Fault Characterization Based on Synchrophasor Data Using Heuristic Approach ()
1. Introduction
The complexity of any power systems is increasing drastically due to various reasons such as ever increasing demand, integration of renewable energy sources, variations in the load and high uncertainty in the fault situations. The disparity in availability of energy resources and the increasing loading demands spread across large geographical regions put an extensive stress over the power grids, tending them to operate closer to their limits. For such diverse operations and dense uncertainty surrounding the operators of the power grids, a strong supervisory control is of utmost importance. Supervisory control and data acquisition (SCADA) is instrumental in this process of grid supervision, even though they only provide the steady state view of the power grid; they were the only solution for grid surveillance. With the onset of synchrophasors, the measurements over wide-area, dynamic real time visualization of power system and monitoring safety and security of the grid seemed to be operative. Synchrophasor [1] measurements are provided by PMUs which are located strategically at certain substations/generating stations in the grid; which enables us to understand the performance of the power system under diverse conditions and with this achieves better utilization of the power system without compromising its reliability.
Wide area measurement system has got a new dimension and is possible only because of the synchrophasor technology coupled with the high-speed communication network established. The major advantage of the PMU is their ability to represent the magnitude along with the angle for each phase voltage and current, rate of change of frequency, and angular separation. This gives an upper-hand to observe the dynamic behavior of the grid in real-time, study the functioning and take the corrective action during the times of critical faults.
PMUs are positioned in various substations, and they constantly deliver measurements of time-stamped positive-sequence voltages and currents of all monitored buses and feeders, as well as frequency and rate of change of frequency. These measurements are stowed in local data storage devices, which can then be accessed from remote locations for analysis. The local storage capacity is usually limited, and the previously stored interesting power system event data must be flagged for permanent storage, to avoid that data being overwritten when the local storage capacity is drained. The phasor data is also available for real-time applications in a steady stream as soon as the measurements are made. This data from several PMUs is made available for various tasks and other applications to the power system stability [2] . This paper aims mainly to illustrate the use of PMUs and the method of fault characterization from the inputs received from the PMU. The next section throws light on some important applications of PMU [3] [4] .
1.1. Disturbance Recording
The events which can’t be termed as faults are often witnessed in the power systems operation [5] [6] ; such as loss of major transmission lines, loss of load etc. documenting such disturbances is gaining more importance for studying the exact causes and gaining insight of what exactly happened, and corrective actions necessary to minimize adverse power system reactions [7] [8] .
1.2. Fault Recording
Digital power system monitoring equipment may be designed differently by storing and processing the captured data to utilize it for additional uses. The same hardware can be made to exploit its use for all other applications. Different programs and software are tailored to meet the different specific applications, which inherently provide the ability to select the type and format of the stored data, which is crucial for providing a background for versatile applications of PMU equipment. One important application is the ability to use PMUs as true fault recorders, as they can capture and display actual 50 Hz wave form and magnitude data on individual channels during power system fault conditions.
1.3. Transmission and Generation Modelling Verification
Computerized power system modeling and analysis are now the normal and accepted ways of guaranteeing that power system parameters have been reviewed before any large capital expenditures on major system changes is approved. Previously, the actual verification of computer models via field tests was considered impractical and seemed a remote possibility. But the PMU class of monitoring equipment can now provide the field parameters’ verification required by using the GPS time synchronization between multiple units.
1.4. Overhead Transmission Line Dynamic Monitoring
This application targets to monitor real time dynamic thermal circuit rating, overload and conductor temperature of the overhead transmission lines for improving/optimizing their utilization. This application development is based on synchrophasor technologies, where two of phasor measurement units are installed at the end of the both lines. The average conductor temperature over the lengths of the line is estimated using the lines impedance, which is calculated based on direct measured voltage and current between two substations. With the conductor temperature known, the dynamic rating, overload and sagging of the overhead transmission lines can be determined [9] .
1.5. Power System Oscillation Monitoring
This application is to facilitate the information on dynamic characteristics of inter-area oscillation, along with the data about the oscillation frequency and damping coefficient; due to which the operator gets to know the dynamic behavior of the grid, and helps in better response during the abnormality.
