[attachment=1423]

AN EXPERT SYSTEM FOR POWER PLANTS
DEPARTMENT OF ELCTRICAL & ELECTRONICS ENGINEERING

Abstract: An intelligent fault diagnosis and operator support system
targeting in the safer operation of generators and distribution
substations in power plants is introduced in this paper. Based on
Expert Systems (ES) technology it incorporates a number of rules for
the real time state estimation of the generator electrical part and the
distribution substation topology. Within every sampling cycle the
estimated state is being compared to an a priori state formed by
measurements and digital signaling coming from current and voltage
transformers as well as the existing electronic protection equipment.
Whenever a conflict between the estimated and measured state arises, a
set of heuristic rules is activated for the fault scenario inference
and report. An included SCADA helps operators in the fast processing of
large amounts of data, due to the user-friendly graphical
representation of the monitored system. Enhanced with many heuristic
rules, being a knowledge based system, the proposed system goes beyond
imitation of expert operatorsâ„¢ knowledge, being able to inference fault
scenarios concerning even components like the power electronic circuits
of generator excitation system. For example, abnormal measurements on
generatorâ„¢s terminals can activate rules that will generate fault
hypothesis possibly related to an excitation thyristors abnormal
switching operation.
Introduction
Artificial Intelligence is a branch of informatics that
was widely adopted in industrial automation during the past fifteen
years. AI programs are developed and used in computer science since the
early days of digital computers. Only during the last two decades
though industry has taken advantage of those special features that make
AI so unique in modeling and representing knowledge, as well as
imitating the common sense reasoning. The continuous augmentation of
available computational strength and the low cost of modern
microprocessors on one hand, and the software tools recently developed
on the other, leaded in a remarkable expansion of AI applications in
the domain of electrical power systems and power electronics.
Expert Systems:
Among others is a very popular AI technique in industry.
According to the working group D10 of the line protection subcommittee
, An Expert System (ES) is a computer program that uses knowledge and
inference procedures to solve problems that are ordinarily solved
through human expertise. The main components of an ES are: a) inference
engine, b) database, c) user-interface. ES incorporate rule kind of
programming. They are currently being used in many applications in the
area of power systems and power electronics. Several systems for the
short or long term load forecasting have been already introduced based
on ES technology .Intelligent SCADA and offline training systems for
non-expert operators is another application where ES are often used.
All these offline applications are nevertheless not critical for the
power system robustness and stability. More and more applications are
currently using ES in real time monitoring and/or control, and AI turns
to be a common practice in industrial automation. Regarding the
category of real time monitoring and control systems, many applications
have already been proposed, focusing mainly on topology estimation and
fault diagnosis in distribution substations , and on the fault
diagnosis and restoration strategies for transmission networks.
Knowledge Based Systems: Go beyond Expert systems in sense that except
for imitating the expertsâ„¢ problem solving behavior, they enrich
problem solving strategy with methods that are not originally employed
by human experts. Systems that use domain knowledge to guide searches
that differ from the expertsâ„¢ are known as Knowledge Based Systems
(KBS).
Intelligent Decision Support Systems: Decision Support Systems (DSS)
are computerized tools derived from decision theory used to enhance
user ability to make decisions efficiently. They are not intended to
offer the final solution, but rather to explore and seek alternative
solutions. The intimate decision is left to the user. Intelligent
Support Systems (IDSS) add intelligence to existing systems to enhance
problem solving
ability and help maintain a broad range of knowledge about a particular
domain. They are used for capturing, organizing and reapplying
knowledge including decision rules and criteria.
Artificial Neural Networks : That simulate the neural activity of the
human brain, deserve the same recognition at the same level as the AI
methodologies mentioned above. ANN have already been broadly classified
under the AI domain. They do not have some of the AI properties but can
be placed under the umbrella of AI technologies. Expert Systems
basically mimic the problem solving behavior of experts using domain
knowledge acquired through interviews during the knowledge acquisition
phase. Knowledge based ES as mentioned go beyond in a sense that they
enrich problem-solving strategy with methods that are not ordinarily
employed by human experts . The proposed system is designed for the
generators and distribution substations protection in power plants.
Especially in weak interconnected power systems, operation of plants
with over than 1000MVA of installed power can be of great importance
for the stability and efficiency of the whole system. An unhandled
fault can have a significant impact on power availability for an
expanded area of the transmission network. Besides, damage on
a generator would add a very high financial overhead, as generators of
this size cost several million Euros. Such unhandled faults have though
been reported in the past and can lead even to human casualties. The
system is designed to instantly recognize and report abnormalities that
can be related to a mechanical equipment failure or to an electrical,
or electronic equipment malfunction, or even to a mistaken human
operator control instruction.
System Overview
Distribution substations are the interlocking
connection points of power plants to the electrical power grid. The
state of all substation components (circuit breakers, disconnectors,
protection relays etc.) is monitored and recorded to Digital Fault
Recorders (DFR) while the electrical values of every circuit breaker,
bus, transformer and generator terminal are measured by ad hoc
installed current and Voltage-transformers.

