AEP Experience with Sub-Synchronous Oscillation Phenomena

Transcription

1 Initially Presented at 72 nd Georgia Tech Protective Relay Conference AEP Experience with Sub-Synchronous Oscillation Phenomena Zachary Campbell Transmission PCE Standards & NERC CIP American Electric Power New Albany, Ohio, United States Kiril Andov Advanced Transmission Studies & Technologies American Electric Power New Albany, Ohio, United States Shawn Coppel Transmission PCE Standards & NERC CIP American Electric Power New Albany, Ohio, United States Abstract Recent installations of renewable energy resources in combination with series capacitor bank installations on AEPs network has renewed concerns about the power system phenomenon known as subsynchronous control interaction (SSCI) and subsynchronous resonance (SSR) on the AEP electrical network. In this paper, historical events which have occurred on the AEP transmission electrical network are illustrated which demonstrate the frequency and magnitude of the subsynchronous oscillation, the network conditions, and the protection system modifications made following the events. These events are further analyzed to support the development of a generic subsynchronous oscillation (SSO) detection relay algorithm. The signal processing techniques used to do this are illustrated, analyzed and the relay detection algorithm is demonstrated through simulation. Keywords sub-synchronous oscillation, American Electric Power, sub-harmonic, oscillations, SSTI, SSCI, SSR I. INTRODUCTION AEP is one of the nation s largest electrical utility providers with more than 26GW of generating capacity, 40,000 miles of transmission lines, and nearly 5.4 million customers operating within 11 states [1]. AEP also operates and maintains 9 series capacitors and dozens of customer operated windfarm and solar interconnections. Traditionally, the power system network combination of a turbine-generator shaft driven system and series compensation would necessitate specific analysis to uncover the potential for forced oscillations of the shaft system, commonly referred to as subsynchronous resonance (SSR) [2]. Most notably, the shaft failures at the Mohave Generating Station in Southern Nevada drew specific attention to this power system phenomenon. Recently, the increased penetration of renewable energy resources in combination with series compensation on the AEP network has renewed similar concerns. This is because of the likelihood for similar problems, like SSR, to exist where certain electrical network configurations or conditions can cause renewable generation resource instability [3]. In these cases, it is not a traditional shaft driven mechanical system that could be damaged due to the subsynchronous content. Instead, the instability of the renewable generation resource can generate a combination of nominal and subsynchronous voltages applied to the power system network which can result in system voltages or currents so large that damage to both network and renewable generation resource occurs. This phenomenon is sometimes referred to as subsynchronous control instability (SSCI). Because of this possibility, there is again a need to study the conditions which can drive this phenomenon. In these cases, however, some of the traditional methods used to analyze the traditional turbine-generator shaft driven system are not applicable. In these cases, the analysis performed in determining the possibility of problems resulting in the combination of renewable energy resources and series compensation uses a black-box model of the renewable generation resource, wherein the understanding of the dynamic behavior of the renewable energy resource is obscured. This is purposeful due to the proprietary nature of the models of the interconnecting entity, but this obscurity limits the understanding of the causes of any potential instability. Regardless, special analysis is performed which attempts to uncover any SSCI phenomena which could occur. While it is not the goal of this paper to inform the user of the analysis methods used in uncovering the system conditions which generate SSCI, it is the goal of this paper to demonstrate various events which have occurred on the AEP network which demonstrate the phenomena. Because it is currently thought that SSR or SSCI is a rare system occurrence [4], it is rarely considered by protective relay engineers who operate within a unique role in which the potential detection of and remedial action to SSCI is within their expertise. A. Event 1 II. AEP SYSTEM HISTORICAL EVENTS On October 22, 2009 a fault on an AEP line in southeastern Texas ultimately resulted in an event in which wind turbine and electrical network damage occurred. In this case, the network in the vicinity of the fault consisted of two 345kV transmission lines, two windfarms, and a series capacitor bank positioned on one of the transmission lines. A simplified network diagram for this system is shown in Fig. 1. The fault caused the station A to station L line to de-energize which resulted in the radial connection of both windfarms infrastructure into the series capacitor bank. The result of this Pg. 1

