Вы находитесь на странице: 1из 6

Assessment of Analysis Techniques used in determining

Grounding System Potential Rise from the Fall of Potential Method


D.J.Woodhouse, Member, IEEE Prof.R.H.Middleton, Fellow, IEEE.
Energy Australia University of Newcastle
Wallsend NSW 2287 Australia Callaghan NSW 2308 Australia

Abstract: The prevalent method used to determine the Ground A. Linear Model
Potential Rise (GPR) of an earthing or grounding system is to
perform a Fall of Potential (FOP) test undertaken during an As the measurement distance increases, the behaviour of
grounding system current injection test. Determining the test GPR the measured voltage can be described by modelling the
requires analysis of FOP test data due to the FOP responses non-
conservative asymptotic behaviour. This paper assesses achievable grounding system as a point source in a resistive media [7], as
tolerance by FOP test analysis including recognised and altemative described in (1). The model derivation assumes the soil
methods. Susceptibility to voltage and distance measurement noise is resistivity is homogeneous, the size of the grounding system
examined by application to an grounding system model to establish is negligible in comparison to the spacings used, and the
minimum error bounds and method estimation variation on GPR measurement reference is the grounding system potential
estimates. The applicability of the FOP test rule of thumb (GPR).
termination condition, to take three to four readings beyond the
knee of the FOP response, is also examined. In particular, the
-
V = GPR ~ 1 4 2 ~ ~ ) (1)
question of whether appropriate information is gathered when this Where V = Voltage between GPR & soil x metres from grid [VI.
condition is met is discussed. p = Soil resistivity [Q.m].
I = Fault current [A].
Keywords: grounding, measurements, impedance, testing, current x = Separation between grounding system & point on soil [m].
injection testing, earthing. r = Equivalent radius of the grounding system [m].

I. INTRODUCTION B. Offset Model

The linear model, as described by (1). is based on a point


The FOP test is used to determine the GPR of a grounding current source and uniform soil resistivity. To improve upon
system during a grounding system current injection test. the estimates of the method a model which better describes
Subsequently the GPR of a grounding system resulting from a the system is required [$I. The offset model, which removes
power system ground fault may be calculated. The GPR of an the negligible grounding system size assumption, was used.
installation is critical in determining accurately the Based on empirical equations used for Mutual Earth
performance of the installations grounding system. This Resistance (MER) and conductive coupling calculations [9].
includes soil potential contours around the installation, the the offset model takes the form described in (2).
impedance of the installation and its components, and the
interaction with a third party system [ 1][2]. Such interactions V = GPR - pI/(2xx+r)
are categorised, nominally in accordance with International (2)
[4] and National [5] standards, using the maximum grounding 111. MANUAL GPR ESTIMATION
system GPR and corresponding fault clearing time.
The usual method for determining GPR from a FOP test is
The FOP method is the most reliable and accurate method to apply linear regression to the FOP response [lo]. A FOP
for determining the GPR of an grounding system. Some response-is typified by two distinct sections, separated by a
installations are located in environments unsuitable to knee, as shown in Fig 1. The response of the FOP test in
performing the FOP test, such as highly urbanised proximity to the grounding system is steep due to its
environments [3]. In these cases the FOP method test dependence on grounding system geometry. requiring
spacings are restricted and altemative methods may be complex analysis to accurately calculate. In contrast, the FOP
required to determine the GPR, such as discussed in [6]. response beyond the knee approaches an inverse (or
hyperbolic) distance relationship as outlined in Section 2. At
11. ANALYSIS MODELS these distances the influence of the grounding system
approaches that of a distant point current source [7].
The FOP test method is based on measurements of soil
potential with respect to GPR at increasing distances from the The linear regression a plot of the FOP
grounding system of the installation under investigation. The response against the inverse of the distance. The points
behaviour of the resulting Voltage Profile Can be described beyond the knee which should appear to be linear, are used to
using either the linear or offset models, as described in the
extrapolate to 0 (or 1/00) in the Ilx plane, being equivalent to
following sections.

(c) 2000 IEEE


0-7803-6420-1/00/$10.00 1153
an extrapolation to remote ground(or 00 m), as illustrated in The following three cases examine the accuracy of the
Fig 2. linear regression analysis method under different soil
structure conditions with a simple groundmat geometry, as
"I CmtRIl
shown in Fig 3, buried at a depth of 05m The three soil
structures examined were:

Case (A) Homogeneous soil, p = 100 a.m.

