摘要
为了解决瓦斯浓度预测使用的单一数据在预测中影响还不够深入的问题,提出基于LSTM神经网络的多源数据融合瓦斯浓度预测模型。模型将上隅角瓦斯浓度、采煤机速度、工作面吨煤瓦斯涌出量等不同数据融合作为输入层参数,使用Adam优化算法更新LSTM网络层参数,利用Attention机制突出关键影响瓦斯浓度的因素,开展多源数据融合的瓦斯浓度预测,结合某矿1008工作面的实际数据,分析不同数据在瓦斯浓度预测中的作用。研究结果表明:单变量下的Attention-aLSTM预测效果相比LSTM提升14.2%;多源数据融合下的Attention-aLSTM相比自身提升了5%。
In order to solve the problem that the single data used in gas concentration prediction has not had enough influence in the prediction,a prediction model of gas concentration with multi-source data fusion based on LSTM network was proposed.Different data such as gas concentration at upper corner,shearer speed and gas emission amount per ton of coal in the working face were fused as the input layer parameters,and the Adam optimization algorithm was used to update the LSTM network layer parameters.The Attention mechanism was applied to highlight the key factors affecting gas concentration,and the gas concentration prediction based on multi-source data fusion was carried out.The role of different data in gas concentration prediction was analyzed combined with the actual data of 1008 working face in a mine.The results showed that the prediction effect of Attention-aLSTM under single variable was 14.2%higher than that of LSTM,and the prediction effect of Attention-aLSTM under multi-source data fusion was 5%higher than itself.
作者
谢谦
董立红
吴雪菲
XIE Qian;DONG Lihong;WU Xuefei(Xi’an Research Institute,China Coal Technology&Engineering Group Corp,Xi’an Shaanxi 710077,China;College of Computer Science&Technology,Xi’an University of Science and Technology,Xi’an Shaanxi 710054,China;College of Energy Science and Engineering,Xi’an University of Science and Technology,Xi’an Shaanxi 710054,China)
出处
《中国安全生产科学技术》
CAS
CSCD
北大核心
2022年第11期71-76,共6页
Journal of Safety Science and Technology
基金
国家重点研发计划项目(2017YFC0804100)
陕西省自然科学基础研究计划项目(2019JLM-11)
天地科技股份有限公司科技创新创业资金专项项目(2020-TD-ZD002)。
关键词
多源数据融合
瓦斯浓度预测
长短期记忆神经网络
注意力机制
multi-source data fusion
gas concentration prediction
long-term and short-term memory neural network
Attention mechanism
