CSEE JOURNAL OF POWER AND ENERGY SYSTEMS, VOL. 2, NO. 3, SEPTEMBER 2016 91 Time Synchronization for Transmission Substations Using GPS and IEEE 1588 Peter A Crossley, Member, IEEE, Hao Guo, Student Member, IEEE, and Zhao Ma, Senior Member, CSEE Abstract—Time synchronization systems that utilize the global navigation satellite systems (GNSS) are widely used in the monitoring, control, and protection of transmission networks. They ensure that phasor measurement units (PMUs) can accurately monitor voltage phase angles, increase the accuracy of fault locators, enhance the capabilities of disturbance recorders, and allow differential feeder protection to use re-routable communication networks. However, concern about the reliability of GNSS receivers used in intelligent electronic devices (IEDs) have been reported; problems include mal-operations of differential protection, erroneous satellite timing/location messages, inappropriate installations, and blocking of satellite signals due to illegal use of GNSS jammers in vehicles. Utilities now require a timing system less dependent on the use of low cost GNSS receivers integrated into IEDs, but one that uses Grandmaster clocks, slave and transparent clocks, and an Ethernet communication network. The IEEE 1588-2008 synchronization protocol uses the Ethernet to disseminate a global time reference around a substation. A future substation will probably include duplicate 1588 grandmasters, each incorporating stable oscillators with GNSS and terrestrial receivers, in conjunction with a 1588 compliant Ethernet data network with slave and transparent clocks, and redundancy boxes for interfacing with IEDs. Although IEEE 1588 protocol is promising for future substation automation systems, its performance and impact has to be fully evaluated before it can be used in real substations. This paper describes how an IEEE 1588 time synchronization testbed is designed, constructed, and tested. Testing involves measuring the time offset when the Ethernet is heavily loaded with other trafﬁc and the holdover capability of 1588 clocks. Additional delay introduced by IEEE 1588 trafﬁc is also measured. As there is limited testing on GPS receivers within the power industry, this paper also uses the testbed to evaluate the steady state and transient behavior of GPS receivers. The results show a 1588 time synchronization system is accurate, secure, and ideally suited for protection and control applications, compared to a timing system merely based on GPS receivers. The information described in this paper should increase a utility’s conﬁdence in applying IEEE 1588 timing in a real substation. Index Terms—GNSS, GPS, IEDs, IEEE 1588, substations, time synchronization, transmission networks. Manuscript received March 11, 2016; revised June 27, 2016; accepted July 27, 2016. Date of publication September 30, 2016; date of current version August 12, 2016. P. A. Crossley (corresponding author, e-mail: email@example.com) and H. Guo are with School of Electrical and Electronic Engineering, The University of Manchester, (Manchester, M13 9PL) UK. Z. Ma is with China Electric Power Research Institute, Bejing 100192, China. DOI: 10.17775/CSEEJPES.2016.00040 I. I NTRODUCTION I NTELLIGENT electronic devices (IEDs) used in substations ensure a transmission network operates in a stable state. But when an abnormal event occurs, the IEDs must rapidly restore the network from the abnormal state back to a normal state. An IED measures power system quantities (voltage, current, frequency, power, etc.) and the operational states of switches and circuit breakers. This data is processed by the local IED and communicated to other IEDs in the substation, remote IEDs, network controllers, and/or system operators where further processing and analysis is carried out. The resulting knowledge is used by one or more IEDs to respond to the event that ﬁrst triggered the change in the measured quantities. The event might be a short-circuit fault or an extreme overload that should be solved locally, or an event such as switching action, generator or feeder tripping, a voltage or frequency excursion, which requires remote responses. These measurements, acquired in each bay of each substation, only describe the actual real time operating state of the power system if they are synchronized to a common time reference. IEDs can be synchronized using either a dedicated timing system based on standalone cabling and signal repeaters, or a network timing system shared with other automation applications and normally based on Ethernet cables and switches . Dedicated timing systems often use one pulse per second (1-PPS) to provide an accurate timing reference at the start of each second, or the IRIG-B time code for time and date information. GPS (or GNSS) receivers convert the time information received from GNSS into a 1-PPS signal and the IRIG-B time code, which are then used to synchronize all the IEDs in a substation . An example of a dedicated timing