- Email: [email protected]
Practical and theoretical approaches to detecting grid dynamics and stability
Control of Transmission and Distribution Smart Grids IFAC and CIGRE/CIRED Workshop on October 11-13, 2016. Prague, Czechon Republic IFAC and Available Control ofCIGRE/CIRED TransmissionWorkshop and Distribution Smartonline Grids at www.sciencedirect.com Control Transmission and Distribution Smart Grids Octoberof 11-13, 2016. Prague, Czech Republic October 11-13, 2016. Prague, Czech Republic
ScienceDirect
IFAC-PapersOnLine 49-27to (2016) 007–011 Practical and theoretical approaches detecting grid dynamics and stability Practical Practical and and theoretical theoretical approaches approaches to to detecting detecting grid grid dynamics dynamics and and stability stability
Dr. Christian Rüster* Dr. Christian Rüster* Dr. Christian Rüster* *A. Eberle GmbH & Co. KG, Nuremberg, Germany (e-mail: [email protected]). *A. Eberle GmbH & Co. KG, Nuremberg, Germany (e-mail: [email protected]). *A. Eberle GmbH & Co. KG, Nuremberg, Germany (e-mail: [email protected]).
Abstract: Electricity supply systems used to be operated in accordance with strict hierarchical principles. Major power stations supply controlled the grid and balanced load aswith wellstrict as stabilized voltage levels. Abstract: Electricity systems usedfrequency to be operated in accordance hierarchical principles. Abstract: Electricity supply systems used toproducers be operated in accordance withon strict hierarchical principles. In such a system the high inertia of power posed a natural limit the amount of observable Major power stations controlled the grid frequency and balanced load as well as stabilized voltage levels. Major power stations controlled the grid frequency and balanced load as well as stabilized voltage levels. dynamic resilience against Blackout phenomena overload situations. In such aeffects systemand the maintained high inertiagood of power producers posed a natural limit onduring the amount of observable In such a system the high inertia ofdistributed power producers posed a natural limit on the amount of observable However, as more and more small, energy producers enter into the picture, inertia is lost and dynamic effects and maintained good resilience against Blackout phenomena during overload situations. dynamic effects and maintained good resilience against Blackout phenomena during overload situations. concern for dynamic instabilities grows. This paper gives examples of some dynamic disturbances of However, as more and more small, distributed energy producers enter into the picture, inertia is lost and However, grids as more andstress more and small, distributed energy producers enter intoquick the picture, inertia isdetection, lost and electrical under points out innovative methods for their and sensitive concern for dynamic instabilities grows. This paper gives examples of some dynamic disturbances of concern for dynamic instabilities grows. This detection. paper gives examples of some dynamic disturbances such as Wavelet-based subharmonic oscillation Furthermore, when electrical grids aredetection, closer of to electrical grids under stress and points out innovative methods for their quick and sensitive electrical grids under stress and points out innovative methods for their quick and sensitive detection, breakdown they can enter a nonlinearoscillation regime where the statistical properties the state grids variables become such as Wavelet-based subharmonic detection. Furthermore, whenofelectrical are closer to such Wavelet-based subharmonic oscillation detection. Furthermore, when electrical grids arebecloser to more as and more We argue that inwhere some situations these statistical can a good breakdown theypronounced. can enter a nonlinear regime the statistical properties of fluctuations the state variables become breakdown theyfor canan enter a nonlinear regimeand where the statistical properties of the statecriticality variables with become early indicator upcoming Blackout propose how to capture breakdown an more and more pronounced. We argue that in some situations these statistical fluctuations can be a good more and more pronounced. We argue thatoninthe some situations these statistics. statistical fluctuations can be a good appropriate measurement algorithm based analysis of voltage early indicator for an upcoming Blackout and propose how to capture breakdown criticality with an early indicator for an upcoming Blackout and propose how to capture breakdown criticality with an appropriate measurement algorithm based on the analysis of voltage statistics. © 2016, IFAC (International Federation Automatic Control) Hosting by Elsevier Ltd.Detection, All rights reserved. Keywords: Mechanical Analogon of ofInter-Area Oscillations, Blackout Early GDASys appropriate measurement algorithm based on the analysis of voltage statistics. Complement to WAMS, HVDC+FACTS devices, Oscillations During a Blackout, FFT – Fast FourierKeywords: Mechanical Analogon of Inter-Area Oscillations, Blackout Early Detection, GDASys Keywords: Mechanical Analogon of Inter-Area Oscillations, Blackout Early Detection, GDASys Transform, Stability – Bifurcation Analysis, Voltage Collapse, Stochastic Processes Complement to WAMS, HVDC+FACTS devices, Oscillations During a Blackout, FFT – Fast FourierComplement to WAMS, HVDC+FACTS devices, Oscillations During a Blackout, FFT – Fast Fourier Transform, Stability – Bifurcation Analysis, Voltage Collapse, Stochastic Processes Collapse, Transform, Stability – Bifurcation Analysis, Voltage Stochastic Processes presence of unusual and poorly damped low frequency 1. INTRODUCTION oscillations and after the blackout. Such presence of before, unusualduring and poorly damped low frequency 1. INTRODUCTION presence of unusual and poorly damped low frequency dynamic oscillatory phenomena are known to cause Wide area synchronous power systems offer great benefits oscillations before, during and after the blackout. Such 1. INTRODUCTION before, duringaction and aftertrippings the blackout. Such unwanted protection which turn for deregulated power energysystems markets. Flexible energy oscillations dynamic oscillatoryrelay phenomenaandare known to incause Widetodays area synchronous offer great benefits dynamic oscillatory phenomena are known to cause can aggravate stability problems in an already stressed Wide area synchronous power systems offer greatas benefits transfer across geographical and political borders well as protection relay action and trippings which in turn for todays deregulated energy markets. Flexible energy unwanted protection relay to action and trippings which in turn system, eventually leading blackouts. for todays through deregulated energy ofmarkets. Flexible energy unwanted lower the pooling generation capacities aggravate stability problems in an already stressed transfercosts across geographical and political borders as wellare as can can aggravate stability problems in an already stressed transfer across geographical and political bordersgrid. as well as some of the key advantages of an interconnected Due to system, eventually leading to blackouts. lower costs through the pooling of generation capacities are system, eventually leading to blackouts. lower costspressures, through the of frequently generation operated capacitiesclose are economic suchpooling grids are some of the key advantages of an interconnected grid. Due to some of stability the key advantages oflimit, an