Information Technology Reference
In-Depth Information
1 Introduction
Traditionally, power system reliability assessments and planning involve deter-
ministic techniques and criteria, which are being used in practical applications even
now, such as WECC/NERC disturbance-performance table for transmission plan-
ning (WECC
2003
; Abed
1999
). But the drawback with deterministic criteria is that
they do not re
fl
ect the stochastic or probabilistic nature of the system in terms of
load pro
les, component availability, failures etc. (Billinton et al.
1997
). Therefore
the need to incorporate probabilistic or stochastic techniques to assess power sys-
tem reliability and obtain suitable indices or guidelines for planning has been
recognized by the power system planners and operators; and several such tech-
niques have been developed (Beshir
1999
; Chowdhury and Koval
2006
; Li and
Choudhury
2007
; Wan et al.
2000
; Xiao and McCalley
2007
).
In this regard, Monte Carlo simulation (MCS) methods lend themselves well by
simulating the actual analytical process with randomness in system states (Billinton
and Li
1994
). In this way, several system effects or process including nonelectrical
factors such as weather uncertainties can be included in a study based on appro-
priate parameter
'
s probability distributions. Figure
1
shows an overview of MCS
based security assessment methodology, which involves two major tasks: database
generation approach and machine learning analysis.
The database generation approach involves the following steps:
Random Sampling: Operating parameters (load, unit availability, circuit outages,
etc.) are randomly selected as per a distribution (e.g., uniform, Gaussian,
exponential, empirical etc.). This process is generally known as Monte Carlo
sampling. Using the generated samples, various base cases are formed.
Fig. 1 Probabilistic reliability assessment based on MCS and data mining (Henry et al.
2004b
)
Search WWH ::
Custom Search