Coal spontaneous combustion temperature prediction model for goaf area based on GAT-Informer
-
摘要:
现有的煤自燃温度预测模型仅考虑监测数据前后的时间关联性，未考虑监测点之间的空间关系，并存在多步长煤自燃温度预测精度低的问题。针对上述问题，提出了一种基于图注意力网络（GAT）和Informer模型（GAT−Informer）的采空区煤自燃温度预测模型。首先，采用随机森林回归法和Savitzky−Golay滤波器对采空区沿空侧煤自燃监测数据中的异常值、缺失值和噪声进行处理，并使用Z−score方法对数据进行标准化。其次，采用GAT提取多个监测点煤自燃监测数据间的空间特征。然后，使用Informer模型的编码器对包含空间特征的数据进行编码，利用多头概率稀疏自注意力机制捕捉数据之间的长期依赖关系和时间特征；解码器通过交叉注意力机制与编码器交互，结合编码器提取的全局特征与目标序列的上下文依赖关系，生成特征矩阵并输入全连接层，得到煤自燃温度预测值。最后，对Informer模型输出的煤自燃温度预测值进行反标准化处理，恢复到原始数据尺度，得到最终的预测结果。实验结果表明，相较于循环神经网络（RNN）、长短期记忆网络（LSTM）、门控循环单元（GRU）和Informer模型，GAT−Informer模型在6个监测点上预测24步长煤自燃温度时，均方误差（MSE）分别平均降低了15.70%，22.15%，25.45%，36.49%，平均绝对误差（MAE）分别平均降低了16.01%，14.60%，20.30%，26.27%，表明GAT−Informer模型能有效提高煤自燃温度多步长预测精度。
Abstract:The existing coal spontaneous combustion temperature prediction models only consider the temporal correlation of the monitoring data, ignoring the spatial relationships between monitoring points, and suffer from low accuracy in multi-step temperature prediction of coal spontaneous combustion. To address these issues, a coal spontaneous combustion temperature prediction model for goaf areas based on graph attention network (GAT)-Informer model (GAT-Informer) was proposed. First, the random forest regression method and Savitzky-Golay filter were used to handle outliers, missing values, and noise in the spontaneous combustion monitoring data along the goaf side, and the Z-score method was applied to standardize the data. Secondly, GAT was employed to extract spatial features from the spontaneous combustion monitoring data at multiple monitoring points. Then, the encoder of the Informer model was used to encode the data containing spatial features, utilizing a multi-head probabilistic sparse self-attention mechanism to capture long-term dependencies and temporal features among the data. The decoder interacted with the encoder through a cross-attention mechanism, combining global features extracted by the encoder with the contextual dependencies of the target sequence to generate a feature matrix, which was then fed into the fully connected layer to obtain the coal spontaneous combustion temperature prediction value. Finally, the predicted temperature value output from the Informer model was de-standardized to restore it to the original data scale, yielding the final prediction results. Experimental results showed that, compared to recurrent neural network (RNN), long short-term memory (LSTM), gated recurrent unit (GRU), and the Informer model, the GAT-Informer model reduced the mean squared error (MSE) by an average of 15.70%, 22.15%, 25.45%, and 36.49%, respectively, and the mean absolute error (MAE) by an average of 16.01%, 14.60%, 20.30%, and 26.27%, respectively, when predicting the coal spontaneous combustion temperature at 24 time steps across six monitoring points. These results indicate that the GAT-Informer model effectively improves the multi-step prediction accuracy of coal spontaneous combustion temperature.