interconnected grid. Due to to their limit. In this the presence of dynamic economic pressures, such grids are frequently operated close economic pressures, such frequency grids are frequently operated weak close effects such aslimit. low oscillations to their stability In this limit, the presence ofondynamic to their stability limit. In this limit, the presence of dynamic interconnections develop into a fundamental effects such as can lowrapidly frequency oscillations on weak effects such as low frequency on weak bottleneck for power grid oscillations stability. interconnections cantransfer rapidlyand develop into a fundamental interconnections can rapidly develop into a fundamental bottleneckoscillations for power transfer and are grid stability.in virtually all Interarea (0,1-1Hz) bottleneck for power transfer and grid present stability. large power systems(0,1-1Hz) on the are transmission Under Interarea oscillations present inlevel. virtually all Interarea oscillationsthese (0,1-1Hz) arewell present in virtually all ambient conditions, modes are damped, amplitudes large power systems on the transmission level. Under large power systems on the transmission level.however Under are negligible. Stability problems become apparent ambient conditions, these modes are well damped, amplitudes ambient these modes are well damped, amplitudes only by conditions, observing the problems system behavior after excitations. are negligible. Stability become apparent however are negligible. Stability problems become apparent however Examples for excitations are power system line only by observing the system behavior after faults, excitations. only by observing the system behavior after excitations. switching, generator disconnection andsystem even the lossline or Examples for excitations are power faults, Examples for excitations are power system faults,into linea application of loads. An unstable system will enter switching, generator disconnection and even the loss or switching, disconnection and even the losswith or period of generator application of machine loads. An rotor unstableangle systemoscillations, will enter into a application of loads. An unstable system will voltage enter into a corresponding power flows and oscillatory and period of machine rotor angle oscillations, with period ofvariations. machineDamage rotorto the angle oscillations, with frequency infrastructure, unwanted corresponding power flows and oscillatory voltage and corresponding flows andeven oscillatory trippings inpower extreme cases blackoutsvoltage are thenanda frequency and variations. Damage to the infrastructure, unwanted frequency variations. Damage to the infrastructure, unwanted Fig. 1. GDASys devices for monitoring oscillatory danger. trippings and in extreme cases even blackouts are then a trippings and in extreme cases even blackouts are then a phenomena in the electrical grid in rack (DMR-D) / mobile danger. Fig. 1. GDASys devices for monitoring oscillatory A very recent and dramatic example for this can be found in Fig. danger. 1. GDASys for grid monitoring version (DA-Box 2000). phenomena in thedevices electrical in rack oscillatory (DMR-D) / mobile an official report to the Indian CERC A very recent andsubmitted dramatic example for this can commission be found in phenomena in the electrical grid in rack (DMR-D) / mobile version (DA-Box 2000). A very recent and dramatic example for this can be found in analyzing widesubmitted area blackouts happened on 30th and version (DA-Box 2000). an officialthe report to the that Indian CERC commission Poorly damped oscillations are thus an important indicator an official report submitted the Indian CERC commission 31st of July 2012 in blackouts the to Northern India electricity analyzing the wide area that happened on 30th grid and for the dynamic stability of the power grid. An early analyzing the wide area blackouts happened and Poorly damped oscillations are thus an important indicator (CERC Summarizing eventsthat during ofon the30th overall 31st of 2012). July 2012 in the Northern Indiaoneelectricity grid Poorly damped oscillations are thus an important indicator detection of these oscillatory phenomena is theAn key to for the dynamic stability of the power grid. early 31st ofblackout July 2012 in theever, Northern Indiaclearly electricity grid largest incidents the report states the for (CERC 2012). Summarizing events during one of the overall the dynamic stability of the power grid. An early detection of these oscillatory phenomena is the key to (CERC 2012). Summarizing events during one of the overall largest blackout incidents ever, the report clearly states the detection of these oscillatory phenomena is the key to largest blackout incidents ever, the report clearly states the Copyright © 2016 IFAC 7 2405-8963 © 2016, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved. Copyright 2016 IFAC 7 Control. Peer review©under responsibility of International Federation of Automatic Copyright © 2016 IFAC 7 10.1016/j.ifacol.2016.10.691
2016 IFAC CTDSG 8 Christian Rüster et al. / IFAC-PapersOnLine 49-27 (2016) 007–011 October 11-13, 2016. Prague, Czech Republic
increased situational awareness. This paper describes two promising approaches for dynamic voltage stability assessment: (1) frequency-domain based Wavelet analysis (2) monitoring the time evolution of the standard deviation of the bus voltage. The authors argue that the corresponding algorithms allow a particularly efficient detection of power system oscillations and stability and give practical measurement examples evidencing their performance. These measurements were taken with GDASys devices as shown in Fig.1.
disturbance. Such an oscillation could for example be related to a malfunction or some dynamical interactions between generators or loads. The damping monitor gives a precise answer as to what kind of disturbance occurred and when, how large it was and how long the disturbance was measurable. The usage of Wavelet- instead of conventional Fourier analysis here is not merely incidental but of utmost importance for true early detection: the choice of a proper Wavelet allows a significantly improved detection time because the time frequency uncertainty relationship is more favourable in online measurement situations. This will be demonstrated below in a measurement where inter-area oscillations were observed in near-real-time after a sudden generation loss.
2. THEORETICAL CONSIDERATIONS: EFFICIENT DETECTION OF GRID OSCILLATIONS Traditionally, grid operators use SCADA systems to monitor and control their network of power lines. This involves measuring average values of important operational parameters with a periodicity of a few seconds. The measurement data is then sent to a centralized supervisory station for review. To enable the analysis of dynamic effects in the grid, one has to drastically increase the measurement data rate. Suitable measurement tools for this purpose are offered by a “Grid Dynamics Analyzing System” as described in W. Haussel et al. (2004) and M. Hofbeck et al. (2008a,b). The variant in the top part of Fig. 1 consists of the core measurement unit DMR-D, a Windows-based industrial PC for data storage and a communication unit supporting a wide range of industry standard protocols. The mobile version of the device and a separate communication device is shown in the lower half of the picture. The devices offer a usable bandwidth between 5 mHz and 98 Hz at a high resolution of 5 mHz.