2 system configuration and initial disturbance is shown in Fig. 2, which illustrates the station Z to station A line relay oscillography data captured. As can be seen in the figure, following the initial fault, system voltages rose to nearly 1.5 PU. However, the frequency content of the currents and voltages during the second half of the event of Fig. 1 was not made up of only 60Hz information. In fact, as shown in Fig. 3, the frequency content of the current was made up of more than 1000A of 25Hz content and less than 500A of 60Hz content, while the voltages were made up of nearly 200kV of 60Hz content and nearly 100kV of 25Hz content. It was thought, at the time that, damage to infrastructure was caused when system voltages far exceeded the electrical ratings of system infrastructure in the area of the event [5]. Windfarm 1 Station Z 13 miles Station A Windfarm 2 7 miles 345kV Fig. 1 Event 1 Simplified Electrical Network 67 miles Station R 37 miles Station L The event ended after nearly 1.5 seconds when the series capacitor bypassed. The event was recreated through simulation using a detailed PSCAD model to attempt to excite to subsynchronous mode of oscillation. Fig. 4 illustrates the PSCAD model simulation results from the development efforts made to replicate the event. The phenomenon was attributed to SSCI between the network and the windfarms [3]. This event demonstrated the need for the detection of the SSCI conditions and justified the installation of a relay specifically dedicated to sensing the modal frequency identified during the analysis of the event. This relay measured the voltages of the system at station Z, filtered the voltages phasor signals coming in from conventional relay signal processing, and operated based on the detected output of the subsequently filtered subsynchronous voltage magnitude. A software flow diagram of the subsynchronous oscillation (SSO) detection algorithm deployed in this case is shown in Fig. 5. [5] discusses this relay in more detail. Fig. 4 Event 1 PSCAD Simulation Results [3] Fig. 2 Event 1 Oscillography Record Fig. 5 Event 1 SSO Detection Software Flow Diagram [5] B. Event 2 On January 8th 2013, a trip occurred on a line in northerncentral Texas from what appeared, at the time, to be a power swing condition. During this operation the line between station M and S, shown in Fig. 8, was open due to temporary local switching. The oscillography record of this event can be seen in Fig. 6, where the FFT of the current and voltage waveforms between 0.0 seconds and 0.6 seconds into the event can be seen in Fig. 7. Fig. 3 Event 1 FFT Spectrum Analysis of Phase A Voltage and Current Pg. 2

3 Event 2A 69kV 35 miles Station P Station M 60 miles Station T Windfarm 6 miles 43 miles Station S Fig. 6 Event 2A Oscillography Record Fig. 7 Event 2A FFT Spectrum Analysis of Phase A Voltage and Current In this case, the only nearby generating resource was a windfarm and there appeared to be no initial disturbance as to the cause of the swing. It is notable that the 69kV bus where this windfarm connects has a low short circuit MVA (less than 160MVA with the station s strongest source out of service) in comparison with the output capability of the windfarm (nearly 50MW). Power swing blocking was enabled on the transmission line relaying that initially tripped due to this event and the condition was considered remedied. A simplified electrical network diagram, illustrating the windfarm connection and surrounding line terminals, is shown in Fig. 8. The line between stations M and P, labeled as Event 2A, is the line which tripped during the event. Event 2B Fig. 8 Event 2 Simplified Electrical Network Additional customer complaints mentioning blinking and dimming of lighting in the area around the facility which the windfarm connects were noted following the initial disturbance. A project was executed which was aimed at capturing any system events which were similar to the nature of event 2A. A relay was conceived of at this time dedicated to monitoring the output of the windfarm. This relay was set to detect for conditions when between 3-7Hz phasor fluctuations occurred on any two of the three voltage or current measurements and the content was above a minimum pickup for a fixed time. A diagram of the approximate signal processing and logic methods used to detect for these conditions are shown in Fig. 9. This relay was configured to trip the line to the windfarm. VAp VBp VCp IAp IBp ICp 3-7Hz BP 3-7Hz BP Fig. 9 Event 2 Installed SSO Detection Algorithm > > > > 2/3 2/3 Fault Detection Timer After only one month in service, in November of 2015, the SSO relay that was installed detected a subsynchronous oscillation and tripped the windfarm. Fig. 10 illustrates the oscillography captured during this event and Fig. 11 illustrates the FFT of the waveforms illustrated from 0.0 seconds to 0.9 seconds. During this event a permanent fault on the station M to station S line of Fig. 8 caused the initial oscillation disturbance and then continued the subsynchronous oscillation after the time-delayed reclose open internal times out and the line circuit breaker closes back into the fault at the 69kV bus to which the windfarm is connected. The windfarm was tripped nearly one-second after the initial fault by the subsynchronous oscillation relay, ending the event. Pg. 3