network that operates independent of the substation communication network is shown in Fig. 1. When the 1PPS signal travels from the GPS receiver to the device, a propagation delay is introduced. During commissioning, this delay must be measured and compensated. Substation applications, such as SCADA or disturbance recorders, requiring timing accuracy in the millisecond range, can use a network time protocol (NTP) system operating over an existing Ethernet communication network, as illustrated in Fig. 2. However, most substation applications require timing accuracies in microseconds and consequently NTP is not suitable. For example, phasor measurement units (PMUs), travelling wave fault locators (TWFL) and IEC61850-9-2 sample value merging units (SV-MU) require an accuracy of c 2016 CSEE 2096-0042 92 CSEE JOURNAL OF POWER AND ENERGY SYSTEMS, VOL. 2, NO. 3, SEPTEMBER 2016 ±1 μs. Whilst current differential protection (CDP) generally requires 20 μs and IEC61850-8-1 GOOSE messaging requires accuracy better than 1 ms. GPS clock IRIG-B over coaxial cable Protection relays Substation HMI IEC 61850, FTP, DNP3, Modbus etc Ethernet switch SCADA gateway Fig. 1. Separate timing and communication networks . GPS clock as NTP master Substation HMI NTP protection relays II. 1588 T IME S YNCHRONIZATION A. Introduction to 1588 Time Synchronization An increasing number of IEDs are using the Ethernet data network for communication purposes  and most experts expect Ethernet to become the communication backbone for future transmission substations. Considering the cost, complexity and reliability, it would be ideal if high accuracy synchronization can utilize the Ethernet. Unfortunately, an Ethernet based NTP timing system only achieves 1–4 ms accuracy, which is inadequate for most applications, including IEC 61850-9-2 SVs, CDP, TWFL, and PMUs. An alternative to NTP is the IEEE 1588 timing standard, which operates over a data network and achieves signiﬁcantly improved timing accuracy. The IEEE 1588-2008 standard (IEEE 1588v2) , also known as the Precise Time Protocol Version 2 (PTPv2), is an IP/Ethernet based time synchronization protocol that realizes sub micro-second timing accuracy. A master-slave IEEE 1588 synchronization network consisting of different types of IEEE 1588v2 devices is shown in Fig. 3, where each 1588v2 device is referred to as a clock. Four types of clock are available in IEEE 1588v2 standard : Other time Ordinary Clock-2 reference (Master as backup for Grandmaster) Ordinary Clock-1 (Grandmaster) GPS input Ethernet switch End-to-end transparent Clock-1 (or non-1588 switch) Peer-to-peer transparent Clock-1 SCADA gateway Fig. 2. Boundary clock Combined timing and communications . The IEEE 1588-2008 (IEEE1588v2) synchronization protocol uses a data network to disseminate the time reference similar to NTP, but delivers timing accuracy better than ±1 μs. Consequently, IEEE 1588 synchronization is expected to become the preferred choice for future Ethernet based protection and control systems. However, before it can be applied in a real substation, various tests should be conducted to help users gain better insight into the feasibility, performance, and limitations of the 1588 protocol. Since testing on an Ethernet system in a real U.K. substation is impossible, the aim of this research is to build a laboratory based testbed to investigate and evaluate the performance and impact of an IEEE 1588 timing system. In addition, many GPS manufacturers claim their GPS receivers can deliver timing accuracy better than one micro-second, but there are few published data for those used in the power industry. Consequently, the performance of GPS receivers is also investigated using the testbed. The goal of this research is to help utilities identify the limitations of timing systems based on GPS and IEEE 1588, and then develop solutions that enhance reliability and increase the conﬁdence in applying the technology. This paper is organized as follows. Section II provides background information on IEEE 1588, followed by test arrangements, results, and discussions in Section III and the conclusions in Section IV. Peer-to-peer transparent Clock-2 Ordinary Clock-3 (Slave) Fig. 3. Ordinary Clock-4 (Slave) End-to-end transparent Clock-2 (or non-1588 switch) Ordinary Clock-5 (Slave) Network of IEEE 1588v2 clocks. 1) An ordinary clock is an end device and can be a grandmaster clock for the whole substation, or a master clock if it supplies the time source for other clocks on a single communication path, or a slave clock if it is synchronised by a grandmaster or master clock. 2) A boundary clock is basically a network bridge or switch and is the combination of slave clocks and master clocks. It will act as a slave clock for the connected upstream master/grandmaster clock and becomes the master clock for its downstream slave clocks. In terms of non-1588 messages and IEEE 1588 management messages, the boundary clock will forward them as a normal Ethernet bridge or switch. On the contrary, IEEE 1588v2 messages related to synchronization and clock selection will terminate in the boundary clock. 