Evaluation of Precision Time Synchronisation Methods for Substation Applications
•Download as PPTX, PDF•
0 likes•66 views
This document evaluates precision time synchronization methods for substation applications. It discusses the synchronization requirements for substation processes and synchrophasors. It then tests the performance of 1 pulse per second (1PPS), IRIG-B, and IEEE 1588 Precision Time Protocol (PTP) using fiber optic cables of different lengths. Statistical analysis shows that while 1PPS has the least jitter, both IRIG-B and PTP can meet the ±1 microsecond accuracy requirements. The conclusions are that 1PPS and IRIG-B are suitable for small substations, while PTP overcomes their shortcomings for large substations without compromising synchronization performance.
Recommended
Recommended
More Related Content
What's hot
What's hot (16)
Similar to Evaluation of Precision Time Synchronisation Methods for Substation Applications
Similar to Evaluation of Precision Time Synchronisation Methods for Substation Applications (20)
Recently uploaded
Recently uploaded (20)
Evaluation of Precision Time Synchronisation Methods for Substation Applications
- 1. CRICOS No. 00213J ISPCS 2012 San Francisco, USA Evaluation of Precision Time Synchronisation Methods for Substation Applications David M. E. Ingram, QUT Pascal Schaub, Powerlink Queensland Duncan A. Campbell, QUT Richard R. Taylor, QUT
- 2. CRICOS No. 00213J ISPCS 2012 San Francisco, USA • High voltage substations – Substation applications – Synchronisation requirements • Synchronisation methods • Performance tests • Discussion of results • Conclusions Presentation Outline 2
- 3. CRICOS No. 00213J ISPCS 2012 San Francisco, USA Transmission Substations 3 350 m 460 m 500 kV 330 kV Aerial photograph from NearMap Pty Ltd (www.nearmap.com)
- 4. CRICOS No. 00213J ISPCS 2012 San Francisco, USA Transmission Substations 4 460 m 540 m 500 kV 330 kV Aerial photograph from NearMap Pty Ltd (www.nearmap.com) 220 kV
- 5. CRICOS No. 00213J ISPCS 2012 San Francisco, USA • Power system incident investigation • Phasor monitoring (magnitude & angle) • Process bus sample synchronisation The Need for Accurate Synchronisation 5
- 6. CRICOS No. 00213J ISPCS 2012 San Francisco, USA Performance Requirements Process Bus IEC 61850-5 timing classes • UCA “9-2LE” requires ±1 µs to meet P2 (overall accuracy). Synchrophasors 6 IEEE Std C37.118.1 ‘Total Vector Error’ <1%. • 26 µs (60 Hz) or 31 µs (50 Hz) if no magnitude error or CT/VT phase error. • Suggested in several papers that ±1 µs be used as synchronisation accuracy. Prot. Class Required Accuracy Timing Class (Ed .1 / 2) P1 ± 25 µs T3 / TS3 P2 ± 4 µs T4 / TS4 P3 ± 1 µs T5 / TS5
- 7. CRICOS No. 00213J ISPCS 2012 San Francisco, USA Test Equipment 7
- 8. CRICOS No. 00213J ISPCS 2012 San Francisco, USA • One pulse per second (1PPS) – The simplest method • IRIG-B – Widely used substation timing • IEEE 1588 Precision Time Protocol (PTP) – The “power profile” specified in IEEE Std C37.238 was used. Time Synchronisation Methods 8
- 9. CRICOS No. 00213J ISPCS 2012 San Francisco, USA PTP Message Parameters 9 Parameter Setting Sync message rate 1 s Announce message rate 1 s Path delay mechanism Peer to peer Path delay message rate 1 s Line rate 100 Mb/s Message type Layer 2 multicast
- 10. CRICOS No. 00213J ISPCS 2012 San Francisco, USA 0 2 4 6 8 10 0.00.20.40.6 1-PPS 0.7 m Fibre Master-Slave Offset (ns) Density 354 356 358 360 362 0.00.20.40.6 1-PPS 66 m Fibre Master-Slave Offset (ns) Density 5050 5054 5058 0.00.20.40.6 1-PPS 998 m Fibre Master-Slave Offset (ns) Density One Pulse per Second 10
- 11. CRICOS No. 00213J ISPCS 2012 San Francisco, USA -200 0 200 400 600 0.000.010.020.030.04 IRIG-B – 0.7 m Fibre Master-Slave Offset (ns) Density Master A Master B 200 400 600 800 1000 0.000.010.020.030.04 IRIG-B – 66 m Fibre Master-Slave Offset (ns) Density Master A Master B 4800 5000 5200 5400 5600 0.000.010.020.030.04 IRIG-B – 998 m Fibre Master-Slave Offset (ns) Density Master A Master B IRIG-B 11
- 12. CRICOS No. 00213J ISPCS 2012 San Francisco, USA Bimodal IRIG-B 12 0 20 40 60 80 100 120 550600650700 IRIG-B (Master B) - 66 m Fibre Time (sec) Master-SlaveOffset(ns)
- 13. CRICOS No. 00213J ISPCS 2012 San Francisco, USA -200 0 200 400 600 0.0000.0050.0100.015 PTP – 0.7 m Fibre Master-Slave Offset (ns) Density Master A Master B -200 0 200 400 600 0.0000.0050.0100.015 PTP – 66 m Fibre Master-Slave Offset (ns) Density Master A Master B -200 0 200 400 600 0.0000.0050.0100.015 PTP – 998 m Fibre Master-Slave Offset (ns) Density Master A Master B Precision Time Protocol 13
- 14. CRICOS No. 00213J ISPCS 2012 San Francisco, USA Statistical Analysis 14
