Você está na página 1de 7

Application of Differential Protection to Long Power Cables-

Cable modelling and charging current compensation methods


A Khanna
Senior Engineer ( Protection and Control ), ATKINS Plc, Leeds, United Kingdom,

Keywords: Differential Protection, Cable Models, Cable b. Cable construction is complex consisting of many layers
Protection, Charging Current Compensation, EMTP-RV of dielectrics and conductors.
c. Cables are generally laid in close proximity and this
Abstract makes interaction between the phases more complex.
Various models have been developed for transmission lines
This paper presents performance of two charging current and HV cables. Most commonly used models are Frequency
compensation methods used in commercially available Dependent (FD) model [1], the Constant Parameter (CP) [2]
differential relays. This paper appreciates importance of cable model, Wide Band (WB) [3] model and the Equivalent PI
models on protection and settings application studies and on model [4]. Each model has advantages and disadvantages. FD
estimation of performance of charging current compensation and WB models are computationally extensive but have a
methods. The problem is split in to a) Significance of Power higher degree of accuracy. Both models show good frequency
Cable Modelling and b) Protection performance evaluations. response ( as they consider frequency dependent parameters )
Models available in EMTP-RV are considered. Transient and are widely used for the transient studies involving cables.
performance of each of models is evaluated to determine Accuracy of the CP model and equivalent PI model is
which are most suitable for protection application studies. dependent on the shift from base frequency. CP and
Single core, 3 phase, 132 kV crossbonded cables of various equivalent PI models are relatively easy to build.
lengths are considered. However, the methodology can be
extended to any cable arrangement including pipe type/ 2 Need for Proper cable model - A Quantitative
Multicores cables. Single and three phase external faults are
considered. Such a study would be useful for long cables i.e.
Insight
in offshore transmission system and in densely populated Shunt capacitance associated with HV cable can create
urban areas where land costs restricts the construction of significant errors if it is not properly accounted. This is
overhead lines. demonstrated through a simple but relevant example. This
example considers a single phase cable. The single line
1 Introduction diagram and cable parameters are shown in the figure -1.

Differential protection is quite widely used in protection of Zs Cable


Power Cables. Cables are expensive piece of equipment and it F
is highly desirable that cable capabilities are fully utilised.
Conversely, the repair and replacement costs are relatively Zs= 0.78+ j x 15
high and the effort is time consuming. These two factors
warrant proper application of cable protection which is fast,
Cable Details
selective and discriminating. To meet these diverse and often Length = 100 KM
contradictory requirements imposed on protection devices, Series Impedance = 0.025+ j x 0.436
significance of accurate cable modelling cannot be ignored.
Cables are generally difficult to model in comparison to Shunt Admittance = 0.119e-3 Mho/ Km
transmission lines for the following reasons
Figure-1: SLD of the example system
a. Shunt capacitance ( admittance ) of the cables is many
orders higher than shunt capacitance of transmission
‘F’ represents the fault point. To quantify the impact, the
lines. The affect of shunt capacitance can be ignored in
shunt admittance is varied by 50% on either side of the base
determination of protection setting for transmission lines
value mentioned in figure-1. Fault resistance was varied in
but it cannot be ignored in determination of the
three steps from 0 to 20 ohms. Variation in sending end and
protection settings for HV cables.
receiving end current magnitudes is plotted in figure -2.
BASE VALUE

NEGATIVE ERROR

POSITIVE ERROR

BASE VALUE
NEGATIVE ERROR

POSITIVE ERROR

Figure -2: Variation of Sending and receiving current magnitudes with the shunt admittance.

