RCA METHOD FOR FAULT DIAGNOSIS IN DIGITAL

SUBSTATIONS OF POWER SYSTEMS

Piao Peng, Zhiwei Liao

School of Electrical Engineering, South China University of Technology, 510640, Guangzhou, China

Fushuan Wen

School of Electrical Engineering, Zhejiang University, 310027, Hangzhou, China

Jiansheng Huang

School of Computing and Mathematics, University of Western Sydney, NSW 2751, Sydney, Australia

Keywords: Digital substation, Fault diagnosis, Root cause analysis, Dempster/shafer theory, Fishbone diagram.

Abstract: The authors present a Cause-Effect fault diagnosis model, which utilises the Root Cause Analysis approach

and takes into account the technical features of a digital substation. The Dempster/Shafer evidence theory is

used to integrate different types of fault information in the diagnosis model so as to implement a

hierarchical, systematic and comprehensive diagnosis based on the logic relationship between the parent and

child nodes such as transformer/circuit-breaker/transmission-line, and between the root and child causes. A

real fault scenario is investigated in the case study to demonstrate the developed approach in diagnosing

malfunction of protective relays and/or circuit breakers, miss or false alarms, and other commonly

encountered faults at a modern digital substation.

1 INTRODUCTION

Fault diagnosis and accident treatment in substations

have become a major challenge of reinforcing power

systems’ safety and reliability. Many integrated

substation diagnosis models and methods have been

proposed to address this challenge by using

information obtained from protective relays and

circuit breakers and employing technologies such as

expert systems (Lee et al., 2000); (Jung et al., 2001);

(Huang, 2002), artificial neural networks (Yang et

al., 1994); (Cardoso et al., 2004), Petri networks

(Huang and Mu, 2006); (Lo et al., 1997), agent

technology (Dong and Xue, 2004), and rough sets

(Hor and Crossley, 2007); (Dong et al., 2002). In

addition, substation diagnosis models and methods

may rely on a single transmission or transformation

equipment as done by the transformer diagnosis

model based on three chromatographic level

correlation analysis (Michel and James, 2005), and

the wavelet theory based transmission line fault

diagnosis model using fault recorders (Silva et al.,

2006). It is observed that the current substation fault

diagnosis models only take into account information

of protective relays and circuit breakers, or fault

features of a single device. In other words, the

existing models and methods, due to employing only

local information, are difficult to diagnose certain

complex faults with uncertainties, including multiple

consecutive failures, malfunctions of protective

relays and/or circuit breakers, missing or false

alarms, and sensor errors, to name a few (Lee et al.,

2000).

With advent of new technologies and tools such

as intelligent primary/secondary equipment and

IEC61850 communication standard, applications of

digital technologies have become the trend in

substation automation, calling for novel fault

diagnosis methods and models with information

sharing and interoperability of intelligent electrical

devices in substations.

In the paper, the authors propose a Root Cause

Analysis (RCA) based Cause-Effect (fishbone

diagram) fault diagnosis model for digital

370

Peng P., Liao Z., Wen F. and Huang J..

RCA METHOD FOR FAULT DIAGNOSIS IN DIGITAL SUBSTATIONS OF POWER SYSTEMS.

DOI: 10.5220/0003627403700378

In Proceedings of the International Conference on Evolutionary Computation Theory and Applications (FCTA-2011), pages 370-378

ISBN: 978-989-8425-83-6

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

substations of power systems. The fusion rule of the

well-developed Dempster/Shafer (D-S) evidence

theory is used to integrate different types of fault

information obtained through monitoring states of

substations, protective relays and circuit breakers.

Based on the logic relationship between the parent

and the transformer/circuit breaker/transmission line

child nodes, and between the root and the child

causes, in the diagnosis model, an hierarchical,

systematic, and comprehensive diagnosis can then

be performed. A software package has been

developed to implement the proposed fault diagnosis

model and deployed in Xingguo Substation, the first

digital substation in Jiangxi Province, China.

2 BASIC PRINCIPLES OF RCA

The RCA, originally applied in organization

management (NELMS, 2007), is a hierarchical and

systematic approach to identify and analyse the root

cause, and then develop the countermeasures

accordingly. A large substation comprises many

components with interactions among them. Through

analysing these interactions, a novel fault diagnosis

model for substation transmission and

transformation systems can be built up according to

the structure and data/information flows of a digital

substation. Based on the theory of RCA, the fault

diagnosis model is formulated to explain the linkage

chain of accident causes so that identify what really

happened and the applicable countermeasures.