Whenever a dynamic deviation of system parameters from normal has been established, the problem to assess the criticality of the state of the system starts. Assessing the actual proximity to voltage breakdown for a complex system such as an interconnected electric grid is a difficult task and so far no single universally efficient and well-accepted assessment scheme is known. However, recent research into complex systems inside and outside the domain of electrical grids indicates that statistical tools such as autocorrelation and variance measurements of key observables can serve as useful indicators during the transition into breakdown. Particularly one body of research brought forth by d. Podolsky et al. (2013), G. Ghanavati (2013, 2014) and E. Cotilla-Sanchez (2012) points out a phenomenon called "critical slowing down" which is characterized by increasing fluctuations in the system, particularly in the nonlinear regime close to breakdown. G. Ghanavati (2013) shows, based on simulations of power systems approaching critical bifurcations, that especially the statistical variance in bus voltages can be a good indicator of the proximity to Blackout. Below the authors of this paper show real data measured in the German transmission system during the blackout which occurred in Turkey in March 2015. While this Blackout did not affect the European or German power system in a critical way, the data employs similar statistical analysis methods and shows that very sensitive event detection is possible. Corresponding measurement devices based on statistical algorithms thus show very good promise as power system instability early warning tools.
Fig. 2. GDASys damping monitor principle. Power systems disturbance model, a damped sinusoid.
The following sections present measurement results obtained during large disturbances in various actual power networks by means of measurement equipment of the A. Eberle GDASys family of devices. The purpose is to demonstrate in practice the efficiency of Wavelet- and statistical variance/standard deviation-based measurement algorithms for power system dynamics assessment.
Another distinguishing characteristic is the usage of highly sensitive methods of frequency domain analysis. Here, the well-known Fourier- as well Wavelet-analysis is available. The Wavelet analysis is employed to realize a function called damping monitor. Its task is to detect oscillatory disturbances on the power line that deviate from nominal system voltage or frequency behavior. By means of high precision Waveletbased Prony analysis, the damping monitor searches the full usable frequency spectrum for power system disturbances matching the fundamental model of a damped sinusoid (Fig. 2). Its output consists of the measured time stamp, frequency, amplitude, duration and the damping factor of a detected
3. PRACTICAL RESULTS: OSCILLATION DETECTION The following two case studies demonstrate the practical application of a Wavelet-based damping monitor during major power system disturbances.
8
2016 IFAC CTDSG October 11-13, 2016. Prague, Czech Republic Christian Rüster et al. / IFAC-PapersOnLine 49-27 (2016) 007–011
3.1 Major oscillation event caused by equipment failure
While the strongest disturbances started at around 70 minutes, it is very interesting to note that the actual cause of the disturbance - the equipment failure and separation of the grid – actually happened an hour earlier at 10 minutes. In the graph this coincides with three smaller amplitude oscillations of 0,2-0,3% in amplitude. While at this point the system was in a weak condition with a poorly damped 0,95 Hz mode waiting to be excited, the actual strong excitation came only an hour later.
In the course of a wide area measurement project aimed at monitoring a substantial part of a large transmission line network, three fixed installation GDASys / DMR-D measurement devices were introduced at strategically chosen locations inside the grid. During one notable incident, an equipment fault on a single phase at a coal fired power station triggered a series of protection device operations. Circuit breakers tripping resulted in a situation where the power station was only connected to the rest of the power system via a weak 200km connection as opposed to the usual 50km line. This effectively constituted a major reconfiguration of the local grid and power flows on the weak remaining lines of the area in which the DMR-D device (named C2) was situated, increased critically. Most importantly, this resulted in the emergence of a new resonance frequency at 0,95 Hz. During the main part of the disturbance, when this new mode was excited by an unknown stimulus, large voltage oscillations were present in the system for about 25 minutes.
The lower panel in Fig. 3 represents an alternate analysis of the criticality of measured oscillations during the event. The plotted data points belong to the same detected oscillation events that are plotted as a time series in the upper panel of the figure. Here, mode oscillation amplitudes are plotted against their duration. Only oscillations falling within the shaded box can be considered as truly safe. Everything outside the box should be regarded as either a warning or even an alarm signal. This includes oscillations either with amplitudes exceeding 0,1% or having a duration longer than 10s. For the 0,95 Hz mode analysed here, a duration of 10s means that the individual disturbance is visible to the system for more than 10 periods. This is a sign of poor damping and thus a warning signal for system stability. Since some of the measured durations even exceed 70 s, they are clearly in the critical regime. Using the data presented as in Fig. 3, it is clear that the disturbance event could have been quickly identified by an operator, especially since such kind of data is available from the Wavelet algorithm in quasi real-time. 3.2 Major Oscillation Event Caused by Sudden Loss of Generation Capacity During another incidence in the same measurement period, multiple generators from a coal fired power station tripped simultaneously. Sudden losses in generation capacity constitute considerable stress for any kind of grid and are prone to excite oscillatory disturbances. This sudden loss of generation capacity affected numerous transmission lines. Notably the disturbance was strong enough so that tie-lines to several neighbouring states with interconnected electricity systems were tripped due to the operation of the Under Frequency Load Shedding relays (UFLS). The idea of tripping the interconnection in this protection scheme was to shed sufficient load to rebalance the grid.
Fig. 3. Time series of 0,95 Hz mode amplitudes measured during a major oscillation event caused by equipment failure (upper panel). Mode criticality analysis of the same event (lower panel).
This oscillation event manifested itself with very large oscillation amplitudes exceeding 2% in the damping monitor of DMR-D device C1. The time series of measured oscillation amplitudes is displayed in Fig. 4 (blue data points). It can be seen that as far as oscillatory disturbances are concerned, the event was over in less than one minute, with amplitudes clearly exceeding the alarm level before returning to ambient conditions. A criticality analysis plot similar to the one shown in the lower panel of Fig. 3 reveals that the large amplitude oscillations are negatively damped and also exceed the duration criterion of 10s. The largest oscillation amplitudes are found for a mode at 0,6 Hz. The longest sustained oscillations however exist for the 0,35 Hz
Device C2 was able to pick up the corresponding instabilities in high resolution using the Wavelet-based damping monitor algorithm. A time series of amplitudes of the 0,95 Hz mode is shown in the upper panel of Fig. 3. The most critical part of the event is centred in the middle of the panel. It can be seen that oscillation amplitudes pick up within a very short time and stay at a very high level of 0,8% to 1% for about 20 minutes before dying off over the course of a few minutes. This amplitude level definitely indicates a highly stressed system and reveals considerable power swings. 9
9
2016 IFAC CTDSG 10 Christian Rüster et al. / IFAC-PapersOnLine 49-27 (2016) 007–011 October 11-13, 2016. Prague, Czech Republic
and 0,7 Hz interarea modes, with oscillation durations exceeding 34 seconds in both cases. The frequencies of the interarea modes are well known in that network. From this it can be concluded that two weak transmission corridors were most strongly affected by the event: one connecting with a neighbouring country and one leading to a remote load center.
