The invention relates to the technical field of autonomous underwater robot fault identification and fault-tolerant control, and specifically relates to an autonomous underwater robot fault identification method based on wavelet energy. The method comprises the following steps: employing a multi-layer wavelet decomposition method to achieve the data decomposition of the sensor and controller of an autonomous underwater robot; employing a method of fault feature extraction to achieve the extraction of fault features of an original signal, a wavelet detail coefficient and a wavelet approximation coefficient; and employing a correlation coefficient method to achieve the fault identification of a to-be-detected fault signal of the autonomous underwater robot. The method provided by the invention effectively solves the problems that an AUV sensor and the controller are affected by external interference and the precision of fault identification is lower. The redundancy description of a fault of an AUV propeller is obtained through the multi-band characteristics of multi-layer wavelet decomposition. Through multi-band fault information, the fault features is extracted at the same time, and a fault feature matrix is built up, thereby improving the identification precision of an AUV fault, and providing accurate fault information for a fault-tolerant controller.