3) An end-to-end transparent clock is a network bridge or switch capable of measuring the time required for a 1588 message to pass through the end-to-end transparent CROSSLEY et al.: TIME SYNCHRONIZATION FOR TRANSMISSION SUBSTATIONS USING GPS AND IEEE 1588 clock. In comparison, this clock passes all non-1588 packets as a normal bridge/switch. For all 1588 “timing” packets, the end-to-end transparent clock measures the residence time and accumulates the value in a special ﬁeld (correctionField) of the IEEE 1588 message. The ﬁnal destination of the message (i.e., a slave clock) can then compensate the residence delay time, as shown in Fig. 4. Message at egress PTP message Network payload protocol Preamble correctionField headers Message at ingress Event message Network payload protocol Preamble correctionField headers 93 hierarchy can be established and ports in the master state will start to send out 1588 messages. After that, the intermediate IEEE 1588v2 transparent clocks (network bridges/switches supporting 1588v2 features) will measure the delay of 1588 messages between the port in the master state and the port in the slave state. This will be used by the port in the slave state to adjust the clock’s local time. Note: ports in the passive state will neither transmit nor receive 1588 messages. Once the master-slave hierarchy is established, the slave clock(s) will estimate the time offset between itself and the Master Clock using data packets containing time information. With reference to Fig. 6; the relationship between timestamps t1 and t4 is: + t1 + time offset + propagation delay = t4 time offset = t4 − t1 − propagation delay + Ingress timestamp Ingress Egress timestamp + Residence time bridge 4) A peer-to-peer transparent clock is a network bridge or switch with the ability to measure the residence time of an IEEE 1588 message and the delay of the link on which the receiving port locates. Similarly, the peer-to-peer transparent clock will forward all non-1588 messages as a normal bridge/switch. For speciﬁc 1588 packets, the peer-to-peer transparent clock will measure the residence time and link delay and then update the value of the “correctionField” so that the receiver of the messages (i.e., a slave clock) can compensate for both the switching delay and the path delay. The residence time and link delay measurement model of a peer-topeer transparent clock is shown in Fig. 5. GPS timing source Master - Master clock is synchronized with clock source (2) Egress Fig. 4. End-to-end transparent clock measuring 1588 message residence time . path + residence time (1) correctionField Transparent clock Slave - Peer to peer delay mechanism measures path propagation time - Transparent clock measures residence and path delay Industrial Ethernet Fig. 5. Peer-to-peer transparent clock measuring 1588 message residence time and path delay . B. Working Principle of 1588v2 The 1588v2 synchronization process requires the selection of the grandmaster or master used to synchronize the slaves and subsequently the establishment of master-slave hierarchy. The best master clock algorithm deﬁned in  can determine which clock in the network is the best clock so that it can be selected as the grandmaster clock. The best master clock algorithm can then determine the state of each port (i.e., master, slave or passive) on a clock. Once the clock selection and state determination are accomplished, the master-slave Hence, it is necessary to measure the propagation delay before calculating the time offset. When using 1588v2 in power industry, the IEEE 1588 Power Proﬁle  speciﬁes the use of Peer Delay Request-Response to measure the propagation delay. More speciﬁcally, this measures the propagation delay between the Master port and associated Slave port(s) using the messages “Sync, Pdelay Req and Pdelay Resp” as shown in Fig. 6. If a 1-step clock is used, there is no Follow Up and Pdelay Resp Follow Up message. If a 2-step clock is used, a Follow Up and a Pdelay Resp Follow Up are required. A 1step clock measures the actual time t1 when the Sync message is sent out and encapsulates the precise timestamp in the Sync message. Whilst the 2-step clock only packs a coarse timestamp in the Sync message and the actual time is carried in the associated Follow Up message later. The synchronization process using the Peer Delay RequestResponse mechanism is described below: 1) Master clock generates the Sync message and transmits it; the time t1 when the Sync message is sent out is carried in Sync message (or in the Follow Up message). 