technologies, improving management measures and optimizing
network structures. But blackout is inherently inevitable because
of unforeseen circumstances, incompleteness of information, treacherous
weather and increasing complication of power systems.
The impact of a blackout increases exponentially with the duration
of restoration, once blackout occurs, it is imperative to take
effective and secure measures to restore the power system as soon
as possible [3–5]. Generally, power system restoration includes
three phases: black-start, network reconfiguration and load
restoration [1]. The objectives of restoration are to take the power
system return to normal securely and rapidly, minimize losses and
restoration duration, and diminish adverse impact on the society
[2]. For network reconfiguration, the purpose is to restore the backbone
network concerned, interconnect relevant subsystems and
finally rebuild a stable skeleton network [6].
Backbone-network reconfiguration can be described as a multistage,
multivariable, multi-objective, combinatorial, nonlinear and
constrained optimization problem. There are no known mathematical
methods for solving such a NP-complete problem exactly
in polynomial time [7]. In order to speed up restoration without
violating security constraints, many methods have been addressed.
Expert system [8] has been employed extensively in making
restoration schemes, it has promising prospects of application.
But the establishment and maintenance of knowledge base is a key problem, especially when the system is becoming larger