Aiming at the problem that the accuracy of hydroelectric generating unit condition monitoring sensor readings and its data fluctuations affect the accuracy and complexity of unit fault diagnosis，a method based on mahalanobis distance-data fusion is proposed，mainly to study and identify the operating state of hydroelectric generating units.This method uses the mahalanobis distance as the calculation basis，fuse the multi-sensor data of the hydropower unit in the continuous time period，and analyze the fluctuation of the index value according to the obtained calculation result value and the actual operation condition of the unit，to achieve the purpose of identifying the normal and faulty operation state of the unit.This method is compared with the data fusion method based on single sensor data and euclidean distance.The results show that the method based on Mahalanobis distance is more excellent in terms of timeliness and accuracy，can more intuitively，accurately and efficiently identify the operating state of the unit，and has significant engineering significance for the actual operation of the unit.