Feature Extraction Of UHF PD Signals And FPGA Implementation

When the insulation failure of high-voltage electrical equipment occurs,it is often accompanied by partial discharge(PD).If it is not prevented,it will lead to short circuit and other serious phenomena,which will harm the safe operation of power equipment and cause huge safety hidden danger.In order to detect the insulation deterioration in high voltage electrical equipment in advance,it is necessary to monitor the insulation equipment.Ultra High Frequency(UHF)sensor has the advantages of wide reception range,strong anti-interference ability and non-contact measurement,etc.,which is widely used in the field of monitoring.There are various kinds of noises in the monitoring site,and the noise will drown the PD signal in it,which is easy to cause misjudgment.To solve this problem,the feature extraction and FPGA implementation in the UHF PD signal monitoring system are deeply studied in this thesis,which provides a practical scheme for the PD monitoring system of high-voltage power equipment.There are a lot of white noise,pulse noise and random noise in PD monitoring field.For the pulse signal within the frequency range of the UHF sensor,the pulse signal is extracted from the white noise.The appropriate pulse features are selected,and the different types of pulse signals are separated by the clustering algorithm,so that the PD pulse signals can be separated from various noises and multi-source PD signals.In order to extract partial discharge pulse signals from white noise,this thesis proposes a sliding window white noise suppression method based on lifting wavelet transform and signal singularity detection to extract pulse signals from white noise.The proposed pulse extraction algorithm was compared with the existing noise reduction algorithm through simulation from two aspects of Root Mean Square Error(RMSE)and Normalized Correlation Coefficient(NCC)to verify the effectiveness of the proposed method.In order to realize the separation of different pulse signals,the characteristics of PD waveform were analyzed.Several pulse signal characteristic parameter points with waveform statistical characteristics are selected,and the two dimensional waveform distribution parameters are finally selected by comparing the characteristic parameters from cohesion and separation.The proposed algorithm and the selected characteristic parameters are implemented on FPGA.In order to make full use of the flexibility of FPGA and the advantages of parallel processing,pipeline and state machine are designed to improve the running speed of the algorithm in FPGA,so that the hardware system can work in real time and effectively.In this thesis,by simulating the feasibility and accuracy of the test system of the UHF signal generator and the artificial partial discharge defect generation device,the experimental results show that the average misjudgment rate less than 6% of this system for the classifying different types of pulse signal test.The proposed algorithm and the characteristics of the selected parameters after implementation in FPGA can separate the different types of pulse signal effectively.It has important reference value to the on-line monitoring system of partial discharge.