As observed from figure -2, error in estimation of the shunt b. Shunt admittance directly proportional to area of cross
admittance can significantly influence sending and receiving section of the cable so a 2500 sq.mm cable was
end current magnitudes. It is also observed that the error has considered.
opposite impact on sending and receiving end currents. This The case study considers single core cables, but the results
means that differential error varies at much faster rate and can and methodology can be applied to Multicores cables. For
become significant consideration when differential protection such long distances cables are cross bonded and are either
is applied. This simple demonstrates the significance of the transposed or are laid in tight trefoil formation. In this paper it
proper cable modelling in determination of appropriate is assumed that cables are laid in trefoil formation as shown
settings. in figure -3

3 Cable Modelling and Simulation


1
It is observed from figure -2, error in estimation of charging
currents at ends of the cable is directly related to error in
estimation of shunt admittance (capacitance). The error can
be reduced by increasing accuracy of model used in the 0.1 Meters
calculations. This section demonstrates performance of the
models available in EMTP -RV which are Frequency
dependent (FD) model, Constant Parameter (CP) model,
Wide Band (WB) Model and the Equivalent PI model. The
accuracy of each the models is analysed. FD model has good
accuracy [5] and is taken as reference for evaluating accuracy 2 3
of other models.
To reduce the number of cases, cable voltage and size was
selected which would have highest shunt admittance. Some
of the factors influencing the shunt admittance are
a. Shunt admittance increases with reduction in insulation 0.12 Meters
thickness i.e. lower voltage cables would have higher Cable numbers as marked inside
shunt admittance. 132 kV is the lowest voltage used for
transmission of Power over moderate distances and is
Cross Bonded; 1 Core/ Phase.
also used for connection to offshore windfarms and so is Ground resistivity 100 Ohms-Meters
selected for this case [5].
Figure-3 Cable Arrangement in trefoil formation-
distances are rounded off to nearest mm
3.1. Approximating the cable construction. EMTP-RV. Transient performance each of the models was
EMTP-RV has a “cable data” utility which is used to generate evaluated for an external fault. Location of the fault is marked
a cable model for simulation. The utility takes as input type of in the figure -1. Fault resistance was varied from 100 to 0
cable i.e. ( single core or pipe type) , number of conductors in ohms in steps of 10 ohms. Relay response was simulated by
each core, type of bonding ,cable arrangements and individual calculating the Fourier Components of the currents, at 50 Hz
conductor details. However this utility does not fully support at two ends of cables. Figure-5 shows results for 150 km long
complex cable construction generally encountered in reality. cable for a 3 phase to earth fault. Format of the legends is
Some degree of simplification/ pre calculation is required so explained below.
that the data can be presented to the utility in required format. I_XX_Y_ZZZ ANG represents angle and MAG
Cable construction parameters and associated assumptions are represents Magnitude of currents
provided in Appendix –A. A is sending end (source end) and B
represents receiving end.
Type of Model; CP Constant
parameter; FD frequency dependent;
WB wide band; PI exact PI model.
I represents current.
Simulations were additionally performed for cables with
lengths of 50 km, 75 km, 100 km and 125 km. In all the
simulations, the results were in close agreement for FD
model, WB model and Exact PI model. Errors were highest
in the CP model and increased with cable lengths.
Performance of the charging current compensation method
was evaluated using FD model.

Note: The cable 4 Evaluation of the Charging current


Actual cross section is compensation methods
not to the scale.
There are atleat two charging current compensation methods
used in the commercially available differential relays. One is
voltage independent and other requires voltage measurement
at cable ends. Dynamic performance of both the methods is
evaluated for single phase and 3 phase faults for cables of
length 50 km, 100 km and 150 km. This variation would
capture the impact of cable lengths.

4.1. Non Voltage Based charging current compensation


method- Algorithm -1
This method is used in one of the commercial products [6].
This algorithm measures the fundamental frequency
differential current under steady state conditions. The prefault
Simplified differential current is continuously subtracted from the
SHEATH subsequent measurements. A condition persisting for more
RADIAL WATER BARRIER than 50 milliseconds gives a new value of compensation
LONGITUDNAL WATER BARRIER current. This compensation is done separately for each phase.
METALLIC SCREEN Main advantages of this algorithm are; it does not need
WATER BARRIER
voltage inputs and does not require determination of cable
INSULATION SCREEN
INSULATION
shunt admittance.
CONDUCTOR SCREEN
CONDUCTOR 4.2. Voltage Based charging current compensation
Figure -4: Cable Construction used in the Model methods - Algorithm -2

