Research On Temperature Prediction And Fault Early Warning Of Distribution Transformer Bushings

As a key component of distribution transformers,the maintenance and management of bushings play an important role in the safe and stable operation of distribution transformers.Due to the large number and wide distribution of distribution transformers,the traditional manual periodic inspection method is difficult to detect bushing faults in time.The abnormal temperature rise of the bushing can reflect its operating state.In order to prevent the power supply interruption and economic loss caused by the shutdown of the distribution transformer caused by the bushing failure,this paper combines the information collected by the monitoring platform to carry out the research on the temperature prediction and fault early warning of the distribution transformer bushing.Aiming at the problem that the accuracy of traditional prediction methods is limited by the distribution law of the sequence itself,A temperature prediction model of distribution transformer bushing based on variational mode decomposition(VMD)and gated recurrent unit(GRU)neural network is established.First,the naive K-nearest neighbor method is used to preprocess the data,and VMD is used to decompose the bushing temperature series into relatively stable sub-sequence components to reduce the influence of different trend information on the prediction model performance;Finally,each sub-sequence are superimposed to obtain the temperature prediction value of bushing.The calculation example results show that,compared with the traditional method,the built model can fully explore the characteristics of the bushing temperature series,more effectively track the changing trend of the bushing temperature,and has better prediction performance.Considering that bushing temperature is affected by multiple factors in complex electromagnetic environment,a prediction model of bushing temperature based on improved PSO-GRU fusion multi-source data is established.First,the heat flow analysis is carried out on the bushing,the multi-source information is fused into the prediction model,and the principal component analysis method is used to reduce the input dimension;Then,the traditional particle swarm optimization algorithm is improved by introducing time-varying learning factor,dynamic inertia weight and dynamic step,and the improved PSO is used to optimize the hyperparameters of GRU.Finally,a distribution transformer bushing temperature prediction model based on multi-source information fusion is established.Compared with the VMD-GRU prediction model and the traditional PSO-GRU prediction model,the calculation example results show that the average relative error of the model is reduced by 36.39%and 21.43% respectively.Combined with the above prediction model,a fault early warning threshold method based on Gaussian distribution theory is proposed.The statistical Gaussian distribution theory is used to fit the predicted residual value,the hierarchical early warning threshold is determined by law,and the hierarchical early warning emergency strategy is formulated to realize the transformation from passive maintenance after failure to active preventive maintenance.The actual case verifies that the early warning model has high sensitivity to judge the operation state of distribution transformer bushing.It can send early warning information in the early stage of fault,reserve sufficient maintenance time for staff,and avoid safety accidents and economic losses caused by bushing fault deterioration.This method does not need to train and mine the prior knowledge of fault data,and have certain engineering practicability.