capacity) and the subsequent loss of Atlas thermal power St. (in Hatay province, at 09:02 am) as well as Dicle hydro (Southeast of the country). The resulting frequency variations reportedly lead to automatic safety measures in which the UCTE grid cut electrical connections with Turkey for safety reasons. The blackout subsequently ensued at 09.36 am. At the time of the separation the net power flow between Turkey and the UCTE grid was fortunately only at a very moderate level. This means that viewed from the European perspective, cutting tie lines to Turkey constituted only a mild strain on the system. During the whole incident, five DMR-D monitoring installations were active in the German transmission grid. A post-event analysis indicates that even though for the far away German grid this was only a very minor disturbance, the DMR-D's statistics based stability exponent measurement algorithm was sensitive enough to pinpoint the dynamics of the grid separation. Taking into account the time evolution of the standard deviation of power system frequency and bus voltage, the stability exponent aims to be a key indicator for dynamic strain in the electrical grid, in a way that is very similar to the theoretical analysis independently suggested by the "critical slowing down" phenomenon particularly as described by E. Cotilla-Sanchez in 2012.
Fig. 4. Time series of damping monitor amplitudes measured with DMR-D device C1 during a major oscillation event caused by generation loss. Despite the event lasting only about a minute, device C1 was able to resolve multiple modes, some of which even happened simultaneously. This superior response time is due to advantages of the employed wavelet analysis algorithms of the device and distinguishes it from typically used Fourier techniques that have difficulties in the subsynchronous range. The high time resolution of the Wavelet analysis is revealed by the comparison with a FFT employing a sliding 100s time window (red data in Fig. 4). The blue Wavelet data rises immediately at what is known to be the onset of oscillations. In comparison, the response of the red 100s FFT is delayed and artificially broadened. In fact, when FFT amplitudes exceed the alarm level, the actual event is already over. Please not that this is actually not due a suboptimal performance of this particular FFT filter implementation but due to a fundamental limitation of any FFT: a filter of sufficient frequency resolution for detecting relevant sub 1Hz oscillations has a specific, corresponding certain minimum window length. Time lag for online measurements is then determined by the window length and cannot be improved without frequency resolution loss.
From a theoretical and algorithmic point of view, increasing power system strain can be understood to cause increasing deviations from the mean value of frequency and / or bus voltage. The stability exponent mirrors the rate of change of the standard deviation of bus frequency and bus voltage hence the value of the stability exponent will also increase. Phases of system stabilization reverse this trend and cause the stability exponent to fall. In practice, the moment of power system separation of the Turkish and the European grid represents just this kind of momentary system disturbance which quickly "heals off" in this case, since the net power flow difference was only quite small.
3.3 March 2015 blackout in Turkey observed from within the German transmission grid On the morning of the 31st of March 2015, a major blackout was reported for the national Turkish electricity grid (ENTSO-E 2015). As a result, 80 out of 81 Turkish provinces were without electricity during busy workday hours, with millions of people stuck on the way to work, as also public train and subway systems were paralyzed.
Fig. 5. Time series of stability exponent measurements in five different locations in German transmission grid, at the time of the Turkey blackout in march 2015. Nevertheless, as is seen in the data shown in Figure 5, all five devices distributed from North to South within Germany showed a distinct and easily recognizable feature in the stability exponent measurement at the exact time of system separation (09:36 am). Here, closer investigation even reveals
According to press reports, the blackout was a consequence of several power stations dropping off the grid, including a major power station in the Aegean region (2000 MW 10
2016 IFAC CTDSG Christian Rüster et al. / IFAC-PapersOnLine 49-27 (2016) 007–011 October 11-13, 2016. Prague, Czech Republic
that the size of the peak correlates with the approcimate "electrical separation" between the incident in Turkey and the measurement location in Germany. The data shows that events happening in the grid can be detected in a very sensitive way. It furthermore is an indication that indices such as the standard deviation based stability exponent which is in line with theoretical conclusions about critical slowing down before blackouts can be a useful early indicator for the proximity of critical states in the grid. Please note however that the rather indirect measurement data shown should not be interpreted as a direct measure to the proximity to actual breakdown.
11
E. Cotilla-Sanchez, P. D. H. Hines, C. M. Danforth "Predicting Critical Transitions From Time Series Synchrophasor Data", IEEE Transactions on Smart Grid, 2012 D. Podolsky, K. Turitsyn, "Critical slowing-down as indicator of approach to the loss of stability", arXiv:1307.4318v1, 2013 G. Ghanavati, P. D. H. Hines and T. I. Lakoba "Investigating early warning signs of oscillatory instability in simulated phasor measurements", arXiv:1312.5371v1, 2013 G. Ghanavati, P. D. H. Hines and T. I. Lakoba, "Identifying Useful Statistical Indicators of Proximity", arXiv:1410.1208v1, 2014
4. CONCLUSIONS DMR-D measurements taken in actual high voltage transmission grids have pointed out the benefits of oscillation monitoring for dynamic stability assessment. Identifying characteristic disturbances in the system enables the user to distinguish normal oscillatory behaviour from significant disturbances. Particularly the employed damping monitor based on Wavelet analysis is an accurate tool for measuring oscillatory modes and allows a straightforward assessment of an oscillation’s criticality. Wavelet based signal conditioning surpasses the capabilities of conventional FFT and was shown to be fast enough to register even very short duration disturbances.
ENTSO-E Report on Blackout in Turkey on 31 March 2015,
available on the ENTSO-E website: https://www.entsoe.eu/Documents/SOC%20do cuments/Regional_Groups_Continental_Europe /20150921_Black_Out_Report_v10_w.pdf
Furthermore, the statistical (standard deviation-based) analysis of the power system observables voltage and frequency has shown good promise to be a sensitive tool for the assessment of proximity to instability in stochastic power systems, i.e. grids which are in a nonlinear regime close to voltage breakdown. This promise roots both in the sensitive detection of a recent blackout in Turkey from afar, when viewed together with the independently derived "critical slowing down" concept.
