Presentation 19.

Testing of the Saturation Detector Algorithm in FRTU

Teratam Bunyagul, King Mongkuts University of Technology North Bangkok Noppadol Charbkaew, King Mongkuts University of Technology North Bangkok Peter Crossley, The University of Manchester Thanachat Thanasettagone, Unipower Engineering Co.,Ltd. In the medium voltage network, metering CTs are sometimes used to detect fault currents. This is to realize the aim of automatically operated distribution feeders. Metering CTs can be saturated during a heavy fault current. To solve the saturation problem, saturation detection and compensation algorithm are designed and evaluated. The article describes the design of the CT saturation detection algorithm. This algorithm does not require CT characteristic from the manufacturer. The algorithm was evaluated using the signal from the simulation model. The algorithm was then programmed into a digital signal controller and tested with the signal generator to confirm the results. Key Word: Saturation Algorithm, Metering CT, Measurement CT, Distribution Feeder, RTU, DAS

A C

F D

Fig. 1 Typical MV network with automation system in Bangkok

1. Introduction
The two most common arrangements of the distribution feeders in Thailand are radial and loop. The arrangements depend on the load density and the system voltage levels. In the rural areas, the majority of the feeders are 22kV overhead insulated lines connected radially. In the capital, the dominant distribution voltage level is 24kV. Most of the feeders are overhead insulated lines with small amount of underground cables in financial districts. These overhead lines are normally operated in loop arrangement with sectionaliser opened and this is referred to as an open-loop or open-ring. A typical feeder has 40-60 distribution transformers. The distribution transformers step down 24kV to 380V to feed the loads. The focus of this paper is on the open-loop distribution feeders in the capital Bangkok as shown in Figure 1. This is because the feeds are typically equipped with pole-mounted Load Break Switches (LBS) remotely controlled by the Distribution Automation System (DAS) in a control room. An operator uses DAS to send a signal to Feeder Remote Terminal Unit (FRTU) via fiber optic cables to control a load-break switch.

In Figure 1, LBS can be automatically controlled by various signaling arrangements that incorporate FRTU and CT, PT from LBS. During a fault at F, the current sensor C detects the upstream side of the fault current and the direction to the fault. The sensor C communicates this information to DAS at a control centre. Simultaneously, and probably before the information from C is received at the control centre, the overcurrent protection at A trips breaker A. After A has been opened, C and D open to isolate the faulty section. Breaker A can then be closed to restore supplies to the loads connected between A and C. With no voltage being detected on the section between the isolated faulted section and the open point E and after acknowledging that the faulty section has been isolated, the open point will be closed to restore supplies to the loads connected between point D and E. This arrangement restores supplies to the non-faulted section from A to C and D and E significantly faster than is achieved on a conventional open-loop network. The outage time of the non-faulted section can be reduced from several hours to seconds or minutes.

OMICRON electronics Asia Limited 2012 Asia-Pacific Protection & Testing Conference

Presentation 19.2

2. Feeder Remote Terminal Unit

The use of tele-control systems not only enables real-time information to be obtained from the supply system, it also permits remote-control operation of switching equipment. A pole-mounted LBS as shown in Figure 2 equipped with sensors, control mechanism for opening/closing the LBS and a communication system can be used to isolate a de-energised feeder section with a permanent fault. This is particularly beneficial on overhead distribution system and will minimise the duration of outages. The inclusion of an FRTU linked to DAS obviously leads to improvement of network security.

The idea of using the data from the saturated CT was originated by the desire to reduce supply interruption of a distribution system. The algorithm utilises the signals derived from measurement or instrument type current sensors (measurement CTs) in LBS. The main problem resulting from the use of measurement CTs (iron-cored CTs) in LBS is that they saturate during a fault. As a result, the secondary current signal is unavoidably distorted. To extract the primary current data from the distorted current signal, the algorithm is designed to overcome this problem. Consequently the algorithm can be used with existing FRTU. It is ideal for retrofitting FRTU with the hardware (dashed box in Figure 2) to increase the accuracy of fault current evaluations.