4 Current (A) be noted that the scenario listed does not illustrate traditional power swing conditions, but rather a SSCI. This is because there is no machinery within the network where slip rate or a rotating mass is the cause of the subsynchronous oscillation. Instead, in this case, the only dynamic electrical facilities present in the electrical vicinity of the event was the windfarm infrastructure. Voltage (V) Fig. 10 Event 2B Oscillography Record Spectrum Analysis of IA 150 Amplitude Fig. 12 Event 2B PSCAD Simulation Results Frequency (Hz) 10 4 Spectrum Analysis of VA Amplitude Frequency (Hz) Fig. 11 Event 2B FFT Spectrum Analysis of Phase A Voltage and Current With this event record captured demonstrating new evidence of the interaction of the electrical network and the windfarm facilities, a PSCAD simulation was worked out to simulate the event conditions. Fig. 12 demonstrates the simulation results from the PSCAD simulation and Fig. 13 illustrates the FFT of the waveforms captured during the entire time window shown. The same network conditions were simulated where the line M to line S opened at time 3.75 seconds into the simulation and at the same windfarm dispatch level as that seen in the November 17 th event. As can be seen, the results fairly closely mirror the event records captured from the facility during the November 17 th event. The FFT results of Fig. 13 demonstrate that the frequency of oscillation was near the event 2B frequency. From this simulation, it can Fig. 13 Event 2B PSCAD Spectrum Analysis of Phase A Voltage and Current C. Event 3 In late 2015 and early 2016, the installation of several series capacitor banks in western Texas near windfarm infrastructure, the historical context within AEP, and the precedent set forth in multiple industry reported events justified the installation of relays installed at the series capacitors which would quickly react to almost any SSCI phenomena. It was at this time that the relay of Fig. 9 was Pg. 4

5 duplicated and modified such that it could act faster and a broader range of frequencies would be detected. Also, the relays were initially set and configured such that they could bypass the series capacitors shown in Fig. 14 in the event of detection of SSO. Initial screening studies identified potential SSCI modes in the range of 30Hz. On August 24 th 2017, a personnel error at station P, shown in Fig. 14, caused an erroneous DTT signal to be sent to station D. This event caused the windfarms (shown as WF V3 and WF V4) connected to station D to become radially connected to the station D to station C line in series with the in-service series capacitor banks shown. 0.3 seconds after the opening of the line between the two stations due to the DTT, the Fig. 15 oscillography was recorded on the line relaying at station D on the station D to station C line relaying. An FFT of the Fig. 15 oscillography waveform content is shown in Fig. 16. After nearly 2 seconds of steady state conditions, similar to that shown in Fig. 15, the series capacitors which were both initially in service bypassed automatically on protective trip of subsynchronous overcurrent function. Nearly 20 minutes passed and dispatch personnel manually reinserted the two series capacitors, at nearly the same time. Station L Current (A) Voltage (V) Fig. 15 Event 3A Oscillography Record Station C 27 miles 82 miles 29 miles WF B Station D 13 miles 11 miles WF V3 WF V4 Station P 20 miles 11 miles 26 miles Station N Fig. 14 Event 3 Simplified Electrical Network WF LM Fig. 16 Event 3A Spectrum Analysis of Phase A Voltage and Current Pg. 5