REFERENCES "Report on the grid disturbance on 30th July 2012 and grid disturbance on 31st July 2012" available on www.cercind.gov.in W. Haussel, M. Hofbeck, T.Sybel, M. Fette, "Collapse Protection System - Basics and Applications", Proceedings of the South African Power System Protection Conference 2004 M. Hofbeck, L. Mayer, T.Sybel, M. Fette, B. Werther, I. Winzenick, "Collapse Prediction Relay CPR-D - Theory and Applications - Implementation into SCADASystems to Support Security Stability Assessment", Proceedings of the South African Power System Protection Conference 2008 M. Hofbeck, T. Sybel, M. Fette, I. Winzenick, "Measurements of the Dynamical Status of HighVoltage-Networks with CPR-D", Proceedings of the South African Power System Protection Conference 2008
11
ScienceDirect
IFAC-PapersOnLine 49-27to (2016) 007–011 Practical and theoretical approaches detecting grid dynamics and stability Practical Practical and and theoretical theoretical approaches approaches to to detecting detecting grid grid dynamics dynamics and and stability stability
Dr. Christian Rüster* Dr. Christian Rüster* Dr. Christian Rüster* *A. Eberle GmbH & Co. KG, Nuremberg, Germany (e-mail: [email protected]). *A. Eberle GmbH & Co. KG, Nuremberg, Germany (e-mail: [email protected]). *A. Eberle GmbH & Co. KG, Nuremberg, Germany (e-mail: [email protected]).
Abstract: Electricity supply systems used to be operated in accordance with strict hierarchical principles. Major power stations supply controlled the grid and balanced load aswith wellstrict as stabilized voltage levels. Abstract: Electricity systems usedfrequency to be operated in accordance hierarchical principles. Abstract: Electricity supply systems used toproducers be operated in accordance withon strict hierarchical principles. In such a system the high inertia of power posed a natural limit the amount of observable Major power stations controlled the grid frequency and balanced load as well as stabilized voltage levels. Major power stations controlled the grid frequency and balanced load as well as stabilized voltage levels. dynamic resilience against Blackout phenomena overload situations. In such aeffects systemand the maintained high inertiagood of power producers posed a natural limit onduring the amount of observable In such a system the high inertia ofdistributed power producers posed a natural limit on the amount of observable However, as more and more small, energy producers enter into the picture, inertia is lost and dynamic effects and maintained good resilience against Blackout phenomena during overload situations. dynamic effects and maintained good resilience against Blackout phenomena during overload situations. concern for dynamic instabilities grows. This paper gives examples of some dynamic disturbances of However, as more and more small, distributed energy producers enter into the picture, inertia is lost and However, grids as more andstress more and small, distributed energy producers enter intoquick the picture, inertia isdetection, lost and electrical under points out innovative methods for their and sensitive concern for dynamic instabilities grows. This paper gives examples of some dynamic disturbances of concern for dynamic instabilities grows. This detection. paper gives examples of some dynamic disturbances such as Wavelet-based subharmonic oscillation Furthermore, when electrical grids aredetection, closer of to electrical grids under stress and points out innovative methods for their quick and sensitive electrical grids under stress and points out innovative methods for their quick and sensitive detection, breakdown they can enter a nonlinearoscillation regime where the statistical properties the state grids variables become such as Wavelet-based subharmonic detection. Furthermore, whenofelectrical are closer to such Wavelet-based subharmonic oscillation detection. Furthermore, when electrical grids arebecloser to more as and more We argue that inwhere some situations these statistical can a good breakdown theypronounced. can enter a nonlinear regime the statistical properties of fluctuations the state variables become breakdown theyfor canan enter a nonlinear regimeand where the statistical properties of the statecriticality variables with become early indicator upcoming Blackout propose how to capture breakdown an more and more pronounced. We argue that in some situations these statistical fluctuations can be a good more and more pronounced. We argue thatoninthe some situations these statistics. statistical fluctuations can be a good appropriate measurement algorithm based analysis of voltage early indicator for an upcoming Blackout and propose how to capture breakdown criticality with an early indicator for an upcoming Blackout and propose how to capture breakdown criticality with an appropriate measurement algorithm based on the analysis of voltage statistics. © 2016, IFAC (International Federation Automatic Control) Hosting by Elsevier Ltd.Detection, All rights reserved. Keywords: Mechanical Analogon of ofInter-Area Oscillations, Blackout Early GDASys appropriate measurement algorithm based on the analysis of voltage statistics. Complement to WAMS, HVDC+FACTS devices, Oscillations During a Blackout, FFT – Fast FourierKeywords: Mechanical Analogon of Inter-Area Oscillations, Blackout Early Detection, GDASys Keywords: Mechanical Analogon of Inter-Area Oscillations, Blackout Early Detection, GDASys Transform, Stability – Bifurcation Analysis, Voltage Collapse, Stochastic Processes Complement to WAMS, HVDC+FACTS devices, Oscillations During a Blackout, FFT – Fast FourierComplement to WAMS, HVDC+FACTS devices, Oscillations During a Blackout, FFT – Fast Fourier Transform, Stability – Bifurcation Analysis, Voltage Collapse, Stochastic Processes Collapse, Transform, Stability – Bifurcation Analysis, Voltage Stochastic Processes presence of unusual and poorly damped low frequency 1. INTRODUCTION oscillations and after the blackout. Such presence of before, unusualduring and poorly damped low frequency 1. INTRODUCTION presence of unusual and poorly damped low frequency dynamic oscillatory phenomena are known to cause Wide area synchronous power systems offer great benefits oscillations before, during and after the blackout. Such 1. INTRODUCTION before, duringaction and aftertrippings the blackout. Such unwanted protection which turn for deregulated power energysystems markets. Flexible energy oscillations dynamic oscillatoryrelay phenomenaandare known to incause Widetodays area synchronous offer great benefits dynamic oscillatory phenomena are known to cause can aggravate stability problems in an already stressed Wide area synchronous power systems offer greatas benefits transfer across geographical and political borders well as protection relay action and trippings which in turn for todays deregulated energy markets. Flexible energy unwanted protection relay to action and trippings which in turn system, eventually leading blackouts. for todays through deregulated energy ofmarkets. Flexible energy unwanted lower the pooling generation capacities aggravate stability problems in an already stressed transfercosts across geographical and political borders as wellare as can can aggravate stability problems in an already stressed transfer across geographical and political bordersgrid. as well as some of the key advantages of an interconnected Due to system, eventually leading to blackouts. lower costs through the pooling of generation capacities are system, eventually leading to blackouts. lower costspressures, through the of frequently generation operated capacitiesclose are economic suchpooling grids are some of the key advantages of an interconnected grid. Due to some of stability the key advantages oflimit, an interconnected grid. Due to to their limit. In this the presence of dynamic economic pressures, such grids are frequently operated close economic pressures, such frequency grids are frequently operated weak close effects such aslimit. low oscillations to their stability In this limit, the presence ofondynamic to their stability limit. In this limit, the presence of dynamic interconnections develop into a fundamental effects such as can lowrapidly frequency oscillations on weak effects such as low frequency on weak bottleneck for power grid oscillations stability. interconnections cantransfer rapidlyand develop into a fundamental interconnections can rapidly develop into a fundamental