-
-
![]()
图 1 GAT-Informer模型架构
Figure 1. Graph attention network（GAT）-Informer model architecture
![]()
图 4 各监测点预测值与真实值对比
Figure 4. Comparison between predicted and actual values at each monitoring point
![]()
图 5 煤自燃温度多步长预测结果
Figure 5. Multi-step prediction results of coal spontaneous combustion temperature
表 1 不同步长下各模型的MSE
Table 1 Mean squared error（MSE） of each model at different steps
模型 步长 1 6 12 24 48 RNN 0.0129 0.0433 0.1911 0.4713 0.7708 LSTM 0.0050 0.0250 0.1929 0.4667 1.5659 GRU 0.0088 0.0233 0.2414 0.5087 1.3844 Informer 0.0511 0.2209 0.3250 0.6589 0.6824 GAT−Informer 0.0176 0.1372 0.1878 0.4037 0.4040 
下载: 导出CSV
表 2 不同步长下各模型的MAE
Table 2 Mean absolute error（MAE） of each model at different steps
模型 步长 1 6 12 24 48 RNN 0.0589 0.1516 0.2939 0.4967 0.6592 LSTM 0.0481 0.1106 0.3102 0.4731 1.0292 GRU 0.0524 0.1045 0.3689 0.5279 0.9279 Informer 0.1707 0.3591 0.4064 0.5837 0.6334 GAT−Informer 0.0907 0.2236 0.2747 0.3929 0.4773
下载: 导出CSV
表 3 不同监测点下各模型的MSE
Table 3 MSE of each model under different monitoring points
模型 监测点 1 2 3 4 5 6 RNN 0.4713 0.1554 0.2429 0.2746 0.3353 0.4727 LSTM 0.4667 0.1572 0.3925 0.2760 0.3927 0.4834 GRU 0.5087 0.1492 0.3346 0.2806 0.3887 0.6372 Informer 0.6589 0.2235 0.3135 0.3831 0.4700 0.5243 GAT−Informer 0.4037 0.1196 0.1392 0.2596 0.3264 0.4440
下载: 导出CSV
表 4 不同监测点下各模型的MAE
Table 4 MAE of each model under different monitoring points
模型 监测点 1 2 3 4 5 6 RNN 0.4967 0.3067 0.3763 0.3685 0.3933 0.4744 LSTM 0.4731 0.2746 0.4609 0.3260 0.3642 0.5145 GRU 0.5279 0.3092 0.3937 0.3635 0.3859 0.5960 Informer 0.5837 0.2709 0.4658 0.4342 0.4847 0.5602 GAT−Informer 0.3929 0.2311 0.2995 0.3017 0.3576 0.4606
下载: 导出CSV
表 5 各监测点上GAT−Informer模型相较于其他模型在MSE和MAE指标上的降低幅度
Table 5 Reduction in MSE and MAE metrics of GAT-Informer model compared with other models at each monitoring point
监测点 RNN GRU LSTM Informer MSE
降低
幅度/%MAE
降低
幅度/%MSE
降低
幅度/%MAE
降低
幅度/%MSE
降低
幅度/%MAE
降低
幅度/%MSE
降低
幅度/%MAE
降低
幅度/%1 14.34 20.90 13.50 16.95 20.64 25.57 38.73 32.69 2 23.03 24.65 23.92 15.84 19.84 25.26 46.49 14.69 3 42.69 20.40 64.53 35.02 58.40 23.93 55.60 35.70 4 5.46 18.13 5.94 7.45 7.48 17.00 32.24 30.51 5 2.65 9.07 16.88 1.85 16.02 7.33 30.55 26.22 6 6.07 2.90 8.15 10.48 30.32 22.72 15.32 17.78
下载: 导出CSV
-
[1] 邓军,白祖锦,肖旸,等. 煤自燃灾害防治技术现状与挑战[J]. 煤矿安全,2020,51(10):118-125. DENG Jun,BAI Zujin,XIAO Yang,et al. Present situation and challenge of coal spontaneous combustion disasters prevention and control technology[J]. Safety in Coal Mines,2020,51(10):118-125.
[2] 杨峰峰,王海晖,苏伟伟,等. 采空区自燃危险区域判识与防控技术研究综述[J]. 煤矿安全,2023,54(6):66-75. YANG Fengfeng,WANG Haihui,SU Weiwei,et al. Review on identification and prevention and control technology of spontaneous combustion risk area in goaf[J]. Safety in Coal Mines,2023,54(6):66-75.
[3] 秦波涛,仲晓星,王德明,等. 煤自燃过程特性及防治技术研究进展[J]. 煤炭科学技术,2021,49(1):66-99. QIN Botao,ZHONG Xiaoxing,WANG Deming,et al. Research progress of coal spontaneous combustion process characteristics and prevention technology[J]. Coal Science and Technology,2021,49(1):66-99.
[4] 谭波,邵壮壮,郭岩,等. 基于指标气体关联分析的煤自燃分级预警研究[J]. 中国安全科学学报,2021,31(2):33-39. TAN Bo,SHAO Zhuangzhuang,GUO Yan,et al. Research on grading and early warning of coal spontaneous combustion based on correlation analysis of index gas[J]. China Safety Science Journal,2021,31(2):33-39.
