RJT WORKSHOP: The Past, Present and Future of the Power Grid Cornell University August 8, 2012 James Thorp

 In a few years after Bob came to Cornell I took a sabbatical leave at AEP in NY. I worked on a digital relaying project aimed at replacing analog relays with new microprocessor based digital relays.  Thanks to Moore’s law it turned out to be a good decision on my part.  Ed Schweitzer was a professor at Washington State who formed a company (SEL). He claims to have sold 400,000 digital relays. He has his own jet. Prof Yang at NCEPU formed a company in China to make PMUs. He is a billionaire.  In 1983 we published a paper [1] that introduced the terms PMU and Synchrophasor. Google Scholar has 33,620 total citations with 20,400 in the last 4 years. ISI reports the majority of citations to PMUs in last 5 years. The majority of citations to my work have been since I turned 70  The DOE Smart Grid Investment Grant (SGIG) is investing in ~ 850 PMUs Phasor (Steinmetz 1897) Phaser (Star Trek 1966) Citations by year

 Synchrophasors use GPS time synchronized sampling to get phasors on a common reference. The phase angle between Maine and Florida can be measured, for example.  Used for monitoring  State estimation  Reconstruction of events  Bob put me on the Data Adequacy Working Group for the 2003 blackout San Diego Blackout vs  Control WAMS (wide area measurement systems)  Almost all control was local  Load frequency an exception (frequency is universal)  Control of Inter-area oscillations is popular today  Protection  Reduce relay involvement in cascading events  Back up projection is the typical culprit  (three primary, a backup, and a backup to the backup)

 In the digital substation it was obvious that all sampling of voltages and currents should be synchronous. Could share current samples between transformer, bus, and line protection processors for backup. A sampling pulse was distributed in the substation.  It was also clear that there was an advantage to sampling synchronously at the other end of the line.  Attempts at Ping-Pong to determine delays, use of GOES satellite signals. None successful. GPS was the answer.  Digital relaying depended on sampling voltages and currents at a nominal rate of n times a 60Hz cycle

Introduction to phasors - Steinmetz  Real Imaginary The starting time defines the phase angle of the phasor. This is arbitrary. However, differences between phase angles are independent of the starting time.  t=0 Samples

Motivation for synchronization By synchronizing the sampling processes for different signals - which may be hundreds of miles apart, it is possible to put their phasors on the same phasor diagram. Angle differences are correct Substation A Substation B At different locations

 “In 1893, Charles Proteus Steinmetz presented a paper on simplified mathematical description of the waveforms of alternating electricity. Steinmetz called his representation a phasor. With the invention of phasor measurement units (PMU) in 1988 by Dr. Arun G. Phadke and Dr. James S. Thorp at Virginia Tech, Steinmetz’s technique of phasor calculation evolved into the calculation of real time phasor measurements that are synchronized to an absolute time reference provided by the Global Positioning System. Early prototypes of the PMU were built at Virginia Tech, and Macrodyne built the first PMU (model 1690) in ”Charles Proteus Steinmetz phasorGlobal Positioning SystemVirginia Tech Macrodyne  It is argued [2] that the investment in improving monitoring of the high voltage transmission network represents the most cost-effective category of smart grid investment.  Jim McIntosh Director of Grid Operations, CAISO said in a JASONS Workshop in 2010 in La Jolla that the stimulus PMUs being installed in California would save California from $200M to $300M a year. Run the system closer to limits  Most vendors of control center software allow PMU data to be integrated with conventional SCADA data From Wikipedia [2] Paul L Joskow, Creating a Smarter US Electricity Grid MIT Center For Energy and Environmental Policy Research, Oct 2011

 “PMUs could improve the performance of energy management systems by providing real-time data to determine system state faster and more accurately than current estimation tools. A more extensive deployment of PMUs is required to make this possible”.  “Automatic control action based on real-time data from a wide-area network of PMUs represents a major change in system operations. Today such system are limited in number and capability. Significant research in control algorithms and improved confidence in the reliability and accuracy of PMU data is needed to make such control more prevalent.” Presented at National Press Club. The second question was whether the federal government could regulate the flow of electricity since there was no electricity when the constitution was written.