Figure 1: Framework of RCA-based fault diagnosis

system for digital substations.

The research tools of RCA include Cause &

Effect/Fishbone Diagram, Brain Storm and WHY-

WHY Diagram. There are three types of Fishbone

Diagrams: arrangement-based, cause-based and

solution-based. As shown in Fig. 1, the cause-based

fishbone diagram is adopted to explain the

philosophy of applying the RCA in fault diagnosis

of digital substations. The following is the

explanation to each component of the diagram:

: a problem node to be solved as a specified

fault in a substation .

c : a child cause of

and a basic reason of a

specified fault.

()

pc denotes the fault probability

caused by

c .

() {, , ,}

SF c c c



 is the set of

child causes which could trig

r : a root cause of

and a fundamental reason

of a specified fault in the power system.

(|)

pr c

denotes the conditional fault probability caused by

with given

c and G(c

)={r

,…,r

} is the root

cause set.

N : the only parent node for the diagnosis

system.

(, ,)

NDMO



, composed of three

elements

D ,

and O , denotes the basic

diagnosis functions.

represents the composition

of the access modes to obtain the required

information from the source

{, , , }

Ddd d 

and

D is the collection of all the n available modes.

M= {met

, met

,…, met

}, denotes the p fault

diagnosis methods applicable at the node.

{[ , ( )] | ( 1, 2, , )}

Ocpci q



  is the diagnosis

output, where



, q is the number of the reasons

{

c }, and ()

pc denotes the fault probability caused

c .

CN and

RN : the child nodes and the root

nodes. Like

N , they are constituted by the three

elements

D ,

, O . Furthermore, they can give a

more detailed diagnosis based on that of

N .

Thereinto, S(CN)={CN

, CN

, …, CN

}  FN, with

()SCN denoting the set of all the child nodes

belonging to

N ; S(RN|CN

) ={RN

, RN

, …, RN

}

 CN

is the set of root node RN belonging to the

child node

CN .

6) From the node definition given in 4), it can be

seen that all nodes, including

N ,

CN and

are independent in obtaining the information needed

by the diagnosis, selecting the appropriate diagnosis

methods, and analysing fault reasons of each node.

3 FAULT DIAGONOSIS OF

DIGITAL SUBSTATION

A root cause analysis based fault diagnosis system

RCA METHOD FOR FAULT DIAGNOSIS IN DIGITAL SUBSTATIONS OF POWER SYSTEMS

371

for digital substations is shown in Fig. 1.

12345

(){ , , , , }SCN CN CN CN CN CN

and

12345

() {, , , ,}SF c c c c c denote the child nodes and the

child causes of transformers, circuit breakers, lines,

bus and secondary system (DC power supply,

network communications and security devices).

3.1 The Modes to Obtain Information

The Substation Configuration Description Language

(SCL) is used to describe the IEC61850 standard

based IED configuration and related parameters,

communication system configuration, substation

system structure, and the relationship among them

for information exchanging. Logical node LN is the

basic function unit of a digital substation to obtain

the needed information. Part of the logic nodes

required in the designed fault diagnosis is listed in

Table 1:

Table 1: Main logic nodes in SCL.

Logical node Explain

1. Pxyz (Protective relay) Protection operation

2. XCBR (Circuit breaker) Switch position

3. RREC (Reclosing) Reclosing operation

4. XSWI (Knife switch) Knife position

5. SMIL (Online monitoring

information of transformer oil

chromatography)

Monitoring value

6. SCBR (Online monitoring

information of circuit

breaker）

Monitoring value

For a comprehensive analysis and diagnosis of

an accident, the required diagnosis information

should also include various electrical and chemical

test results of the equipment. The diagnosis

information is divided into three types as following:

1) The location variant information, i.e., the remote

information with time stamps.

2) The section information, including the

information of remote communication and remote

measurement at a certain time point.

3) The data files, including various electrical and

chemical test results of equipment, chemical

experiment results, overhaul history, waveform files

of breaker and recorded faults via on-line

monitoring.

3.2 The Parent and Child Nodes

The information access mode

{}Dd of a parent

node

(, ,)

NDMO is a passive one in obtaining

location variant information of Pxyz(Protective

relay), XCBR(Circuit breaker), RREC(Reclosing)

and secondary equipment. The diagnosis method M

= {m

} based on the optimization algorithm

developed in (Guo, Wen, Liao, Wei, and Xin, 2010),

is to diagnose substation faults with information

obtained from protective relays, circuit breakers and

secondary equipment. The output of the diagnosis is

1122334455

{[ , ( )], [ , ( )],[ , ( )], [ , ( )],[ , ( )]}O cpc cpc cpc cpc cpc

where

c ,

c respectively denote

transformer faults, malfunction of protective relays

and/or circuit breakers, line faults, bus faults, and

malfunctions of secondary equipment (Fig. 2).