Approximation of cable construction requires some correction This method is used in atleast three of the commercial relays
factors to be applied which are [5]. [7],[8],[9]. Charging current drawn by the cable is related to
voltage at cable ends and the shunt admittance of the cable.
a. Correction for Resistivity. The simplified mathematical model is explained
b. Correction for Permittivity. Let VA be voltage at the end ‘A’ ; VB be voltage at the end ‘B
These construction details were passed on to EMTP-RV ; yC be shunt admittance per unit length; L be the length of the
utility to generate models. Same input details were used to cable; the charging current flowing into cable at each end is
generate the four different types of models allowed by calculated by following relations
Figure -5: Transient Response of the Models

resistance varied from 100-10 ohm in steps of 10 ohms and


‫ܮ‬ then from 10 ohms to 8 ohms in 2 ohms step. Internal fault
‫ܫ‬஼஺ ൌ  ܸ஺ ൈ ‫ݕ‬஼ ൈ ሺͳሻ was simulated at centre of cable with charging current
ʹ
compensation active. Parameter used in evaluation of
‫ܮ‬ charging current compensation methods is Ratio of
‫ܫ‬஼஻ ൌ  ܸ஻ ൈ ‫ݕ‬஼ ൈ ሺʹሻ
ʹ differential current- with charging current compensation; to
restraint current ( restraint current is average of the currents at
ICA1 and ICB are the charging currents drawn by ends A and B ends) for an external and an internal fault. Ratio of
respectively. Let IA and IB the total current flowing into cable differential current- without charging current compensation;
at ends A and B respectively. The actual differential current, to the restraint current was also plotted in the same figure for
net of charging currents can be determined as comparison.
The above ratios should be low for an external fault ( i.e.
‫ܫ‬஽ூிிି஼ ൌ  ‫ܫ‬஺ െ ‫ܫ‬஼஺ ൅  ‫ܫ‬஻ െ ‫ܫ‬஼஻ ሺ͵ሻ algorithms with low ratio are better for an external fault ) and
high for internal fault ( i.e. algorithms with high ratio are
In this paper a simplified version of the algorithms is used. better for an external fault ). Figure 6 and 7 show results for
Mathematical treatment adopted in this paper is expected to an external fault for algorithm-1 and 2 respectively. Figure -7
be pessimistic and actual performance of relays is expected to demonstrates impact of algorithm-1 on sensitivity for an
be better. Cable data sheet indicates shunt admittance of internal fault. COMP and UN_COMP imply with and without
119.38 μMHO/KM ( shunt capacitance at 0.38 μF/kM). charging current compensation; 1_PH and 3_PH represent
Admittance value used in evaluating the performance was type of fault ( 1phase or 3 phase ); “xxx km” is length of the
determined by running simulation on the FD model with cable.
shunt compensation in service.
5. Results
4.3. Simulation set up and performance evaluation. Algorithm -1
Algorithms were evaluated for an external fault and for a. Algorithm is simple and adaptive as it adjusts according to
internal fault. Shunt compensation at each end of cable was cable parameters. No additional settings are required.
kept at value to keep terminal voltages within limits. Fault b. Its performance is good when ratio of charging current to
short circuit current for an external fault is low.
1
Current flowing into the cable is treated as Positive and all Effectiveness of this algorithm falls with increase in
the currents are phasors. source impedance and cable impedance.
COMP : Implies With Charging Current Compensation

REDUCING FAULT RESISTANCE

UN_COMP : Implies Without Charging Current Compensation

Figure -6: Ratio of Differential Currents to Restraint currents for an external fault for algorithm-1

1_PH and 3_PH: Imply one phase to earth and 3 phase earth fault
ti l

Figure -7: Ratio of Differential Currents to Restraint currents for an external fault for algorithm-2
c. Efficacy of this algorithm is also dependent on the a. This algorithm is very effective in compensating cable
system conditions at which charging current was charging current.
determined. Error increases if pre fault system b. This algorithm needs VT inputs in addition to current
conditions are different from conditions at which inputs.
charging current is determined. c. Effectiveness/ accuracy of this algorithm is
d. Sensitivity of the alogorithm-1 is much better then independent of the length of cable.
uncompensated case but lower than algorithm -2. d. Accuracy of the algorithm is directly related to the
Algorithm -2 accuracy with which shunt admittance is calculated.
Figure -8: Ratio of Differential Currents to Restraint currents for an internal fault for algorithm-1