 Too few PMUs. “Transmission 101” [4] estimates there are 15,700 transmission substations in the US. A PNNL study [5] has 48,000 nodes. That’s Transmission not EHV but even 1,000 PMUs is probably below the threshold  We are interested in state estimation and control with measurements of less than 2% of the states  Lack of confidence in the reliability and accuracy of PMU by some (reason for standards) Would you write a paper with a fix for a 1985 computer problem?  Persistent concerns about PMU locations and latency. NIST, Virginia Tech, Georgia Tech, and Texas A&M are testing PMUs

 DOE Demonstration Project: Dynamic State estimation of Dominion Virginia Power 500 kV network at 1/30 sec interval. Linear, three phase, PMU only measurements communicated by a Sonnet network to control center Provide measurements for control applications with latency of < 30ms, i.e. appropriate for bandwidths of 5 or 6 Hz.  Emphasis on renewable energy implies more energy storage.  Energy storage is a new potential control means added to the power electronic arsenal Linear since only PMU measurements of voltage and current

Time Aug 6, 2012 “Utilities now receive updates on transmission lines 30 times a sec instead of every two seconds”

Substation PMU Deployment Strategy PDC phasor data concentrator – take data with the same time-tag make one vector with one time tag SEL Dual-Use Line Relay/PMU also give time tagged breaker status

 China has decreed that there will be a PMU in every high voltage substation. Much cheaper than retrofitting. They now have ~2000  China’s growth rate is such that fixed frequency PSSs are unworkable  The demand has a doubling time of ~ 7.5 years, they are commissioning a 1000MW plant a week.  Frequency of modes change rapidly.  They designed a WAMS Based Wide- area Coordinated Modulation Control of Multi-infeed HVDC assuming they knew the system. When a mode is observed the operator inserts the system which does a real-time Prony to determine the frequency of the wave. Then the system adjusts parameters in the controller to match the frequency and the loop is closed. The operator removes the controller after the mode is successfully damped [18]

YN GZ GX GZ Control signals control server PMU signal Control Unit Increase the damping of inter-area oscillation in CSG ~ 10 cycles Control Applied

Not continuously adaptive control but a control that is engaged by an operator like a SIPs system and observes and adjusts before acting. SIPs as we know them do not change parameters before engaging. Adaptive SIPs? (SIPS system integrity protection system) From observations of the actual event the system selects parameters of the control or even the controller structure from a large predetermined list and then engages. At least one author has labeled things like this as “enumeration based robust control” Perhaps “Synchrophasor Aided Gain Scheduling”? 900 papers in gain scheduling in power systems but many are for a power plant and none mentions WAMs 15

 Create a large data base of situations (~contingencies).  Augment with actual archived PMU data.  Use Data Mining with lots of options to find out what works, Done off-line (cancer researchers are succeeding in finding DNA markers for various cancers with data bases of millions of cases)  We have used simulation to create a data base with more than 15,000 cases for a 4000+bus model.[9] Hydro Quebec has used 60,000 cases  Use data mining to select the scheduling variables. Give CART* the option of using many possible scheduling variables. CART will select the best. There are solvable technical issues with complex measurements in CART *CART Classification And Regression Trees. There is also a BART which is Bayesian Additive Regression Trees 16

 University contributions in adding PMUs to existing estimators  Large number of papers in Control  Improve PSSs with a few remote PMU measurements  Robust control applied to reduced order models with full state feedback. Software used to model large, real system > 10,000 buses does not (cannot) find eigenvalues or A matrices. How do you control something with no model? MIT finding 2

 Universities have made contributions in finding locations for PMUs (there are never enough so you must put the few in the best places)  Best locations for:  observability (measure local voltage and line currents-learn ~4 voltages from 1 PMU – only need PMU at ~ 1/3 of the buses)  Redundancy to loss of PMUs or line opening or topology changes…  Phased installation-optimum per year  Techniques  Topological – enumerating trees,  binary integer programming,  data mining Unfortunately the utilities ignore these results and use their own criteria ($)

Synchrophasor Detectives Just as new telescopes (Hubble) or new microscopes (STEM) have shown us thing we did not predict, wide area - time synchronized - measurements have produced a few surprises. The next two slides are from Mack Grady UT Austin (with his permission) The title comes from his second slide Load- Dallas “Central ERCOT” Austin UT monitor at Austin Energy Harris substation Wind Country McDonald Observatory 400 Miles