The child nodes such as

CN ,

CN and

CN are

defined below:

1) The child node of transformer

The information access mode

{}Dd of child

node

1111

(, ,)CN D M O



is an active mode in

obtaining online monitoring information of the

transformer oil chromatography. M = {m

} is the

method to analyse the gas in the oil so as to diagnose

transformer faults using the improved three-ratio

method.

{[ , ( | )] | 1,2, ,9}

Orprcj



 

is the

output of the diagnosis, where

r is partial

discharge,

r is type-1 low-temperature

overheating(below 150

C),

r is type-2 low-

temperature overheating (150

C-300

C),

r is

medium-temperature overheating,

r is high-

temperature overheating,

r is low-energy

discharge,

r is low-energy discharge and

overheating,

r is arc discharge, and

r is arc

discharge and overheating.

(|)

pr c is the fault

probability caused by



with given

c .

The child node of circuit breaker

2123

{, , }Dddd



is the information access mode of

child node

2222

(, ,)CN D M O



, and

d is a passive

mode in obtaining location variant information of

XCBR,

d is an active mode in obtaining online

monitoring information of SCBR,

d is an active

FTP mode in obtaining online monitoring waveform

files of circuit breakers. Based on the Dempster’s

Fusion Rule and expert knowledge-base,

M = {m

}is

the method to establish the set of state sign with

online monitoring information, including switching

coil current, switch waveform file, storage time of

energy-storage motor, and current curves. The

method diagnoses the faults of circuit breakers

according to the coil switching current RMS and the

FCTA 2011 - International Conference on Fuzzy Computation Theory and Applications

372

elapsed time, the energy-storage motor storage time,

the total distance of a circuit breaker’s operation, the

instantaneous and the average switching speed of a

circuit breaker.

()

ata

Figure 2: The functional diagram of FN.

{[ , ( | )] | 1, 2, ,10}

Orprcj is the output

of the diagnosis,

r : mismatch of the switching coil

core and over resistance of switch operation,

r :

short circuit of the switching coil,

r : burn or break

of the switching coil,

r : deformation or

displacement of latch and valve connected to the

core mandrel,

r : poor contact and operation of

auxiliary switch and closing contactor,

r : fault of

DC power or system auxiliary power,

r : fault of

operating mechanism,

r : fault of energy-storage

motor,

r : mechanical failure, such as deformation

and displacement of linkage unit, and latch failure,

and

r : short residual life.

(| )

pr c denotes the

fault probability caused by

r with given

c .

The child node of line

{}Dd is the information access mode of child

node

3333

(, ,)CN D M O ,

d is the active FTP

mode to obtain recorded line fault files (Contrade).

M = {m

}is the method that utilizes the sudden-

change of the phase current difference to select the

phase and then locate the fault by estimating the

distance with the sampled data from the recorded

fault curve.

{[ , ( | )]| 1,2,3, 4}

Orprcj is the

output of the diagnosis, with

r as single phase

grounding fault,

r as double phase grounding fault,

r as inter-phase short circuit fault,

r as three

phase short circuit fault.

(|)

pr c denotes the fault

probability caused by

with given

c . In addition,

the identified fault location is included in the

diagnosis output (Fig. 3).

4 THE FAULT DIAGONOSIS

FLOWCHART BASED ON RCA

As illustrated by the flowchart in Fig. 4, the RCA

based fault diagnosis includes two cases.

4.1 Without Operation of Protective

Relaying, Circuit Breaking and

Reclosing

This case is mainly for monitoring and evaluating

the status of transmission and transformation

equipment. Each child node (

and O ) is

started periodically with a timer interval

int erval

t .

According to the output of child nodes

CN ,

and

CN , the state of transmission and

transformation equipment of the substation is

evaluated, with the evaluation results

O ,

O and R as given in Eqn. (1).

1235

[( | ), ( | )]

[(|),(|)]

[( | ), ( | )]

[(|),(|)]

















rc prc

RO O O O

rc prc

(1)

4.2 With Operation of Protective

Relaying, Circuit Breaking and

Reclosing

The major analysis and diagnosis procedure of this

case is as following:

Once the protective relays, circuit breakers and

reclosers operate, the diagnosis

M of

N is started,

to obtain the location variant information of Pxyz,

XCBR, RREC and secondary equipment by mode

d . In the diagnosis, the optimization technology is

employed to identify the faulty components and

provide the child cause set of fault as shown in Eqn.

