254results about How to "Implement fault detection" patented technology

The invention discloses a winding fault locating method for a power transformer based on fusion of the deep convolutional neural network and visual identification. The method comprises the steps of 1)enabling a power transformer to be equivalent to a two-port network, establishing a winding equivalent circuit, and calculating a transfer function of the winding equivalent circuit; 2) for a circuitwith normal circuit parameters, setting a sine wave excitation source with frequency change at one end of a power transformer winding, and obtaining an amplitude-frequency characteristic curve of each section of winding nodes in the normal state; 3) carrying out sweep frequency response analysis on the circuit in various fault states to extract amplitude-frequency characteristics, and calculatingto obtain the amplitude-frequency characteristics of each section of winding node; 4) establishing a characteristic matrix for the acquired amplitude-frequency characteristics; 5) carrying out frequency sweep response analysis on the to-be-diagnosed power transformer to form a characteristic matrix, (6) converting the characteristic matrix into an image, taking simulation and historical detectiondata as a training set, taking fault types and positions as labels, and inputting the training set and the labels into the deep convolutional neural network for training, and (7) carrying out fault classification and location on the to-be-diagnosed transformer.