Page 6 - CE_Industral_Journal_Jan2008
P. 6
A Novel Approach for Ground Fault Detection
high impedance fault detection study [9], the results of
years of experience with high impedance fault detec-
tion and testing are summarized and the formal evalu-
ation of the performance of a randomness based high
impedance fault detection algorithm are disclosed.
This technique was implemented in a product which
uses multiple algorithms based mainly on energy, ran-
domness, interharmonics, etc. in layers as described in
[10].
There are many other attempts to detect high imped-
ance ground faults and published in literature. They High impedance ground
are reported in some of the references provided. It can fault detection system
be appreciated that conventional means for detecting
high impedance faults in electrical power lines are Figure 1. Electrical power distribution network having
typically not always conclusive and/or reliable and a composite high impedance ground fault detection
can be expensive. Therefore, a need exists for a new system.
reliable and economic solution for detecting high im- The composite high impedance detection system in-
pedance faults in electrical power lines which address- cludes three individual high impedance ground fault de-
es the engineering and legal ramifications of detecting tection systems which are not shown in Figure 1 but are
and determining what to do once a high impedance shown in Figure 2. Also shown in Fig. 1 are the potential
fault is detected. transformer PT and the current transformer CT which
This paper describes a novel approach to detect high provide the analog inputs for the high impedance ground
impedance ground faults. The approach uses only fault detection system which could be implemented in a
three algorithms based on higher order statistics, wave- protective relay used for protecting the line.
lets and neural networks, and uses decision logic to These individual high impedance fault detection systems
determine the detection of the high impedance ground have individual algorithms for individually detecting high
faults. Some of the laboratory test results are also in- impedance faults. These algorithms are based on wavelet,
cluded in the paper.
higher order statistics and neural network. The individual
high impedance fault detection algorithms can each have
Novel High Impedance Ground Fault a different confidence level. A fault is identified as a high
Detection Approach impedance fault once it is detected independently by the
The work presented in this paper relates to a new ap- algorithms and processed through a decision logic.
proach to high impedance ground fault detection that Figure 2 shows a composite high impedance ground fault
includes a multi-scheme high impedance ground fault detection system including a higher order statistics based
detection scheme employing individual ground fault high impedance ground fault detection system identified
detection systems each having their own algorithm in Figure 2 as a 2nd order statistical system, a wavelet
application that use various features of phase and/or based high impedance ground fault detection system,
ground currents to individually detect a high imped- and a neutral network based high impedance ground
ance ground fault. Suitable features of the currents fault detection system. As shown in Figure 2, power
include their waveform signatures, their sample values system signals are acquired, filtered and then processed
etc. Figure 1 shows a schematic diagram of an elec- by individual high impedance ground fault detection
trical power distribution system having an electrical algorithm and results of these individual algorithms are
power distribution line and a composite high imped- further processed by a decision logic to provide the
ance ground fault detection system. detection decision. The decision logic can be modified
depending on application requirement. For example,
one-out-of-three logic will provide maximum depend-
ability, whereas three-out-of-three logic will provide
maximum security and two-out-of-three will be some-
where in-between.
Industry Journal 5