6 of the simulation trip. Fig. 20 shows a FFT of the Fig. 19 time series data, from 0.0 seconds to 0.25 seconds. As can be seen, the simulation resembles the actual event waveforms, though slightly shorter, less severe, and with slightly different SSO mode frequency than the actual event. This simulation also illustrates an event in which SSCI conditions existed due to the combination of radially connected windfarm infrastructure and a heavily series compensated electrical network. Fig. 17 Event 3B Oscillography Record Fig. 19 Event 3B PSCAD Simulation Results Fig. 18 Event 3B Spectrum Analysis of Phase A Voltage and Current Approximately 0.4 seconds after the first of the two series capacitor banks was reinserted, the second of the two series capacitor banks was reinserted and caused the creation of the oscillography shown in Fig. 17 was recorded when the second of the two series capacitors was re-inserted. A FFT of the oscillography waveform captures of Fig. 17, from 0.0 seconds to 0.65 seconds, is shown in Fig. 18. The event stopped when the windfarms at V3 and V4 tripped. Fortunately, no infrastructure damage was noted by personnel at either the windfarm or on the electrical network. Following the event, PSCAD simulations were developed which were aimed at replicating the event. These simulations resulted in the development of Fig. 19 and Fig. 20 demonstrating the last part of the Aug 24 th event. Fig. 19 shows that at time 0.0 seconds the station D to station P line opens, which immediately increases the amount of subsynchronous frequency content in the waveforms. At 0.25 seconds in the figure the windfarms Fig. 20 Event 3B PSCAD Simulation Spectrum Analysis of Phase A Current and Voltage III. AEP SUBSYNCHRONOUS OSCILLATION DETECTION RELAY As mentioned in each of the events, AEP has made use of several different SSO detection relays. Due to the fact that there is an increasing amount of renewable generation resource connection requests to the AEP network an attempt has been made to standardize a detection algorithm which would work to detect a large range of subsynchronous frequencies with a high degree of reliability. Also, while there are a few out-ofthe-box solutions readily available for use in the detection of SSO events, none of the solutions demonstrated the level of customization or flexibility desired by AEP. Therefore, AEP developed a solution using off-the-shelf components but on a Pg. 6

7 Fig. 21 Event 1, 2B, 3B Multimodal Symmetrical Component Analysis platform which was customizable enough that the degree of flexibility and the functions AEP sought to employ were possible. As can be seen in each of the events demonstrated earlier, SSCI conditions are clearly dominated by a three phase system phenomenon where there exists 60Hz waveform content and subsynchronous waveform content in both the voltage and current signals. To demonstrate whether the subsynchronous waveform content generated during the system conditions of the three AEP events previously shown demonstrated a balanced subsynchronous three phase source, the event records had their individual analog voltage or current channel data filtered according to either a 60Hz FIR filter, or a subsynchronous filter designed to extract only the subsynchronous content from the waveform data. Then the filtered data was turned into a phasor using a discrete Fourier transform, where, for 60Hz filtered data the DFT was fitted to one cycle of the 60Hz period, or, for subsynchronous waveform content, the DFT was fitted to one cycle of the subsynchronous frequency period. Then, the symmetrical components of the three phase current or voltage phasors were calculated and displayed as shown in Fig. 21. Figure data labels shown as XXN illustrates 60Hz symmetrical component information (0 as zero sequence, 1 as positive sequence, 2 as negative sequence), while the XXS labels refer to subsynchronous sequence components. Fig. 21 illustrates that in each of the previously illustrated events (events 1, 2B and 3B) there does exist a large amount of positive sequence subsynchronous content. Using the observation that each of the events demonstrates a system condition where there are well-balanced nominal frequency phasors and well-balanced subsynchronous frequency phasors, it is possible to build a model of power system voltages and currents operating under SSCI conditions where the relationships of Eq. s 0A thru 0F are true. Additionally, [6] suggests the use of this signal model for multimodal signal representation. (0A) (0B) (0C) (0D) (0E) (0F) Defining as the positive sequence current magnitude of the nominal frequency, as the subsynchronous positive sequence current, as the positive sequence voltage magnitude of the nominal frequency, as the subsynchronous positive sequence voltage, ias 120, as the phase difference between the nominal frequency voltage and current, β as the phase difference between the subsynchronous frequency voltage and current, as the nominal frequency, and as the subsynchronous frequency. If one were to calculate the instantaneous power of the hypothetical power system operating under these conditions, the following would result. (1) Under very balanced system conditions, like those observed in Fig. 21, the following would be true. Pg. 7