24kV

LBS

CT PT

PT CT LBS

FRTU
Fiber Optic

FRTU

Fiber Optic

Communication Module

Circuit Breaker Controller

Digital I/O

Hardware with Algorithm

Fig. 2 The installation layout of Load Break Switch (LBS) and Feeder Remote Terminal Unit (FRTU)

The idea of using the data from the saturated CT was originated by the desire to reduce supply interruption of a distribution system. The algorithm utilises the signals derived from measurement or instrument type current sensors (measurement CTs) in LBS. The main problem resulting from the use of measurement CTs (iron-cored CTs) in LBS is that they saturate during a fault. As a result, the secondary current signal is unavoidably distorted. To extract the primary current data from the distorted current signal, the algorithm is designed to overcome this problem. Consequently the algorithm can be used with existing FRTU. It is ideal for retrofitting FRTU with the hardware (dashed box in Figure 2) to increase the accuracy of fault current evaluations.

OMICRON electronics Asia Limited 2012 Asia-Pacific Protection & Testing Conference

Presentation 19.3

3. Saturation Detector
Extracting the fundamental magnitude of the primary fault current is crucial. The fault current seen at the FRTU suffers from the distortion caused by the saturation of magnetic iron cores. This makes it a difficult task to derive the fundamental magnitude from the severely distorted signal. The short data windows 1 shown in Figure 3 extracts the fundamental magnitude from the healthy parts of the distorted current signal [1]. The discrete conversion make it possible to implement the equations into the digital algorithm, i.e. :-

fundamental magnitude is the result of the correlation over the selected data window. In the digital relay, new input values are repeatedly sampled and fed into the relay. Effectively the sampled input signal is shifted by a short period of time and the correlation process repeated using the data over the new window. The period of time depends on the sampling rate of the relay. The relay algorithm takes a newly sampled current signal into the data window and removes the first sample point from the data window. For example, if the data window contains the current signal ranging from t1 to tm, when the new value of the current signal comes in, the data window will contain the current signal ranging from t2 to tm+1. This way the data window slides point by point along the secondary current signal. Every new point fed into the algorithm provides the new magnitude value of the fundamental frequency component. The output of the sliding data window is the magnitude of the fundamental frequency component. In addition, filtering based on the evaluation of the mean current magnitude over the chosen data window alleviates the effects of noise and harmonics on the calculation of the current magnitude [2]. Therefore we use the mean value of the current magnitude over a time period to represent the magnitude replica of the primary current signal at the fundamental frequency. The mean value of the current magnitude or Imean can be evaluated from

Where I1 =

the

fundamental

magnitude

is

and m is the number of sampling

points in the data window.

where m is the number of samples in the data window.

Fig. 3 Sliding data window in discrete form

Figure 3 illustrates the evaluation of the fundamental magnitude over a short data window of interval 1. The discrete values i[1] to i[m] correspond to the instantaneous samples at times t1 to tm respectively. The discrete values of a and b are then correlated with i[n] in accordance with (8) and (9). A single discrete value of the peak

Fig. 4 The result of the algorithm with X/R=20

OMICRON electronics Asia Limited 2012 Asia-Pacific Protection & Testing Conference

Presentation 19.4

The result from the analysis technique tested with data from EMTDC software with the fundamental frequency of 60Hz. The mean current magnitude (Imean) as shown with the thick solid line is almost constant within the healthy parts, i.e. it can be seen as a series of straight horizontal lines. The result from the analysis technique tested with data from EMTDC software with the fundamental frequency of 60Hz, X/R=20 is shown in Figure 4. The mean current magnitude (Imean) as shown with the thick solid line is almost constant within the healthy parts, i.e. it can be seen as a series of straight horizontal lines. Detecting the series of straight horizontal lines from the current magnitude is an aim for the saturation detector. The test can be achieved by a simple statistic formula for determining variance. The variance provides the difference between each point in the circle and its mean value:

When the current transformer is in deep saturation, the secondary current signal deviates from a sinusoid. This behaviour provides a vital clue to indicate the saturated periods. By investigating how much the current signal deviates from sinusoid, this information can be used in conjunction with Imean and Ivariance to set the more accurate condition of non-saturation. We define meanError to indicate the deviation of the secondary current signal from sinusoid. Having taken the derivative of the sine function twice, we have the negative sine function divided by 2. Exploiting this sinusoidal property [3], meanError is defined as the average of the error caused by the second derivative of the current signal deviates from a sinusoid. The equation to calculate meanError will be