2. Case Studies
Important case studies [10] have been discussed in subsequent sections of the paper from which fault characteristics are studied and MATLAB programs are developed for event detection from the inputs received from various PMUs. The heuristic way suggests the approach of studying the existing phenomenon in detail to develop enhanced operations based on them. The data was received from POSOCO and this PMU data available was plotted for analysis and based on these results the consequences of fault on the system were studied. The fault type and fault clearing time was also evident from those graphs. Every single fault in power system is reverberated in the complete synchronized grid. Due to which analyzing the fault at any location is possible by analyzing PMU data of any other substation which is connected in that same grid. By means of voltage and current 40 ms data plots, we can find out the type of fault, its duration, information about successful or un-successful autore- closure and operation or maloperation of protection system.
2.1. Single Phase Fault: CGPL-JETPUR Line 1
The PMU data obtained from POSOCO [11] is plotted in Figure 1 and analysis is done based on these graphs.
B-phase voltage suddenly drops at time instant 19:35:55:240 which is of magnitude 7.6 kV (phase value). This drop is comparatively higher. We can observe that the fault clearing time (19:35:55:240 to 19:35:55:440) is 200 msec. the oscillations in the system are damped out later on, thus, clearing the fault successfully. There is no autoreclosure action taken.
2.2. LL Fault on 400 kV Ranchi Sipat 1 Circuit
R-phase to Y-phase to phase fault had occurred on 400 kV Ranchi Sipat circuit 1.
![]()
Figure 1. Voltage profile of Dehgam PMU.
The voltage plot for phase and zero sequence, corresponding to the Bhadravati PMU is shown in Figure 2 and Figure 3 given below. Phase voltage plot from Bhadrawati PMU shows that dip is observed in R- and Y-phase indicating phase to phase fault while the zero sequence voltage shows marginal decrease indicating no ground was involved.
![]()
Figure 3. Bhadrawati zero sequence voltage.
2.3. Three-Phase Fault at 400 kV Dadri
This is rather easy to observe as the severe voltage dip can be observed in all the phases simultaneously. The Figure 4 shows the 400 kV Agra bus voltage profile (Phase to Earth) during the three phase fault at 400 kV Dadri at 17:37:920 hrs on 13th March 2012. The sharp dip in voltage in all the three phases is quite evident.
2.4. Loss of Generation at Rihand STPS
Loss of generation or load will cause considerable disturbance in the line. The incidence at 400 kV Vindhyachal bus during incident of generation loss at Rihand STPS in Northern region is shown at Figure 5. It can be seen that the fault got cleared in around 320 milli. seconds indicating the operation of Local breaker Backup protection.
![]()
Figure 4. Dip in all three-phase voltages at 400 kV line.
![]()
Figure 5. Voltage profile at Vindhyachal due to loss in generation.
3. Fault Analysis
3.1. Type of Fault
Event detection [12] logic might facilitate the operator for responding to only the contingent situations with utmost awareness instead of keeping a watch on all PMU data received. Event detection logic will be used to determine the data window within which event has occurred so that the identified data window can be used for other functionalities such as fault classification and fault location identification. This technique might help in increasing the overall preparedness levels of the system operators for the fault occurrence.
The detection method utilizes three ranges of equal duration, target detection, prior detection range (ending t2 seconds before the target range) and a posterior range (beginning t3 seconds after the target range). This range formation is shown in Figure 6, where, t1 is the target detection range, which includes the overlapping range for the sliding window and the detection range, and t2 and t3 are the interval times between the target and other detection ranges. The event detection rules and thresholds that can be used are as given. Here, a mean value is used for the detection of step change type events; variance value is used for the detection of oscillation type events between the different range data.
Transmission line faults consist of 85% - 87% of the total number of faults occurring in power system. These faults are classified as single line-to-ground faults, line-to-line faults, double line-to-ground faults and three- phase faults. Any fault occurring in the power system can be classified by using the sequence component data which is available from the PMU. Table 1 summarizes the occurrences of phase components during various faults.
