AbstractPower distribution systems are very vulnerable during hurricane events. Failure of power distribution systems could bring significant disruptions to the community’s daily activities. Sparse historical hurricane data are insufficient to establish hurricane risk models. Therefore, it has been challenging to evaluate the pole-wire system’s performance during hurricane events under strong winds. In the present study, a probabilistic framework integrating hurricane risk modeling and physics-based analysis is proposed to assess the reliability of the power distribution system subjected to hurricane winds. Based on historical hurricane data, hurricane tracks are simulated using a modified statistical method by matching the synthetic data with the statistical characteristics from historical hurricanes facilitated by a copula model. Using a novel statistical model that implements a machine learning (ML) algorithm hurricane intensities are predicted. A hurricane risk model is established using the synthetic hurricane data. Fragility curves for each pole are obtained by physics-based Monte Carlo simulations facilitated by ML-based regression models instead of the empirical fitting model in order to incorporate the most influential factors. A surrogate model trained by the ML algorithm is employed to obtain the system fragility curve with a low computational cost. Finally, the annual failure probability of the pole-wire system could be obtained by integrating the annual hurricane wind speed probability density and the pole-wire system fragility curve.