0.3 0 1 2 3 4 5 6 0 1 2 3 4 5 6

Comparison of LQR Controller Performance Performance of RSCAD-based LQR-WAC compared against a cloudcomputing implementation using the ExoGENI Network. 90

PMU Lag Contingency One PMU in the Hardware- ExoGENI test-bed was delayed by eight milliseconds, or roughly half a cycle. The resulting destabilization that occurs is presented. 91

ExoGENI Packet Loss Contingency A loss of packets between t=6.5 and t=9.5 was injected into the test-bed and the resulting disturbances recorded during transience. 92

Conclusions A working wide-area, reduced-order model of WECC was created Wind penetration studies revealed need for improvements in power system stability improvements. We developed SVC-based and PSS-based widearea controllers that were capable of damping otherwise unstable WECC power system model under various operating conditions. Hardware-in-the-loop and cloud-in-the-loop testbeds demonstrated controller functionality in simulated real-world settings and equipment. 93

Future Works Update the WECC Model Series FACTS devices not possible currently due to fictitious lines Issues regarding performance loss or total failure of controller when delay or PMU mismatch is observed PSS is currently hypothetical. How to implement on an aggregate of machines 94

95 Works Cited

96 Questions?