Supervised and unsupervised learning in composite reliability evaluation

Reliability analysis of composite generation and transmission systems is very computationally expensive. This paper presents supervised and unsupervised learning, and explains how these methods can reduce the computational requirements of such reliability analyses. Using learning algorithms will induce an (additional) error in the reliability indices. The type of error the different methods produce and the expected severity of the error is discussed. A case study is included to illustrate how the methods can be used in practice, and exemplifies the theoretical discussions.