| 李占利,邢金莎,靳红梅,李洪安,张蕴.基于EMD和时序注意力机制的明渠流量预测模型[J].高技术通讯(中文),2022,32(2):122~130 |
| 基于EMD和时序注意力机制的明渠流量预测模型 |
| Open channel flow prediction model based on EMD and temporal attention mechanism |
| |
| DOI：10.3772/j.issn.1002-0470.2022.02.002 |
| 中文关键词: 煤矿； 明渠流量； 经验模态分解(EMD)； 长短时记忆网络(LSTM)； 注意力机制 |
| 英文关键词: coal mine, open channel flow, empirical mode decomposition (EMD), long short-term memory (LSTM), attention mechanism |
| 基金项目: |
|
| 摘要点击次数: 1254 |
| 全文下载次数: 801 |
| 中文摘要: |
| 为了预防煤矿水害事故的发生，本文提出将经验模态分解(EMD)算法与时序注意力机制（TA LSTM）结合的明渠流量预测模型，通过对明渠流量的实时预测来反映矿井涌水量的变化情况。模型首先通过EMD将明渠流量分解为多维子分量，充分提取明渠流量本身的波动特征和趋势特征；然后以长短时记忆网络(LSTM)为基础，融入注意力机制增强历史时间点对当前时刻的信息表达，构造时序注意力机制模型；最后通过该模型分别训练学习EMD分解后各分量的时序规律并进行预测，将各分量预测结果融合得到最终的明渠流量预测值。将此模型与现有其他模型进行了对比实验，其均方根误差和平均绝对百分比误差均小于其他模型。该研究为进一步预测矿井明渠流量提供了有效依据。 |
| 英文摘要: |
| In order to prevent the occurrence of water disaster in coal mines, an open channel flow prediction model based on empirical mode decomposition (EMD) algorithm and temporal attention based on long short-term memory （TA-LSTM） model is proposed. Through the real-time predictions of open channel flow, the change of mine water inflow can be reflected. Firstly, the open channel flow is decomposed into multi-dimensional sub-components by EMD, and the fluctuation characteristics and trend characteristics of open channel flow are fully extracted. Then, based on long short-term memory network (LSTM), the TA-LSTM model is introduced by integrating attention mechanism to enhance the information expression of historical time points to the current moment. Finally, the time series rules of each component after EMD decomposition are trained and learned by the proposed model, and the final prediction value of open channel flow is obtained by fusing the prediction results of each component. Comparative experiments show that the root mean squared error and mean absolute percentage error of this model are smaller than those of other models. This study provides an effective basis for further prediction of mine open channel flow. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