2) When a Sync message arrives at a peer-to-peer transparent clock, timestamp t2 is recorded by the clock as the ingress timestamp. When the Sync message is forwarded, timestamp t3 is generated as the egress timestamp. The time difference between t3 and t2 is the residence time that is accumulated in the correctionField of Sync (or Follow Up). The peer-to-peer transparent clock also accumulates the link delay in the correctionField of Sync. If there are multiple peer-to-peer transparent clocks, the residence time and link delay measurement and accumulation process is repeated in each clock. 3) Slave clock receives the Sync message (and Follow Up message), records the receiving time t4 , and extracts the timestamp t1 and correctionField data. Once the slave clock receives the sending and receiving timestamps with propagation delay of Sync, it can calculate the time offset between the master clock and the slave clock. Finally, once the time offset is calculated, the local time of the slave clock is adjusted to follow the master clock time. 94 CSEE JOURNAL OF POWER AND ENERGY SYSTEMS, VOL. 2, NO. 3, SEPTEMBER 2016 ts1 Pdelay_Req ts2 ts3 Pdelay_Req Pdelay_Resp ts4 ts6 ts7 Pdelay_Resp ts8 Pdelay_Resp_Follow_Up t1 Port in Master state ts5 Pdelay_Resp_Follow_Up Sync t2 Peer-to-peer transparent t3 clock Follow_Up Port in Slave state Sync t4 ts9 Initially, slave time = master time + time offset Clock output: 1-PPS or IRIG-B Follow_Up Pdelay_Req ts10 ts11 Pdelay_Resp ts13 ts12 Pdelay_Req Pdelay_Resp Pdelay_Resp_Follow_Up Timestamps known by Slave Clock t4, also t1 if 1-step Master in use t1, t4 ts14 ts15 ts16 Pdelay_Resp_Follow_Up correctionField of Sync / Follow_Up = fractional ns = CF0 correctionField of Sync / Follow_Up = CF0 + linkDelay 1 = CF1 correctionField of Sync / Follow_Up = CF1 + residenceTime = CF2 correctionField of Sync / Follow_Up = CF2 + linkDelay 2 Fig. 6. Peer delay request-response mechanism. C. Previous Application and Research of IEEE 1588 In China, IEEE 1588 has been employed in real substations since December 2009 and details of network design and engineering are provided in . However, no results on the performance of IEEE 1588 timing have been given. IEEE 1588 devices including clocks and Ethernet switches from various manufacturers with power proﬁle were tested in the IEEE 1588 power proﬁle plugfest hosted by the IEEE Power & Energy Society (PES) Power System Relaying Committee (PSRC) in early 2010 . Results indicated the achievable timing accuracy was in the range between a few hundred nano-seconds and a few micro-seconds when there was no other trafﬁc in the network. It was also discovered the best master clock algorithm would select a non-qualiﬁed clock as the grandmaster clock during the transient state. The report also suggested a multi-vendor testbed would be useful in identifying implementation issues. Authors in  integrated IEEE 1588 timing in a real substation automation system in Italy in 2011. In this system, the Ethernet switches did not support 1588 and the delay request-response mechanism had to be used instead of the Peer Delay Request-Response mechanism that is speciﬁed by the 1588 power proﬁle. Experimental results showed the time of the slave clock drifted even when the background trafﬁc only occupied 1% of the bandwidth. It was advised that IEEE 1588v2 Ethernet switches would be necessary to achieve sub micro-second timing accuracy. A hardware testbed integrating IEEE 1588 and IEC 61850 was set up during the UCAIug Network Interoperability Demonstrations at CIGRE 2012 . It was reported that good interoperability between different vendors was achieved (detailed results were not published); however, utilities may not obtain sufﬁcient conﬁdence on IEEE 1588 timing because the testing period was too tight. In Australia, a full IEEE 1588 hardware testbed with Peer Delay Request-Response mechanism was built in 2012 . Test results indicated that timing accuracy better than 500 ns could be obtained when three transparent clocks were used between grandmaster and slaves. However, all tests were conducted with no background trafﬁc and the testing period was only 1800 s, which was relatively short. Therefore, this paper expands the research work by integrating assessment of long-term stability of GPS receivers and IEEE 1588 slaves and investigation of impact of IEEE 1588 trafﬁc on network latency. III. P ERFORMANCE A NALYSIS FOR GPS & 1588 T IMING A. Timing Stability of GPS The timing system for a transmission substation needs to maintain a high level of accuracy over a long period. However, direct use of GPS is now considered unreliable because of problems related to climatic conditions and signal CROSSLEY et al.: TIME SYNCHRONIZATION FOR TRANSMISSION SUBSTATIONS USING GPS AND IEEE 1588 interference. Hence, it is necessary to investigate the longterm time differences when using different GPS receivers. This is critically important since automation systems use phasor measurement information from widely separated substations and this relies on time references derived from different GPS receivers. The laboratory system used for measuring the time offset between different GPS receivers is shown in Fig. 7. One of the GPS receivers is selected as the reference and its 1-PPS is used as the