热力发电2025，Vol.54Issue(12)：67-75,9.DOI:10.19666/j.rlfd.202503051
基于FFCM-MHDA-iTransformer的汽轮发电机定子铁芯温度预测模型
Prediction model for stator core temperature of turbo generator based on FFCM-MHDA-iTransformer
摘要
Abstract
Abnormal stator core temperatures in generators can lead to serious issues such as aging of insulating materials and winding shorts,thereby affecting the overall performance and lifespan of the generator.This study presents a stator core temperature prediction model for turbo generators based on FFCM-MHDA-iTransformer.It leverages an improved Transformer architecture,namely the inverted Transformer(iTransformer)model,which adopts an inverted time-series encoding approach to address the limitations of the standard Transformer in handling multivariate variable correlations.The model employs fused Fourier convolution mixer(FFCM)to enhance and extract local features from time-series data.Furthermore,the model replaces conventional self-attention with multi-head differential attention(MHDA),effectively reducing attention noise and directing the model's focus towards critical information.After training and validation,the proposed model demonstrates higher prediction accuracy compared to other mainstream prediction models.It facilitates timely detection of potential faults,preventing shutdowns for maintenance,and holds significant application value for ensuring stable operation of turbo generators.This approach effectively enhances the accuracy and practicality of temperature prediction technology.关键词
汽轮发电机/定子铁芯温度预测/iTransformer模型/融合傅里叶卷积混合器/多头差分注意力机制Key words
turbo generator/stator core temperature prediction/iTransformer model/fused Fourier convolution mixer/multi-head differential attention引用本文复制引用
HOU Yifan,YUAN Hejin,SUN Fan..基于FFCM-MHDA-iTransformer的汽轮发电机定子铁芯温度预测模型[J].热力发电,2025,54(12):67-75,9.基金项目
宁夏自然科学基金项目(2024AAC03758)Natural Science Foundation of Ningxia(2024AAC03758) （2024AAC03758）