bottleneckoscillations for power transfer and are grid stability.in virtually all Interarea (0,1-1Hz) bottleneck for power transfer and grid present stability. large power systems(0,1-1Hz) on the are transmission Under Interarea oscillations present inlevel. virtually all Interarea oscillationsthese (0,1-1Hz) arewell present in virtually all ambient conditions, modes are damped, amplitudes large power systems on the transmission level. Under large power systems on the transmission level.however Under are negligible. Stability problems become apparent ambient conditions, these modes are well damped, amplitudes ambient these modes are well damped, amplitudes only by conditions, observing the problems system behavior after excitations. are negligible. Stability become apparent however are negligible. Stability problems become apparent however Examples for excitations are power system line only by observing the system behavior after faults, excitations. only by observing the system behavior after excitations. switching, generator disconnection andsystem even the lossline or Examples for excitations are power faults, Examples for excitations are power system faults,into linea application of loads. An unstable system will enter switching, generator disconnection and even the loss or switching, disconnection and even the losswith or period of generator application of machine loads. An rotor unstableangle systemoscillations, will enter into a application of loads. An unstable system will voltage enter into a corresponding power flows and oscillatory and period of machine rotor angle oscillations, with period ofvariations. machineDamage rotorto the angle oscillations, with frequency infrastructure, unwanted corresponding power flows and oscillatory voltage and corresponding flows andeven oscillatory trippings inpower extreme cases blackoutsvoltage are thenanda frequency and variations. Damage to the infrastructure, unwanted frequency variations. Damage to the infrastructure, unwanted Fig. 1. GDASys devices for monitoring oscillatory danger. trippings and in extreme cases even blackouts are then a trippings and in extreme cases even blackouts are then a phenomena in the electrical grid in rack (DMR-D) / mobile danger. Fig. 1. GDASys devices for monitoring oscillatory A very recent and dramatic example for this can be found in Fig. danger. 1. GDASys for grid monitoring version (DA-Box 2000). phenomena in thedevices electrical in rack oscillatory (DMR-D) / mobile an official report to the Indian CERC A very recent andsubmitted dramatic example for this can commission be found in phenomena in the electrical grid in rack (DMR-D) / mobile version (DA-Box 2000). A very recent and dramatic example for this can be found in analyzing widesubmitted area blackouts happened on 30th and version (DA-Box 2000). an officialthe report to the that Indian CERC commission Poorly damped oscillations are thus an important indicator an official report submitted the Indian CERC commission 31st of July 2012 in blackouts the to Northern India electricity analyzing the wide area that happened on 30th grid and for the dynamic stability of the power grid. An early analyzing the wide area blackouts happened and Poorly damped oscillations are thus an important indicator (CERC Summarizing eventsthat during ofon the30th overall 31st of 2012). July 2012 in the Northern Indiaoneelectricity grid Poorly damped oscillations are thus an important indicator detection of these oscillatory phenomena is theAn key to for the dynamic stability of the power grid. early 31st ofblackout July 2012 in theever, Northern Indiaclearly electricity grid largest incidents the report states the for (CERC 2012). Summarizing events during one of the overall the dynamic stability of the power grid. An early detection of these oscillatory phenomena is the key to (CERC 2012). Summarizing events during one of the overall largest blackout incidents ever, the report clearly states the detection of these oscillatory phenomena is the key to largest blackout incidents ever, the report clearly states the Copyright © 2016 IFAC 7 2405-8963 © 2016, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved. Copyright 2016 IFAC 7 Control. Peer review©under responsibility of International Federation of Automatic Copyright © 2016 IFAC 7 10.1016/j.ifacol.2016.10.691
2016 IFAC CTDSG 8 Christian Rüster et al. / IFAC-PapersOnLine 49-27 (2016) 007–011 October 11-13, 2016. Prague, Czech Republic
increased situational awareness. This paper describes two promising approaches for dynamic voltage stability assessment: (1) frequency-domain based Wavelet analysis (2) monitoring the time evolution of the standard deviation of the bus voltage. The authors argue that the corresponding algorithms allow a particularly efficient detection of power system oscillations and stability and give practical measurement examples evidencing their performance. These measurements were taken with GDASys devices as shown in Fig.1.
disturbance. Such an oscillation could for example be related to a malfunction or some dynamical interactions between generators or loads. The damping monitor gives a precise answer as to what kind of disturbance occurred and when, how large it was and how long the disturbance was measurable. The usage of Wavelet- instead of conventional Fourier analysis here is not merely incidental but of utmost importance for true early detection: the choice of a proper Wavelet allows a significantly improved detection time because the time frequency uncertainty relationship is more favourable in online measurement situations. This will be demonstrated below in a measurement where inter-area oscillations were observed in near-real-time after a sudden generation loss.
2. THEORETICAL CONSIDERATIONS: EFFICIENT DETECTION OF GRID OSCILLATIONS Traditionally, grid operators use SCADA systems to monitor and control their network of power lines. This involves measuring average values of important operational parameters with a periodicity of a few seconds. The measurement data is then sent to a centralized supervisory station for review. To enable the analysis of dynamic effects in the grid, one has to drastically increase the measurement data rate. Suitable measurement tools for this purpose are offered by a “Grid Dynamics Analyzing System” as described in W. Haussel et al. (2004) and M. Hofbeck et al. (2008a,b). The variant in the top part of Fig. 1 consists of the core measurement unit DMR-D, a Windows-based industrial PC for data storage and a communication unit supporting a wide range of industry standard protocols. The mobile version of the device and a separate communication device is shown in the lower half of the picture. The devices offer a usable bandwidth between 5 mHz and 98 Hz at a high resolution of 5 mHz.
Whenever a dynamic deviation of system parameters from normal has been established, the problem to assess the criticality of the state of the system starts. Assessing the actual proximity to voltage breakdown for a complex system such as an interconnected electric grid is a difficult task and so far no single universally efficient and well-accepted assessment scheme is known. However, recent research into complex systems inside and outside the domain of electrical grids indicates that statistical tools such as autocorrelation and variance measurements of key observables can serve as useful indicators during the transition into breakdown. Particularly one body of research brought forth by d. Podolsky et al. (2013), G. Ghanavati (2013, 2014) and E. Cotilla-Sanchez (2012) points out a phenomenon called "critical slowing down" which is characterized by increasing fluctuations in the system, particularly in the nonlinear regime close to breakdown. G. Ghanavati (2013) shows, based on simulations of power systems approaching critical bifurcations, that especially the statistical variance in bus voltages can be a good indicator of the proximity to Blackout. Below the authors of this paper show real data measured in the German transmission system during the blackout which occurred in Turkey in March 2015. While this Blackout did not affect the European or German power system in a critical way, the data employs similar statistical analysis methods and shows that very sensitive event detection is possible. Corresponding measurement devices based on statistical algorithms thus show very good promise as power system instability early warning tools.
