The discontinuities are widely distributed in the rock mass and are difficult to analyze one by one. Therefore, it is of great engineering value and scientific significance to carry out the dominant grouping of the discontinuities. The existing research methods are sensitive to the initial information, the grouping results are not reliable, and it is difficult to accurately group the discontinuities with similar discontinuity orientations. To address these problems, this paper proposes a method for grouping discontinuities in rock masses based on the pelican optimization algorithm (POA). The POA algorithm is used to globally find the optimal initial clustering center and combine with the fuzzy C-mean algorithm (FCM) to fully group the discontinuity orientations. A Monte Carlo simulation technique is used to generate discontinuity orientations that conform to the Fisher distribution. Based on the orthogonal design, using the recognition error rate as the index, the new algorithm was compared with the traditional FCM algorithm, and the variation regulation of grouping accuracy was investigated under different number of discontinuities, number of discontinuities groups, clustering centers and dispersion degree. The results indicate that the cluster centers have a significant impact on grouping accuracy, and the proposed method is capable of effectively grouping structural data with unclear boundary of geological features, thereby improving the accuracy and reliability of the grouping results. Based on the data of the slope structural plane of a reservoir in Dalian, it is grouped and processed to verify the engineering practicability of the new method. This study can provide a basis for the three-dimensional network computer simulation of structural plane and the stability analysis of rock mass engineering.