ARIMA−SVM组合模型驱动下的瓦斯浓度预测研究

范京道; 黄玉鑫; 闫振国; 李川; 王春林; 贺雁鹏

doi:10.13272/j.issn.1671-251x.2022030024

ARIMA−SVM组合模型驱动下的瓦斯浓度预测研究

doi: 10.13272/j.issn.1671-251x.2022030024

范京道^{1, 2,},
黄玉鑫²,
闫振国^2, ,,
李川¹,
王春林³,
贺雁鹏²

1.
陕西延长石油(集团)有限责任公司, 陕西西安　710075
2.
西安科技大学安全科学与工程学院, 陕西西安　710054
3.
陕西延长石油巴拉素煤业有限公司, 陕西榆林　719000

基金项目: 国家自然科学基金项目（52074214）；陕西省自然科学基础研究计划资助项目(S2019-JC-LH-QY-SM-0065)。

详细信息

作者简介:
范京道（1965—），男，陕西蒲城人，高级工程师，博士研究生导师，博士，主要从事煤矿智能开采技术研究工作，E-mail：fanjd@126.com

通讯作者:
闫振国（1975—），男，山西交城人，讲师，博士，主要从事智能通风与安全技术方面的研究工作，E-mail：yanzg@xust.edu.cn

中图分类号: TD712
计量
- 文章访问数: 392
- HTML全文浏览量: 58
- PDF下载量: 50
- 被引次数: 0
出版历程
- 收稿日期: 2022-03-08
- 修回日期: 2022-08-24
- 网络出版日期: 2022-05-19

Research on gas concentration prediction driven by ARIMA-SVM combined model

1.
Shaanxi Yanchang Petroleum(Group) Co., Ltd., Xi'an 710075, China
2.
College of Safety Science and Engineering, Xi'an University of Science and Technology, Xi'an 710054, China
3.
Shaanxi Yanchang Petroleum Balasu Coal Industry Co., Ltd., Yulin 719000, China

摘要

摘要: 针对单一瓦斯预测模型挖掘矿井瓦斯浓度时间序列全部特征能力较弱的问题，提出了一种基于自回归滑动平均模型（ARIMA）和支持向量机（SVM）模型的组合预测模型，并采用该模型对瓦斯浓度进行预测。首先，分别应用ARIMA模型和SVM模型对实验数据进行预测分析，得到2种单一模型预测结果。其次，结合自相关函数和偏自相关函数及贝叶斯准则，得到最优ARIMA模型为ARIMA（1,1,2），通过核函数等参数寻优，确立最优SVM模型，从而建立ARIMA−SVM组合模型。利用ARIMA模型处理瓦斯浓度时间序列的历史数据，得到相应的线性预测结果和残差序列，利用SVM模型进一步对数据残差序列中的非线性因素进行分析，得到非线性预测结果，将2个模型的预测结果进行组合，得到目标瓦斯时间序列最终预测结果。实验结果表明：① ARIMA−SVM组合模型预测结果与矿井实际数据的拟合度优于ARIMA模型和SVM模型。② 相对于ARIMA模型、SVM模型，ARIMA−SVM组合模型的误差大幅度减小，且预测结果明显优于单一模型。③ ARIMA−SVM组合模型的平均绝对误差、平均绝对百分比误差及均方根误差均为最小，表明ARIMA−SVM组合模型预测精度更高。
- 瓦斯浓度预测 /
- 瓦斯浓度时间序列 /
- 自回归滑动平均模型 /
- 支持向量机 /
- ARIMA−SVM组合模型
Abstract: The single gas prediction model has weak capability in mining all characteristics of the mine gas concentration time sequence. In order to solve the problem, a combined prediction model based on autoregressive intergrated moving average (ARIMA) model and support vector machine (SVM) model is proposed. The model is used to predict gas concentration. Firstly, the prediction results of the two single models are obtained by using the ARIMA model and the SVM model to predict and analyze the experimental data respectively. Secondly, combining the autocorrelation function, partial autocorrelation function and Bayesian criterion, the optimal ARIMA model is obtained as ARIMA(1,1,2). According to the optimization of kernel function and other parameters, the optimal SVM model is established, and then the ARIMA-SVM combined model is established. The ARIMA model is used to process the historical data of the gas concentration time series and obtain the corresponding linear prediction result and the residual sequence. The SVM model is used to further analyze the nonlinear factors in the data residual sequence and obtain the unlinear prediction result. The prediction results of the two models are combined to obtain the final prediction result of the target gas concentration time series. The experimental results show the following results. ① The fitting degree of the prediction results of the ARIMA-SVM combined model is better than that of the ARIMA model and SVM single model. ② Compared with the ARIMA model and SVM model, the error of the ARIMA-SVM combined model is greatly reduced, and the prediction result is obviously better than that of the single model. ③ The mean absolute error, mean absolute percentage error and root mean square error of the ARIMA-SVM combined model are the smallest. This result indicates that the prediction precision of the ARIMA-SVM combined model is higher.
- gas concentration prediction /
- gas concentration time series /
- autoregressive integrated moving average model /
- support vector machine /
- ARIMA-SVM combined model

