Abstract
--
This paper introduces a new framework for Expert
Systems, fully integrated into a SCADA/EMS real time control
software, whose main goal is to support the creation of new
applications without demanding the developing of additional rule
processing software. In order to validate this framework, two
applications that address the daily needs of control centers were
developed: SISPRO (Alarm Processing System) and RECOMP
(Power Systems Restoration Support System). Results show that
this tool allows for quicker problem solving, error reduction and
operator capacity increase.
Index Terms
-- Artificial intelligence, Computer applications
Computer fault diagnosis, Expert systems, Power system
restoration
I. I
NTRODUCTION
TILITIES
have invested heavily in control and
supervision equipment, but operation is still dependent on
peopleware. All knowledge on the operation and the
planning of the power sector are still property of an exclusive
group of experts. These experts are capable of making
decisions based on logic, heuristics, experience and non-
formalized knowledge, such as intuition. Nevertheless, the
growth seen in the power companies and the loss of experts,
due to retirement, have made utilities understand that it is
imperative that they develop new techniques for control,
supervision and automation that can preserve corporate
intelligence and memory. Hence, it will be possible to use it in
the future to train novice specialists so that they can
familiarize themselves with situations they have never faced in
order to be prepared to solve them faster whenever they may
arise.
These needs make us consider techniques that can learn and
incorporate previously available knowledge, an area in which
Artificial intelligence (AI) excels. AI branches into Fuzzy
Logic [1], Genetic Algorithms [2], Neural Networks [3], and,
specially, Expert Systems [4], whose main characteristic is to
simulate the human expert thought process. Taking those
needs in consideration, CEPEL used Expert Systems to
1
CEPEL (Electric Power Research Center), Av. Horácio de Macedo 354,
CEP 21941-911, Rio de Janeiro-RJ, Brazil
2
FSMA (Salesian College of Macaé), R. Monte Eliseo SN, Visconde de
Araujo, CEP 27943-180, Macaé-RJ, Brazil
E-mails:
• V. N. A. L. da Silva – navarro@cepel.br
• R. Linden – rlinden@cepel.br
• G. F. Ribeiro – gfr@cepel.br
develop the project “Expert Systems for Control Centers in
Real Time - SAGE–EXPERT”.
Expert Systems are a well proven technology for Energy
Management Systems (EMS) applications, having decades of
application in this area. They were chosen because their
solutions are fully explainable, can be easily documented and
incorporate previous knowledge, abilities that neural networks
lack. Operators have detailed knowledge of previous faults
and there are well documented manuals on the operation under
faulty conditions. Eventhough those bodies of knowledge may
not cover all the possible power system faults, they may not be
discarded beforehand.
Fuzzy logic was not chosen because alarms have no
inherent imprecision, even the numeric ones, which are
considered in relation to crisp boundaries. Hence, fuzzy
logic’s ability to treat numbers as linguistic concepts would
not improve the system’s performance. Genetic algorithms
could be considered as a tool to develop rules for the expert
systems. Similar approaches, even though directed to fuzzy
logic rules, were adopted in several papers for different
application areas, such as [5]. This option was not considered
in this paper due to the facts that previous knowledge was
extensive and an association rule technique had already been
used to mine for prospective rules.
In addition to all those features, Expert Systems have a
near-human reasoning process and a long successful history in
applications installed in Power System Control Centers.
Nevertheless, for years now, fewer people have been
considering Expert System as a useful tool in EMS because
when compared to other AI approaches such as ANN and GA,
Expert System suffers from disadvantages, mainly in
maintenance. The idea behind this work is also to simplify the
maintenance tasks introducing a powerful graphical interface
that makes it easier to create and reuse generic topological
rules that apply to all power stations that share the same
configuration. This way, expert systems can be seen as more
competitive when compared to the other AI techniques that
have automatic but slow training.
SAGE-EXPERT is a framework that is completely
integrated to SAGE, CEPEL’s own SCADA/EMS system. It
allows for the creation of new software applications without
the development of additional modules for rule processing and
communication with SAGE. It is implemented on a
standardized programming language and can execute on
multiple operating systems without loss of either performance
V. N. A. L. da Silva
1
, R. Linden
1,2
, G. F. Ribeiro
1
A FRAMEWORK FOR EXPERT SYSTEMS
DEVELOPMENT INTEGRATED TO A
SCADA/EMS ENVIRONMENT
U