- 15. CRICOS No. 00213J ISPCS 2012 San Francisco, USA • 1-PPS provides the least jitter: – No ‘time of day’ information Discussion – Performance 15 • IRIG-B can meet ±1 µs requirements – No compensation for propagation delay • PTP with C37.238 meets ±1 µs requirements – Compensates for propagation delay – Supports redundancy of grandmasters
- 16. CRICOS No. 00213J ISPCS 2012 San Francisco, USA • Optical cable used to take signals out to the switchyard. • 1-PPS and IRIG-B require multiport repeaters – Introduces error – OTDR cannot see through repeater • PTP with C37.238 requires transparent clocks – Can share Ethernet with process bus or PMU connection. Discussion – Signal Distribution 16
- 17. CRICOS No. 00213J ISPCS 2012 San Francisco, USA • 1-PPS is most compatible, but least information • IRIG-B clients can have mutually incompatible requirements – Local time vs UTC • PTP + C37.238 has limited options – Improves compatibility of grandmasters and slave clocks – Must use TAI as time reference Discussion – Compatibility 17
- 18. CRICOS No. 00213J ISPCS 2012 San Francisco, USA • 1-PPS and IRIG-B can be used for small substations, or where merging units are in the control room. • PTP overcomes shortcomings of 1-PPS and IRIG-B for large substations. • PTP benefits are not at the expense of synchronising performance. – Similar performance between IRIG-B and PTP from the same clock hardware Conclusions 18
- 19. CRICOS No. 00213J ISPCS 2012 San Francisco, USA Project Sponsorship & Funding • Project sponsorship by Powerlink Queensland. • Funding support 19
- 20. CRICOS No. 00213J ISPCS 2012 San Francisco, USA • Equipment support from is appreciated. Acknowledgments 20
Editor's Notes
- Hi, I’m David Ingram and I am a PhD candidate at the Queensland University of Technology in Brisbane, Australia. I am studying the support infrastructure required for to implement transmission substation process buses based on IEC 61850. Time synchronisation is a critical component of a “whole of station” process bus, and I am presenting in this paper the results of benchmarking experiments I conducted.
- In this presentation I will describe transmission substations, the need for synchronisation and the level of performance that is required. I will then present the experimental method that I used to evaluate each timing method and describe the sync methods I tested. I’ll describe the results and discuss the implications of these and then finish with some conclusions.
- Transmission substations are generally considered to be those that operate at 100kV and above. Each network has a range of standard voltages, with 110kV and 132kV common in Australia. In Queensland we use both. The size and topology depends on the role the substation plays. Substations can occupy a large area, and cable distances of 500 m or more not unusual. This substation in Western Sydney is well laid out, with two operating voltages, 500kV and 330kV.
- This is substation north of Melbourne is more haphazard and has been developed over time. As result, it is much more spread out, with separate switchyards for the three voltages in use.
- Sequence of event logging is commonly used. Protection relays record when states change, such as a circuit breaker trip, in their internal logs. The time stamps need to be aligned so the logs can be compared. Currently IRIG-B or NTP is used for this. Some applications, in notably phasor monitoring require greater time stamp accuracy than NTP or modulated IRIG-B can provide. 1PPS inputs commonly used in combination with IRIG-B or NTP. The absolute time is in the IRIG-B or NTP message. The precise second boundary is the 1PPS edge. IEC61850 part 9-2 specifies a means of transmitting instantaneous data, typically currents and voltages, over Ethernet. The ‘merging unit’ digitises the instantaneous current and voltage signals. 1PPS over fibre specified in the UCA Implementation Guideline, based on the IEC60044-8 specification. If samples come from more than one source and need to be compared, the samples must be synchronised. The example on the left, the transformer protection relay is subscribing to sampled value messages from the high voltage side and from the low voltage side. When separate merging unit are used, the samples need to be taken at the same time, otherwise phase errors will occur. The right example is hybrid protection, where the high voltage side current signals are digitised by a merging unit and sent to the protection relay over Ethernet. The low voltage side current signals are sent using conventional analogue cabling and the relay must synchronise its sampling to be the same as the merging unit’s. This configuration is likely when one part of a substation is upgraded to digital process bus protection, as was done with Powerlink Queensland’s Loganlea substation.
