Research On Prediction Of Power Transformer Operation State Based On GRU And SVM Model

Power transformers are extremely critical power equipments that undertake transformation and distribution work in the power system,which play a particularly important role in prediction and diagnosis of power transformer state.This article will use dissolved gas in oil analysis as a basis for determining future operational state of the power transformer through a combination of advanced intelligent algorithms running state of the transformer prediction and diagnosis.It is of great significance to the healthy and stable operation of the entire power system,which can provide new ideas for the realization of smart grids.In the aspect of prediction,this article uses gated recurrent unit neural network.In this paper,the input sequence processing and optimization parameters are improved,and a method for predicting future gas data based on historical gas content data is proposed.This method uses complementary set empirical mode decomposition to process the gas content data sequence to reduce the interference of data noise,nonlinearity and other factors on the prediction model.At the same time,the gradient optimization algorithm Adam is used to optimize the hyperparameters of the neural network model to obtain optimal prediction model.Through the prediction method proposed in this paper,experiments are carried out on seven characteristic gases,and the results show that this method can accurately predict the future content of characteristic gases and provide important reference value for transformer state diagnosis.In terms of diagnosis,this paper uses the "no coding ratio" of the seven characteristic gases as the characteristic quantity and presents an improved gray wolf optimization algorithm support vector machine transformer operating state diagnosis method.This method is mainly based on statistical principles combined with the principle of structural risk minimization to establish a diagnostic model.The kernel principal component analysis method is used to reduce the dimensionality and retain the non-linear information of the original features,which can better mine the feature indicators in the data.The improved gray wolf optimization algorithm is used to optimize the kernel parameters,which improves the convergence speed and ergodicity to a certain extent,avoids to fall into the local optimum during the training process,and improves the diagnosis effect of the support vector machine model.The experimental results show that the proposed method has good accuracy and make a good prediction of the changing trend of the transformer operation status in the future.Finally,an intelligent operating state prediction system based on STM32F107 was designed in this acticle.Through analysis of examples,the proposed methods are used to effectively verify the overheating faults and discharge faults,which have high application value in ensuring the safe and stable operation of the transformer and the power system.