Detection And Signal Analysis Of Characteristic Gases In Transformer Oil Based On Raman Spectroscopy

At present, sustainable development of ultra-high voltage transmissiontechnology makes the operational environment of power grid hub status of powertransformer become more complex. The effect of operational state detection and faultdiagnosis becomes more and more important. DGA (Dissolved Gas Analysis) is bestway to monitor the operation status of transformer and to diagnose potential fault.However, DGA has disadvantage such as poor accuracy, low sensitivity and high costof operation and maintenance. For the efficient implementation of DGA, the newmethod must be taken into the field of gas detection.Raman spectroscopy is used in detection of gases dissolved in transformer oil.And this method requires only one frequency light source to measure the qualitativeidentification of the main characteristics gas in oil. It is a unique advantage. WhenRaman spectroscopy is used in quantitative analysis of gases dissolved in transformeroil, the technical characteristics are similar to spectroscopy. In summary, it is anexcellent detection technology.This paper summarizes the influence factors of operation and fault of powertransformer on the transformer oil. Firstly, the paper studied on the process of gasesproduction in oil and the method to separate the gases. At the optimum temperature of50-60℃, hollow tube type high molecular polymer osmotic membrane can separatethe gases and oil.Secondly, the paper explores the light scattering principle based onthe summary and research of the current method for the detection of characteristicgases in transformer oil. It leads to the final design——the detection system ofcharacteristic gases in transformer oil based on Raman spectrum.Because of interference factors during the process, the pretreatment of spectradata have been revealed for Raman spectroscopy data acquisition, and this paper getsthe real spectral information which has low interference.This paper researches Raman spectrum and gets feature extraction of Ramanspectrum by data fitting, such as height, Raman shift and line width. Compare withthe result of data fitting and experimental data shows data fitting is the best method toget feature extraction of Raman spectrum.This paper also researches the role of feature extraction for qualitative analysisand quantitative analysis the Raman spectrum. Through a series of characterization of the gases sample detection and quantitative determination of concentration limitjudgment, and determine the value and advantage of Raman spectroscopy detectionmethod in monitoring of power transformer gas dissolved oil. Mainly embodied in:widely applicable, no light selective, high sensitivity.