Improving Availability of Distributed PMU in Electrical Substations using Wireless Redundant Process Bus

Transcription

1 Improving Availability of Distributed PMU in Electrical Substations using Wireless Redundant Process Bus Paolo Castello 1, Paolo Ferrari 2, Alessandra Flammini 2, Carlo Muscas 1, Paolo Attilio Pegoraro 1, Stefano Rinaldi 2 1 Department of Electrical and Electronic Engineering, University of Cagliari, Cagliari, Italy. Tel: (+39) Department of Information Engineering, University of Brescia, Brescia, Italy. Tel: (+39) paolo.castello@diee.unica.it, stefano.rinaldi@ing.unibs.it Abstract Protection and measurement systems in electrical substations are requested to have high availability. When considering an all-digital substation protection system, all the components and infrastructure devices (instrument transformers, processing units, merging units, intelligent electronic devices, communication network and synchronization source) contribute to the overall availability level. In particular, in this paper the availability of a distributed phasor measurement unit inspired by the standards IEC 61850, is investigated. A cost-effective solution to enhance distributed PMU reliability by means of redundancy using an IEEE wireless network is proposed. The process bus has a wireless backup for both time synchronization protocol and sampled values packets carrying measurement information. Results are reported for a wireless network that uses software based timestamping for PTP synchronization. Keywords; phasor measurement unit; process bus; wireless LAN; availability; hybrid network; synchrophasor; substation; WiFi I. INTRODUCTION The reliability of the protection and measurement system in electrical substations strongly depends not only on the instrument transformers or processing unit but on the overall system that includes several electronic devices: merging units (MUs), intelligent electronic devices (IEDs), communication network and synchronization source. In particular, the availability of the distribute phasor measurement unit (PMU), designed for power substations (as presented in [1], [2]) and inspired by the standards IEC 61850, has to be discussed following the scheme proposed in the literature for all-digital substation protection systems. In [3], for instance, it has been underlined that the Ethernet switches in the communication infrastructure of the process bus can become the bottleneck of the system reliability and Ethernet redundancy has a great impact on the mean time between failures (MTBF). A cost-effective solution to enhance distributed PMU reliability could be increasing the redundancy in the communication infrastructure by means of an IEEE wireless network. The process bus could thus have its wireless counterpart that serves as a backup for both precision time protocol (PTP) packets for MU synchronization and sampled values (SVs) packets (directed to the IED enabled to behave as a PMU). In this paper the possibility to use a Wireless backup communication infrastructure for distributed PMU application is investigated. Simulation results are reported for a wireless network that uses software based timestamping for PTP synchronization. II. THE POINT OF FAILURES OF A DISTRIBUTED PMU In the architecture of a distributed PMU, proposed in [1], [2], the different functionalities required by the PMU (voltage and current sampling, filtering, estimation, time synchronization and communication) are distributed in different nodes in the electrical substation, as it is shown in the generic scheme in Figure 1. The MUs collect the data from the sensors connected on the power grid and transfer this information on the network. Then, these data are collected and processed by an IED, which implements the synchrophasor estimation algorithm and produces the results. The correctness of the estimation relies on the time synchronization accuracy which can be achieved, as well as on the algorithm used for synchrophasor evaluation. Generally speaking, the time information is provided to the distributed nodes by a time reference source. Typically, the IEEE 1588 protocol is used to distribute the time information because of the strict synchronization accuracy, in the order of the microsecond, required by synchrophasor estimation. This distributed architecture has more points of failures than a centralized solution, whose analysis has been presented in [4]. In particular, the typical failures of the distributed PMU are: failure of the hardware platform of the nodes (IED, MUs, sensors,..); failure of the synchronization system; failure of the communication infrastructure (switches, routers,..). A possible solution could be the duplication of all the devices (sensors, MUs and IEDs,..) but this solution is, generally, not feasible from the economic point of view. As

