Transcript nlp.fi.muni.cz

Multi-Agent Framework for Power
Systems Simulation and
Monitoring
Miroslav Prýmek
and
Aleš Horák
Faculty of Informatics, Masaryk University
Brno, Czech Republic
Introduction – the project
Academy of Sciences of Czech Republic, project T100300414
Intelligent methods for increasing of reliability
of electrical networks
The main objective of the research project is an increase in safety and
reliability of operation of the electric power system by developing new
intelligent methods. A reduction of black-out risk in the Czech
Republic is very topical as well.
Participants – Institute of Informatics of AS CR, Faculty of
Electrical Engineering and Computer Science VSB TU Ostrava
and Faculty of Informatics MU Brno
Introduction – the aim
Reliability oriented maintenance
●possible replacement for the time oriented maintenance
●decreased expenses, increased effectivity and flexibility of the maintenance
●new methods for failure monitoring of power network elements are necessary
●simulation, prediction
Simulation methods for reliability computation
●failure rate data acquisition
●development of adequate tools
●utilization of existing approaches (database of failures) + their extension with
simulation processes
Remote control elements
●tools for system integration (monitoring and control)
ANALYSIS
●
●
●
●
●
development of a communication framework based on the multi-agent
approach that allows:
simulation of EPN processes
on-line data acquisition
interaction with monitoring and control systems
prototype implementation (named Rice)
Requirements
●
●
●
●
●
●
●
decentralization
platform independence
performance scaling
modularity and extensibility
open standards
security
low-cost practical application
IMPLEMENTATION - AGENTS
Multi-agent approach
independent units with local interactions
●decisions based on their own knowledge about the
state of the world
●querying and coordination of other agents
●
Agents in the Rice system
agent interface defined with message types
●similar to OOP principles – can use OOP models
●
IMPLEMENTATION - AGENTS
IMPLEMENTATION - AGENTS
IMPLEMENTATION – COMMUNICATION
Agent communication
4-layers:
●the network layer: TCP/IP
●the agent level delivery layer: CORBA
(independence on the network layer)
●
the message purpose layer: KQML (virtual
knowledge base + ontology)
●
the message subject itself
This is not the only possibility, but it offers the best
flexibility
IMPLEMENTATION - EXAMPLE
Agent A – transformation station transforming
very high voltage to high voltage
Agent B represents high voltage line
Agent C represents a distribution point
IMPLEMENTATION - EXAMPLE
IMPLEMENTATION - EXAMPLE
Agent B asks agent A for a notification for each state change of A and
each output voltage change of A:
(subscribe
:sender
B
:receiver
A
:timestamp 1113340454
:reply-with query_1
:language
KQML
:ontology
KQML_ontology
:content (ask
:sender B
:receiver A
:reply-with 2
:language Prolog
:ontology Power_system
:content out_voltage(X),state(Y)
)
)
IMPLEMENTATION - EXAMPLE
Event - there is an outage on line B and the current supply is
cancelled, B sends to C a message:
(tell
:sender B
:receiver C
:timestamp 1113341454
:in-reply-to query_2
:reply-with
query_3
:language
Prolog
:ontology Power_system
:content out_voltage(0), state(fatal_failure)
)
C can react to this message in several ways - it can set its output
voltage to 0V too or it can switch to another (backup) input line
VIEWER PROTOTYPE
CONCLUSIONS
●
●
●
●
events can result in different actions (data
acquisition, simulation viewer, ...)
separate development of different parts of the
system is possible
enough flexibility to all possible development
plans
with new modules – a complex system for
electrical power networks monitoring and control