reference input to the measurement server. The other GPS receivers, A and B, also feed their own 1-PPS to the measurement server. The server then measures the time difference between the rising edge of the 1-PPS from reference GPS receiver and receiver A and B under test. Note that the mask angle for GPS signal reception is only conﬁgurable on GPS receiver A and the device manual suggests increased mask angle reduces the ﬁeld of view and lowers the timing error caused by multi-path satellite signals situated low in the sky. 1-PPS GPS Antenna Measurement server 1-PPS GPS Antenna 1-PPS Reference GPS receiver 95 an increased mask angle can improve the timing accuracy by reducing the occurrence of spikes in the timing signal. A reason for the time offset spikes is the location of a high wall to the south of the GPS antennas. Results indicate satellites are lost at the southern edge and the signal quality might not be sufﬁcient for ideal synchronization; see Fig. 9. Fig. 9. Log of GPS satellites using reference receiver. According to Fig. 10, the accuracy of the 1-PPS from receiver B is worse than receiver A; occasionally the time offset of GPS receiver B exceeds the threshold value ±1 μs, which might lead to mal-operation if an application requires an accurate time reference from multiple GPS receivers. The minimum and maximum time offset of receiver B measured during the tests are −677 ns to 1372 ns. Time Offset for GPS Receiver B GPS Antenna GPS receiver B Fig. 7. Measurement of time offset between receivers. The time offset between GPS receiver A and reference GPS receiver is shown in Fig. 8, and Fig. 10 shows the time offset between receiver B and the reference. The time offset for receiver A is always signiﬁcantly less than ±1 μs and the time offset spikes are reduced when the mask angle is increased. More speciﬁcally, the minimum offset value is −379 ns and the maximum value is 456 ns when the mask angle is less than 8◦ . When the mask angle is increased to above 15◦ , the timing offset varies between −346 ns to 318 ns. This veriﬁes Time Offset (ns) Time Offset for GPS Receiver A 900 800 700 600 500 400 300 200 100 0 −100 −200 −300 −400 −500 Receiver A with 5 degree mask angle Receiver A with 8 degree mask angle Receiver A with 15 degree mask angle Receiver A with 20 degree mask angle 0 5,000 10,000 15,000 20,000 25,000 30,000 35,000 40,000 45,000 Time (s) Fig. 8. Time offset between receiver A and reference. Time Offset (ns) GPS receiver A 1500 1400 1200 1000 800 600 400 200 0 −200 −400 −600 −800 −1000 Receiver B with unconfigureable mask angle 0 10,000 20,000 30,000 40,000 50,000 60,000 Time (s) Fig. 10. Time offset between receiver B and reference. B. Timing Stability of IEEE 1588 Synchronization System Unlike direct use of GPS, 1588 timing relies on a wired data network to accomplish time synchronization within a substation. This means it is more controllable as the data network can be carefully designed, engineered and monitored so that it could always operate correctly. The lab setup to measure the time offset between the reference GPS receiver (1588 grandmaster) and the 1588 slaves is indicated in Fig. 11. The reference GPS receiver can transmit 1588 timing packets as well as output 1-PPS to the measurement server. The 1588 slaves are synchronized and then feed 1-PPS to the measurement Server for time offset measurement. The Grandmaster and Slave clocks use the 1588 Power Proﬁle and all Ethernet switches utilize the Peer Delay Request-Response Mechanism as required by the 1588 Power Proﬁle. The background trafﬁc in the data network occupies 85% of the bandwidth and is multicast trafﬁc with priority 4. The bandwidth of the network is 100 Mb/s. Note: 1588 96 CSEE JOURNAL OF POWER AND ENERGY SYSTEMS, VOL. 2, NO. 3, SEPTEMBER 2016 Power Proﬁle requires the worst-case time offset to be less than ±1 μs with a network load up to 80% of total bandwidth and the priority of 1588 trafﬁc is 4. GPS Antenna 1-PPS 1588 switch (peer-to-peer transparent clock) Grandmaster clock (reference GPS receiver) Fig. 11. slaves. Measurement server 1-PPS 1-PPS 1588 switch (peer-to-peer transparent clock) 1588 switch (peer-to-peer transparent clock) Slave Clock-1 Slave Clock-2 1588 network Measurement of time offset between 1588 grandmaster and 1588 The measured time offsets for both 1588 Slave clocks are shown in Fig. 12. The offsets are always less than ±150 ns and the ﬂuctuation is signiﬁcantly less than achieved with individual GPS receivers. Note: during the measurement period, a single 1588 packet is randomly lost due to deﬁciencies in the Ethernet switch. However, the 1588 slaves always maintain synchronization. C. Resynchronization of GPS Timing In a real substation, a local clock may lose global/national/ regional synchronization due to a power supply failure or the loss of an external synchronization signal. For example, U.K. National Grid reported a number of differential protection maloperations resulting from corruption of the 1-PPS