2 well discussed in [5], some of the elements that compose a distributed measurement and control system in process bus are more important than others from the point of view of the availability of the system. In particular, the time distribution system is one of the weaknesses of the distributed PMU, since it could affect the accuracy of the estimation of the synchrophasor. The analysis of the behavior of the synchronization system in case of fault is out of the scope of the paper since it is well described in [6]. The other weakness, typical of any distributed system, is the network infrastructure, used in the considered PMU architecture to transfer the data from MUs to the IED. In case of a network failure, the measurement data transmitted by the MUs are lost and the IED cannot estimate the synchrophasor. In addition, a failure of the network affects also the time synchronization of the distributed measurement devices: the loss of synchronization packets inevitably affects the synchronization performance. For these reasons it is important to improve the availability of network infrastructure. In the following section, solutions typically adopted to improve the availability of network infrastructure are presented and analyzed. Figure 1. Architecture of a IEC distributed measurement system. III. THE COMMUNICATION INFRASTRUCTURE A. IEC Process Bus The standards IEC [7] define two logical communication levels: the station bus and the process bus. The former connects the SCADA (Supervisory Control And Data Acquisition) system with the control at bay level. The latter links the controller of the bay (the IED) with sensors and protections installed at field level. As mentioned in the previous section, distributed PMUs make use of the process bus to transfer the information from voltage and current sensors to the IED, where the information is processed and elaborated to estimate the synchrophasor. The performance of the process bus must be higher than the performance of the station bus because the information exchanged at field level is more critical. Among the requirements, those to be mentioned are: low communication latency; accurate time synchronization; bandwidth; availability and reliability. As mentioned in [8], these requirements can be satisfied together using advanced redundant network infrastructures and dedicated protocols [9]. In particular, the proper configuration of switched Ethernet network (like the use of message priorities, V-LAN and multicast filtering, as suggested in [10]) allows reducing the communication latency and jittering. The most demanding time synchronization requirements can be satisfied using GPS receiver, whose time information could be distributed in the substation using a dedicated IRIG-B fiber optic network. Nevertheless, today this solution has been progressively replaced by PTP (Precision Time Protocol), which allows an accurate time distribution using the communication infrastructure [8]. The transmission of the raw sampled measurement data, the so called Sampled Value (SV), requires an adequate transmission bandwidth: in fact, each merging unit can generate a stream of data on the order of some megabits per second. Typically, 1 Gbit/s or 10 Gbit/s fiber optic Ethernet network are used at process bus level. The high availability and reliability of the network is typically obtained adopting physically redundant network topology (e.g. ring structure) managed by redundancy protocols, like RSTP (Rapid Spanning Tree Protocol) which is able to reconfigure a ring network in less than 100 ms. Although RSTP is faster than traditional STP protocol, it cannot guarantee a zero-recovery time, required by the most demanding applications, like protections. Recently, new solutions have been proposed to satisfy such requirements: PRP (Parallel Redundancy Protocol) and HSR (High- Availability Seamless Redundancy). The former guarantees the zero-recovery time by transmitting a copy of each message over two separate networks, the latter sends a copy of each message on both the two direction of a ring network. In any case, both solutions imply a relatively high increase of the overall installation costs. In the following section, an alternative approach to improve the reliability of the process bus by mean of the use of Wireless links, as suggested in [9], is described and applied to improve the availability of distributed PMUs. B. Improving process bus availability As mentioned in the previous section, the most demanding applications