- The two applications requiring very accurate synchronisation are sampled value process buses and phasor monitoring. The source of the requirements differs, however ±1µs is a benchmark that both have. IEC 61850-5 specifies requirements for synchronising accuracy, based on the protection class, which is in turn based on the application. Class P2 covers standard transmission substations, and the corresponding time class is T4 in edition 1 of the standard, or TS4 in the draft edition 2. T4 requires sampling be accurate to within 4µs. The UCAI implementation guideline for 61850-9-2, often called 9-2 light edition, specifies a 1µs accuracy for the sync source and for synchronising to use 1PPS over glass fibre. This allows for up to 2µs propagation delay of the signal, and for some sampling error in the merging unit, so that the 4µs requirement of class T4 is met. 2µs is approximately 400 ms for fibre optic cable with a refractive index of 1.5. If the propagation delay exceeds 2µs compensation at the “end user” device. The C37.118.1 requirement is not ‘hard and fast’, but is suggested in several papers covering this application and is considered to be a good engineering judgement.
- I have developed a process bus test bed to test sampled value process bus protection, along with other process bus applications including circuit breaker control and transformer tap change control. The two racks shown here contain timing equipment and protection relays, and represents the ‘control room’ of a substation. A variety of clocks have been purchased, donated or lent for the test bed. This allows established synchronising methods, including one pulse per second and IRIG-B, to be tested alongside the precision time protocol. The experiments for this paper use the same clocks for IRIG-B and PTP, however there is significant variation in the performance of PTP devices. The application requirement for synchronisation is 1PPS, and therefore I generated this signal at the master clock and the slave clock. A digital oscilloscope was used to measure the delay between the rising edges, and the screen shot shows a sample of the waveforms. The four channel oscilloscope enabled three slave signals to be compared to the master, and sampled at 1 gigasample/sec. 1PPS delays were collected for 30 minutes, giving 1800 measurements for each test. Each measurement was collected by the PC for further statistical analysis. Fibre optic cables are preferred for substation applications due to the inherent galvanic isolation they provide. I used three lengths of multimode glass fibre optic cables for these experiments. The lengths were 0.7m, 66m and 998m. The 0.7m cable provided a baseline delay through the optical converters, which was less than 10ns. The 66m cable represented connections within a substation control room or an indoor substation, and the 998m cable represented a large outdoor substation. Some clocks had optical ports for pulse outputs, and some had optical Ethernet ports. Simple fibre optic pulse interfaces and 100BASE-FX transparent clocks were used to convert optical and electrical signals.
- The Precision Time Protocol, IEEE 1588 or PTP, is recommended by IEC and NIST smart grid standards roadmaps for synchronisation in transmission smart grids. Two widely used precision time synchronisation methods, 1PPS and IRIG-B, were compared to PTP to provide a performance benchmark. NTP is used in substations, but does not have the accuracy required to meet the 1µs requirements of 61850-9-2 and synchrophasors. 1PPS is a rising edge pulse, and in for these experiments was an optical signal, as that is required by 9-2 light edition, and is good practice in high voltage substations. The IRIG-B I used is unmodulated, as the 1kHz AM signal does not have sharply defined edges for clock recovery.
- The PTP message parameters from the Power Profile, C37.238, were used for testing. Sync, Announce and Peer Delay messages were all sent once per second. Layer 2, raw Ethernet, messages were used and the Ethernet network operated at 100 Mb/s.
- One of the master clocks had two pulsed outputs: one was electrical and provided the reference signal, and that other was optical which drove the fibre optic cable. A fibre optic receiver was connected to the remote end of the fibre, and provided the ‘slave’ 1PPS signal for comparison purposes. The 0.7m fibre test shows that the offset between the electrical & optical outputs, even with propagation delay through the fibre optic receiver, was less than 10ns. The spread with 66m of fibre is the same as for the 0.7m fibre, however the kilometre of fibre did have more spread. Modal dispersion is the most likely cause of this, and is often the limiting aspect for fibre optic communications, rather than signal level. Compensation would be required for the 1km cable, as the mean delay exceeds 2µs.
- This graph shows the IRIG-B synchronising performance with two master clocks. As with 1-PPS, the mean delay varies linearly with cable length. The optical receiver was used to generate an electrical IRIG-B signal for the slave clock. There is more jitter, and the standard deviation with IRIG-B is approximately 40-90 times that of 1-PPS ,depending on the master clock. The spread with IRIG-B did not increase with the 998m fibre, which suggests a clock recovery algorithm is being used by the slave clock. A second IRIG-B master clock was used with the original slave to look for device dependent performance variation. Master B has less jitter in the observed delay than Master A, however the distribution is bimodal.
