| Studies have shown that winding deformation is the main cause of transformer faults,so it is very important to detect winding health in a timely and effective manner.In this paper,when the frequency response method is used to detect the health of the windings online,the response signal is submerged by noise and the amplitude of the target sinusoidal signal cannot be accurately obtained,and the related detection methods are studied.First,according to the characteristics of low signal-to-noise ratio of noisy signal,complex noise background and known frequency of target signal,a method to detect the amplitude of weak sinusoidal signal by distinguishing the boundary between chaos and periodic state of Duffing oscillator system is proposed.Research shows that:Duffing vibrator can effectively detect the amplitude of each frequency target signal under the interference of mixed noise,and the lowest signal-to-noise ratio range of its detection is-25.52~27.26dB.At the same time,in order to reduce the influence of the signal phase deviation on the amplitude detection in the actual winding deformation detection process,a method of detecting the signal phase first using a vibrator array composed of Duffing vibrators and then detecting the amplitude is proposed.Research shows that the method in this paper can effectively detect the amplitude of the target signal under mixed noise interference with a signal-to-noise ratio of about-26dB,but the detection error of direct amplitude detection(even when there is no noise interference)still exceeds the maximum error of the power industry Allowed standard(12.2%).Secondly,in order to solve the problem of low accuracy of Duffing oscillator amplitude detection caused by strong noise,a method of preprocessing the noisy signal using wavelet transform is proposed.This method reduces the noise power of the noisy signal and improves the amplitude detection accuracy.The research shows that:compared with the detection of a single Duffing vibrator,the detection error of the method in this paper is reduced by 2%~5%,and the minimum detection signal-to-noise ratio threshold of each frequency signal is widened by-2~4dB.Finally,the method proposed in this article is used to process the three sets of measured response signals of the online detection.The results show that the maximum error between the detected signal amplitude and the standard signal is 5.69%,which meets the requirements of the power industry standard,which proves that the method of this article has practical value in the detection of winding deformation. |