Transformer surface temperature based on VMD-BiGRU-MHA prediction method research
In order to improve the accuracy of transformer surface temperature prediction,a prediction method based on Variational Modal Decomposition,Bi-Directional Gated Recurrent Unit and Multi-Head Attention is proposed.In order to increase prediction accuracy,the VMD-BiGRU-MHA model first uses VMD to break down the original data into multiple subsequences and produce more stable individual components.These subsequences are then fed into BiGRU for training.Lastly,MHA is added to mine the time-series long-distance data features of the transformer surface temperature.Comparing the model suggested in this research to BiGRU and VMD-BiGRU,the findings demonstrate that it has a smaller prediction error and a faster prediction speed.
万方数据
维普
