Transformer noise prediction method based on CEEMDAN-VMD-TCN

Abstract

Abstract To solve the problem of weak precision of traditional prediction methods in transformer noise prediction, a novel prediction approach by CEEMDAN-VMD-TCN is presented in this paper. Firstly, the noise sequence is preliminarily disaggregated by Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN). Then, the fuzzy entropy value of each modal component is calculated. The genetic algorithm is adopted to find the optimum parameter group of Variational Mode Decomposition (VMD), with which the components with higher fuzzy entropy values are decomposed again by VMD. Finally, the decomposed components and related feature sequences are input into the Temporal Convolutional Network (TCN) for prediction. The prediction results are superimposed and reconstituted to yield the end results. The results show that the proposed method has greater prediction accuracies compared with other prediction ways.