8 and and and Through simplification, Eq. 1 would result in the following. (2) Where: (4) (3) frequency SSO modes solves the first part to the problem. The solution involves separating the DC mode of the phasor magnitude from the AC part of the phasor magnitude using more conventional signal processing techniques. This is simpler to accomplish than separating 55Hz waveform content from 60Hz waveform content because DC mode rejection is simpler than AC mode rejection. Obtaining the magnitude of the AC mode provides the ability to extract the magnitude of the third term of Eq. 2. Eq. 2 illustrates that there are two DC modes to the instantaneous power represented in Eq. 2. The first two terms are generated purely by the balanced three phase content of the nominal system frequency waveforms and the subsynchronous frequency waveforms. But, because the first two terms are not discernable from one another in the frequency domain, because they are both made up purely of DC content, one cannot ascertain only the subsynchronous instantaneous power from the nominal frequency power. However, the last term of Eq. 2 illustrates that instantaneous power oscillates with a radian frequency of. As can be seen in Fig. 21, it is clear that when using the output of a phasor calculation when observing subsynchronous oscillation phenomenon using conventional relay signal processing techniques, there is clear indication that the magnitude of a phasor oscillates with the same radian frequency noted above as. This fact illustrates that the use of the phasor magnitude of voltages and currents within a relay to ascertain whether subsynchronous power is present, is clearly possible. In fact, [5] suggests the use of power, as others have done in a similar manner [7], as input to an operating quantity in a protection element for use in the SSCI detection application space. There is one problem however, and that is that the conventional signal processing techniques used in conventional protective relaying platforms attenuates waveform information in the range of subsynchronous frequencies. Fig. 22 illustrates this phenomenon well, where, from 0Hz 40Hz, depending on what type of conventional signal processing filter is used, 70% of waveform attenuation can occur [8][9]. Having less than 70% of the original signal available to determine whether SSCI conditions exist using the phasor data processed from these filtering techniques is clearly a problem. Fortunately, this problem can be broken down into two parts. The first part of the solution to the problem can be illustrated considering the SSCI phenomenon of Fig. 10, where it is clear that the subsynchronous mode of oscillation is on the order of 55Hz. A 55Hz subsynchronous oscillation will generate a oscillation in the magnitude of the phasors used in a conventional relay. As can be seen in Fig. 22, 55Hz signal data is not attenuated significantly by the conventional filtering methods. This means that the 5Hz phasor oscillation data is also not attenuated significantly. Additionally, because it is very challenging to distinguish between 55Hz and 60Hz frequency due to the close proximity of the frequencies using digital filter techniques, the use of phasor magnitude, which contains a DC mode meant to capture 60Hz content and an AC mode meant to capture the subsynchronous mode, to detect near nominal Gain COSINE FILTER SINE FILTER Frequency (Hz) Fig. 22 Digital filtering frequency response used in Conventional Relaying The second part of the problem is defined by the fact that there is significant loss of subsynchronous information that requires conventional signal processing embedded within protection relays, and even PLCs, be bypassed altogether and an alternative method meant to extract only the subsynchronous content be used in its place. To accomplish this, a team at AEP developed a simple analog filter meant to extract subsynchronous content from the current and voltages being monitored and completely rejects 60Hz waveform content. The filtered information is then scaled appropriately, fed back into a relay, filtered using only a DC mode filter, and instantaneous power is then calculated from the current and voltage channel data. This signal, when in the presence of a balanced subsynchronous source, is a DC mode with magnitude using the same term definitions discussed previously. Using the solutions discussed, these two methods of SSO detection were then deployed in a modern PLC where the signal processing methods, hardware, and a small portion of the logic used for SSO detection is shown at a high level, within Fig. 23. This figure illustrates the combined concept deployed to detect generic SSO conditions which can have a wide range of subsynchronous frequencies that can be detected. Pg. 8

