Transformer Branch Parameter Identification Based on Macaron Sequence Decomposition
WANG Linpeng
SONG Gongfei
WANG Menglong
Abstract：Parameter identification plays an important role in power system.As a long-term prediction problem of time series,in order to explain the complex time pattern,a Transformer branch parameter identification method based on Macaron sequence de-composition is proposed.Among them,the sequence decomposition module is regarded as the internal block of the deep model.Dur-ing the whole prediction process,the hidden sequence is gradually decomposed,including the past sequence and the intermediate results of the prediction.At the same time,the Macaron network is used to replace the original feedforward layer in Transformer with two half-step feedforward layers,and the self-attention module and the sequence decomposition module are placed between them.The experimental results show that the proposed algorithm has higher prediction accuracy and is significantly better than other ma-chine learning algorithms and deep learning algorithms.
Keywords：Macaron networksequence decompositionself-attention mechanismparameter identificationdeep learning
Publication Date：2025-10-20
Online Publishing Date：2026-01-16(First online date of this platform, not the publication date of the document)
Pages：7( 2677-2682,2738 )