Research Of Acoustic Method For Insulation Detection In High Voltage Switchgear

In this thesis, partial discharge causing insulation fault is researched and the executing process for ultrasonic detection of partial discharge is elaborated,which includes structural design and parameters selection for hardware detection circuit and software program and data processing method.Based on analyzing technical features of ultrasonic signals from partial discharge for switchgear, a detection circuit is designed for ultrasonic signals.The design process and its functions are explained in detail. The Hilbert-Huang Transform (HHT) theory is adopted for data processing, including data decomposition and feature selection. An experiment platform based on simulated switchgear is built for practical analysis.The technical features of ultrasonic signals from partial discharge for switchgear are analyzed.The partial discharge always occurs during the half period of power frequency, which is near the peak point for resistive circuit and between the zero point and peak point for capacitive circuit.The duration time of discharge becomes longer as the applied voltage increases.According to the shape of discharge area,the discharge may occur in every half period or just in the positive half one,which leads to the characteristic of periodic outbreak by power frequency.The outbreak can cause a series of high frequency oscillatory waves with random frequency,duration time and amplitude.A detection circuit is designed for ultrasonic signals, including power supply circuit,ultrasonic sensor, preposition signal process circuit and signal conditioning circuit.The center frequency of ultrasonic sensor is set to 40kHz,and 20-60kHz for bandwidth.Amplification factor for preposition amplifier circuit is 80dB.The cut-off frequency of high-pass filter is 20kHz.The pass-band for digital band-pass filter is 20 kHz-60kHz.The minimum attenuation for stop-band is 80dB and the maximum attenuation for pass-band is 1 dB.Some typical time-frequency analysis methods are briefly introduced and their pros and cons are compared,which shows that the HHT performs very well for nonlinear and non-stationary data.The conceptions for Hilbert-Huang Transform (HT),Instantaneous Frequency (IF) and Intrinsic Mode Function (IMF) are introduced in detail.The features and procedures of Empirical Mode Decomposition (EMD) and its improved method (EEMD) are presented.The physical sense of HHSP is revealed.The signals are supposed to be processed by band-pass filtering and EEMD before using HT.An experimental platform is built and HHT method is used for data processing and features selection. Obvious difference between discharge signal and noise signal spectrums has been found.The energy is very small for frequency spectrum of noise signal in the vicinity of 40kHz while it is the contrary case for discharge signal.A guideline based on HHT is given for feature selection of ultrasonic signal,i.e. selecting IMF1 as the feature.A preliminary estimation of time delay based on selected feature is tested and it is showed that the results are basically accurate.