Figure 1. Snapshot of the system GUI applied on a 350MVA unit of a
thermoelectric plant.
From the operator perspective an alarm situation arises
when a monitored value exceeds a predefined upper or lower limit,
activating a sound or light alert on control panel. An expert operator
would handle this situation by first checking the control panel
indications, trying then to locate the faulted area, according to the
theoretical state of the switching equipment and the current values of
the measurement points. This procedure may take some time especially
when operators act under stress conditions. On the other hand inference
process can be a very complicated task when some input data or
measurements are faulted. For example, a very difficult fault to
diagnose has been reported in the past, when after a voltage
transformer explosion a bypass switch broke and caused short-circuit,
supplying the generator with an unbalanced load. In this case the
switch position was mistakenly reported and the operator could not
easily detect the real current flow path.

Figure 2. Fault recognition and analysis algorithm
The time between the fault appearance and its recognition
and restoration inference can be critical for the equipment and
personnel safety.
A sophisticated fault diagnosis and monitoring system can
detect similar contradictions and point out the optimal restoration
sequence. The proposed expert system uses a dedicated module for the
topology and state estimation of the generator and the distribution
substation. This module considers as known inputs the voltages and
currents measured on the arriving from the network transmission lines,
as well as the generator and transformer current and voltage. Also
known is considered the state of the circuit breakers, disconnectors,
protection relays etc. Based on the above values the system composes an
estimated state regarding the voltage and current flow at all measuring
points. Another module composes the same state based on the acquired
measurements at the same points. The estimated and measured states are
being compared till a conflict arises between the estimated and
measured values of a certain measurement point. Then the fault locating
module locates the faulted area, and the fault scenario module
inferences the fault hypothesis. The system then activates the
restoration module in order to propose the restoration sequence
bringing the process back to its normal operation.

Figure 3. Basic system architecture diagram
System Architecture
The proposed knowledge based expert system runs on a
dedicated x86 based computer. Extra data acquisition and digitization
hardware is required connected to the PCI bus for fast data acquisition
of the various measured or reported values of generator and substation
components. The core of the system is the running software. It is
consisted of three main subprograms running simultaneously and using
three different threads
Data acquisition and monitoring System: This program is responsible for
the data acquisition, interfacing the external acquisition hardware. It
passes all acquired information to the inference engine and displays
some defined data to the system monitor. It also displays some selected
by the operator data, implementing thus the system GUI input and
output. Selected data are sent to the system Data Base for history
logging.
Data Base: The system database is consisted mainly by two modules:
-The knowledge database keeps all the knowledge acquired during
the system design phase via exhausting interviews with the station
expert operators. This database is designed in a way that allows
knowledge modification and update, offering to the system flexibility
and upgrade capability.
-The history recording and logging data base which is used for the
storage of selected values that can be accessed by the inference engine
in real time, or can be even used offline for data further processing
and evaluation.
Inference Engine: This program is the heart of the whole system. It is
an intelligent function based on rule-base programming. Using the
current data values of the data acquisition module and the knowledge
stored in the knowledge base, it inferences knowledge imitating the
expert operator reasoning. In the same time it performs advanced checks
that an operator cannot do in real time, using special rules that offer
a quality process monitoring and analysis. When a fault is diagnosed
the engine inferences the fault scenario and proposes the necessary
restoration actions. Alternatively, the inference engine can produce
not only message output but control signaling as well.
Conclusion:
This work introduces a knowledge based expert system for
the generator and substation monitoring and fault diagnosis in power
plants. The fault detection is based on a comparison algorithm polling
for specific measurement values, comparing them to the corresponding
estimated values, according to the system current inputs, and then
checking for possible conflicts. Whenever a conflict arises the system
uses rule-based reasoning to inference the fault scenario and the
optimal restoration sequence, which is fed back to the control room
operator for further action. The knowledge based expert system
efficiency is based on, but not limited to, the expert operators
reasoning.
It can report and analyze faults, even having
received partially mistaken input data, something that for a human
operator is very difficult or impossible in real time, especially under
emergency situations. The knowledge base can be continuously updated
with rules, offering thus a learning capability that enriches the
system with new, recent experience. Based on some advanced rules the
system can offer fault scenario inference performing multiple input
calculations, even with strictly restrictive complexity for the human
operator real-time processing. This can lead to a detailed fault
diagnosis even when the cause is indirect. For example, a failure of
power semiconductor elements of the generator field excitation
rectifier, can be recognized and be classified indireclty, according to
its effects on the measured and estimated parameters.
References:
[1] M.S Kandil-N.E.Hasanien: Long-Term Load Forecasting for fast
Developing utility using a knowledge based expert system, IEEE
Transactions on Power Systems, vol7, No2, May, 2002
[2] M.Negnevitsky: A knowledge based tutoring system for teaching fault
analysis. IEEE
Transactions on Power Systems, vol13, No1, May 1998
[3] M.Kezunovic-Z.Ren-D.R.Sevcik-J.Lucey: An. expert system for
automated analysis of circuit breaker operations. ISAP03, Lemnos August
2003
[4] H.Lee-B.AhnY.Park:Afault diagnosis expert system for distribution
substations, IEEE
Transactions on Power Systems, vol15, No1,January 2000
[5] H.Lee- D.Park- B.shin- Y.Park- J.Park- S.Venkata: A fuzzy expert
system for the integrated fault diagnosis, IEEE Transactions on Power
Delivery, vol5, No2 April 2000