Fig. 2. GDASys damping monitor principle. Power systems disturbance model, a damped sinusoid.
The following sections present measurement results obtained during large disturbances in various actual power networks by means of measurement equipment of the A. Eberle GDASys family of devices. The purpose is to demonstrate in practice the efficiency of Wavelet- and statistical variance/standard deviation-based measurement algorithms for power system dynamics assessment.
Another distinguishing characteristic is the usage of highly sensitive methods of frequency domain analysis. Here, the well-known Fourier- as well Wavelet-analysis is available. The Wavelet analysis is employed to realize a function called damping monitor. Its task is to detect oscillatory disturbances on the power line that deviate from nominal system voltage or frequency behavior. By means of high precision Waveletbased Prony analysis, the damping monitor searches the full usable frequency spectrum for power system disturbances matching the fundamental model of a damped sinusoid (Fig. 2). Its output consists of the measured time stamp, frequency, amplitude, duration and the damping factor of a detected
3. PRACTICAL RESULTS: OSCILLATION DETECTION The following two case studies demonstrate the practical application of a Wavelet-based damping monitor during major power system disturbances.
8
2016 IFAC CTDSG October 11-13, 2016. Prague, Czech Republic Christian Rüster et al. / IFAC-PapersOnLine 49-27 (2016) 007–011
3.1 Major oscillation event caused by equipment failure
While the strongest disturbances started at around 70 minutes, it is very interesting to note that the actual cause of the disturbance - the equipment failure and separation of the grid – actually happened an hour earlier at 10 minutes. In the graph this coincides with three smaller amplitude oscillations of 0,2-0,3% in amplitude. While at this point the system was in a weak condition with a poorly damped 0,95 Hz mode waiting to be excited, the actual strong excitation came only an hour later.
In the course of a wide area measurement project aimed at monitoring a substantial part of a large transmission line network, three fixed installation GDASys / DMR-D measurement devices were introduced at strategically chosen locations inside the grid. During one notable incident, an equipment fault on a single phase at a coal fired power station triggered a series of protection device operations. Circuit breakers tripping resulted in a situation where the power station was only connected to the rest of the power system via a weak 200km connection as opposed to the usual 50km line. This effectively constituted a major reconfiguration of the local grid and power flows on the weak remaining lines of the area in which the DMR-D device (named C2) was situated, increased critically. Most importantly, this resulted in the emergence of a new resonance frequency at 0,95 Hz. During the main part of the disturbance, when this new mode was excited by an unknown stimulus, large voltage oscillations were present in the system for about 25 minutes.
The lower panel in Fig. 3 represents an alternate analysis of the criticality of measured oscillations during the event. The plotted data points belong to the same detected oscillation events that are plotted as a time series in the upper panel of the figure. Here, mode oscillation amplitudes are plotted against their duration. Only oscillations falling within the shaded box can be considered as truly safe. Everything outside the box should be regarded as either a warning or even an alarm signal. This includes oscillations either with amplitudes exceeding 0,1% or having a duration longer than 10s. For the 0,95 Hz mode analysed here, a duration of 10s means that the individual disturbance is visible to the system for more than 10 periods. This is a sign of poor damping and thus a warning signal for system stability. Since some of the measured durations even exceed 70 s, they are clearly in the critical regime. Using the data presented as in Fig. 3, it is clear that the disturbance event could have been quickly identified by an operator, especially since such kind of data is available from the Wavelet algorithm in quasi real-time. 3.2 Major Oscillation Event Caused by Sudden Loss of Generation Capacity During another incidence in the same measurement period, multiple generators from a coal fired power station tripped simultaneously. Sudden losses in generation capacity constitute considerable stress for any kind of grid and are prone to excite oscillatory disturbances. This sudden loss of generation capacity affected numerous transmission lines. Notably the disturbance was strong enough so that tie-lines to several neighbouring states with interconnected electricity systems were tripped due to the operation of the Under Frequency Load Shedding relays (UFLS). The idea of tripping the interconnection in this protection scheme was to shed sufficient load to rebalance the grid.
Fig. 3. Time series of 0,95 Hz mode amplitudes measured during a major oscillation event caused by equipment failure (upper panel). Mode criticality analysis of the same event (lower panel).
This oscillation event manifested itself with very large oscillation amplitudes exceeding 2% in the damping monitor of DMR-D device C1. The time series of measured oscillation amplitudes is displayed in Fig. 4 (blue data points). It can be seen that as far as oscillatory disturbances are concerned, the event was over in less than one minute, with amplitudes clearly exceeding the alarm level before returning to ambient conditions. A criticality analysis plot similar to the one shown in the lower panel of Fig. 3 reveals that the large amplitude oscillations are negatively damped and also exceed the duration criterion of 10s. The largest oscillation amplitudes are found for a mode at 0,6 Hz. The longest sustained oscillations however exist for the 0,35 Hz
Device C2 was able to pick up the corresponding instabilities in high resolution using the Wavelet-based damping monitor algorithm. A time series of amplitudes of the 0,95 Hz mode is shown in the upper panel of Fig. 3. The most critical part of the event is centred in the middle of the panel. It can be seen that oscillation amplitudes pick up within a very short time and stay at a very high level of 0,8% to 1% for about 20 minutes before dying off over the course of a few minutes. This amplitude level definitely indicates a highly stressed system and reveals considerable power swings. 9
9
2016 IFAC CTDSG 10 Christian Rüster et al. / IFAC-PapersOnLine 49-27 (2016) 007–011 October 11-13, 2016. Prague, Czech Republic
and 0,7 Hz interarea modes, with oscillation durations exceeding 34 seconds in both cases. The frequencies of the interarea modes are well known in that network. From this it can be concluded that two weak transmission corridors were most strongly affected by the event: one connecting with a neighbouring country and one leading to a remote load center.
capacity) and the subsequent loss of Atlas thermal power St. (in Hatay province, at 09:02 am) as well as Dicle hydro (Southeast of the country). The resulting frequency variations reportedly lead to automatic safety measures in which the UCTE grid cut electrical connections with Turkey for safety reasons. The blackout subsequently ensued at 09.36 am. At the time of the separation the net power flow between Turkey and the UCTE grid was fortunately only at a very moderate level. This means that viewed from the European perspective, cutting tie lines to Turkey constituted only a mild strain on the system. During the whole incident, five DMR-D monitoring installations were active in the German transmission grid. A post-event analysis indicates that even though for the far away German grid this was only a very minor disturbance, the DMR-D's statistics based stability exponent measurement algorithm was sensitive enough to pinpoint the dynamics of the grid separation. Taking into account the time evolution of the standard deviation of power system frequency and bus voltage, the stability exponent aims to be a key indicator for dynamic strain in the electrical grid, in a way that is very similar to the theoretical analysis independently suggested by the "critical slowing down" phenomenon particularly as described by E. Cotilla-Sanchez in 2012.
