Evaluation Of Aging State Of Transformer Oil Paper Insulation Based On Raman Spectroscopy

As a common power equipment,transformer plays a key role in the power system.The safe and stable operation of transformer is related to the safety and stability of the whole power grid.In order to ensure the long-term stable operation of the power grid,it is necessary to evaluate the aging state of the transformer regularly to ensure that the equipment is in good working condition.Long term research shows that the oil paper insulation system inside the transformer largely reflects the service life of the transformer.It is of great significance to study and develop the evaluation method of aging degree of oil paper insulation for the safe use of transformers,and provide theoretical guarantee for the continuous operation of power grid.The internal environment of transformer is characterized by high temperature corrosion,narrow space and isolation from the outside.The judgment of the aging degree of transformer is mostly based on the observation of the aging degree of external devices and empirical evaluation,which needs an accurate and effective evaluation method.Laser Raman technology has obvious advantages in detecting the aging of oil paper insulation of transformer,which is convenient and fast.The spectrogram of insulating oil contains rich information of transformer,and does not need to screen the characteristic quantity of various oils separately.It is a new research direction of transformer aging evaluation.In previous studies,Raman aging model can not well judge the aging state of different oil paper ratio insulating oil.In this paper,the related research is carried out:(1)According to IEEE standard for accelerated aging test of transformer oil,several samples of oil paper insulation at different oil paper ratios at 130 ? were prepared.The polymerization degree of sample insulating paper with aging time was analyzed.The polymerization degree of samples with different oil paper ratio was compared and the aging state was divided according to the polymerization degree of samples.(2)The optimized scheme of Raman detection platform for different oil paper ratio samples was studied.By comparing the data performance of Raman spectrum of samples under different parameters,the instrument characteristic parameters which can show the difference between different samples to the greatest extent are selected.Under the condition of 250 MW laser power and 0.5s integration time,the information of Raman spectrum data is complete and clear,which is convenient for further study.(3)A multidimensional scaling method is proposed to reduce the dimension of Raman data and extract the characteristic components of Raman data.Compared with the previous Raman spectrum feature extraction method,this method has more effective feature extraction for different oil paper than samples.The results show that the cumulative contribution rate of four characteristic components extracted from a single oil paper ratio sample is more than 95%;the cumulative contribution rate of five characteristic components extracted from a mixed oil paper ratio sample is 95.32%.The results of feature extraction of different oil paper ratio samples by multidimensional scaling method are similar,which ensures the reliability and stability of feature components.The potential relationship between spectral feature components and traditional aging feature quantity is analyzed.(4)According to the characteristics of the obtained characteristic quantity,the improved BP neural network and the improved KNN aging evaluation model are constructed.The data results show that the accuracy of single oil paper ratio sample in the improved BP neural network model is higher than 92%,but it has obvious bias for S1 and S4 aging stage;different oil paper ratio samples are not suitable for improving BP neural network model;the improved KNN algorithm is suitable for different groups of test samples The accuracy of evaluation is relatively stable,the accuracy of single oil paper ratio sample is higher than 92%,and the accuracy of different oil paper ratio samples is 91.5%,which is suitable for the Raman spectrum aging evaluation of transformer oil.Finally,the comparison between the traditional aging feature detection and the evaluation results of the model is carried out to verify the feasibility,reliability and accuracy of the Raman aging evaluation model.