in electrical substations, like protections, have strict availability requirements. A zero-recovery time can be, for instance, satisfied using dedicated network solution, like PRP or HSR, highly increasing the overall installation costs. As demonstrated in [9], the failure of the network may cause the loss of packets up to several hundreds of milliseconds - the typical recovery time of an RSTP network- that does not allow the distributed PMU to compute synchrophasors in the meanwhile. In this case, the availability of the process bus can be augmented by means of an alternative solution, as proposed in [9]: besides the performing, Ethernet-based, process bus an additional WLAN network infrastructure can be installed. In the case a fault affects the wired network, a copy of the information can be transferred to the end user using the backup WiFi connection, achieving a zero-recovery time. The improved architecture of the distributed measurement system over the process bus is shown in Figure 2. The IED, which collects the SVs packet and performs the synchrophasor algorithm, is connected to the distributed MUs by a main Ethernet process bus and by a back-up WiFi connection. The MUs send simultaneously copies of the SV packet on both the connections: in the case of failure in the wired network, the information is transmitted over the back-up connection. The IED normally receives two copies of the SV message from the two connections; it accepts the first copy and discards the second one. The main advantage of this solution, compared to

3 PRP and HSR solutions, is the lower installation costs: the installation of a WiFi network is typically cheaper than the installation of a second, additional, cabled infrastructure or the use of specific network hardware interfaces. As demonstrated in [9], the WiFi communication can be advantageously adopted also in electrical substations, despite the high level of electromagnetic noise. Two possible operative modes of WiFi network can be used (see [11]): peer to peer communication (ad hoc mode) or communication via access point station (infrastructure or mode). As secondary benefit, the use of a back-up connection also improves the synchronization availability: in the case of a fault on the cabled network, a copy of the time information is transferred to the end nodes using the WiFi connection. Nevertheless, the quality of service provided by the WiFi connection is typically lower than the one of a wired connection: the communication latency is higher and the synchronization performance is lower, a situation that may affect the overall performance of the distributed PMU. For instance, in the proposed approach, the distributed MUs are synchronized using PTP protocol; this means that the clock synchronization is affected by the additional latency variability of the WiFi connection. In the rest of the paper, the analysis is focused on performance which can be obtained by the distributed PMU in the case information coming from WiFi connection is used by both the measurements algorithm and the synchronization protocol. IED Backup WiFi channel SV MU 3i WiFi Process Bus IEEE 1588 Router Figure 2. Improving the distributed measurement architecture based on IEC process bus. IV. THE IMPACT OF WIRELESS COMMUNICATION ON DISTRIBUTED PMU PERFORMANCE When the wired Ethernet communication network undergoes temporary outages, as described in previous sections, the IED receives SV packets only from the backup IEEE wireless network. Thus, the time tagging of the received packets presents an uncertainty related to the synchronization achievable by PTP algorithm using the wireless network. A low-cost software solution based on linux operating system and general purpose processors, obtained by timestamping the SV at drivers level, can reach a jitter in synchronization offset in the order of a few microseconds as shown in [11]. In particular, a b WLAN at 11 Mbps has an average offset deviation of 4.6 µs, with a variance equal to MU Switch SV 3v Cbrk Station Bus GOOSE 2.5 µs 2, in infrastructure mode ( mode) and 0.66 µs, with variance 0.2 µs 2, in ad-hoc mode. The synchronization offset, as pointed out in [2], impacts directly particularly on the estimation of the phase angle and thus affects the Total Vector Error (TVE). In fact, the TVE can be represented in terms of relative amplitude estimation error a and of phase angle error ϕ a as follows: j ϕ 2 p p ae a a TVE = = + p a a ( ϕ ) where p represents the complex phasor, with amplitude a and phase φ (in radians), the tilde indicates the estimated counterparts, and the approximation is valid for small errors. The phase angle estimation error