output of GPS receivers upon GPS restoration. To understand the issue, the test facility shown in Fig. 7 is used to monitor the behavior of two GPS Receivers after the loss of GPS synchronization and then its subsequent restoration. To emulate the loss of the GPS signal, the GPS Antennas connected to receivers A and B are temporarily disconnected. The measurement results are illustrated in Fig. 13 and 14. When receiver A loses the GPS signal, the time offset between itself and the reference GPS receiver drifts at 2 ns/s. The mask angle of receiver A is 20◦ and the worst-case initial time offset is 150 ns. Therefore, receiver A can maintain micro-second accuracy for 425 s. When receiver A regains GPS signal, an 8 μs timing spike in the time offset occurs and its duration is several seconds. This could cause a problem for an IED if precise timing is required . Time Offset for 1588 Slaves Time Offset for GPS Receiver A 200 1588 Slave C 1588 Slave D 100 Time Offset (ns) Time Offset (ns) 150 50 0 −50 −100 −150 0 5,000 10,000 15,000 20,000 25,000 30,000 35,000 40,000 Time (s) Time offset between 1588 slaves and grandmaster. The results presented in this paper demonstrate IEEE 1588 delivers long term timing accuracy within ±150 ns when 1588 Power Proﬁle is enabled on 1588 Grandmaster and Slave clocks and all Ethernet switches support 1588 Peer Delay Request-Response mechanism, regardless of the network load (85% bandwidth in this test) and the loss of 1588 packets. The variation in the timing accuracy is −79 ns to 107 ns for slave C and −109 ns to 72 ns for slave D. In comparison, conventional GPS timing system delivers inferior long term timing accuracy as factors such as signal interference, antenna positions, and system installation techniques degrade the performance of the system. The situation often becomes worse when a large number of distributed GPS antennas and receivers are used and each IED obtains the time signal from its own GPS Receiver. In addition, the clock adjustment algorithm used in different GPS receivers is not the same, and in some cases the time offset between GPS receivers can fail to satisfy the ±1 μs requirement. However, when 1588 timing is used, only one GPS receiver operates as the 1588 grandmaster for the whole substation at any point in time, and all 1588 Slaves follow this grandmaster and deliver ±1 μs accuracy more easily. Loss of GPS Restoration of GPS 100 200 300 400 500 600 700 800 900 Time (s) Fig. 13. Receiver A time offset: loss and recovery of GPS. When receiver B loses the GPS signal, the time offset between itself and the reference receiver drifts at 1.33 ns/s. When GPS synchronization is restored, Receiver B recovers synchronization very quickly, and no signiﬁcant time offset spike is observed. Hence, the transient performance of receiver B is acceptable for IEDs that require precise timing. Time Offset for GPS Receiver B 600 Time Offset (ns) Fig. 12. 2,000 1,000 0 −1,000 −2,000 −3,000 −4,000 −5,000 −6,000 −7,000 −8,000 −9,000 −10,000 0 Loss of GPS 400 200 0 Restoration of GPS −200 0 100 200 300 400 500 Time (s) Fig. 14. Receiver B time offset: loss and recovery of GPS. 600 700 CROSSLEY et al.: TIME SYNCHRONIZATION FOR TRANSMISSION SUBSTATIONS USING GPS AND IEEE 1588 D. Resynchronization of 1588 Timing The test facility shown in Fig. 11 is used to investigate how 1588 slave clocks behave when 1588 timing packets are lost and then recovered. The loss is emulated by disconnecting the 1588 slaves from their adjacent Ethernet switches. The impacts on slave C and D are shown in Fig. 15 and 16 respectively. When slave C is disconnected from the data network, it drifts at 0.8 ns/s, resulting in a 1 μs accuracy holdover time of about 1116 s. Once the connection is restored and it starts to receive the 1588 packets, it immediately synchronizes to the grandmaster. In comparison, when slave D is disconnected, it drifts at −3.6 ns/s and the holdover time is about 248 s. Again, it rapidly synchronizes to the grandmaster when the communication is restored. Time Offset for 1588 Slave C 150 Recovery of 1588 Packets Time Offset (ns) 100 Two extremely important components of the network trafﬁc are the IEC 61850 9-2 sample values (SV) and the IEC 61850 GOOSE messages. These maintain the correct operation of the primary plant used in a substation and ensure the power system operates in a stable state. As deﬁned in the IEC 61850 standards and IEEE 1588 Power Proﬁle, 1588, SV and GOOSE trafﬁc all have the same priority value of 4. This means a particular packet from any of these sources would be randomly placed in the output queue as shown in Fig. 17, which means the residence time within a switch is un-deterministic. If a 1588 Ethernet switch is used, un-deterministic residence time values will not affect the 1588 timing accuracy since the delays are measured and compensated and this is previously proved by the tests described in Section B. However, a data packet such as a SV message is very sensitive to the network latency . Hence, it is necessary to investigate how the 1588 trafﬁc affects the data trafﬁc from other applications that share the same Ethernet network. 