You must become a synchrophasor detective

 On Feb. 26, 2008, a short circuit in a Miami electric power substation and an operator's error gave managers of the nation's electrical grids a glimpse of an uneasy future. The events triggered a chain reaction of power plant and transmission line outages in the state, unleashing sharp swings in voltages and power frequency that blacked out power for nearly 1 million customers in southern and central Florida for up to four hours.  A video depicting the Florida incident's rippling spread has been created by Virginia Polytechnic Institute and State University's electrical and computer engineering department, which caught the disturbance on its first-generation grid frequency monitoring network. Some grid executives have downloaded the video on their laptops as a kind of horror flick for engineers of what could happen.

What terminal? What measurements? Determination of triggering logic Performance evaluation System State Assessment PMU data Critical System Locations Supervisory signals Adjustment of Dependability- Security balance under stressed system conditions. Relay 1 Relay 2 Relay 3 OR VOTE AND Supervisory signal See detail below Adjusting balance of security-dependability Adjusting balance of security-dependability One of the five DOE Demonstration projects JST Pacific Gas & Electric Southern California Edison Three parallel 500Kv lines between Midway and Vincent path 25 2 out of 3 is voting

Voting Scheme Adaptive Voting Scheme with three relays. State of the System –StressedSecurity = Vote –Safe Dependability = Don’t Vote We are NOT changing relay settings neither during, before or after a fault. 25 JST

Data Mining: CART* Proposed Decision Tree: –Nodes = 7. –Probability of correctly distinguishing 1 and a 0: 0.99%. Splitting NodesPMU placement Terminal Nodes Vote Decision 26 *Classification And Regression Trees JST Utility demand for We studied rare events, half of which would have caused a major disturbance. By voting we reduced the number of major disturbances from ~7,500 to 75

PMU Placement: 27 Line Current TESLALOSBANOS MIDWAY N.GILAIMPRLVLY TRACYTESLA LUGOVINCENT METCALFMOSSLAND LUGOMOHAVE MALINROUND MT PMU ROUND MT TESLA VINCENT LUGO IMPLVLY JST

Decision Trees -Location of PMUs and logic for PMU inputs to real-time, discrete-event control [5] Predicting cascading events [6] voltage security [7-8] transient stability [9-10] detection of islanding [11] processing post disturbance records [12] security assessment [13-14] adaptive security dependability of relays [14] Proposed NASPI’s Planning and Implementation Task Team (PITT) has made base lining of phase angle differences their highest priority. Set angle thresholds and operator alerts Estimation of line flows and voltages after an outage Correlate PMU data with GIC monitoring VAR issues –return to Pilot Points with PMUs and CART Equipment monitoring and asset management At 60 times a second can monitor I 2 t in transformers due to faults

 There is a more limited history of data mining of archived PMU data. Fifteen months of PMU data of 54 angle differences has also been subjected to statistical analysis to detect abnormal power system behavior using software developed for NASA by PNNL [15-17].  [16] presents results using state estimator data sampled every few minutes. Part of [16] is to identify data as atypical [17] rather than typical.  Plans to expand this by WECC using OSIsoft. They will ultimately archive 150,000 measurements per second

 More PMUs  All high voltage buses>100kV ~50,000 substations  Operator alarms and alerts  Oscillation  Angles  Voltages  More adaptive relays  Possibly universal  Coordinated SIPs  WECC has >100. One went wrong in San Diego blackout  Control  Inter-area oscillations  Out of step SIPs System Integrity Protection Schemes Remedial Action Schemes Special Protection Schemes Control actuated by protection: If one of two parallel lines trips shed load and reduce generation

Given we are likely to have to make due with the existing wires and towers as AC transmission lines for decades, PMU technology, advances in computing, and communications offer the best hope for a Smart Transmission Grid. Returning to the MIT Findings  PMUs could improve the performance of energy management systems by providing real-time data to determine system state faster and more accurately than current estimation tools.  Automatic control action based on real-time data from a wide-area network of PMUs represents a major change in system operations. Significant research in control algorithms and improved confidence in the reliability and accuracy of PMU data is needed to make such control more prevalent.” Terry Boston, President and CEO of PJM is reported to have said ”running the nation’s power grid isn’t rocket science- it’s harder” Thank You