9 Voltages Currents CTs PTs HS- XDCR HS- XDCR Analog Filtering G f PLC CT/PT Card CT Inputs VT Inputs a/d a/d LEA Inputs a/d Scaling cos cos DC Blocking Fig. 23 Generic SSO Relay Detection Algorithm Signal Processing and Logic Flowchart Initial observations shown in Fig. 21 provided some insight on how to set the pickup settings for the two elements onboard the relay. The observation noted that the SSO power delivered during the event was approximately 100% of the power output rating of one of the two facilities. Therefore, precedent was established to pick a power level that is somehow proportional to the power output of the facilities being monitored by the relay. This being the case, and in a similar manner to the way in which overcurrent elements are selected, the suggestion is made that a selected level of pickup of the output capacity of the facilities being monitored be configured. For the setting2 value this is a simple exercise to perform where the pickup of the element is simply the power rating of the facility divided by the pickup level desired and then reflected thru the CT, PT, and analog scaling ratios. The calculation can be shown in Eq. 5. This pickup value ensures a dependable but also secure response to the SSO condition. The setting1 value of the element is more difficult to set. This is because the amplitude of the subsynchronous mode has a magnitude of which is made up of both nominal frequency voltages and currents and subsynchronous frequency voltages and currents. Also, the magnitude of this term is made up of a phase angle difference between the nominal frequency current and voltage and the subsynchronous frequency current and voltage. To simplify the task of attempting to determine all of these quantities in real time, extracting their values, and then determining the subsynchronous power from then, the authors first propose to assume that the term of Eq. 3 is zero. Doing so will assume the smallest amount of signal availability for use, which allows for a dependable manner to (5) dft dft [A]^-1 Power Calc Power Calc RMS1 1 RMS2 2 Sfi I0 I1 Sfp > > PT2 20% > Zero Crossing Run Zero Crossing Run PT1 X2 X1 X1 X2 0 0 SEL Frequency Calculator Frequency Calculator (1/ƒ1)*3 (1/ƒ2)*4 SSO Detect Logic detect SSO conditions. Also, the authors propose that be set to nominal system voltage reflected to the relay input as is similar to the conditions shown in each of the events shown earlier. Additionally, using the fact that the SSO mode voltages that started out during each of the events were only a small fraction of the nominal voltages, the authors propose to use an assumed = 0.15 (15% of nominal voltage). the pickup value in this way ensures a dependable but also secure response to the SSO condition. Using each of these assumptions, Eq. 3 simplifies to the following Equation, Eq. 6. As shown in Fig. 23, the timers of the elements (3 and 4) are dynamic in such a way that oscillations resulting in low frequency disturbances are responded to by the SSO detection element at a slower rate than those at a higher frequency. This allows for a secure response to any SSO condition. To illustrate the detection element, MATLAB simulations were developed which analyzed the reaction of the algorithm and simulated hardware when exposed to each of the three system phenomena as well as external faults. Table I, below, illustrates the settings of the elements as developed from the settings guidance mentioned earlier. The following figures illustrate the response of the simulated relay to each of the events. (6) Pg. 9

10 TABLE I. SETTINGS OF SIMULATED GENERIC SSO DETECTION ALGORITHM FOR SIMULATED EVENTS OF FIG. S 24, 25, 26 AND 27 Event Ratings, CTR, PTR, Standard Pickup, Vnom,Inom 1 FacilityRating = 150MW, CTR=400, PTR=3000, Pickup=5, Vnom=67, Inom=5 2B 3B Ext. Flt FacilityRating = 50MW, CTR=120, PTR=600, Pickup=5, Vnom=67, Inom=5 FacilityRating = 200MW, CTR=600, PTR=3000, Pickup=5, Vnom=67, Inom=5 FacilityRating = 200MW, CTR=600, PTR=3000, Pickup=5, Vnom=67, Inom= / / / /3 3/4 The simulated relay response for event 1 is shown in Fig. 24. As this figure shows, the relays hardware filters quickly respond to the subsynchronous frequency injection into the currents and voltages, as can be seen in the two time series plots below the raw event time series plots. The digital filters then begin to pass through the frequency content of the instantaneous powers that they are each designed to detect. The signal SFi, in the middle figure, is the subsynchronous power calculated using the hardware filtered channels. The signal SFp is the power calculated using the phasor data amplitudes. Because the frequency of the SSO content in this event is near the cutoff frequency of both the hardware filters and the phasor data channels, the SSO event is detected by both protection algorithms filters. The time series plot below the filtered operating quantity amplitudes illustrates the instantaneous RMS content of each of the signals, as well as the settings of 1 and 2. As can be seen in the last time series plot in Fig. 24, when the RMS content is above the pickup level of the settings, the variables PT1 or PT2 will assert, and the frequency tracking function will begin to detect the frequency of oscillation of the SSO modes. Once this occurs, the variables X1 and X2 in Fig. 23 show that the period timers (1/f1*setting and 1/f2*setting) begin to change from their initial values of 3 and 4 to the actual oscillation period multiplied by these values. Once the timers recognize that the inputs to the timers have been on longer than this value, a trip occurs. As Fig. 24 shows, a trip occurs at 0.78s into the event; which is nearly 0.35 seconds after the initial fault occurs. This is illustrated by the signal T2OUT changing from a logical zero to a logical one in the figure. In this case, the hardware filtered detection algorithm detected the SSO event before the phasor based method. Fig. 24 Generic SSO Relay Response to Event 1 Pg. 10