50 Loss of 1588 Packets 0 −50 −100 0 100 200 300 400 500 1588 with priority = 4 8 7 6 5 4 3 2 1 61850 9-2 SV with priority = 4 Mixed traffic Time (s) Fig. 15. 97 Switch Time offset slave C: Loss-recovery 1588 packets. 87 6 5 4 3 2 1 1 2 3 4 5 6 7 Time Offset for 1588 Slave D 100 Loss of 1588 Packets Time Offset (ns) 0 2 6 2 5 4 1 3 1 2 1 61850 GOOSE with priority = 4 −100 Fig. 17. Interaction of trafﬁc with same priority. −200 −300 Recovery of 1588 Packets −400 0 100 200 300 400 500 600 700 Time (s) Fig. 16. Time offset slave D: Loss-recovery 1588 packets. Based on these observations, re-synchronization of 1588 timing ensures 1588 slave clocks follow the grandmaster in a fast and secure manner. In comparison, restoration of the GPS signal in a conventional GPS receiver might introduce a time offset spike that corrupts the output of the receiver and could cause mal-operation of an IED. The test setup illustrated in Fig. 18 is used to investigate the impact of 1588 trafﬁc on the latency of SV packets through an Ethernet switch. The grandmaster injects 1588 messages into the switch with a rate up to 128 packets/s and the latency of SV packets can be measured under various 1588 trafﬁc load conditions. More speciﬁcally, a MU produces the SV stream which will be copied by an Ethernet tap before entering the 1588 switch. The copied SV packet is captured Wiring Legend Original SV stream Copied SV stream 1588 traffic E. Impact of 1588 Trafﬁc The IEEE1588 protocol uses data packets within an Ethernet network to disseminate the time reference from the grandmaster. Sharing an Ethernet network between 1588 trafﬁc and trafﬁc related to other IED applications can result in an inevitable interaction. This might affect a critical IED especially if the trafﬁc is not carefully managed and the network is not appropriately engineered. Grandmaster clock (1588 traffic generator) MU Capture card Ethernet tap 1588 switch T2 T1 Fig. 18. Measurement of latency of SV packets. 98 CSEE JOURNAL OF POWER AND ENERGY SYSTEMS, VOL. 2, NO. 3, SEPTEMBER 2016 and timestamped by an Ethernet capture card at T1 . After the original SV packet is sent out from the switch, it is captured and timestamped by the capture card at T2 . Hence, the latency of a SV packet is (T2 −T1 ). The measurement error introduced by the Ethernet tap and the capture card is < ±60 ns, which is negligible for the network latency measurement. The measurement results are shown in Fig. 19; the average latency for a SV packet is about 18 μs. This is because the size of the SV packet from the MU is 141 bytes or 1128 bits and the transmission delay within the switch is 1128/(100 × 106 ) = 11.28 μs @ 100 Mb/s. According to IEC 61850-904 , there is a minimum switching delay of 8 μs within an Ethernet switch and thus the total latency for a SV packet is about 19.28 μs, which is similar to the measurement results. It is also observed latency can increase to about 24 μs when there is no 1588 trafﬁc. The reason is that an Ethernet switch has to periodically send a bridge protocol data unit (BPDU) according to the Ethernet standard . A BPDU contains 64 bytes or 544 bits, which causes additional latency of 544/(100 × 106 ) = 5.44 μs @ 100 Mb/s if it is transmitted before a SV packet. Density SV Latency Within an IEEE 1588 Ethernet Switch under Different 1588 Traffic 0.020 0.018 0.016 0.014 0.012 0.010 0.008 0.006 0.004 0.002 0 SV latency without 1588 traffic SV latency with 96 Sync/second SV latency with 128 Sync/second 1818.5 1919.5 2020.5 21 21.5 2222.5 23 23.5 2424.5 2525.5 2626.5 2727.5 28 28.5 2929.5 30 Latency (μs) Fig. 19. Impact of 1588 on SV latency. When 1588 trafﬁc is present in the network, the maximum latency for a SV packet increases from 24 μs to 29 μs. Considering a 1588 Sync message contains 72 bytes or 576 bits, the transmission latency is thus 576/(100 × 106 ) = 5.76 μs. 29 μs indicates a BPDU and a 1588 message are transmitted before a SV packet. Consequently, if 1588 trafﬁc shares the data network with critical automation applications, it will introduce additional delay of about 5.76 μs at a data rate of 100 Mb/s or 0.576 μs at 1000 Mb/s. This has to be considered when planning and designing the data network. IV. C ONCLUSION Centralized time synchronization based on the IEEE1588 protocol has numerous advantages, including: the ability to use high quality antennas and receivers; simpliﬁed installation requirements; automatic compensation of the propagation delay from the time reference; removal of the need for manual measurement of signal delay; operation over general substation Ethernet network; removal of the need to construct a dedicated time distribution network; reduced requirements on number of output modules on GPS receivers; future scalability; immunity to GPS synchronization degradation and loss; and ease of