Case (B) Two layer soil model - high on low.


= 100 Q.m Upper soil layer resistivity, depth 2m.
p2 = 10 Q.m Lower soil layer resistivity.

Case (C) Two layer soil model - low on high.


pI = 10 Q.m Upper soil layer resistivity, depth 2m.

91 p2 = 47 Qm Lower soil layer resistivity.

3 : ,L 1 IS
, *.
Distance{om Earlhgdd[m]
tin

Fig. 1: Typical fall of potential response.

The linear regression method relies on the user selecting the


appropriate data points from either the linear or inverse
distance FOP responses. For FOP tests containing a

I
p ~ o ~ p o . l o , o . 5 - - - 0.l0,O.S
significant number of data points there is no quantitative
method for selecting the optimal 'inverse' subset of data as
there is no measure of true error in the estimate.

~ 0.0.5 lO,O,O.s 20,0,05

Fig 3: Grounding System used for Assessment

The GPR in these cases was calculated using a least squares


fit to the subset of points from the FOP response which
produced the greatest GPR estimate. The points chosen were
usually those points at the greatest spacings. The maximum
GPR was used as it was the result closest to the actual GPR
for each example. The analysis of an ideal FOP response in
the (l/x,V) plane using this manual method as illustrated in
Fig 2, produces GPR estimates approaching the actual GPR as
underestimates. This behaviour is due to the inclusion of non-
OlOS Rzoa
Inverse Btance to Earth Grid [llm] ideal data within the FOP response to the response expected
by an ideal point model, with data points at smaller spacings
Fig 2: Determining GPR using linear regression. having the most significant error. As these points are removed
from the subset used to calculate an GPR estimate, the
IV. MANUAL GPR ESTIMATION ASSESSMENT estimate approaches the actual GPR. This method is only
suitable for use on near 'ideal' data. The results of this
Assessment of manual estimation methods, which are based method are summarised in Table 1.
on manual point selection, is difficult due to inherent bias
incurred by the analyst's experience and result expectation. A. Linear Model
This investigation restricted the scope of the manual
estimation method assessment to near ideal data and the 'best' To achieve an estimate with accuracy better than 1% using
result ie the nearest result to actual GPR. Further development the linear model, without detailed knowledge of the soil
is being undertaken to extend the methods described here to model, the FOP test would have had to have been performed
field data. The assessment considered the dependency of the to a distance of 6 times Dsub'to provide sufficient data. To
method on soil resistivity and grounding system geometry.
For brevity and consistency between results, only results of I Dvlb- the largest dimension of the earthing system. In a
the soil resistivity investigation are outlined here. rectangular system it is approximately the length of the system
diagonal.

(c) 2000 IEEE


0-7803-6420-1/00/$10.00 1154
increase estimate accuracy measurements are required at effort required to achieve an equivalent result from the linear
greater spacings. If the soil model is known, the number of model is approximately 3 times that of the offset model. This
readings and effort required can be significantly reduced. represents a significant saving in time and effort, if the same
From Table 1 it is also apparent that selection of the accnracy is required or a significant increase in accuracy for
maximum spacing by choosing four points past the knee the same expenditure of effort.
point of the FOP response appears adequate for accurate
analysis. V. AUTOMATED GPR ESTIMATION
Table 1:
Linear Regression Response on FOP Test Analysis Every GPR estimate is dependent on the data set selected
Dependence on Soil Conditions
for parameter estimation. There are at least two aspects of
I I Soil Soil Soil I data set determination to be considered. The first is estimation
of the knee point, since we know from Section 3, inclusion of
data points before the knee frequently do not allow good
curve fitting, particularly for linear model based methods.
3 140 I Additionally, measurement errors, or local soil or ground
variations, may give outliers in the data which need to be
4 80 2.8 1.7 0.4 3.5 eliminated. At present, data set selection is performed
5 160 5.6 OS* 0. I I .2 manually. Clearly, to improve the objectivity of the tests, and
reduce the time taken to perform the analysis, automation of
this process is desirable. Two methods are proposed, which
use different methods to assess the estimate error, for
determining the appropriate data set:
300 10.6 0.2 0.3
a. First Order Least Squares Minimisation Method
9 400 14.1 0.1 0.3
b. Second Order Least Squares Minimisation Method
* - Estimate of nading when knee OCCUR.
As part of the automation process they calculate for all
B. Offset Model suitable data sets in the FOP response an GPR estimate and
use the set with the smallest error as the optimal GPR
The accuracy of GPR estimates derived using the manual
estimate. Both methods performance are compared in the
selection method based on the offset model is summarised in
following sections using both the linear and offset models of
Table 2. To achieve estimates with accuracy better than 1%.
the FOP response.
without any knowledge of the soil model, requires the FOP
test to be performed to a minimum distance of 2 times Dsub.
.A. First Order Least Squares Error Minimisation
This is a significantly smaller requirement, and hence better
result than for estimates based on the linear model.
This method assumes that the best GPR estimate to the
Table 2: FOP response has the least deviation from a linear model. It
Offset Model Linear Regression Response on FOP Test Analysis also describes a linear fit to the data when modelled in the
I r
Dependence on Soil Conditions
I I I k
(I/x,V) p l m , as descFibed in Section 3 and shown in Fig 2.
Case (A) Case (B) Case (C) The method involves determining the least squares best fit to
y,-5 Maximum
Spacing (x)
[ml %E 4 %E %E
a line for each set of points in the response, where the set
contains
basis of:
at least three points. The estimate is selected on the