11 Fig. 25 Generic SSO Relay Response to Event 2B The simulated relay response for event 2B is shown in Fig. 25. As this figure shows, the relays hardware filters do not respond to the subsynchronous frequency injection into the currents and voltages, as can be seen in the two figures below the raw event captures. This is because the SSO mode frequency falls far outside the cutoff frequency of the hardware filtering. Because of this, the signal SFi, being the instantaneous subsynchronous power, is zero during the entire event. The signal SFp is the power calculated using the phasor data amplitudes. Because the frequency of the SSO content in this event is captured only in the phasor data channels, the SSO event is seen by only the phasor based portion of the protection algorithms. The fourth time series plot in Fig. 25 illustrates the instantaneous RMS content of each of the signals, as well as the settings of 1 and 2. As can be seen in the last capture in Fig. 25, when the RMS content is above the pickup level of the settings, the frequency tracking function begins to detect the frequency of oscillation of the SSO mode. Once this occurs, the variable X1 in Fig. 23 show that the period timer (1/f1*setting) begin to change from its initial value of 3 to the actual oscillation period multiplied by these values. But, in this case, because the subsynchronous mode is highly damped, the RMS content falls below the pickup setting before timer1 reaches its terminal value. This drops out the timer which was timing to trip. Once the second fault occurs, the SSO RMS content grows again. This time, however, the event ends due to the line tripping from to the previously installed SSO relay at the windfarm terminal monitoring this event. If the new SSO relay algorithm had been in place during this event, it is highly likely that the SSO event would have continued to decrease in amplitude and completely be attenuated in magnitude, resulting in no SSO event detection. This is acceptable because for highly damped SSO modes, there is little desire to trip electrical infrastructure. The simulated relay response for event 3B is shown in Fig. 26. This event is very similar to the response of the simulated relay of event 1. The main difference is that in this event playback, there is pronounced subsynchronous voltage and current waveform content as soon as the event starts. As this figure shows, the relays hardware filters are responding to the subsynchronous waveform content before the 1.7s unstable SSO event into the currents and voltages, as can be seen in the two time series plots below the raw event captures. The signal SFi, in the middle figure, is the subsynchronous power calculated using the hardware filtered channels and quickly grows immediately following the unstable SSO event which occurs at approx. 1.7s in the event playback. The time series plot below the filtered operating quantity amplitudes in Fig. 26 illustrates the instantaneous RMS content of each of the signals, as well as the settings of 1 and 2. As can be seen in the last time series plot in Fig. 26, when the RMS content is above the pickup level of the settings, the frequency tracking function begins to detect the frequency of oscillation of the SSO modes. Once this occurs, the variables X1 and X2 in Fig. 23 show that the period timers (1/f1*setting and 1/f2*setting) begin to change from their initial values of 3 and 4 to the actual oscillation period multiplied by these values. Because the frequency of the SSO content in this event is near the cutoff frequencies of both the hardware filters and the phasor data channels, the SSO event is detected by both protection algorithms filters, but the phasor based algorithm frequency tracking function cannot determine a frequency of oscillation and therefore does not operate first. Pg. 11