Fig. 4. Time series of damping monitor amplitudes measured with DMR-D device C1 during a major oscillation event caused by generation loss. Despite the event lasting only about a minute, device C1 was able to resolve multiple modes, some of which even happened simultaneously. This superior response time is due to advantages of the employed wavelet analysis algorithms of the device and distinguishes it from typically used Fourier techniques that have difficulties in the subsynchronous range. The high time resolution of the Wavelet analysis is revealed by the comparison with a FFT employing a sliding 100s time window (red data in Fig. 4). The blue Wavelet data rises immediately at what is known to be the onset of oscillations. In comparison, the response of the red 100s FFT is delayed and artificially broadened. In fact, when FFT amplitudes exceed the alarm level, the actual event is already over. Please not that this is actually not due a suboptimal performance of this particular FFT filter implementation but due to a fundamental limitation of any FFT: a filter of sufficient frequency resolution for detecting relevant sub 1Hz oscillations has a specific, corresponding certain minimum window length. Time lag for online measurements is then determined by the window length and cannot be improved without frequency resolution loss.
From a theoretical and algorithmic point of view, increasing power system strain can be understood to cause increasing deviations from the mean value of frequency and / or bus voltage. The stability exponent mirrors the rate of change of the standard deviation of bus frequency and bus voltage hence the value of the stability exponent will also increase. Phases of system stabilization reverse this trend and cause the stability exponent to fall. In practice, the moment of power system separation of the Turkish and the European grid represents just this kind of momentary system disturbance which quickly "heals off" in this case, since the net power flow difference was only quite small.
3.3 March 2015 blackout in Turkey observed from within the German transmission grid On the morning of the 31st of March 2015, a major blackout was reported for the national Turkish electricity grid (ENTSO-E 2015). As a result, 80 out of 81 Turkish provinces were without electricity during busy workday hours, with millions of people stuck on the way to work, as also public train and subway systems were paralyzed.
Fig. 5. Time series of stability exponent measurements in five different locations in German transmission grid, at the time of the Turkey blackout in march 2015. Nevertheless, as is seen in the data shown in Figure 5, all five devices distributed from North to South within Germany showed a distinct and easily recognizable feature in the stability exponent measurement at the exact time of system separation (09:36 am). Here, closer investigation even reveals
According to press reports, the blackout was a consequence of several power stations dropping off the grid, including a major power station in the Aegean region (2000 MW 10
2016 IFAC CTDSG Christian Rüster et al. / IFAC-PapersOnLine 49-27 (2016) 007–011 October 11-13, 2016. Prague, Czech Republic
that the size of the peak correlates with the approcimate "electrical separation" between the incident in Turkey and the measurement location in Germany. The data shows that events happening in the grid can be detected in a very sensitive way. It furthermore is an indication that indices such as the standard deviation based stability exponent which is in line with theoretical conclusions about critical slowing down before blackouts can be a useful early indicator for the proximity of critical states in the grid. Please note however that the rather indirect measurement data shown should not be interpreted as a direct measure to the proximity to actual breakdown.
11
E. Cotilla-Sanchez, P. D. H. Hines, C. M. Danforth "Predicting Critical Transitions From Time Series Synchrophasor Data", IEEE Transactions on Smart Grid, 2012 D. Podolsky, K. Turitsyn, "Critical slowing-down as indicator of approach to the loss of stability", arXiv:1307.4318v1, 2013 G. Ghanavati, P. D. H. Hines and T. I. Lakoba "Investigating early warning signs of oscillatory instability in simulated phasor measurements", arXiv:1312.5371v1, 2013 G. Ghanavati, P. D. H. Hines and T. I. Lakoba, "Identifying Useful Statistical Indicators of Proximity", arXiv:1410.1208v1, 2014
4. CONCLUSIONS DMR-D measurements taken in actual high voltage transmission grids have pointed out the benefits of oscillation monitoring for dynamic stability assessment. Identifying characteristic disturbances in the system enables the user to distinguish normal oscillatory behaviour from significant disturbances. Particularly the employed damping monitor based on Wavelet analysis is an accurate tool for measuring oscillatory modes and allows a straightforward assessment of an oscillation’s criticality. Wavelet based signal conditioning surpasses the capabilities of conventional FFT and was shown to be fast enough to register even very short duration disturbances.
ENTSO-E Report on Blackout in Turkey on 31 March 2015,
available on the ENTSO-E website: https://www.entsoe.eu/Documents/SOC%20do cuments/Regional_Groups_Continental_Europe /20150921_Black_Out_Report_v10_w.pdf
Furthermore, the statistical (standard deviation-based) analysis of the power system observables voltage and frequency has shown good promise to be a sensitive tool for the assessment of proximity to instability in stochastic power systems, i.e. grids which are in a nonlinear regime close to voltage breakdown. This promise roots both in the sensitive detection of a recent blackout in Turkey from afar, when viewed together with the independently derived "critical slowing down" concept.
REFERENCES "Report on the grid disturbance on 30th July 2012 and grid disturbance on 31st July 2012" available on www.cercind.gov.in W. Haussel, M. Hofbeck, T.Sybel, M. Fette, "Collapse Protection System - Basics and Applications", Proceedings of the South African Power System Protection Conference 2004 M. Hofbeck, L. Mayer, T.Sybel, M. Fette, B. Werther, I. Winzenick, "Collapse Prediction Relay CPR-D - Theory and Applications - Implementation into SCADASystems to Support Security Stability Assessment", Proceedings of the South African Power System Protection Conference 2008 M. Hofbeck, T. Sybel, M. Fette, I. Winzenick, "Measurements of the Dynamical Status of HighVoltage-Networks with CPR-D", Proceedings of the South African Power System Protection Conference 2008
11