i. The result with the smallest error estimate and


ii. The GPR estimate exceeds the largest voltage
measurement in the FOP profile.

B. Second Order Least Squares Error Minimisation

The premise of the second order (or parabolic) method is


that the best estimate is the set of data points which
10.6 0.0 0.0 0.0
minimise the second order content of the estimation in the
* ~ Eslunate of reading when knee. occurs. (l/x, V) plane. Firstly, the estimate is assumed to be of the
form
From Table 2 it is apparent that selection of the maximum V = ay2+ by f c (5)
spacing by choosing four points past the knee point of the
FOP response appears excessive to produce an GPR estimate where y = x-l
and a,b & c are constants.
of the required accuracy of 1% in this investigation. The

0-7803-6420-1/00/$10.00 ( c ) 2000 IEEE 1155


All possible combinations of points containing a minimum
of three points are used to generate a least squares tit using
the least squares form given by (6), as outlined in [ 1 t]:

e = (YTY)-'YTv 16)

where 8 = matrix of parameters a,b.c.


Y = matrix of y measurements. U
V = vector of v meawfements.
pr;
Each combination of points produces an GPR estimate and
the amount of curvature in the estimation. The 'best' result is
5 0.3 0.2 0.5 0.0
chosen on the basis of:
1. Curvature of the result being less than that of any 6 0.1 0.2 0.5 0.0
other result.
ii. The calculated GPR is greater than the maximum 7 0.0 0.2 0.0 0.0
measurement in the FOP response. I 8 I 0.0 I 0.2 I 0.0 I 0.0

C. Automation Methods I 9 I0.0 I 0.0 I 0.0 I 0.0 I


Table 5:
Combining the two automation methods with the two Estimates using Automated Calculators of FOP Test Analysis
system models results in four automated estimation methods.

0 First Order Linear Model (LR)


0 First Order Offset Model (OR)
0 Second Order Linear Model (PLM)
0 Second Order Offset Model (POM)

The four automated estimation methods were assessed for


accuracy for three cases of soil resistivity models. The results
of the tests are summarised in Tables 3.4 and 5.
Table 3:
Estimate.. using Automated Calculators of M P Test Analysis 8 0.2 0.0 0.2 . 0.0

9 0.1 0.0 0.1 0.0

The methods based on the linear model compare


unfavourably with the. corresponding methods based on the
offset model. This is explained by the better match the offset
model has to a typical FOP response. The responses from the
methods are dependant on the type of soil structure the
grounding system is situated in. Soil structures which extend
the zone of influence of the grounding system, such as a low
on high soil resistivity structure, require greater spacings to be
made than would a homogeneous or high on low soil
resistivity structure.