[1] A.G. Phadke, J. S. Thorp and M. G. Adamiak, "A New Measurement Technique for Tracking Voltage Phasors, Local System Frequency and Rate of Change of Frequency," IEEE Trans on PAS, PAS 102, vol. 5, , May [2] The Future of the Electric Grid, An MIT Interdisciplinary Study, December 2011 [3] Paul L Joskow, Creating a Smarter US Electricity Grid, CEEPR WP , MIT Center For Energy and Environmental Policy Research, Oct 2011 [4] Silverstein, A., “Transmission 101”, ECEP Transmission Technologies Workshop, April, 2011 [5] S. M. Rovnyak, C. W. Taylor, and J. S. Thorp, “Performance index and classifier approaches to real-time, discrete-event control,” Control Engineering Practice, vol. 5, no. 1, pp. 91–99, [6] Kamwa I.; Samantaray, S.R.; Joos, G., “On the Accuracy versus Transparency Trade-Off of Data-Mining Models for Fast-Response PMU- Based Catastrophe Predictors”, IEEE Transactions on Smart Grid, Vol.: 1, Issue: 2, pp.: 144 – 158, [7] Nuqui, R.F.; Phadke, A.G.; Schulz, R.P.; Bhatt, N., “Fast on-line voltage security monitoring using synchronized phasor measurements and decision trees”, IEEE PES Winter Meeting, Vol.: 3, pp.: 1347 – 1352, [8] Ruisheng Diao; Kai Sun; Vittal, V.; O'Keefe, R.J.; Richardson, M.R.; Bhatt, N.; Stradford, D.; Sarawgi, S.K., “Decision Tree-Based Online Voltage Security Assessment Using PMU Measurements”, IEEE Transactions on Power systems Vol.: 24, Issue: 2, pp.: 832 – 8, [9] Sun, K., Likhate, S., Vittal, V, Mandal, S, “An Online Dynamic Security Assessment Scheme Using Phasor Measurements and Decision Trees”, Trans Power Systems, Vol. 22, No pp [10] Jan Ma; Makarov, Y.V.; Miller, C.H.; Nguyen, T.B.; “Use multi-dimensional ellipsoid to monitor dynamic behavior of power systems based on PMU measurement”, IEEE PES General Meeting, Conversion and Delivery of Electrical Energy in the 21st Century, pp.: 1 – 8, 2008 [11] Rui Sun; Zhongyu Wu; Centeno, V.A, “Power system islanding detection & identification using topology approach and decision tree” IEEE PES General Meeting, pp.: 1 – 6, 2011 [12] Kamwa, I.; Samantaray, S.R.; Joos, G., “Development of Rule-Based Classifiers for Rapid Stability Assessment of Wide-Area Post- Disturbance Records”, IEEE Transactions on Power systems, Vol.: 24, Issue: 1, pp.: 258 – 270, [13] Zhiyong Li; Weilin Wu,”Phasor Measurements-Aided Decision Trees for Power System Security Assessment”, ICIC '09. Second International Conference on Information and Computing Science, Vol.: 1, pp.: 358 – 361, [14] Bernabeu, E.E.; Thorp, J.S.; Centeno, V., “Methodology for a Security/Dependability Adaptive Protection Scheme Based on Data Mining”, IEEE Transactions on Power Delivery, Vol: 27, Issue: 1, pp 104 – 111, [15] Ferryman, TA, and Amidan, BG “Statistical Analysis of Abnormal Electric Power Grid Behavior” 2010 Hawaii International Conference on System Sciences [16] Ferryman, TA, and Amidan, BG,. “Investigation of Phase Angle Differences Using Statistical Analysis of Real World State Estimator Data,” 2012 Hawaii International Conference on System Sciences [17] Amidan, B.G. and Ferryman, T.A., “ Atypical Event and Typical Pattern Detection within Complex Systems”, IEEEAC paper #1200, version 3 Dec 9, 2004 [18] Lu, Chao; Wu, Xiaochen; Wu, Jingtao; Li, Peng; Han, Yingduo; Li, Licheng,”Implementations and Experiences of Wide-Area HVDC Damping Control in China Southern Power Grid” PESGM2012 References