Research And Application Of Oil Chromatographic Data Prediction Method Based On WATL

As the key core equipment of power system,the safe and stable operation of transformer is of great significance.Oil chromatogram data is an important index to measure the operation state of power transformer.It is widely used in engineering practice to evaluate the operation state of transformer through the trend change of oil chromatogram data.In this paper,based on wavelet decomposition,attention mechanism,temporal convolutional network and long short term memory are fused by using encoder decoder principle,and a prediction method of oil chromatogram data based on watl is proposed.Firstly,aiming at the problems of random oscillation and complex time dependence of oil chromatographic data,this paper extracts low-frequency trend component and high-frequency fluctuation component from oil chromatographic sequence by wavelet decomposition,and uses time convolution network(TCN)as encoder to encode low-frequency component and high-frequency component respectively,At the same time,the coding vector of the last hidden layer of TCN is used as the input of the decoder’s long-term and short-term memory network.Secondly,attention mechanism is introduced in the prediction process to increase the weight of features that have a great impact on the current output,and enhance the information expression of key time points.The final output is obtained by wavelet reconstruction of the prediction results of components.The experimental results show that,compared with the long-term memory network,the average autoregressive model and other algorithms,under the condition of taking the calibration coefficient(R2),mean square error(MSE)and absolute error(MAE)as the evaluation indexes,the performance of the method in the three indexes is improved by 4.50%,1.10% and 13.37% on average.Finally,based on the simulation experiment of watl model,a visual oil chromatogram data prediction system is designed and developed.The system realizes the transformer condition evaluation through the trend analysis of oil chromatography data,and has been applied in a power enterprise.