Considering the rare events in construction of the Bayesian Network associated with power systems

In this paper a simple novel approach is presented to construct the Bayesian Network (BN) associated with a power system. In the approach, assuming independent outage events, a general structure is considered for BN which is then modified by using the Mutual Information (MI). Therefore the BN associated with the system is constructed easily and it is not required to use the common structure learning algorithms or use from the physical topology and cause-effect relations that their using for complex and large systems is intractable. The required training data is provided by the state sampling method of Monte Carlo (MC) simulation. Most of the transmission system components have relatively low failure probabilities and so their outages are rare events. Thus, for a more accurate impact analysis of transmission system components the Importance Sampling (IS) scheme is employed in the generation of data. The proposed method that is applicable to large and complex power systems is employed for detailed reliability assessment of IEEE Reliability Test System (IEEE-RTS) and its usability and efficiency is verified.