Abstract: This paper proposes a new multi-label database query method for combined power quality disturbances (PQDs) recognition and classification, aiming at the problems of the complexity of combined PQDs and the insufficient accuracy of current classification. This new method can be used to recognize combined PQDs more scientifically and accurately, which can provide powerful decision-making assistance for PQ management and disturbance event accountability. First, this method employs the proposed feature extraction method based on the tunable Q-factor wavelet transform (TQWT) and time-varying root mean square (RMS) to effectively extract the fundamental time domain features from the PQDs, which is an effective way to overcome the current difficulties of insufficient accuracy in extracting fundamental amplitude features. Next, the proposed frequency-domain characteristic curve segmentation method is used to extract the high-frequency characteristic curve of the PQDs effectively. Then, the fundamental frequency amplitude feature database and the high-frequency characteristic curve database are established. The multi-label database query with fast dynamic time warping (DTW) is used to classify the combined PQDs effectively. The simulation results show that the new method has the following advantages: It is hardly affected by the fundamental frequency deviations within the range specified in the GB/T standard, and it has not only good noise tolerance capability, but also high classification accuracy for 27 kinds of PQDs, including single, double, triple, and quadruple disturbances. Finally, its effectiveness is further verified by the actual disturbance data collected from the power grid.