Computational Intelligence - November 2012 - 67

A fully decentralized multi-agent system based on
bus agents (BAs) is proposed to tackle the complex
distribution system restoration problem.
[24], the authors proposed a multi-agent
system based on P2P communication
mode and applied the contract network
protocol (CNP) to the negotiation process between agents. The parameters of
CNP are optimized by a genetic algorithm (GA). In [25], the authors adopted
a spectral clustering algorithm to partition a large distribution network into
logically related clusters, and used agents
to manage these clusters.
However, the multi-agent systems
proposed in the afore mentioned
research are not fully decentralized or
are not appropriate for complex distribution systems. In this work, we propose
a fully decentralized multi-agent system
based on bus agents (BAs) to tackle the
complex distribution system restoration
problem. Our approach uses a depthfir st communication mechanism
(DFCM) for effective and efficient information exchange among agents. In
brief, the proposed multi-agent system
has the following characteristics.
1) The system is able to handle service
restoration of complex distribution
system with DGs. The agents can
constitute island system flexibly
under different fault conditions.
2) The system can provide restoration
strategies without considering the
initial topology of the distribution
network. Therefore, it can effectively
handle cascading failure and enhance
system reliability.
3)The system is fully decentralized,
enabling a high degree of scalability
for different distribution network
topologies.
The rest of this paper is organized as
follows. Section II presents a brief introduction to agent and multi-agent system.
Section III describes the formulation of
restoration problem of distribution system. The restoration MAS framework is
presented in detail in Section IV. Finally,
the feasibility and performance of the
proposed MAS is demonstrated on a

complex distribution network in Section
V, and a conclusion is given in Section VI.
II. Agent and Multi-Agent System

There is no precise definition of agent
at present. Artificial intelligence (AI)
researchers generally consider an agent
as a form of entity with sensing, decision making, and actuation capabilities
[26]. In this article, an agent has all the
above characteristics, where it can collect data and make effective decisions
according to the sensory information.
Multi-agent system is an extension of
the agent technology where a group of
loosely connected autonomous agents act
in an environment to achieve a common
goal [27]. Through a composed agent's
behaviors, an MAS is capable of providing different design benefits such as parallelism, robustness, scalability, geographic
distribution and cost effectiveness [28]
[29]. In MAS, each agent merely collects
local information, while communicating
with other agents to realize information
sharing. The task can be divided into a
series of simple subtasks, and then be
assigned to different agents. A multi-agent
system has the following advantages over
a single agent or a centralized approach:
1) The centralized system, requiring a
powerful central computer, may lead
to a single point of failure. The
whole system would be out of control once the central computer is
down. The MAS does not suffer
from the single point of failure
problem since it is decentralized;
therefore significantly increasing the
system's reliability and security.
2) The MAS can improve system efficiency because of its parallel computation and asynchronous operation.
3) The MAS has a high degree of scalability and flexibility, as such agents
can be added or removed as needed.
In this paper, the multi-agent system
is implemented using JAVA agent
development environment (JADE).

Following the FIPA specifications, the
JADE is a middleware for the development of multi-agent system with necessary tools to facilitate the debugging
process and to track the actions of the
agents. Agents in JADE communicate
by message passing, where the messages
adhere to ACL (Agent Communication
Language) standards. For instance, an
ACL message has several attributes such
as performative, receiver, sender, content and conversation ID.
III. Distribution System
Restoration Model

The objective of the mathematical
model of a distribution system restoration is to maximize the supply of power
for the deenergized loads, giving priority to vital loads
N

/ bi xi Li ,

max

(1)

i =1

where L i is the load at bus i , x i is the
decision variable representing its status
(x i = 1: restored; x i = 0: not restored),
b i is the weight coefficient associated
with the load L i , depending on its priority, and N is the number of the buses
in system.
Typical constraints need to be considered for service restoration of distribution network include [19]-[21].
1) Capacity limitation constraint

Pi # C j,

(2)

id Bj

where C j is the capacity that bus j can
provide, B j is the set of branches which
absorb power from bus j, and Pi is the
power flow of branch i .
2) Limits on branch power flow
Pi # Pi max ,

(3)

where Pi is the power flow on the
branch i , and Pi max is the maximum
capacity of branch i .
3) Power balance constraint

i d Tj

Pi =

Pi + x j L j ,

(4)

idBj

where T j is the set of branches injecting
power to bus j, B j is the set of branches
absorbing power from bus j, and L j is
the load at bus j.

november 2012 | Ieee ComputatIonal IntellIgenCe magazIne

Table of Contents for the Digital Edition of Computational Intelligence - November 2012