maintenance and replacement. However, if high timing accuracy (< 1 μs) is required from IEEE 1588, two extra requirements are required, which signiﬁcantly increase the capital cost: all Ethernet switches must be 1588 compliant and additional 1588 slave clocks are required to translate 1588 and provide 1-PPS and IRIG-B signals for non-1588 IEDs. In comparison with timing approaches directly using 1-PPS and/or IRIG-B from GPS receivers, 1588 timing over a fully compliant 1588 network has the following advantages: better long term accuracy; no time offset spike when synchronization is restored; correct operation under heavy network load conditions and during the loss of 1588 packets; negligible bandwidth consumption and negligible impact on sample values or GOOSE messages. In conclusion, time synchronization system based on IEEE 1588 is an ideal method for accurate timing of IEDs in a transmission substation. R EFERENCES  D. Ingram and B. Smellie. (2014, Oct.). Solving electrical substation timing problems: A white paper on the use of the precision time protocol for substation protection and control systems. [Online]. Available: http:// www.chronos.co.uk/ﬁles/pdfs/tek/SOLVING ELECTRICAL SUBSTATION TIMING PROBLEMS.pdf  W. An, N. Tart, D. Barron, M. Bingham, and A. Hackett, “A transmission utility’s experience to date with feeder unit protection systems,” in 2012 International Conference on Developments in Power Systems Protection (DPSP), 23–26 Apr. 2012, pp. 1–6.  C. Strauss. Practical Electrical Network Automation and Communication Systems. Oxford, UK: Newnes, 2003.  IEEE Standard for a Precision Clock Synchronization Protocol for Networked Measurement and Control Systems, IEEE Standard 1588TM 2008, 2008.  Siemens AG. IEEE 1588 precision time synchronization solution for electric utilities. [Online]. Available: http://w3.siemens.com/mcms/ industrial-communication/en/rugged-communication/technologyhighlights/ieee-1588-precision-time-synchronization-solution-forelectric-utilities/Pages/ieee-1588-precision-time-synchronizationsolution-for-electric-utilities.  IEEE Stanadard Proﬁle for Use of IEEE 1588 Precision Time Protocol in Power System Applications, IEEE Standard C37.238-2011, 2011.  R. Moore, R. Midence, and M. Goraj, “Practical experience with IEEE 1588 high precision time synchronization in electrical substation based on IEC 61850 process bus,” in 2010 IEEE General Meeting, July 2010, pp. 1–4.  C. M. De Dominicis, P. Ferrari, S. Rinaldi, and M. Quarantelli, “On the use of IEEE 1588 in existing IEC 61850-based SASs: Current behavior and future challenges,” IEEE Transactions on Instrumentation and Measurement, vol. 60, no. 9, pp. 3070–3081, Sep. 2011.  M. Goraj, H. Kirrmann, R. Mackiewcz, and C. Hoga. (2012, Dec.). UCAIug at CIGRE 2012. PAC World. [Online]. December 2012 Issue (Industry Reports). Available: https://www.pacw.org/no-cache/issue/ december 2012 issue/industry reports/do we have a conscience. html  D. M. E. Ingram, P. Schaub, and D. A. Campbell, “Use of precision time protocol to synchronize sampled-value process bus,” IEEE Transactions on Instrumentation and Measurement, vol. 61, no. 5, pp. 1173–1180, May 2012.  Communication Networks and Systems for Power Utility Automation – Part 5: Communication Requirements for Functions and Device Models, IEC 61850-5: 2013 (E), 2013.  Communication Networks and Systems for Power Utility Automation – Part 90-4: Network Engineering Guidelines, IEC 61850-90-4: 2013, 2013.  IEEE Standard for Ethernet – Section 1, IEEE Standard 802.3-2015, 2016. CROSSLEY et al.: TIME SYNCHRONIZATION FOR TRANSMISSION SUBSTATIONS USING GPS AND IEEE 1588 Peter A. Crossley (M’95) received the B.Sc. degree from UMIST, Manchester, U.K., in 1977 and the Ph.D. degree from the University of Cambridge, U.K., in 1983. He is a Professor of electrical power systems engineering at the University of Manchester, U.K. and Director of the EPSRC Centre for doctoral training in power networks. He has published over 250 technical papers on protection and is an active member of CIGRE, IEEE and IET. Hao Guo (S’12) received the B.Eng. degree from the University of Manchester, Manchester, U.K. in 2012. He is currently a Ph.D. student at The University of Manchester. His research interests include time synchronization, Ethernet redundancy and engineering within power substations. 99 Zhao Ma rreceived the B.Eng. degree from Xian Jiaotong University in 1982 and the Ph.D. degree from the Staffordshire University, U.K., in 1996. He is National Distinguished Expert of “1000 Elite Program” and Chief Expert – Smart Power Distribution of China Electric Power Research Institute (CEPRI). He is CEng. FIET, current CIGRE China and the Asia Paciﬁc Region SC6 Chairman. His main work areas include: smart distribution network planning and asset management; intelligent T&D equipment, in particular for design and development; technical consulting; and MVDC and Energy Internet.
* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project