The invention discloses an electric vehicle emergency power supply method based on a k-means algorithm. An electric vehicle emergency power supply model comprises multi-target optimization of miniaturization of important user power-failure loss cost, electric vehicle operation, scheduling and maintenance cost and the like, by taking the feature of random dispersion of electric vehicles into consideration, the electric vehicle emergency power supply clustering method based on a k-means algorithm, provided by the invention, has the functions of improving the active aggregation capability of the electric vehicles and reducing the aggregation time of the electric vehicles, can bring the mobile energy storage advantage of city electric vehicles into full play, reduces the quantity of other emergency power supply devices of a city, decreases power-failure loss of users, and improves power consumption reliability of the city.