Where i[n] is sampled secondary current, is fundamental angular frequency. where I1[n] is the current magnitude of the fundamental frequency component of the secondary current signal Imean is the mean of current magnitude in the data window and m is the number of samples in the data window. The results of Imean and Ivariance are illustrated in Figure 5. During the non-saturated periods, when Imean is a straight horizontal linethe value of Imean is greater than Ivariance. This can be seen as the grey areas indicated by numbers (1), (2), (3), (4) and (5). On the other hand, during the saturated periodswhen Imean is not a straight horizontal linethe value of Ivariance is greater than Imean. Figure 6 illustrates that the meanError obtained from (12) increases when the current signal severely deviates from a sinusoid (saturated periods) as shown by the grey areas (1) and (2), and decreases when the current signal is a sinusoid (non-saturated periods). Furthermore, meanError is greater than Imean during saturated periods and less than Imean during non-saturated periods. This leads to a complete condition to distinguish the saturation stage of current signal. The flag can be set when

is true [4]. The employment of (13) to set the flag gives an acceptable accuracy to detect saturation. The saturation detector is now able to distinguish all possible behaviour of the saturated current signal and correctly exclude the non-saturated periods out of the distorted current signal. Figure 6 shows how the saturation detector uses the complete condition (13) to set the flag. Only the regions that the flag is set provides useful information about the current signal. This can be seen as the grey areas indicated by numbers (1), (3) and (5).
Fig. 5 The comparison of Imean and Ivariance in a severe case

OMICRON electronics Asia Limited 2012 Asia-Pacific Protection & Testing Conference

Presentation 19.5

Fig. 6 The comparison of Imean, Ivariance and meanError

4. Magnitude Evaluator
When the non-saturation flag is set the current magnitude correctly represents the magnitude of the primary current signal. When reset the current magnitude is erroneous. The concept of the magnitude evaluator is to utilise the current magnitude only when the flag is set. Then the most recent magnitude value determined when the flag is set is held when the flag is reset. When the flag is set again, a new magnitude value is calculated and updated every sample. Therefore, the magnitude evaluator continuously produces the discrete estimation of the primary current magnitude in real time every new sample.

line shows the calculated current magnitude of the fundamental component of the primary current, i.e. the magnitude of the signal if the CT had been an ideal (non-saturated) 1:1 transformer and after the removal of the exponential offset. The calculated magnitude fluctuates significantly during the saturated periods but remains reasonably constant during the non-saturated periods. As the magnitude evaluator works, the current magnitude calculated during the non-saturated and saturated periods is converted into a magnitude result that is only updated within the non-saturated periods. During each saturated period the magnitude evaluator holds the last magnitude value calculated during the previous non-saturated period. Hence the current magnitude result (shown with the thick solid line) shows the peak magnitude of the fundamental component of the primary fault current.

5. Test results
The saturated currents were obtained from an analogue test bench using real CTs. The algorithm was tested using current signals captured by a fault recorder at the output of a saturated CT. The test bench is capable of supplying 200A into a short circuit fault for a period of a few second. User selectable resistors and inductors determine the actual fault currents and the X/R ratios. The fault recorder is wired in series with a 1 Ohm resistor and the secondary output of the CT. A fault was applied by closing a fault switch on the test bench at a random point on wave. The captured signals as COMTRADE files were transferred to a personal computer (PC). The PC was then played back the COMTRADE files using TransPlay software that connected to OMICRON CMC256+. Then the CMC256+ supplied the saturated current signals to the hardware with the algorithm. The diagram of the test procedures are illustrated in Figure 8.
Analog Test Bench
Saturated Current COMTRADE Files Saturated Current

Fault Recorder

PC

OMICRON CMC 256+

Hardware with Algorithm

Fig. 8 Test Setup

Fig. 7 The result from the magnitude evaluator

Figure 7 shows the flag used to detect when the signal is not saturated (non-saturation flag) and the calculated peak primary current magnitude. Although the current is severely distorted by the effect of CT saturation the saturation detector is able to detect non-saturated periods. These periods occur when the non-saturation flag (shown with the thin solid line) is set to high. The dotted

A total of 2,100 secondary current signals were created by closing a fault switch on the test bench at a random point on wave. The results of the test can be summerised in the Table 1.