6. Conclusions Disclaimer
Cable shunt capacitance can significantly influence The views presented in this paper are that of author and
through fault current. Protection application and setting not of organisation for which the author is working.
studies should consider shunt admittance of the cable. For
long cables, total shunt admittance may not be simple References
product of length of the cable and shunt admittance per
unit length. Cable shunt admittance should be determined [1]. L. Marti.: “ Simulations of Transients in underground
either by running measurements during the first cables with frequency dependent modal
energisation or by using suitable modelling techniques. transformation matrices”. IEEE transactions on
Transient performance of four different cable models was Power Delivery, Volume 3, issue 3 July 1988, pp
considered. Results from PI, FD and WB model were in 1099-1110
closed agreement. Constant Parameter model appears to be [2]. H.W.Dommel. : “Digital Computer Solution of
least accurate for single core cables in trefoil formation Electromagnetic Transients in singe and multiphase
and flat formation. networks”. IEEE Transactions, Vol PAS-88, Pages
Two different algorithms for cable charging current 388-399, April 1969.
compensation were considered in this paper. Both the [3]. Morched A, Gsutaven B. And Tartibi M.: “An
algorithms improve performance of Differential relays. universal model for accurate calculation of
Algorithm -2 which requires cable end voltages has much Electromagnetic transients on overhead lines and
better accuracy for cable lengths of 50 km or more. It can underground”. IEEE transactions on Power Delivery,
neutralise the charging current to practically zero. It does Volume 14, issue 3 July 1999, pp 1032-1032
exhibit some errors during transients. When algorithm -1 is [4]. P.Kundur.: “ Power System Stability and control”.
used on cables with lengths of 50 km or more, slope of EPRI, Power System Series 1994, pages 206-209.
differential characteristic should be increased to [5]. C.F.Jensen et al.: “Switching studies for the Horns
compensate for error. Rev 2 windfarm main cable”. IPST-11, paper no.
For cable lengths of 50 km or less, both the algorithms 182.
perform as expected. For cable lengths of 50 km or less [6]. ABB, “Application Manual RED 670”, 2012, pages
algorithm –1 has advantage as it does not require voltage 109-110.
inputs. This may be significant advantage in cases where [7]. Alstom.: “ Technical Reference Manual MiCOM
VT inputs are not available. P545”, 2011, Section 6.2.2.
Settings should be worked out using relay, cable and [8]. GE.: “L90 Line Differential Relay UR Series
detailed system models including the source impedances. Instruction Manual”. 2006, Section 8.1.18.
Some feasible system depletion scenarios should be [9]. Siemens.: “ Application Manual for 7SD 522/523”.
considered for determination of appropriate settings. 2009, Page 94
Appendix – A 132 kV cable parameters1

Parameter Range1 Value Units


Cross-section of the Cable 2500 sq.mm
Diameter of the Conductor 62 mm
Correction Factor for Resistivity 0.8281
Resistivity 1.68E-08 Ω.meter
Corrected Resistivity 2.03E-08 Ω.meter
Radius of the Conductor 31.00 mm
Conductor Screen thickness 1.5-2.5 2.00$ mm
Total Radius Over Screen 33.00
Insulation Thickness 15.00 mm
Total Radius Over Insulation 48.00 mm
Insulation screen Thickness 1-1.5 1.252 mm
Total Radius Over Insulation Screen 49.25 mm
Longitudinal Water Barrier 0.6-1.2 1.002 mm
Total radius Over the Water Barrier 50.25 mm
Inner Radius of the Screen 50.25 mm
Area of the Cooper Screen 95 mm
Outer Radius of the Screen 50.55 mm
Longitudinal Water barrier 0.6-1.2 12 mm
Radial Water Barrier 0.2 mm
Radius Over the Water Barriers 51.75 mm
Total radius 57.15 mm
Thickness of the Outer Sheath 5.40 mm
Notes
1. Source - ABB XLPE land cable systems – User Guide.
2. Range is applicable for cases where exact data was not available.
3. Probable value.

Você também pode gostar