HTML全文

图 1 原始瓦斯浓度时间序列

Figure 1. Original time sequence of gas concentration

下载: 全尺寸图片幻灯片

图 2 非平稳瓦斯浓度时间序列的一阶差分结果

Figure 2. Result of first-order difference for time series of nonstationary gas concentrations

下载: 全尺寸图片幻灯片

图 3 非平稳瓦斯浓度时间序列的二阶差分结果

Figure 3. Results of second-order difference for time series of nonstationary gas concentrations

下载: 全尺寸图片幻灯片

图 4 自相关与偏自相关函数

Figure 4. Autocorrelation and partial autocorrelation functions

下载: 全尺寸图片幻灯片

图 5 BIC图

Figure 5. BIC diagram

下载: 全尺寸图片幻灯片

图 6 ARIMA模型的瓦斯浓度预测结果

Figure 6. Gas concentration prediction results by ARIMA model

下载: 全尺寸图片幻灯片

图 7 SVM模型的瓦斯浓度预测结果

Figure 7. Gas concentration prediction results by SVM model

下载: 全尺寸图片幻灯片

图 8 ARIMA−SVM组合模型的瓦斯浓度预测结果

Figure 8. Gas concentration prediction results by ARIMA-SVM combined model

下载: 全尺寸图片幻灯片

图 9 瓦斯浓度预测结果

Figure 9. Prediction results of gas concentration

下载: 全尺寸图片幻灯片

表 1 9月1日采集的部分瓦斯浓度数据

Table 1. Part of the gas concentration data collected on September 1

时间	瓦斯体积分数/%	时间	瓦斯体积分数/%	时间	瓦斯体积分数/%	时间	瓦斯体积分数/%	时间	瓦斯体积分数/%	时间	瓦斯体积分数/%
00:00	0.15	01:00	0.21	02:00	0.15	03:00	0.11	04:00	0.15	05:00	0.20
00:05	0.15	01:05	0.19	02:05	0.13	03:05	0.13	04:05	0.17	05:05	0.18
00:10	0.13	01:10	0.17	02:10	0.13	03:10	0.09	04:10	0.17	05:10	0.22
00:15	0.17	01:15	0.17	02:15	0.13	03:15	0.15	04:15	0.17	05:15	0.18
00:20	0.21	01:20	0.19	02:20	0.11	03:20	0.19	04:20	0.19	05:20	0.20
00:25	0.19	01:25	0.21	02:25	0.13	03:25	0.15	04:25	0.17	05:25	0.22
00:30	0.17	01:30	0.19	02:30	0.11	03:30	0.13	04:30	0.11	05:30	0.2
00:35	0.15	01:35	0.19	02:35	0.13	03:35	0.15	04:35	0.13	05:35	0.2
00:40	0.17	01:40	0.17	02:40	0.11	03:40	0.15	04:40	0.11	05:40	0.22
00:45	0.11	01:45	0.15	02:45	0.13	03:45	0.11	04:45	0.09	05:45	0.18
00:50	0.15	01:50	0.15	02:50	0.13	03:50	0.19	04:50	0.16	05:50	0.20
00:55	0.17	01:55	0.15	02:55	0.11	03:55	0.19	04:55	0.18	05:55	0.20

下载: 导出CSV

表 2 ADF检验结果

Table 2. Result of ADF test

临界值			P值	T值
1%置信度	5%置信度	10%置信度	P值	T值
−3.43	−2.86	−2.57	0.5	−6.22

下载: 导出CSV

表 3 Ljung−Box检验表

Table 3. Ljung-Box inspection table

lag滞后阶数	自相关系数	P值
1	0.004	0.877
2	0.007	0.944
3	−0.022	0.839
4	−0.040	0.363
5	−0.050	0.397
6	−0.053	0.156
7	−0.040	0.112
8	0.041	0.078
9	0.021	0.096
10	0.041	0.068

下载: 导出CSV

表 4 各模型预测结果分析

Table 4. Prediction results analysis of each model

模型	MAE	MAPE	RMSE
ARIMA	0.028 4	0.047 6	0.075 4
SVR	0.025 1	0.032 5	0.056 3
ARIMA−SVM	0.015 8	0.019 3	0.010 3

下载: 导出CSV

参考文献(15)

