摘要
针对传统人工经验的煤气化炉剩余寿命预测方法存在精度不足、实时性差的问题,提出一种变分模态分解-改进蜣螂算法-三重注意力双向长短期记忆网络(Variational Mode Decomposition-Improved Dung Beetle Optimizer-Triplet Attention Bidirectional Long Short-Term Memory, VMD-IDBO-TABiLSTM)的煤气化炉剩余寿命预测模型.利用VMD充分挖掘数据中的隐藏时序特征,通过结合人工蜂群算法与动态权重系数对IDBO算法进行全局策略的超参数优化,采用三重注意力机制的TABiLSTM学习复杂的时间依赖关系.以机械结构寿命损失分数为准则,对模型的可靠性进行分析,结果表明:在变工况的条件下,VMD-IDBO-TABiLSTM模型对3台煤气化炉的剩余寿命预测值与实际值的拟合度分别为93.88%, 91.75%, 96.72%,显示出模型在泛化能力和鲁棒性方面的显著优势.
In response to the issues of insufficient accuracy and poor real-time performance in traditional artificial experience-based prediction methods for the remaining service life of gasifiers,this study proposes a novel prediction model for the remaining service life of gasifiers,which integrates Variational Mode Decomposition(VMD),Improved Dung Beetle Optimizer(IDBO),and Triplet Attention Bidirectional Long Short-Term Memory(TABiLSTM).The model first employs VMD to thoroughly extract hidden temporal features from the data,followed by a global strategic hyperparameter optimization of the IDBO algorithm through the integration of artificial bee colony algorithms and dynamic weight coefficients.Furthermore,the model utilizes TABiLSTM with triple attention mechanisms to learn complex temporal dependencies.Using the mechanical structure service life loss score as the criterion,comparative analysis of model reliability indicates that the VMD-IDBO-TABiLSTM model achieves a high degree of fit with the actual values in varying conditions for the remaining service life prediction of three gasifiers,with fit rates of 93.88%,91.75%,and 96.72%,respectively,demonstrating the model's significant advantages in terms of generalization and robustness.
作者
高林
李瑶
李捍东
熊国江
GAO Lin;LI Yao;LI Handong;XIONG Guojiang(School of Electrical Engineering,Guizhou University,Guiyang 550025,China;State Grid Tianfu New Area Power Supply Company,Chengdu 610000,China)
出处
《湖南科技大学学报(自然科学版)》
北大核心
2025年第2期80-91,共12页
Journal of Hunan University of Science And Technology:Natural Science Edition
关键词
煤气化炉
剩余寿命预测
变分模态分解
深度学习
coal gasifier
remaining life prediction
variational modal decomposition
deep learning