These results produce the following conclusions about the


VI. WHITE NOISE ERROR PERFORMANCE
analysis methods examined:
A number of simulations, using Monte Carlo Analysis,
The low on high soi1 resistivity case requires the most
were performed to determine the performance of the
effort to achieve accurate results, in agreement with the
automated methods when white noise is introduced into the
previous results.
FOP response. A set of near ideal data points (correct to three
0 The two estimation methods using the offset model are significant figures) for a FOP response curve were used as the
superior to the methods using the linear model for these basis of the analysis. Each point in the response was varied
ideal conditions. The offset model based methods randomly from the ideal data set within defined tolerance
achieve near perfect results with only 5 readings with the limits for each run of the Monte Carlo analysis. The variation
largest required reading at a distance less than 5 times for each point was independent of the variation made at any

0-'7803-6420-1/00/$10.00 (c) 2000 IEEE 1156


other point. An GPR estimate was calculated with each run are summarised in Table 7, which confirms that the effect of
and the process repeated. the distance error is linear. However, in comparison to the
effect of voltage measurement error is almost negligible. The
A. Variation in Voltage Tolerance ratio of the relative effects of distance and voltage
measurement errors is estimated at lo3.
The first set of runs examined the effect of varying the
tolerance in the voltage measurements, as summarised in Table 7:
Effect of distance measurementerror tolerance on
Table 6, using the Parabolic Minimisation Method based on GPR estimate using the Parabolic Minimisation Method
the linear model. Table 6 summarises the average GPR and (True GPR = 2.565V)
standard deviation [o]for the estimates made within the given
voltage tolerance range. These results indicate that the error in
the estimate is proportional to the voltage measurement error.
This is a significant result considering that the effect is not
modelled as a linear relationship but acts to minimise
distortion in the linearised form of a rational equation.

B. Variation in Distance Tolerance


Analyses where carried out to investigate the effect of error
in the distance measurement component of the FOP data
points, again using the Parabolic Minimisation method based
on the linear model. The results of the analyses are 5.0 2.563 9.6 2.541 7.3
summarised in Table 7.
10.0 2.563 9.9 2.541 10.0
Table 6:
Effect of voltage measurementerror tolerance on 20.0 2.562 10.7 2.554 21.5
GPR estimate using Parabolic Minimisation Method
(Ax = 0. Im, True GPR = 2.565V) C. Comparison of Minimisation Methods
I I I I

The effect of white noise in voltage and distance


measurements on GPR estimates for all four automated
methods was examined. The soil model used was Case C the
23 2.563 2.6 low on high soil resistivity structure from Section 3.
5.0 2.562 4.9
7.5 2.563 7.1
0.655
10.0 2.561 9.3
E
c
12.5 2.562 12.0 B
25.0 2.565 22.5
.g
w E.65 ,
e L..

250.0 2.754 151.6


U
a
Lu
., . .
--
+ Case1
0.645
0
d
-9- Case2 *
case3 \
d
- Actual EPR
0.64
LR OR PLM POM
Method
Fig 3 Average GPR estimate for minimisation methods

It was found that the First Order Offset Method gave the
best average estimate to the GPR, with h e Seoond Order
Offset Method having the smallest standard deviation. All
estimates are within 1.6% of the correct GPR,as shown in Fig
3. It is reasonable for either of the methods based on the
offset model be recommended as the most accurate method
when the effects of white noise are considered. Under the
various conditions examined no methods consistently
produced the smallest error. Consequently no method could

