摘要
矿压失衡引起的顶板事故是煤矿重大灾害之一,矿压的精准预测对保证煤层的安全开采具有重要意义。为提高矿压的预测精度,提出了一种基于堆叠LSTM的多源矿压预测模型。首先,采用灰色关联度对煤矿工作面多源矿压进行分析排序并进行数据预处理;其次,采用堆叠式网络结构,确定每一个LSTM层的隐藏节点数、迭代次数等参数;最后,采用Adam优化算法对模型进行优化,从而对工作面矿压进行预测。采用均方根误差作为评价指标对预测模型性能进行评估,实验结果表明:相较于BP模型,堆叠LSTM多源矿压预测模型在训练集和测试集上RMSE分别减少了49.15%和51.26%;相较于LSTM,分别减少了45.37%和46.61%;相较于GRU,分别减少了44.66%和45.89%。因此,堆叠LSTM多源矿压预测模型在工作面矿压预测方面具有更高的精确性。
The roof accident caused by the unbalanced mine pressure is one of the major disasters in coal mines.The accurate prediction of the mine pressure is of great significance to ensure the safe mining of coal seams.In order to improve the prediction accuracy of mine pressure,a multi-source mine pressure prediction model based on stacked-LSTM was proposed.Firstly,the gray correlation degree was used to analyze and sort the multi-source pressure of the coal mining face,as well as data pre-processing.Secondly,the stacked network structure was used to determine the parameters of each LSTM layer,such as the number of hidden nodes and the number of iterations.Finally,the Adam optimization algorithm was used to optimize the model,so as to realize the prediction of the mine pressure at the working face.The root mean square error(RMSE)was used as an evaluation index to evaluate the performance of the prediction model.The experimental results show that,for the multi-source mine pressure prediction model based on stacked-LSTM,the RMSE on the training set and the test set was reduced by 49.15%and 51.26%respectively compared with the BP model,was reduced by 45.37%and 46.61%compared with LSTM,and was reduced by 44.66%and 45.89%compared with GRU.Therefore,this new prediction model has a higher accuracy in mine pressure prediction at the working face.
作者
贾澎涛
苗云风
JIA Pengtao;MIAO Yunfeng(College of Computer Science and Technology,Xi'an University of Science and Technology,Xi'an,Shaanxi 710054,China)
出处
《矿业研究与开发》
CAS
北大核心
2021年第8期79-82,共4页
Mining Research and Development
基金
国家重点研究发展计划项目(2018YFC0808303)
西安市科技计划项目(2020KJRC0069)。
关键词
深度学习
堆叠式网络
长短时记忆网络
多源矿压
Deep learning
Stacked network
Long-short term memory network
Multi-source mine pressure
