The invention discloses a power network fault early warning method based on a short-term random failure model of equipment, which comprises the following steps: 1, establishing a dynamic load calculation model of equipment evolving in time and space; 2, establishing a failure probability model of electric power equipment; 3, an accurate forecasting method of power outage being proposed for extremeweather conditions. In the manner, the invention relates to a power network fault early warning method based on a short-term stochastic failure model of equipment, based on the non-sequential Monte Carlo method, an accurate prediction method of distribution network fault and user outage is established, which realizes fault prediction of distribution network equipment and user outage prediction, and helps distribution network operation management develop from passive rush repair management to active preventive management, thus shortening rush repair time of distribution network equipment and reducing user outage probability and duration.