Enhanced Power System Security Assessment Through Intelligent Decision Trees - Complex System Modelling and Control Through Intelligent Soft Computations - page 338

contained training database. Later in the numerical results section, the entropy
measure introduced in this section will be used to measure the information content
in the training database used for producing the decision trees.

4 Technical Approach

The overall flowchart of risk-based planning approach is shown by Fig. 3 , along

with the proposed ef

fl

cient sampling approach. The proposed algorithm consists of
two stages, where stage I utilizes a form of strati ed sampling to approximately

identify the boundary region and stage II utilizes importance sampling to bias the
sampling towards the boundary region. The database generation is performed for
every critical contingency or a group of critical contingencies screened, as depicted
by the left-side loop. The right-side loop feeds back information about the region of
sampling state space requiring more emphasis in the training database, in order to
reduce decision tree misclassi cations and improve the accuracy. This chapter

primarily focuses on the proposed ef

cient sampling method.

Fig. 3 Proposed approach