[1]	齐黎明,卢云婷,关联合,等. 煤与瓦斯突出预测敏感指标确定方法探索及应用[J]. 矿业安全与环保,2021,48(3):85-89. QI Liming,LU Yunting,GUAN Lianhe,et al. Exploration and application of determination method of sensitive index for prediction of coal and gas outburst[J]. Mining Safety & Environmental Protection,2021,48(3):85-89.
[2]	林海飞,周捷,高帆,等. 基于特征选择和机器学习融合的煤层瓦斯含量预测[J]. 煤炭科学技术,2021,49(5):44-51. LIN Haifei,ZHOU Jie,GAO Fan,et al. Coal seam gas content prediction based on fusion of feature selection and machine learning[J]. Coal Science and Technology,2021,49(5):44-51.
[3]	邵阳,康宇. 基于多元线性回归预测采煤工作面瓦斯涌出量[J]. 煤炭技术,2021,40(4):118-121. SHAO Yang,KANG Yu. Prediction of coal face gas emission based on multiple linear regression[J]. Coal Technology,2021,40(4):118-121.
[4]	郑晓亮,来文豪,薛生. MI和SVM算法在煤与瓦斯突出预测中的应用[J]. 中国安全科学学报,2021,31(1):75-80. ZHENG Xiaoliang,LAI Wenhao,XUE Sheng. Application of MI and SVM in coal and gas outburst prediction[J]. China Safety Science Journal,2021,31(1):75-80.
[5]	王彦彬. 基于PCA−PSO−ELM的瓦斯涌出量预测[J]. 湖南科技大学学报(自然科学版),2020,35(4):1-9. WANG Yanbin. Prediction of gas emission based on PCA-PSO-ELM[J]. Journal of Hunan University of Science & Technology (Natural Science Edition),2020,35(4):1-9.
[6]	刘莹,杨超宇. 基于多因素的LSTM瓦斯浓度预测模型[J]. 中国安全生产科学技术,2022,18(1):108-113. LIU Ying,YANG Chaoyu. LSTM gas concentration prediction model based on multiple factors[J]. Journal of Safety Science and Technology,2022,18(1):108-113.
[7]	吴奉亮,霍源,高佳南. 基于随机森林回归的煤矿瓦斯涌出量预测方法[J]. 工矿自动化,2021,47(8):102-107. WU Fengliang,HUO Yuan,GAO Jianan. Coal mine gas emission prediction method based on random forest regression[J]. Industry and Mine Automation,2021,47(8):102-107.
[8]	张震,朱权洁,李青松,等. 基于Keras长短时记忆网络的矿井瓦斯浓度预测研究[J]. 安全与环境工程,2021,28(1):61-67,78. ZHANG Zhen,ZHU Quanjie,LI Qingsong,et al. Prediction of mine gas concentration in heading face based on Keras long short time memory network[J]. Safety and Environmental Engineering,2021,28(1):61-67,78.
[9]	周松元,罗渭,马伟东,等. 基于TreeNet算法的煤与瓦斯突出预测模型构建研究[J]. 矿业研究与开发,2022,42(5):190-196. doi: 10.13827/j.cnki.kyyk.2022.05.002 ZHOU Songyuan,LUO Wei,MA Weidong,et al. Research on coal and gas outburst prediction model construction based on TreeNet algorithm[J]. Mining Research and Development,2022,42(5):190-196. doi: 10.13827/j.cnki.kyyk.2022.05.002
[10]	林海飞,高帆,严敏,等. 煤层瓦斯含量PSO−BP神经网络预测模型及其应用[J]. 中国安全科学学报,2020,30(9):80-87. LIN Haifei,GAO Fan,YAN Min,et al. Study on PSO-BP neural network prediction method of coal seam gas content and its application[J]. China Safety Science Journal,2020,30(9):80-87.
[11]	郭一鹏,郭晓丽,郭晓红,等. 利用时间序列分析法预测工作面瓦斯涌出规律[J]. 矿业安全与环保,2012,39(1):33-35,39. doi: 10.3969/j.issn.1008-4495.2012.01.011 GUO Yipeng,GUO Xiaoli,GUO Xiaohong,et al. Using time series analysis method to predict the law of gas emission from working face[J]. Mining Safety & Environmental Protection,2012,39(1):33-35,39. doi: 10.3969/j.issn.1008-4495.2012.01.011
[12]	汤进. 基于瓦斯浓度时间序列的大数据预测[D]. 徐州: 中国矿业大学, 2018. TANG Jin. Big data prediction based on methane concentration time series[D]. Xuzhou: China University of Mining and Technology, 2018.
[13]	邵良杉,詹小凡. 煤与瓦斯突出missForest− EGWO−SVM预测模型[J]. 辽宁工程技术大学学报(自然科学版),2020,39(3):214-218. SHAO Liangshan,ZHAN Xiaofan. MissForest-EGWO-SVM prediction model for coal and gas outburst[J]. Journal of Liaoning Technical University (Natural Science),2020,39(3):214-218.
[14]	吴雅琴,李惠君,徐丹妮. 基于IPSO−Powell优化SVM的煤与瓦斯突出预测算法[J]. 工矿自动化,2020,46(4):46-53. doi: 10.13272/j.issn.1671-251x.2019110018 WU Yaqin,LI Huijun,XU Danni. Prediction algorithm of coal and gas outburst based on IPSO-Powell optimized SVM[J]. Industry and Mine Automation,2020,46(4):46-53. doi: 10.13272/j.issn.1671-251x.2019110018
[15]	李欢,贾佳,杨秀宇,等. 煤矿综采工作面瓦斯浓度预测模型[J]. 工矿自动化,2018,44(12):48-53. doi: 10.13272/j.issn.1671-251x.17364 LI Huan,JIA Jia,YANG Xiuyu,et al. Gas concentration prediction model for fully mechanized coal mining face[J]. Industry and Mine Automation,2018,44(12):48-53. doi: 10.13272/j.issn.1671-251x.17364