Study On Inter-turn Short Circuits Detect In Rotor Winding Of Generator Based On DSP

In this paper the inter-turn short circuits detect system in rotor winding of generator based on DSP is studied. The detect system include four departments, Data Acquisition, transmission control, Algorthm analysis of the data, implement the matlab algorithm.Based on analysis the mechanism of the inter-turn short circuits in rotor winding of generator are studied , Determine the air-gap electromotive force is the information that is need to collect when detect inter-turn short circuits. Changes in air-gap electromotive force caused by air-gap magnetic field, therefore, take advantage of the differential detection-coil to collect the electromotive force signal between stator and rotor air-gap. Determine the location of inter-turn short circuits faults according to the electromotive force transient signal in the coil.Collect analog signal by the system which hardware is designed based on TMS320C6203B; A/D conversion module ADS5422 as the foundation in order to convert the analog signals; Use CY7C68013A as the data transfer control chip to realize data transport by USB mode. Use the 32bit expansion bus of TMS320C6203B connect the FIFO and read data form FIFO; Use the timer output signal of TMS320C6203B to control the sampling frequency of AD; Use EMIF of TMS320C6203B connect Flash program unit and USB interface device; In addition, also include Voltage conversion circuit, clock circuit, reset circuit.Based on data acquisition foundation, inter-turn short circuit fault detection system in rotor winding use algorithms to analyze and treatment data that was collected by hardware. Because the Fourier transform separates the signal only in Time Domain, through the spectrogram, we only discover frequency components, can not refect the fault character information. Utilizing wavelet transform distinguish the distortion point in signals, however, the fault position is not accurate enough from the wavelet transform.so the algorithm of wavelet packet was use to extract energy characteristic information,which is handled by normalized as the input data of the neural network.through training the neural network obtain the perfect mechanism of training and testing. The collection data handled by normalized as the test data input neural network. Finally get the accurate location of the fault notch. Experiments show the efficiency and accuracy of this method.