New insights into wiener deconvolution and optimum filtering of space charge signal in pulsed electroacoustic system

Abstract

Insulation design of High Voltage Direct Current (HVDC) cables at high voltages is quite challenging. At higher electric fields space charge formation takes place in polymeric insulation, which results into distortion of the electric filed distribution that may lead to high electric stress, locally. The local stresses may further lead to local damage, degradation or initiate breakdown of insulation. Therefore, measurement and estimation of space charge is considered very important. There are many non-destructive techniques available in literature. Out of these, Pulsed Electroacoustic (PEA) is widely used. Although the system is popularly used by research community, it does not give charge profile directly due to limited frequency band of the transducer-amplifier system connected at ground electrode. Therefore it requires signal processing technique like deconvolution to compensate the effect of transducer-amplifier transfer characteristics and estimation of actual charge profile. From literature it seems that, apart from lot of researchon space charge and PEA method, research on deconvolution is scantily available. It was observed that, there is a need of investigating the deconvolution process on space charge signal, so that optimum space charge profile can be obtained. In the present work, the author thoroughly examines the existing iterative and numerical optimum parameter estimation methods like GCV method, L-curve method, normalized periodogram method and proposes simple, robust and novel “U-curve” method for optimum parameter estimation, for Wiener Deconvolution. The author, further, presents a comparison between proposed method with existing numerical method on experimentally obtained raw charge signal from Poly methyl methacrylate (PMMA) sample as well as low density polyethylene (LDPE) sample. The experiments for LDPE film were conducted at 25ºC on 100µm LDPE film samples under applied electric fields of 10kV/mm for one hour. In addition to space charge deconvolution, certain analytical insights into Wiener deconvolution propose optimum parameter estimation, explicitly. By using proposed method computation part required for optimum parameter estimation is eliminated completely. The results were verified experimentally for complex higher order PEA system as well as for a RC network and found to be accurate. In this thesistwo IEEE Transactions papers (on Instrumentation and Measurement, on Dielectrics and Electrical Insulation) are publishedanda few IEEE conference proceedings (CMD, CATCON, ISEIM, CONECCT, ISPCC).