Roof pressure prediction method of coal working face based on spatiotemporal correlation analysis
-
摘要: 顶板压力一般通过液压支架工作阻力进行度量，基于深度学习的顶板压力预测方法效果受训练样本集影响极大，而训练样本集的构建依赖于时间窗口的选择和紧密关联液压支架群的识别，但现有方法依靠人工经验来确定时间窗口，且忽略了不同液压支架之间的关联性，严重阻碍了顶板压力预测精度的提高。针对上述问题，提出了一种基于时空关联分析的采煤工作面顶板压力预测方法。首先，通过计算同一液压支架工作阻力序列在时间维度上的灰色关联度，选择最优时间窗口。然后，通过计算不同液压支架工作阻力序列在空间维度上的灰色关联度，获得最优辅助矩阵，识别出紧密关联液压支架群。最后，基于最优时间窗口和最优辅助矩阵，确定每个训练样本的标签和对应特征，完成训练样本集构建，以对长短时记忆(LSTM)模型进行训练来预测顶板压力。实验结果表明，与依赖人工经验构建训练样本集完成LSTM模型训练的方法相比，本文方法有效降低了顶板压力预测误差。Abstract: The roof pressure is usually measured by the hydraulic support working resistance, and the effect of the roof pressure prediction method based on depth learning is greatly affected by the training sample set. The construction of the training sample set depends on the selection of the time window and the identification of the closely related hydraulic support group. However, the existing methods rely on manual experience to determine the time window, and ignore the correlation between different hydraulic supports, which seriously hinders the improvement of the roof pressure prediction precision. In order to solve the above problems, a roof pressure prediction method of coal working face based on spatiotemporal correlation analysis is proposed. Firstly, the optimal time window is selected by calculating the grey correlation degree of working resistance series of the same hydraulic support in the time dimension. Secondly, the optimal auxiliary matrix is obtained by calculating the grey correlation degree of working resistance sequences of different hydraulic support in spatial dimension, and the closely related hydraulic support group is identified. Finally, based on the optimal time window and the optimal auxiliary matrix, the label and corresponding characteristics of each training sample are determined, and the training sample set is constructed to train the long short time memory (LSTM) model to predict the roof pressure. The experimental results show that the proposed method can reduce the prediction error of roof pressure effectively compared with the method which relies on manual experience to construct training sample sets to complete LSTM model training.
-
图 1 基于时空关联分析的顶板压力预测方法流程
Figure 1. Flow of roof pressure prediction method based on spatiotemporal correlation analysis
表 1 部分液压支架工作阻力序列时间关联度
Table 1. Time correlation degrees of working resistance sequences for partial hydraulic supports
