Enhanced Power System Security Assessment Through Intelligent Decision Trees - Complex System Modelling and Control Through Intelligent Soft Computations - page 346

and inter-dependencies. Again setting p = 0.75, 75 % of the points is expected from
N-dimensional boundary region S, as the probability distribution is altered to produce
more samples from S. Figure 9 depicts the probability reorientation by importance
sampling process towards the boundary region in a 2-dimensional state space. The
parameter p serves as a sliding parameter that controls the extent of biasing between a
completely operational study with p = 1 to investment planning study with p = 0.

5 Case Study

The proposed sampling approach is applied for a decision tree based security
assessment study for deriving operating rules against voltage stability issues on
SEO region (Syst

ctrique Ouest, West France, Brittany), a voltage security-
limited region of the French EHV system containing 5,331 buses with 432 gen-
erators supplying 83,782 MW. Figure 10 shows 400 kV network of the French
system, where it can be seen that the Brittany region (highlighted in pink) is pretty
weakly interconnected. During winter periods, when demand peaks, the system gets
close to voltage collapse limits. Moreover the local production capabilities being far
lower than the local consumption, it puts the EHV grid under pressure as the needed
power comes from remote location, eventually leading to cascading phenomenon at
the sub voltage levels. The busbar fault at 225 kV Cordemais bus is the most
credible contingency in the Brittany region during winter period.
So in order to avoid the risk of collapse situations under such contingency
events, the operator may have to resort to expensive preventive measures such as
starting up close yet expensive production units. It is therefore very important to
assess the risks of a network situation correctly considering uncertainties in oper-
ating conditions and obtain operating rules built with decision trees, that aid to take
right decision at right time.
Section 5.1 describes the study speci

me El

è é

cations in terms of historical data used in
this study, the sampling parameters and assumptions, and tools and methods used to
perform power system assessments. Section 5.2 provides the numerical illustration,
presenting the systematic application of stages 1 and 2 of the ef

cient sampling

approach in Sects. 5.2.1 and 5.2.2 respectively, and finally discussing the results

from the proposed method and their signi cance in Sect. 5.2.3 in terms of operating

rule ' s classi cation accuracy and economic bene ts.

5.1 Study Speci cations

Data preparation: The historical database of French EHV power grid system for the
study is extracted from records made every 15 s on the network by SCADA. The
load in the SEO region starts to increase at the end of October, as the winter comes
closer, and decreases in February. The heavily loaded period is the winter, during