Explainable AI Enabled Conductor Galloping Predictor Design

Conductor galloping, often happening at extreme weather conditions, poses significant challenges on the reliable operations of the power grid, as conductor galloping may lead to serious mechanical damages to the system. Hence, it is of crucial importance to accurately predict such events in advance. Unfortunately, this is a delicate task in that its associated data is hard to collect and its physical dynamic is challenging to decipher. In this work, we propose to use a variation of the decision tree, Classification and Regression Tree (CART), to deal with the imbalanced dataset for conductor galloping prediction. Compared with more advanced machine learning techniques, its tree structure may help illuminate the physical dynamics with a remarkable performance guarantee.