k 40号 45号 50号 55号 60号 65号 70号 75号 80号 85号 90号 95号 100号 105号 110号 1 0.834 0.830 0.802 0.836 0.807 0.816 0.816 0.809 0.820 0.838 0.798 0.857 0.851 0.857 0.863 2 0.817 0.817 0.777 0.821 0.788 0.793 0.792 0.788 0.795 0.816 0.774 0.832 0.840 0.845 0.857 3 0.800 0.791 0.757 0.800 0.761 0.771 0.765 0.761 0.772 0.793 0.754 0.811 0.825 0.832 0.848 4 0.784 0.778 0.737 0.784 0.742 0.753 0.747 0.745 0.752 0.777 0.745 0.800 0.816 0.823 0.847 5 0.770 0.766 0.727 0.764 0.726 0.741 0.731 0.730 0.737 0.768 0.731 0.790 0.810 0.823 0.847 6 0.756 0.754 0.720 0.754 0.713 0.729 0.724 0.726 0.726 0.758 0.718 0.784 0.810 0.820 0.846 7 0.752 0.745 0.707 0.746 0.705 0.724 0.723 0.723 0.725 0.756 0.712 0.780 0.803 0.821 0.847 8 0.749 0.743 0.700 0.745 0.700 0.720 0.719 0.718 0.719 0.754 0.712 0.781 0.802 0.817 0.844 9 0.743 0.741 0.699 0.737 0.695 0.718 0.718 0.717 0.719 0.755 0.708 0.778 0.800 0.816 0.845 10 0.742 0.739 0.694 0.738 0.694 0.717 0.720 0.721 0.719 0.756 0.706 0.779 0.798 0.816 0.844 11 0.742 0.740 0.695 0.737 0.693 0.720 0.718 0.724 0.725 0.754 0.704 0.777 0.797 0.815 0.842 12 0.743 0.741 0.696 0.732 0.693 0.723 0.720 0.727 0.724 0.753 0.703 0.775 0.796 0.816 0.841 13 0.745 0.746 0.696 0.733 0.699 0.722 0.726 0.728 0.728 0.758 0.704 0.774 0.796 0.813 0.840 14 0.752 0.747 0.705 0.738 0.703 0.722 0.728 0.727 0.731 0.755 0.701 0.773 0.794 0.813 0.842 15 0.754 0.750 0.706 0.739 0.707 0.728 0.729 0.731 0.728 0.761 0.702 0.773 0.798 0.814 0.845 16 0.759 0.755 0.711 0.745 0.711 0.733 0.733 0.729 0.729 0.764 0.705 0.777 0.799 0.814 0.840 17 0.759 0.759 0.714 0.749 0.716 0.735 0.736 0.725 0.731 0.762 0.708 0.778 0.801 0.815 0.841 18 0.765 0.763 0.718 0.756 0.723 0.737 0.735 0.732 0.735 0.769 0.716 0.784 0.801 0.816 0.840 19 0.768 0.767 0.722 0.756 0.729 0.742 0.743 0.737 0.737 0.768 0.715 0.788 0.805 0.818 0.844 20 0.770 0.772 0.721 0.766 0.731 0.744 0.744 0.745 0.742 0.769 0.719 0.790 0.808 0.818 0.847 
下载: 导出CSV
表 2 部分液压支架工作阻力序列时间关联度排名
Table 2. Ranks of time correlation degrees of working resistance sequences for partial hydraulic supports
k 40号 45号 50号 55号 60号 65号 70号 75号 80号 85号 90号 95号 100号 105号 110号 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 4 4 4 4 4 4 4 4 4 4 4 4 4 4 5 5 5 6 7 5 6 7 7 10 9 6 8 5 5 5 4 7 6 11 11 8 9 10 11 14 14 14 12 7 9 6 7 8 7 13 15 12 11 13 13 15 17 16 14 11 11 9 6 4 8 15 16 15 12 15 17 18 19 19 19 10 10 10 10 11 9 17 17 16 18 17 19 19 20 20 17 13 14 13 13 10 10 20 20 20 16 18 20 16 18 18 15 14 12 15 12 12 11 19 19 19 17 19 18 20 16 15 18 17 15 17 16 15 12 18 18 17 20 20 14 17 12 17 20 18 17 19 14 16 13 16 14 18 19 16 16 13 11 13 13 16 18 18 20 19 14 14 13 14 15 14 15 12 13 9 16 20 20 20 19 14 15 12 12 13 14 12 12 11 8 12 11 19 19 16 17 9 16 9 10 11 13 11 10 9 10 11 9 15 16 14 18 18 17 10 9 10 10 9 9 7 15 10 10 12 13 12 15 17 18 8 8 9 7 8 8 8 7 8 5 8 8 11 11 20 19 7 6 6 8 6 6 6 6 7 7 9 7 8 9 13 20 5 5 7 5 5 5 5 5 5 6 6 6 7 8 6
下载: 导出CSV
表 3 部分液压支架最优时间窗口
Table 3. Optimal time windows for partial hydraulic supports
液压支架 40号 45号 50号 55号 60号 65号 70号 75号 80号 85号 90号 95号 100号 105号 110号 最优时间窗口 5 5 6 6 6 5 5 5 5 5 6 6 8 8 7
