Given the problems of multiple types, strong feature correlation and the high recognition error rate of hybrid Power Quality (PQ) disturbances, a hybrid PQ disturbance recognition method based on a chaos ensemble decision tree is proposed. First, from the IEEE standard, the common signal models of 7 kinds of single PQ disturbances and 16 kinds of hybrid PQ disturbances are obtained, and disturbance waveform samples are generated in batches. Then through S-transform time-frequency domain analysis, 9 features of disturbance in the time-frequency domain are designed and extracted according to the difference of these disturbances. Finally, taking advantage of the collective ability of ensemble learning and chaotic search, a chaotic ensemble decision tree is constructed, and the identification of hybrid PQ disturbances is effectively completed. Simulation experiments and a 142 field data test show that for 23 types of disturbances, the recognition accuracy of the proposed method is higher than that of basic decision trees, complex decision trees and weighted nearest neighbor method, and has good application prospects. This work is supported by the National Natural Science Foundation of China (No. 51577086), Jiangsu “Six Talents Peaks” Project (No. TD-XNY004), Scientific Research Major Project of Jiangsu Universities (No. 19KJA510012), and Science and Technology Project of State Grid Jiangsu Electric Power Co., Ltd. (No. J2021041).