During fault, the voltage in the faulted phase/phases will dip at a comparatively faster rate than the voltage dip in healthy phases. This information can be used to classify the faulted phases. A fault is said to have occurred in the phase when the ratio of dv/dt of that phase to the maximum dv/dt among all the three phases is greater than a given threshold.
P = max (phase A dv/dt, phase B dv/dt, phase C dv/dt)
Ratio A = (phase A dv/dt)/P
Ratio B = (phase B dv/dt)/P
Ratio C = (phase C dv/dt)/P
The above ratios are compared against the set threshold and if violations seem to occur, respective flag is set high. Fault is classified using Table 2.
![]()
Figure 6. Bina phase B sample: Type of fault detection.
![]()
Table 1. Presence of sequence components of various faults.
![]()
Table 2. Rate of change of voltage based fault detection.
Also dv/dt of zero sequence voltage is measured and if it exceeds the threshold, fault is said to have involved ground. Calculation of total fault clearing time.
The time interval between the fault inception and the fault clearance is known as fault clearance time. This time is the longest fault current interruption time of the associated circuit-breaker (s) for elimination of fault current on the faulty item of plant. The dv/dt is very high during fault occurrence and fault clearing and hence this phenomenon can be used to determine the time of occurrence of fault and the time of fault clearing by the tripping of breakers. An additional parameter called recovery from fault time can also be measured by calculating the time required for the voltage to recover more than the specified voltage threshold after clearing of fault. The previous sample from which high negative dv/dt is detected is considered as fault initialization time (ti). During the fault period due to action of automatic voltage regulator (AVR), a slight change in voltage can be observed and on clearance of fault, dv/dt is again high (more than threshold dv/dt Fault Recover) and hence the previous sample at which high positive dv/dt is detected is considered as fault removal time (tr). On the removal of fault, the voltage tries to recover and the instant at which the voltage is greater than or equal to “Fault Recover Voltage”, this instant is considered as recovery time (trv). The fault clearing time and recovery from fault time is calculated as given below, and shown in Figure 7;
t0―Fault clearing time = tr − ti
t1―Recovery from fault time = trv − tr
3.2. Autoreclosure Operation Verification
Whenever a fault occurs, we can tentatively assume that the fault is a transient fault. Because generally statistical evidence shows that about 80% of the faults are not permanent. A small time delay is given before closing the circuit breakers, which is known as reclosure. The reclosure can be manual or automatic. If the fault was indeed a transient one then the reclosure would be successful, or else the breakers will trip again implying that the fault is permanent.
The unsuccessful reclosure operation is depicted in Figure 8 and Figure 9. A second single line to ground (SLG) fault occurs on the same phase within auto reclose time from the occurrence of first SLG fault. Two faults are observed within a time gap of one second which shows that the line has a persistent fault.
For detection of successful and unsuccessful auto re-closure operation following an SLG fault, the same algorithm that is used for detection of fault can be used.
3.3. Loss of Load/Generation Detection
Significant df/dt is observed in system during disturbance like tripping of load/generation, fault and reclosure
![]()
Figure 7. Bina phase B sample: Calculation of total fault clearing time.
![]()
Figure 8. Voltages from the PMU for unsuccessful auto reclosure operation.
![]()
Figure 9. Zero sequence voltages for unsuccessful autoreclosure operation.
action. During disturbance like fault and subsequent reclosure action, the system frequency pre disturbance and post disturbance is more or less the same, but during an event involving loss of load or generation, the frequency post disturbance will increase or decrease until it reaches a new steady state value depending on load generation balance. This phenomenon can be used to detect and effectively distinguish between loss of generation/load and any other event. Algorithm will first compute the sample where the df/dt crosses set threshold and then calculates the average frequency in the pre disturbance period (fprior) and post disturbance frequency average (fpost).
Figure 10 and Figure 11 depict the corresponding frequency profiles. Loss of generation/load event is confirmed if “delta frequency” exceeds threshold.