thus contributes in a quadratic way to the overall TVE and it is linearly related to the time synchronization error when a stationary test signal is considered. On the contrary, when dynamic signals, as the modulation or frequency ramp test signals prescribed by the IEEE C synchrophasor standard [12], are considered, the time-shift caused by synchronization offset has a more complex impact on TVE. For instance, when amplitude and phase modulated signals are considered the timestamping error affects both amplitude and phase estimation. The maximum frequency estimation errors obtained for tests under off-nominal conditions or in presence of stationary harmonics and interharmonics are, in general, not affected by synchronization errors, since time-shifted analyzed signals present the same time-invariant spectral content. The results of frequency ramp tests are instead prone to the synchronization error, due to the direct impact of the time-shift on the frequency variation. However, since the rate of change of frequency used in [12] is equal to 1 Hz/s, even an offset in the order of 10 µs results in a very low frequency shift of 10-5 Hz. Wireless communication also introduces higher propagation delays for SV packets with respect to the wired Ethernet process bus. In [9] an analysis of the propagation performance of WiFi for different substation application is reported. In particular, under noisy substation environments, a propagation delay in the order of milliseconds is considered for SV messages going from MU to an IED designed for overcurrent protection applications. This could be a drawback for latency constraints stated by the synchrophasor standard for PMUs suitable for protection application. Thus, the distributed PMU algorithm should keep also into account the worst case latency due to WiFi communication and, possibly, commute to dedicated procedures for this emergency state. V. TEST RESULTS In this section the results on the performance of a distributed PMU relying on a wireless communication are reported. The architecture proposed in [2] for a PMU and equipped with the enhanced synchrophasor estimation algorithm presented in 2 (1)

4 [13], which aims at being simultaneously compliant with the classes P and M of the standard, is tested. The idea is to verify the performance of the distributed PMU that is compliant with the standard under normal operative conditions, when only the wireless network is working. This approach will allow defining the limits of the given design under failure conditions. The tests prescribed by the synchrophasor standard have been performed for both the ad-hoc and infrastructure WLAN () modes discussed in [11]. The following values have been used for maximum synchronization offset: 9.4 µs in mode and 2 µs for ad-hoc mode. Such values have been obtained using the aforementioned averages and variances of offset reported in [11] for software based timestamping and PTP synchronization, and considering a Gaussian distribution with a 3σ maximum deviation. All the simulated test results are obtained with a sampling frequency F s =4000 Sample/s (80 Sample/cycle at 50 Hz) as suggested by the standard IEC for protection purposes. In the following, the obtained measurement errors are compared with the limits given by the synchrophasor standard [12], for the measurement reporting rate of 50 Sample/s. TABLE I. MAXIMUM TVE % FOR STEADY-STATE TESTS Signal Frequency (range: f 0 ± 5 Hz) Harmonic distortion (10 % each Harmonic) Out-of band interference (10% of signal magnitude with f 0 at 50 Hz) TVE % nd rd th Hz Hz Hz Hz Table I shows results in terms of maximum TVE % for synchrophasor estimations and steady state tests. The algorithm provides a good performance in presence of offnominal frequencies and harmonic distortion (the limit provided by the standard [12] is 1 % in both cases). Different results are obtained in presence of out-of-band interfering signals, when the required TVE limit is 1.3%. In particular, the table reports the values obtained for four specific interharmonic frequencies chosen in the required test range, that is [10, 25] Hz and [75, 100) Hz. The results suggest that compliance with the M-class of the standard could be obtained only in the ad-hoc WiFi mode, whereas mode does not allow meeting all the specification. TABLE II. MAXIMUM TVE % FOR DYNAMIC TESTS Modulation (f m = 5 Hz) Frequency ramp (f 0 ± 5 Hz, ±1 Hz/s ) TVE % k x, k a = k a = Hz/s Hz/s The test results in dynamic cases (modulation and frequency ramp) are shown in Table II. In such cases, the TVE limit is 3% for modulations and 1% for frequency ramps. Compliance with the standard is easily matched. Table