中国电机工程学报2025，Vol.45Issue(24)：9644-9654,中插14,12.DOI:10.13334/j.0258-8013.pcsee.242238
基于高斯混合模型聚类和注意力机制的过热器壁温集成预测模型
Integrated Prediction Model for Superheater Wall Temperature Based on Gaussian Mixture Model Clustering and Attention Mechanism
摘要
Abstract
Aiming at the over-temperature problem that is prone to occur in the rear screen superheater piping in the boiler during the peaking and frequency adjustment process of coal power units,this paper proposes an integrated reheater wall temperature prediction model based on Gaussian mixture model(GMM)clustering and attention mechanism.First,the key feature variables are screened by maximum information coefficient(MIC)and correlation-based feature selection(CFS)algorithms Then,the dataset is divided into subsets corresponding to different operational conditions by using the GMM clustering algorithm.Independent prediction models are constructed by using the parallel architectural-temporal convolutional network-long short-term memory(PA-TCN-LSTM).Finally,an integrated prediction model is developed by combining the attention mechanism with the fusion of a single prediction model.Simulation results show that the integrated model he root mean squared error(RMSE)and mean absolute error(MAE)by 20%and 18%compared to single models,with a 66%improvement in generalization ability across different operational conditions.The average absolute error across all conditions is 1.44℃,meeting industrial precision requirements.This scheme effectively improves the adaptability of the prediction model to changes in working conditions,provides technical support for the safe and stable operation of coal power units under complex working conditions,and has important application value.关键词
集成模型/深度学习/壁温预测/注意力机制/后屏过热器Key words
integrated model/deep learning/wall temperature prediction/attention mechanism/rear screen superheater分类
能源科技引用本文复制引用
孙凯,郝勇生,华山,孙立..基于高斯混合模型聚类和注意力机制的过热器壁温集成预测模型[J].中国电机工程学报,2025,45(24):9644-9654,中插14,12.基金项目
国家自然科学基金项目(52276003).Project Supported by National Natural Science Foundation of China(52276003). （52276003）
