基于MHAGRU模型的重介选煤分选密度预测方法  

Method for Predicting Separation Density in Heavy-medium Coal Preparation Based on MHAGRU Model

作　　者：侯晓松 纪玉华 高奎 郭莹 于刚 鲁法明 HOU Xiaosong;JI Yuhua;GAO Kui;GUO Ying;YU Gang;LU Faming(Fucun Coal Industry Co.,Ltd.,Zaozhuang Mining(Group),Jining,Shandong 277600,China;Coal Washing and Dressing Center,Zaozhuang Mining(Group),Zaozhuang,Shandong 277000,China;School of Computer Science and Engineering,Shandong University of Science and Technology,Qingdao,Shandong 266590,China)

机构地区：[1]枣庄矿业(集团)付村煤业有限公司,山东济宁277600 [2]枣庄矿业(集团)煤炭洗选加工中心,山东枣庄277000 [3]山东科技大学计算机科学与工程学院,山东青岛266590

出　　处：《数学建模及其应用》2024年第4期40-46,共7页Mathematical Modeling and Its Applications

基　　金：山东省自然科学基金重大基础研究项目(ZR2024ZD22)。

摘　　要：重介选煤过程中的分选密度是影响选煤效率的关键因素之一,准确预测分选密度可帮助优化选煤工艺,提高产品品质.本文提出了一种结合多头注意力机制和门控循环神经网络的分选密度预测模型.传统的分选密度预测方法难以处理非线性、多维度和时间序列相关性的问题,而本文提出的模型通过引入多头注意力机制,能够有效捕捉不同时间步长和特征维度之间的依赖关系,从而提升模型的预测精度.同时,门控循环神经网络在处理长时间序列数据时具有较好的记忆能力,能够有效避免梯度消失问题.实验结果表明,与传统模型相比,基于多头注意力机制和门控循环神经网络的分选密度预测方法在预测精度和鲁棒性上均有显著提升,适用于实际生产中的分选密度预测任务.The separation density in the heavy medium coal preparation process is one of the key factors affecting coal preparation efficiency.Accurate prediction of separation density can help optimize the coal preparation process and improve product quality.This paper proposes a separation density prediction model that combines a multi-head attention mechanism with a Gated Recurrent Unit.Traditional separation density prediction methods struggle to handle the nonlinearity,multidimensionality,and time-series correlations inherent in the data.By introducing the multi-head attention mechanism,the proposed model can effectively capture dependencies across different time steps and feature dimensions,thereby enhancing prediction accuracy.Meanwhile,the gated recurrent unit′s strong memory capability for long-term time-series data helps avoid the vanishing gradient problem.Experimental results show that,compared with traditional models,the multi-head attention mechanism with a gated recurrent unit-based separation density prediction method significantly improves prediction accuracy and robustness,making it suitable for practical production scenarios in separation density prediction.

关 键 词：分选密度 时间序列 多头注意力 门控循环单元 

分 类 号：O29[理学—应用数学]