- This two minute time series shows that the 1-PPS delay between the slave and master B periodically increases by 50-100 ns. The mechanism for this bimodality is unknown, as the design of the IRIG-B master device is not published by the manufacturer. A possibility is a periodic correction of a phase locked loop.
- The two master clocks and the slave clock used for IRIG-B testing also supported PTP over Ethernet. One of the master clocks did not have a 100BASE-FX port, so a peer-peer transparent clock with copper and fibre ports was used to drive the fibre. So comparisons could be made between masters, the transparent clock was also used with the master that did have a fibre optic port. The PTP slave had an optical Ethernet port and connected directly to the fibre. This shows the synchronising performance with Master A and Master B. As was the case with IRIG-B, Master B gave better performance than Master, with reduced jitter. Master A has an uncompensated crystal oscillator, while Master B has a temperature compensated crystal oscillator. The lower phase noise and higher stability of the TCXO gives better performance for IRIG-B and PTP. The path delay compensation can be seen here to be effective, but there were device dependent fixed offsets in observed delays.
- Delta td bar is the mean delay time corrected based on the corresponding mean 0.7m fibre measurement. S-t-d is the sample standard deviation of the 1PPS delays. 1800 measurements were used for each mean and standard deviation. The predicted delay is based on an estimate of the refractive index, n=1.5, for the fibre optic cable. 1PPS and IRIG-B used 850nm wavelength, as required by IEC 60044-8, while PTP used 1300nm wavelength required by 100BASE-FX.
- 1PPS over short dedicated links is an effective means of synchronising sampling, and is the method used in the first process bus substation in Australia, Loganlea. Where time of day, or absolute time, is required then additional information is required and this could be via NTP or IRIG-B. IRIG-B performed surprisingly well, but as with 1PPS, propagation delay was an issue for the longest fibre optic cable. PTP, as tested using C37.238 parameters, meets the 1µs timing requirements of 9-2 light edition and synchrophasors. The biggest advantage for transmission substations is that the protocol automatically compensates for path delay. PTP supports multiple master clocks, with a prescribed fail-over mechanism, the Best Master Clock Algorithm. This is a big improvement on IRIG-B which only supports a single master clock with optical transmission.
- If multiple slave clocks are present, a means of distributing the 1PPS or IRIG-B is required, which is not straight forward. The traditional approach is multi-drop coax and AM IRIG-B, with transformers providing isolation at each device. With electrical IRIG-B high current drive outputs are required, and for optical IRIG-B separate drivers are required. Where compensation is required, the total path length from the master to the slave must be known. This could be determined from “as built” cable schedules, or measured using an optical time domain reflectometer, or OTDR. Unfortunately OTDRs cannot measure past a repeater or distributor, so a treed optical network is challenging to compensate. PTP allows transparent clocks to distribute the signal in a treed arrangement, with the transparent clocks measuring path delay for each segment. C37.238 requires that 16 transparent clocks do not introduce more than 800ns of error, allowing 200ns of error from the grandmaster to give an overall error of 1µs. Where Ethernet queuing delays do occur due to other traffic, these are compensated for by the transparent clock. This means that high levels of other network traffic does not affect PTP performance. Shared networks reduce design, construction and maintenance costs. This is good as there is increasing pressures on utilities to reduce costs.
- The simplicity of 1PPS is both a blessing and a curse. There is no incompatibility between 1PPS sources and slaves, provided the same medium is used. Unfortunately the signal does not convey time of day information. IRIG-B is the next step, where the pulse train encodes absolute time data. Unfortunately there are options, such as local time or UTC, and some slave devices have mutually incompatible settings. Some substations require two separate IRIG-B sources to accommodate this. PTP is the most complex of the systems, but by using the Power Profile we are almost certain to have interoperability. [Mention Plug-fest here if relevant] TAI is the time reference, although offsets to local time and UTC can be provided for local display.
- The established synchronising methods, 1PPS and IRIG-B, have been shown to be effective for smaller substaitons, or where the merging units or secondary converters are installed in a control room. Large substations, such as the ones I showed earlier, is where PTP provides a significant improvement. The synchronising performance is similar, if not better than, IRIG-B while compensating for propagation delay. More and more protection relays and other IEDs will natively support PTP over the same interface as the process bus connection. PTP will be the simplest method to implement at this time.
- I would like to acknowledge project funding by Powerlink Queensland, the high voltage transmission operator in Queensland, as well as personal support by the Australian Government and QUT.
- Donations and long-term loans of PTP hardware from Hirschmann, Meinberg and Cisco are appreciated and have enhanced the performance of the test bed.
- Thank you for your attention.