下载: 导出CSV
表 4 部分液压支架工作阻力序列空间关联度
Table 4. Spatial correlation degrees among working resistance sequences for partial hydraulic supports
液压支架 液压支架 40号 45号 50号 55号 60号 65号 70号 75号 80号 85号 90号 95号 100号 105号 110号 40号 1.000 0.902 0.863 0.877 0.872 0.862 0.856 0.849 0.855 0.854 0.842 0.854 0.853 0.854 0.849 45号 0.895 1.000 0.863 0.876 0.875 0.864 0.855 0.840 0.852 0.852 0.835 0.844 0.845 0.841 0.836 50号 0.852 0.860 1.000 0.860 0.855 0.842 0.834 0.826 0.832 0.830 0.826 0.827 0.824 0.827 0.820 55号 0.878 0.881 0.868 1.000 0.899 0.876 0.867 0.854 0.863 0.852 0.849 0.854 0.849 0.845 0.839 60号 0.868 0.873 0.862 0.898 1.000 0.900 0.883 0.859 0.869 0.857 0.850 0.854 0.850 0.847 0.838 65号 0.860 0.869 0.852 0.878 0.902 1.000 0.902 0.874 0.878 0.867 0.855 0.855 0.853 0.850 0.843 70号 0.852 0.859 0.842 0.866 0.883 0.901 1.000 0.886 0.886 0.874 0.863 0.862 0.858 0.855 0.844 75号 0.846 0.846 0.836 0.854 0.860 0.873 0.887 1.000 0.888 0.876 0.862 0.861 0.854 0.849 0.844 80号 0.850 0.856 0.840 0.861 0.868 0.876 0.886 0.886 1.000 0.904 0.875 0.880 0.872 0.865 0.849 85号 0.852 0.857 0.840 0.853 0.858 0.867 0.876 0.876 0.906 1.000 0.886 0.890 0.878 0.868 0.855 90号 0.841 0.843 0.839 0.852 0.854 0.856 0.866 0.865 0.876 0.888 1.000 0.887 0.870 0.858 0.845 95号 0.852 0.851 0.839 0.856 0.857 0.856 0.864 0.862 0.883 0.890 0.886 1.000 0.901 0.886 0.865 100号 0.839 0.840 0.823 0.840 0.840 0.842 0.849 0.843 0.865 0.869 0.857 0.892 1.000 0.894 0.867 105号 0.839 0.834 0.823 0.833 0.835 0.836 0.844 0.836 0.855 0.856 0.843 0.874 0.892 1.000 0.886 110号 0.838 0.834 0.822 0.832 0.831 0.834 0.838 0.836 0.843 0.847 0.835 0.857 0.869 0.890 1.000
下载: 导出CSV
表 5 不同方法预测误差对比
Table 5. Comparison of prediction errors under different methods
% 方法 40号 45号 50号 55号 60号 65号 70号 75号 80号 85号 90号 95号 100号 105号 110号 传统方法 12.90 11.90 18.26 12.53 11.96 13.96 13.81 13.86 12.89 12.25 14.43 10.88 11.71 8.82 6.99 本文方法 11.99 11.44 17.34 12.17 10.93 12.24 12.52 13.07 12.57 11.57 14.24 9.99 11.08 8.31 6.99
下载: 导出CSV
-
[1] 宁小亮. 2013—2018年全国煤矿事故规律分析及对策研究[J]. 工矿自动化,2020,46(7):34-41.NING Xiaoliang. Law analysis and counter measures research of coal mine accidents in China from 2013 to 2018[J]. Industry and Mine Automation,2020,46(7):34-41. [2] 徐刚, 黄志增, 范志忠, 等. 工作面顶板灾害类型、监测与防治技术体系[J]. 煤炭科学技术,2021,49(2):1-11.XU Gang, HUANG Zhizeng, FAN Zhizhong, et al. Types, monitoring and prevention technology system of roof disasters in mining face[J]. Coal Science and Technology,2021,49(2):1-11. [3] 尹春雷, 魏文艳, 何勇华, 等. 基于大数据分析与线性回归模型的工作面顶板压力研究[J]. 自动化应用,2021(5):46-50.YIN Chunlei, WEI Wenyan, HE Yonghua, et al. Research on roof pressure of working face based on big data analysis and linear regression model[J]. Automation Application,2021(5):46-50. [4] 程敬义, 万志军, PENG Syd S, 等. 基于海量矿压监测数据的采场支架与顶板状态智能感知技术[J]. 