12 Once the timers recognize that the inputs to the timers have been on longer than this value, a trip occurs. As Fig. 26 shows, a trip occurs at 1.94s into the event; which is nearly 0.25 seconds after the SSO event becomes unstable. This is illustrated by the signal T2OUT changing from a logical zero to a logical one in the figure. In this case, the hardware filtered detection algorithm detected the SSO event before the phasor based method. To further illustrate the detection elements capability, faults were simulated to demonstrate the security of the relay algorithm to fault conditions external to system infrastructure where the SSO algorithm would be applied. As can be observed in Fig. 23, the relay detection algorithm is somewhat immune to large phase to ground faults because a large zero sequence current will continue to feed the previously measured phasor power delivered into the bandpass filter of the algorithm, which will allow the relay algorithm to effectively ignore the fault. Because of this, a permanent three phase fault with a typical high speed reclose external to the relay was simulated such that this immunity would be bypassed. The simulation results of this scenario with the settings used in the Event 3 simulation are seen in Fig. 27. As can be seen in the figure, the three phase currents increase rapidly during the faults, while the voltages decrease in magnitude. Each fault is nearly 4 cycles long, and the open internal time of the high speed reclose is a typical AEP value of twenty cycles. The hardware filtered channels detect these large step changes in magnitude but quickly settle out. However, the bandpass filtered output of the phasor based instantaneous power calculation increases rapidly following each of the faults and then falls after each of the faults. These rapid rises and falls each increase the magnitude of the phasor based RMS power above the pickup setting of the algorithm detection element. This releases the frequency detection algorithm. Because the frequency detection algorithm does not detect an SSO mode, the timer value used to detect the SSO condition never times out and the SSO detect condition is not declared. This is a proper operation of the relay element and demonstrates the security of the detection algorithm in the presence of a challenging fault against which the relay should hold. Fig. 26 Generic SSO Relay Response to Event 3B Pg. 12

13 Fig. 27 Generic SSO Relay Response to Three Phase Permanent Fault Conditions with High-Speed Reclose IV. CONCLUSIONS Several SSCI events that have occurred on the AEP system have been illustrated. It has been demonstrated using historical event record capture signal analysis that SSO can be detected by modeling the waveforms captured during the event using balanced nominal frequency voltages and currents in sum with balanced subsynchronous voltages and currents. Demonstrations have been provided using a detection algorithm which is both secure and dependable in the presence of common power system phenomenon and SSO event phenomenon. s guidance has been provided which would assist with setting this detection algorithm. REFERENCES [1] [2] P.M. Anderson, and R.G. Farmer, Series Compensation of Power Systems, PBLSH! Inc, pp , [3] G. Irwin, Simulation and Analysis Methods for SSR/SSTI/SSCI, PUCT Panel Session, Nov 19th Available online: PROJECT4350/AnalysisMethods_SSR_SSCI_SSTI.ppt [4] S.H. Horowitz, A.G. Phadke, Power System Relaying, 3rd ed., New Jersey: Wiley, pp [5] L.C. Gross, Sub-Synchronous Grid Conditions: New Event, New Problem, and New Solutions, 37th Annual Western Protective Relay Conference, Oct 19-21, [6] P.M. Anderson, Power System Protection, Vol.1, New Jersey: Wiley, 1999, pp [7] Z. Zhang, P.Eng., Ilia Voloh, J. Cardenas, I. Antiza, F Iliceto. Inter- Area Oscillation Detection by Modern Digital Relay. CIGRE St. Petersburg, May [8] V.K. Ingle, and J.G. Proakis, Digital Signal Processing using MATLAB, Brooks/Cole, 2000, pp [9] E.O. Schweitzer, and D. Hou, Filtering for Protective Relays, 19 th Annual Western Protective Relay Conference, Spokane, Washington, Oct 20-22, BIOGRAPHIES Zachary P. Campbell received his bachelor s degree from the University of Akron, in Akron, Ohio, in 2008, and his master s degree from The Ohio State University, in Columbus, Ohio, in Zak is a principal engineer at American Electric Power (AEP) where he has worked in various protection and control departments since He is a member of IEEE, CIGRE and is a registered professional engineer in the state of Ohio. zpcampbell@aep.com Kiril Andov received his bachelor s degree from the University of St Cyril & Methodius in Skopje, Macedonia, Currently, Kiril works at AEP as senior engineer in system performance analysis group. In his current role, Kiril performs various EMT studies and provides support to various departments. Kiril is AEP s subject matter expert of insulation coordination studies area and subsynchronous oscillations studies area. kandov@aep.com Shawn Coppel graduated Magna Cum Laude from DeVry Institute of Technology in 1996 with a bachelor s degree in Electronics Engineering Technology. Shawn is a Senior Engineering Technologist for American Electric Power (AEP) where he has worked in the transmission protection & control laboratory since Previously, Shawn worked in telecommunications prototype development at Lucent Technologies. Shawn is a Veteran of the U.S. Army and specialized in radio systems repair. scoppel@aep.com Pg. 13