The invention discloses an optical access network ODN intelligent clustering planning method for sparse user distribution, and belongs to the technical field of communication. The method specifically comprises the following steps of: firstly, for a certain planning area, randomly generating position coordinates of k initial centroids as second-level light splitting points on a wiring segment of an ODN by taking the geographic position of a user in the area as input; then, constructing a cost function T, and optimizing the number k of the initial centroids and position coordinates; thirdly, determining the position coordinates of first-level light splitting points and optimizing the position coordinates; according to the optimized positions of the second-level light splitting points and the first-level light splitting points, calculating the reserved length of an optical fiber, and completing the planning of the wiring segment; for feeder line segment planning of the ODN , determining a weight connectivity graph through a connectivity relationship between field areas, and searching for a minimum spanning tree by adopting a minimum spanning tree algorithm; finally, completing optical access network ODN planning of sparse users according to planning results of the wiring segment and the feeder line segment. According to the method, the top and the bottom are combined, and the optical fiber planning cost is reduced while optimal planning is realized.