Delta frequency (del fdet) = fpost − fprior
If delta frequency is positive then the event is loss of load else event is loss of generation. During the period when df/dt exceeded threshold, the corresponding df/dt value is used for computation of quantum of load/gene- ration loss.
3.4. Event Localization
During fault the decrease in voltage can be captured by PMU located at various locations [13] . Any fault occurring
![]()
Figure 11. Frequency profile during loss of load.
system will have its effect at all the locations but the effect will decrease as the distance of the location from actual fault increases. The fault voltage will decrease more at the location nearer to the fault location. Figure 12 shows this phenomenon, the voltages prior to fault are all synchronized and the system appears to be stable providing the constant average voltage of 244 volts. At the instant of fault, there is major voltage drop at PMU located at Bina. The maximum intensity of the fault is reflected at this location. We can conclude that fault must be nearer to Bina line.
4. MATLAB Simulation Results
Based on these case-studies and the results deduced above, the algorithms and the results were verified on the data for phase B to ground fault which occurred on Bina Seoni line on 07-01-2014. Due to these algorithms, the event occurrence can trigger an alarm and make the operators aware about the system tending towards instability. The thresholds can be pre-decided to increase the margin of time available to bring the system back to stability by taking appropriate restorative action. This overall helps in enhancing the grid visualization and increases the awareness which directly helps in saving the system to become instable.
The results of the MATLAB codes are as given below. The event window was detected correctly. Event occurred between the time instants t = 4.8 and t = 6.36. This gave us the data window (of 40 samples) within which the event had occurred.
In Figure 13, the detected target window is shown in red color. Thus, we detected the occurrence of an event if the rate of change of frequency (df/dt) exceeded a particular threshold value.
The fault was accurately classified as phase B to ground fault at time instant 5.12 sec. Thus, we obtained a particular time instant at which event had occurred and the type of fault was also correctly displayed.
The time intervals were accurately calculated by the program. The results obtained were as follows: Fault initiation time = 5.12 msec; Fault clearing time = 5.2 msec; Breaker opening time = 80 msec.
Successful autoreclosure operation was displayed. Thus, auto reclosure operation is verified. If voltage
![]()
Figure 12. Voltage profile of phase B at different substations.
recovers within auto reclose time, we can consider that the auto reclosure operation was successful else not. Subsequently, the results obtained gave the instant at which df/dt crosses the threshold point, and it displayed the time instant 14.24 sec. fprior, fprior and fdet were also shown and finally it was detected that loss of generation had occurred. Event was located on B-phase nearer to Bina Seoni PMU. Thus, by considering the maximum drop in voltage event was localized.
The results that were obtained from the generalized program which detected the occurrence of the event, classified it, displayed whether auto reclosure operation was successful or not, detected the fault initialization time, fault clearing time and breaker opening time. Moreover it also indicated the location where the event had occurred. Such an algorithm if implemented in the data centers where the PMU data is received, might help to segregate the data containing detailed information about the event occurrence in much faster way which not only will be helpful to the grid operators but will save the grid from concurrent cascading fault occurrences.
5. Conclusion
Synchrophasor data can be used for a variety of applications, as presented in this paper, which eventually make the power system more reliable. Moreover, it also increases situational awareness of the grid operators. The main objective of this paper is quick post-fault analysis of abnormal conditions. Documenting such events for further analysis and detailed post mortem of the event is necessary for gaining insight in the power systems behavior. This paper presents the logical approach to characterizing the faults and acquiring maximum information through the PMU data that has been obtained. The PMU will be used for real time applications in near future. The threshold values used in the programs if decided more appropriately will give more accurate results. It is seen that most of the blackouts occur due to lack of situational awareness. Easily understandable visualizations can be developed for the operators. The PMUs will be able to create situational awareness of the system in real time in order to make the operator to take preventive actions.
Acknowledgements
We sincerely acknowledge and thank for the treasure of data received from the Power System Operation Corporation Limited (POSOCO). The nation will always be indebted to the efforts of people from POSOCO.