Study On Programmable Device Application In PD Online Monitoring System

Nowadays partial discharge (PD) On-line monitoring has becoming a hot focus in research on safe operation of large electrical equipment, because it can supervise their working condition and also provide reliable reference for accidental diagnose. However, there are two defects in present PD online monitoring scheme: 1.since the present scheme includes data sampling, preprocessing, remote transmission, post processing and so on, system will be complex and slow echo; 2.since the sampling frequency of PD signal is higher and higher, which brings mass data, it will make great press to data processing, store and transmission by the traditional embedded equipment. Because of these defects, author propose a novel PD Online Monitoring scheme based on FPGA, which integrates the denoising, data compression and pattern recognition into one module, therefore the scheme can reduce the data remote transmission and post processing procedure, even improve the real-time performance.This thesis focused on the novel scheme based on FPGA, which include four main contents as follows:1. Study of the de-noise processing and data compression by using wavelet package soft threshold algorithm: study of the selection on the best wavelet package bases with cost function, study of the selection on the best depth to wavelet package by the correlativity analysis, the result of simulation show that the algorithm is effective.2. Study of the implementation of the wavelet package transform on FPGA: propose a pipelined and modularized design scheme by using the DSP Builder as development platform; propose a novel test method by using text reading and writing program in VHDL language to simulate the FPGA system; the simulation result prove that the FPGA system is precise and with a high real-time performance.3. Study of the selection of the best feature combination by using Gene Algorithm (GA): since the confine from field equipment and the demand of the recognition rate, present a new evaluation standard to evaluate the feature, base on which, the best combination is found with GA, the simulation show that the recognition rate to the best combination is higher and even with a low error.4. Study of the application of RBF ANN in the PD pattern recognition field and the design of RBF ANN on FPGA: optimize the RBF algortithm and parameter to hasten the recognition rate of the RBF; build up the RBF hardware architecture which consisted of feedforward block, feedback block and modification block, study of the design of these three block independently; simulation proved that the feedforward block could work as off-line-learning RBF ANN, and it can recognize the pattern of PD correctly.The novel PD Online Monitoring Scheme, which is provided in this paper, would overcome two defects of the traditional project, and has a good future in industrial application. Even more, the wavelet package transform system based on FPGA is also with application value in high speed digital signal processing field, the best feature combination and RBF ANN also have signification in other pattern recognition field.