煤炭学报,2020,45(6):2090-2103.CHENG Jingyi, WAN Zhijun, PENG S S, et al. Technology of intelligent sensing of longwall shield supports status and roof strata based on massive shield pressure monitoring data[J]. Journal of China Coal Society,2020,45(6):2090-2103. [5] 屈世甲, 李鹏. 基于支架工作阻力大数据的工作面顶板矿压预测技术研究[J]. 矿业安全与环保,2019,46(2):92-97. doi: 10.3969/j.issn.1008-4495.2019.02.021QU Shijia, LI Peng. Research on prediction technology of roof mining pressure based on big data of support resistance[J]. Mining Safety & Environmental Protection,2019,46(2):92-97. doi: 10.3969/j.issn.1008-4495.2019.02.021 [6] 尹希文, 徐刚, 刘前进, 等. 基于支架载荷的矿压双周期分析预测方法[J]. 煤炭学报,2021,46(10):3116-3126.YIN Xiwen, XU Gang, LIU Qianjin, et al. Method of double-cycle analysis and prediction for rock pressure based on the support load[J]. Journal of China Coal Society,2021,46(10):3116-3126. [7] 吴宛容. 基于改进灰色神经网络模型的顶板压力预测研究[D]. 徐州: 中国矿业大学, 2014.WU Wanrong. Research on hybrid grey-neural network for roof pressure forecasting in coal mine[D]. Xuzhou: China University of Mining and Technology, 2014. [8] 曾庆田, 吕珍珍, 石永奎, 等. 基于Prophet+LSTM模型的煤矿井下工作面矿压预测研究[J]. 煤炭科学技术,2021,49(7):16-23.ZENG Qingtian, LYU Zhenzhen, SHI Yongkui, et al. Research on prediction of underground coal mining face pressure based on Prophet+LSTM model[J]. Coal Science and Technology,2021,49(7):16-23. [9] ZHANG Tong, ZHAO Yixin, ZHU Guangpei, et al. A multi-coupling analysis of mining-induced pressure characteristics of shallow-depth coal face in Shendong mining area[J]. Journal of China Coal Society,2016,41(S2):287-296. [10] WU Yuting, YUAN Mei, DONG Shaopeng, et al. Remaining useful life estimation of engineered systems using vanilla LSTM neural networks[J]. Neurocomputing,2018,275:167-179. doi: 10.1016/j.neucom.2017.05.063 [11] KONG Weicong, DONG Zhaoyang, JIA Youwei, et al. Short-term residential load forecasting based on LSTM recurrent neural network[J]. IEEE Transactions on Smart Grid,2017,10(1):841-851. [12] 赵毅鑫, 杨志良, 马斌杰, 等. 基于深度学习的大采高工作面矿压预测分析及模型泛化[J]. 煤炭学报,2020,45(1):54-65.ZHAO Yixin, YANG Zhiliang, MA Binjie, et al. Deep learning prediction and model generalization of ground pressure for deep longwall face with large mining height[J]. Journal of China Coal Society,2020,45(1):54-65. [13] 王志奎. 基于支架工作阻力大数据的工作面区域矿压预测技术研究[D]. 青岛: 山东科技大学, 2018.WANG Zhikui. Research on prediction technology of mining area pressure based on large data of support working resistance[D]. Qingdao: Shandong University of Science and Technology, 2018. [14] 刘思峰, 党耀国, 方志耕, 等. 灰色系统理论及其应用[M]. 5版. 北京: 科学出版社, 2010: 17-29.LIU Sifeng, DANG Yaoguo, FANG Zhigeng, et al. Grey system theory and its application[M]. 5th ed. Beijing: Science Press, 2010: 17-29. -
点击查看大图
图(1) / 表(5)
计量
- 文章访问数: 125
- HTML全文浏览量: 85
- PDF下载量: 21
- 被引次数: 0