0-7803-6420-1/00/$10.00 (c) 2000 IEEE 1157


be considered outstanding and analysis should rely on a Directives concerning the protection of Telecommunications
combination of methods rather than a singular method. Lines against harmful Effects from Electricity Lines, CCITT
Geneva, 1%3.
Earth Potential Rise - Code for Practice for the Protection of
VIICONCLUSIONS Personnel and Equipment Against Earth Potential Rises Caused by
High Voltage Power System Faults. Telecom Australia -
ESAA( 1984).
Two models have been used to determine the GPR from the IEEE Std81.2-1991. IEEE Guide for Measurement of Impedance
results of a fall of potential test. Analysis of these models and Safety Characteristics of Large, Extended or Interconnected
using manual point selection of the data sets was examined Grounding Systems, June 25. 1992.
and found to be: E.D.Sunde, Earth Conduction Effects in Transmission Systems.
Dover, London,1968. Revision of publication by Bell Telephone
Laboratories 1949.
0 Lacking in consistency through dependence on user Ljung, L,System Identification - Theory for the user, Prentice
experience. Hall, 1987.
J.M.Nahman,Fximity effect? on the ground fault current
0 To achieve 1% accuracy the FOP test using the linear distribution within the earthing system formed by a substation and
model requires measurements to a distance of 6 times the associated transmission lines, IEE Proceedings, Vol. 135. Pt.
Dsubfor the test case examined. Using the offset model a C, No.6, November 1988, pp.497-502.
distance of 2 times Dsubwas sufficient. Jacques Fortin, Guide for Measuring Hydro-Quebecs Grounding
System Installations, Master of Applied Science Thesis,
Selection of the maximum spacing, by choosing four Electrical Engineering Department, Ecole Polytechnic. University
points past the knee point of the FOP response, appears of Montreal, CP 6079 Succ. A. Montreal, Canada H3C 3A7.
to be adequate for accurate analysis. December 1985. Report No. 85102.
R.H.Middleton, G.C.Goodwin, Digital Control and Estimation -
0 Due to the asymptotic nature of a fall of potential A Unified Approach, Prentice-Hall Intemational Inc.. 1990.
response, manual methods tend to result in non-
conservative GPR estimates. VI. BIOGRAPHIES

The four methods automating the GPR estimation process, Darren.J.Woodhouse (M2000) was bom in
were examined, and found to be: Maidand, NSW, Australia, on May 25 , 1969.
He received his B.E.(Elec.)(Hons I) (1993) and
BMaths (1994) from the University of
0 Two automated methods employing offset models were Newcastle, Australia. He joined Energy
superior to the linear model based methods. Near perfect Australia, then Shortland Electricity. in 1988 as
results were achieved with only 5 readings with the a cadet engineer. Since 1993 he has been
Development Engineer in Safearth Engineered
largest required reading distance being less than 5 times Solutions, a specialist engineering group
Dsub. focussed on earthing, including research and
The error in the GPR estimate is proportional to and development. He is currently working towards
dominated by the voltage measurement error. This his PhD in the area of power system earthing
system testing.
indicates that GPR analysis accuracy is limited by the test
methodology and instrumentation used to perform the
FOP test. Professor Richard H. Middleton was born on
10th December 1961 in Newcastle Australia. He
received his B.Sc. (1983), B.Eng. (Hons-
These methods have been found to be very accurate in 1)(1984) and Ph.D. (1986) from the University
simulated testing. When used with field data the analysis of Newcastle, Australia. He has had visiting
results are significantly better than the present methods, appointment3 at both the University of Illinois at
thereby reducing safety risk exposure due to incorrect GPR Urbana-Champaign and at the University of
Michigan. ln 1991 he was awarded the
determination. Australian Telecommunications and Electronics
Research Board Outstanding Young Investigator
award (national -annual award). In 1994 he was
Further research is being conducted into increasing GPR
analysis technique robustness, accuracy and error $\
$ 1[ awarded the Royal Society of New South Wales
Edgeworth-David Medal (annual award for an
identificationand correction for field data. outstanding young scientist). He has served as an associate editor of the E E E
Transactions on Automatic Control and as Head of Department of Electrical
VXEFERENCES and Computer Engineering at the University of Newcastle. He was elected to
the grade of Fellow of the IEEE starting 1999. He is currently an Associate
Editor for the IEEE Transactions on Control System Technology and also of
[I] ANSVIEEE Std81.1-1983, IEEE Guide for Measuring Earth Automatica, a panel member for the Australian Research Council, a Professor
Resistivity, Ground Impedance, and Earth Surface Potentials of a
in the Department of Electrical and Computer Engineering at the University
Ground System, March I I , 1983. of Newcastle, and is also associate director of the Centre for Integrated
[2] ESAA-EGI (97) ESAA, Substation Earthing Guide, Carman, Dynamics and Control (A Commonwealth Special Research Centre). His
W.D. et. al. 1997. research interests include a broad range of Control Systems n e o r y and
[3] Carman, W.D., Woodhouse, D.J., Poon, P.W.Y. Measuring the Applications.
performance of Earthing Systems in Cable Fed Systems in highly
Urbanised Environments, IO CEPSl Conference, Christchurch,
1994, Vol. 111, pp.101-112.

0-7803-6420-1/00/$10.00 (c) 2000 IEEE 1158

Вам также может понравиться