(2).

112 2

334455

( ) {[ , ( )], [ , ( )],

[ , ( )], [ , ( )], [ , ( )]}

SF O c pc c pc

cpc cpc cpc

(2)

RCA METHOD FOR FAULT DIAGNOSIS IN DIGITAL SUBSTATIONS OF POWER SYSTEMS

373

2) Due to the lag in acquiring various waveform

files compared with obtaining the location variant

information, a delay of

delay

t is introduced to start

each child node.

To avoid conflicts between the parent node and

child node due to potential errors existing in

information source, the set of all possible faults is

used as the basis of the diagnosis, and each

symptom of faults is used as the evidence in

conducting the comprehensive analyse to the output

N , O ,

O ,

O and

O . The frame of

discernment is the basic concept of D-S evidence

theory. For a judgment problem, all possible results

that can be recognized are expressed by Θ, a non-

empty set known as the frame of discretion. The

frame consists of a number of mutually exclusive

and exhaustive elements.

12345

{, , , , }qqqqq ,

where

q is a transformer fault,

q is malfunction of

protective relays and/or circuit breakers,

q is a line

fault,

q is a bus fault and

q is malfunction of

secondary equipment.

()

mq , the assigned value to function m by

proposition

q , meets the following conditions:

0)(



m

(3)

1)(且,0)(，







iii

qmqmq

(4)

()

mq is known as the basic probability assignment

function (BPAF) of

q , which reflects the belief to

the accuracy of

q , i.e., the direct support to

q but

no support to any subset of

q . Furthermore, ()

is defined as the focus element of evidence if

q is a

subset of Θ and

()

mq >0. Φ represents an empty set

in Eqn. (3). Here the diagnosis result of

N is taken

as Evidence-1 corresponding to the BPAF

()

mq ,

and the diagnosis result of

CN is taken as

Evidence-2 corresponding to the BPAF

()

mq.

()

mq and

()

mq are supposed to be the two

BPAF of independent evidence in the same frame of

discernment Θ. While

m is the BPAF of Evidence-

1 with

() ()

mq pc



m is the BPAF of Evidence-

2 with

() (|)

lji

mq pr c



The D-S Fusion Rule is to reflect the joint effect

of the evidences in the same frame of discernment

through calculating a single BPAF with the BPAFs

of different evidences. By applying the rule, the joint

effect of Evidence-1 and Evidence-2 is evaluated in

Eqn. (5).

)()(

)(

qqq

qmqm











(5)

where ()mq is the orthogonal sum of

()

mq and

()

mq, denoted by

mm m



 .

() ()1

qql

mqm q

mqm q k









(6)

where

() ()

kmqmq







expressing the

conflict degree resulted in the fusion course of the

evidences, and 0≤

k≤1. In general, the larger the k,

the more intense conflicts are among the evidences.

Figure 3: The functional diagram of CN.

FCTA 2011 - International Conference on Fuzzy Computation Theory and Applications

374

()

ata

Figure 4: The RCA based fault diagnosis.

Figure 5: The main connection scheme of the 110 kV Xingguo digital substation.

RCA METHOD FOR FAULT DIAGNOSIS IN DIGITAL SUBSTATIONS OF POWER SYSTEMS

375

Table 2: Location variant information obtained by mode

d .

Time Alarm ID Alarm value Alarm description

2009-12-20

15:20:12 50ms

PCOS_PZB1H/Q0PTOC3

$ST$Op$general

Operation of overcurrent, segment-2, 1

time,

limit of high reserve for transformer 1#

2009-12-20

15:20:13 150ms

PCOS_P110LINE1/Q0XC

BR1$ST$Pos$stVal

1 Operation of circuit breaker numbered 111

2009-12-20

15:20:13 260ms

PCOS_PZB1L/Q0XCBR1

$ST$Pos$stVal

1 Operation of circuit breaker numbered 901

2009-12-20

15:20:13 327ms

PCOS_PZB1M/Q0XCBR

1$ST$Pos$stVal

1 Operation of circuit breaker numbered 301

2009-12-20

15:20:13 383ms

PCOS_

P110LINE3/Q0XCBR1$S

T$Pos$stVal

1 Operation of circuit breaker numbered 131

Table 3: Online monitoring information of transformer oil chromatography obtained by mode

d .

Time Alarm ID Alarm value Alarm description

2009-12-20