III reports the results in terms of maximum response time and maximum under/overshoot for the amplitude and phase step tests. The limits of the standard are 34 ms and 5 % for response time and under/overshoot, respectively. Thus, compliance is obtained, but it is interesting to note that the phase error due to synchronization offset can directly affect also the response to abrupt phase changes, thus leading to longer transients. TABLE III. STEP TESTS RESULTS: MAXIMUM RESPONSE TIME AND MAXIMUM UNDER/OVERSHOOT Amplitude Step (±10 %) Phase Step (±10 ) Max Response time (ms) Max Under/Overshoot (%) Max Response time (ms) Max Under/Overshoot (%) Frequency estimation results are shown in Table IV and V. Such results are reported, in terms of frequency error (FE), to highlight that the differences between the ad hoc and infrastructure mode are negligible and the impact of synchronization offset is very small for frequency estimation. In fact, as aforementioned, the time-shift does not have any impact in frequency estimation under steady-state conditions, while for dynamic signals, its impact is small because of the low rate of change of frequency. TABLE IV. MAXIMUM FE FOR STEADY-STATE TESTS Signal Frequency (range: f 0 ± 5 Hz) Harmonic distortion (10 % each Harmonic) FE [Hz] nd rd th

5 TABLE V. MAXIMUM FE FOR DYNAMIC TESTS Modulation (f m = 5 Hz) Frequency ramp (f 0 ± 5 Hz, ±1 Hz/s ) FE [Hz] k x, k a = k a = Hz/s Hz/s The results for rate of change of frequency (ROCOF) estimations are shown in the Tables VI and VII, in terms of maximum ROCOF error (RFE). As happens for frequency estimation, ROCOF evaluation is only slightly affected by synchronization offset in the dynamic cases and differences between ad hoc and infrastructure mode are negligible. As a consequence, the accuracy of the PMU in frequency and ROCOF estimations is almost the same as in the Ethernetbased architecture. TABLE VI. MAXIMUM RFE FOR STEADY-STATE TESTS Signal Frequency (range: f 0 ± 5 Hz) Harmonic distortion (10 % each Harmonic) TABLE VII. Modulation (f m = 5 Hz) Frequency ramp (f 0 ± 5 Hz, ±1 Hz/s ) RFE [Hz/s] nd rd th MAXIMUM RFE FOR DYNAMIC TESTS RFE [Hz/s] k x, k a = k a = Hz/s Hz/s The results show that, when the wired process bus is unavailable and only the WiFi is working, the distributed PMU could measure synchrophasors in compliance with the M class, a part from the case of out-of-band conditions in mode. However, it should be considered that the higher latency, typically in the order of milliseconds (see [9]) for WiFi communication channel, could prevent the PMU from respecting also the latency compliance required for P class. In contingency condition, it seems reasonable to safeguard protection functionalities, thus, a possible solution is to design a very responsive P-class compliant algorithm, which can be used during communication network emergencies. As a possible example of low latency P-class algorithm, a two cycle algorithm obtained by changing the duration and parameters of the P-channel TFT filter in [13] (see [14] for the TFT mathematical details) has been implemented. In particular order K=1 has been used for Taylor expansion, to reduce the over/undershoot in presence of step changes, and a Kaiser window with β=5 has been adopted. With such a configuration all the synchrophasor measurement requirements of the standard are respected while keeping the latency to 20 ms (the limit is 40 ms for the P class). Table VIII reports, as an example, the response times when step changes occur, thus showing the fast dynamic behavior of the algorithm, both with ad hoc and infrastructure mode WLAN. As aforementioned, the under/overshoots are negligible, while delay time is far smaller than the required 5 ms. TABLE VIII. Amplitude Step (±10 %) Phase Step (±10 ) MAXIMUM RESPONSE TIME FOR STEP TESTS Response Time (ms) Frequency is obtained, in this case, by derivative from the phase angle of the estimated phasor. An additional one cycle boxcar filter is used to obtain a better rejection of harmonics in frequency estimation, while ROCOF is obtained by derivative from subsequent frequency values. The additional filter would give a further half-cycle delay to frequency and ROCOF computation. However, to keep the latency limited to 20 ms (half the duration of the synchrophasor filter), frequency and ROCOF are given by the current estimations corresponding to the synchrophasor time tag. Frequency value is thus corrected using ROCOF estimation to compensate the lack of alignment, as in [15], and full P-class compliance is reached. VI. CONCLUSIONS One of the key requirements of instruments and devices adopted for the automation of primary