OMICRON electronics Asia Limited 2012 Asia-Pacific Protection & Testing Conference

Presentation 19.6

Case 1 2 3 4 5 6 7

Current Magnitude (Apeak) 20 20 30 30 55 55 75

X/R 10 40 10 40 10 40 20

Number of Current signals 300 300 300 300 300 300 300

Average Error (%) 2.37 3.29 3.61 4.62 3.85 5.29 6.07

Confidential Interval (%) 95 95 95 95 95 95 95

[3]

Kasztenny B,, Rosolowski E., I.ukowicz M., Izykowski J., 1997, Current Related Relaying Algorithms Immune to Saturation of Current Transformers, Proc, IEE International Conference on Developments in Power System Protection, Vol. 434, 365-368. Bunyagul T, Crossley PA, Gale P Design and evaluation of an overcurrent relay suitable for operation with measurement current transformers. In: Developments in Power System Protection, 2001, Seventh International Conference on (IEE), 2001. pp 201-204

[4]

Table 1: The result from the hardware testing

6. Conclusion
The paper proposes a specifically designed algorithm suitable for protecting a distribution feeder that utilizes the current signals derived from iron-core measurement type CTs. The algorithm successfully solves this problem by selecting and processing only the healthy parts of the saturated current signals. As a result, the magnitude of the primary currents can be predicted. The design of the algorithm is aimed to be simple, fast and reliable. It is also provides major advantages over conventional CT compensation techniques. Firstly, the algorithm does not require a detailed model of CT, i.e. it is not necessary to know the magnetisation and hysteresis curves, and the CT dimensions and properties. Secondly, the saturation detector uses relative values based on secondary current. By comparing relative values, the saturation detector does not need to set thresholds and it is not necessary to design the system to operate with particular CTs. This is important because each CT has different physical properties and the properties may change with time due to remanence. By not requiring knowledge of the physical properties of the CTs, the saturation detector is significantly more practical.

8. About the Authors

Teratam Bunuyagul received B.Eng degree in electrical engineering from King Mongkut's University of Technology North Bangkok in 1995. He obtained his M.Sc and Ph.D degrees from University of Manchester Institute of Science and Technology (UMIST), UK in 1998 and 2003 respectively. His main activity during academic career has involved research into new techniques in power system protection and control, the design and development of new types of protection relays, phasor measurement unit and control systems. Noppadol Charbkaew graduated with B.Eng in electrical engineering from King Mongkuts University of Technology North Bangkok (KMUTNB), Thailand in 2001 and M.Eng in Elektro- and Informationtechnik from Rosenheim University of applied sciences, Germany in 2003. And he obtained his Ph.D. degree at KMUTNB in 2010. His research activities and interests are implementation of signal processing in power system protection, control and monitoring.

7. Literature
[1] Wiszniewski A. and Szafran J,, 1989, Distance Digital Algorithm Immune to Saturation of Current Transformers, Proc. IEE Fourth International Conference on Developments in Power System Protection, Vol. 302, pp 196-199. Bunyagul T, Crossley P, Gale P Overcurrent protection using signals derived from saturated measurement cts. In: Power Engineering Society Summer Meeting, 2001. IEEE, 2001, Vol.101, pp 103-108.

[2]

OMICRON electronics Asia Limited 2012 Asia-Pacific Protection & Testing Conference

Presentation 19.7

Peter Crossley joined UMIST, now The University of Manchester, in 1991 where he is currently a Professor specialising in Power System Protection and Control. His previous career included 13 years with ALSTOM T&D where he was involved in the design and application of protection relays. He graduated from UMIST in 1977 with a BSc degree and the University of Cambridge in 1983 with a PhD degree. He has authored/co-authored over 100 technical papers and is a member of various IET, IEEE and CIGRE committees on Protection. Thanachat Thanasettagone graduated as Bachelor degree of Electrical Engineering from Chulalongkorn University in 1992. He has been working as Sales and marketing Manager of Unipower Engineering Co.,Ltd in Thailand since 2000. The company specialises in test equipment for high voltage and protection systems.

OMICRON electronics Asia Limited 2012 Asia-Pacific Protection & Testing Conference