A Bayesian Network Approach to Predicting Transmission Line Down Times

Extraordinary events in the power system, such as major blackouts, can have severe consequences for society. Resilience as a concept is used to systematically enhance the capability of the power system to mitigate such extreme events, characterized by their rare occurrence yet significant consequences. While there has been a substantial amount of work on the failure probability of transmission lines in severe weather, more detailed work on the restoration time has been limited. The latter is of high importance as extreme weather can cause large damage to power system infrastructure and delayed access to the failure site, subsequently leading to increased down times and long-lasting power supply interruptions. Limited historical data of varying quality challenge the ability to generalize fixed statistical down time distributions to different transmission lines under different conditions. An alternative to relying on (lack of) historical data is to build a logical model. This paper proposes a Bayesian Network (BN) model to predict transmission line down times. Parameter values are obtained through structured expert elicitation and historical fault data. Although lack of data makes it difficult to validate the model, the results show that a BN approach has potential applicability for predicting down time distributions for individual transmission lines considering time-varying inputs.