electrical substation is the availability. In order to improve this aspect, in this paper the failure points of distributed PMU system have been analyzed. The attention has been focused on network infrastructure, that is on the main weakness point of any distributed system. The availability of the process bus can be improved using a wireless back-up connection in addition to wired infrastructure: in case of fault on the cabled connection, the measurement information can be transferred using the wireless link. Nevertheless, the performance of a wireless infrastructure is lower compared to wired network: the time synchronization protocols are able to provide a synchronization uncertainty on the order of few microseconds and the communication latency could rise up to some milliseconds. The influence of these limits on the performance of distributed PMU has been analyzed. The analysis, performed considering a wireless network that uses software based timestamping for PTP synchronization, shows that the distributed PMU, with the algorithm proposed in [13] for synchrophasor measurement, could operate in compliance with the M class also during fault condition (and, then, using the wireless connection), a part from the case of out-of-band conditions, where the worst

6 estimated TVE is on the order of 1.46 %. However, during fault conditions, a low latency P-class compliant PMU, obtained modifying the design parameters, is used to preserve protection functionalities and overcome additional latency due to wireless link delay. REFERENCES [1] P. Castello, P. Ferrari, A. Flammini, C. Muscas, S. Rinaldi, "An IEC Compliant distributed PMU for electrical substations", 2012 IEEE International Workshop on Applied Measurements for Power Systems (AMPS), Aachen, Germany, Sept , 2012, pp [2] P. Castello, P. Ferrari, A. Flammini, C. Muscas, S. Rinaldi A New IED With PMU Functionalities For Electrical Substations, Instrumentation and Measurement, IEEE Transactions on, vol.62, no.12, pp , Dec [3] P. P. Parikh, T. S. Sidhu, A. Shami, "A Comprehensive Investigation of Wireless LAN for IEC Based Smart Distribution Substation Applications," Industrial Informatics, IEEE Transactions on, vol.9, no.3, pp , Aug [4] Peng Zhang; Ka Wing Chan, "Reliability Evaluation of Phasor Measurement Unit Using Monte Carlo Dynamic Fault Tree Method," Smart Grid, IEEE Transactions on, vol. 3, no.3, pp , Sept [5] Zhang, Peichao, Levi Portillo, and Mladen Kezunovic. "Reliability and component importance analysis of all-digital protection systems." Power Systems Conference and Exposition, PSCE' IEEE PES. IEEE, [6] G. Gaderer, S. Rinaldi, N. Kero, "Master failures in the Precision Time Protocol", IEEE International Symposium on Precision Clock Synchronization for Measurement, Control and Communication, ISPCS 2008, Ann Arbor, MI, USA, September 22-26, 2008, pp [7] IEC Communication networks and systems for power utility automation, IEC Ed. 2, [8] C. M. De Dominicis, P. Ferrari, A. Flammini, S. Rinaldi, M. Quarantelli, "On the Use of IEEE 1588 in Existing IEC Based SASs: Current Behavior and Future Challenges", IEEE Trans. Instrumentation and Measurement, September, 2011, Vol. 60, N. 9, pp [9] P. Ferrari, A. Flammini, S. Rinaldi, E. Sisinni, G. Prytz, "Advanced Networks for Distributed Measurement in Substation Automation Systems", 2013 IEEE International Workshop on Applied Measurements for Power Systems (AMPS), Aachen, Germany, September 25-27, [10] IEC Communication networks and systems for power utility automation - Part 90-4: Network engineering guidelines, [11] J. Kannisto, T. Vanhatupa, M. Hannikainen, T.D. Hamalainen, "Software and hardware prototypes of the IEEE 1588 precision time protocol on wireless LAN," Local and Metropolitan Area Networks, (LANMAN) 2005., the 14th IEEE Workshop on, Sept [12] "IEEE Standard for Synchrophasor Measurements for Power Systems," IEEE Std C (Revision of IEEE Std C ), vol., no., pp.1-61, Dec [13] P. Castello, J. Liu, A. Monti, C. Muscas, P. A. Pegoraro, F. Ponci, Toward a class P + M Phasor Measurement Unit, IEEE International Workshop on Applied Measurements for Power Systems, Sept [14] M. A. Platas-Garza, J. A. de la O Serna, "Dynamic Phasor and Frequency Estimates Through Maximally Flat Differentiators," Instrumentation and Measurement, IEEE Transactions on, vol.59, no.7, pp , July [15] A. J. Roscoe, "Exploring the Relative Performance of Frequency- Tracking and Fixed-Filter Phasor Measurement Unit Algorithms Under C Test Procedures, the Effects of Interharmonics, and Initial Attempts at Merging P-Class Response With M-Class Filtering," Instrumentation and Measurement, IEEE Transactions on, vol. 62, no.8, pp , Aug