[5] 邓军,雷昌奎,曹凯,等. 采空区煤自燃预测的随机森林方法[J]. 煤炭学报,2018,43(10):2800-2808. DENG Jun,LEI Changkui,CAO Kai,et al. Random forest method for predicting coal spontaneous combustion in gob[J]. Journal of China Coal Society,2018,43(10):2800-2808.
[6] 王怡,谢军,任广意. 采空区遗煤自然发火的指标气体研究[J]. 矿业研究与开发,2020,40(10):118-122. WANG Yi,XIE Jun,REN Guangyi. Study on index gas in spontaneous combustion of goaf residual coal[J]. Mining Research and Development,2020,40(10):118-122.
[7] 文虎,赵向涛,王伟峰,等. 不同煤体自燃指标性气体函数模型特征分析[J]. 煤炭转化,2020,43(1):16-25. WEN Hu,ZHAO Xiangtao,WANG Weifeng,et al. Analysis on characteristics of indicator gases of spontaneous combustion of different coals[J]. Coal Conversion,2020,43(1):16-25.
[8] 赵琳琳,温国锋,邵良杉. 不均衡数据下的采空区煤自燃PCA−AdaBoost预测模型[J]. 中国安全科学学报,2018,28(3):74-78. ZHAO Linlin,WEN Guofeng,SHAO Liangshan. PCA−AdaBoost model for predicting coal spontaneous combustion in caving zone with imbalanced data[J]. China Safety Science Journal,2018,28(3):74-78.
[9] 昝军才,魏成才,蒋可娟,等. 基于BP神经网络的煤自燃温度预测研究[J]. 煤炭工程,2019,51(10):113-117. ZAN Juncai,WEI Chengcai,JIANG Kejuan,et al. Prediction of coal spontaneous combustion temperature based on BP neural network[J]. Coal Engineering,2019,51(10):113-117.
[10] 郑学召,李梦涵,张嬿妮,等. 基于随机森林算法的煤自燃温度预测模型研究[J]. 工矿自动化,2021,47(5):58-64. ZHENG Xuezhao,LI Menghan,ZHANG Yanni,et al. Research on the prediction model of coal spontaneous combustion temperature based on random forest algorithm[J]. Industry and Mine Automation,2021,47(5):58-64.
[11] 汪伟,梁然,祁云,等. 基于PSO−BPNN的煤自燃危险性预测模型[J]. 中国安全科学学报,2023,33(7):127-132. WANG Wei,LIANG Ran,QI Yun,et al. Prediction model of coal spontaneous combustion risk based on PSO-BPNN[J]. China Safety Science Journal,2023,33(7):127-132.
[12] 贾澎涛,林开义,郭风景. 基于PSO−SRU深度神经网络的煤自燃温度预测模型[J]. 工矿自动化,2022,48(4):105-113. JIA Pengtao,LIN Kaiyi,GUO Fengjing. A temperature prediction model for coal spontaneous combustion based on PSO-SRU deep artificial neural networks[J]. Journal of Mine Automation,2022,48(4):105-113.
[13] 刘永立,刘晓伟,王海涛. 基于LSTM神经网络的煤矿火灾预测[J]. 黑龙江科技大学学报,2023,33(1):1-5. DOI: 10.3969/j.issn.2095-7262.2023.01.001 LIU Yongli,LIU Xiaowei,WANG Haitao. Coal mine fire prediction based on LSTM neural network[J]. Journal of Heilongjiang University of Science and Technology,2023,33(1):1-5. DOI: 10.3969/j.issn.2095-7262.2023.01.001
[14] 王斌,贾澎涛,郭风景,等. 基于多特征融合的煤自燃温度深度预测模型[J]. 中国矿业,2024,33(2):84-90. DOI: 10.12075/j.issn.1004-4051.20230635 WANG Bin,JIA Pengtao,GUO Fengjing,et al. Deep prediction model of coal spontaneous combustion temperature based on multi-feature fusion[J]. China Mining Magazine,2024,33(2):84-90. DOI: 10.12075/j.issn.1004-4051.20230635
[15] 王树斌,王旭,闫世平,等. 基于Transformer的矿井内因火灾时间序列预测方法[J]. 工矿自动化,2024,50(3):65-70,91. WANG Shubin,WANG Xu,YAN Shiping,et al. Transformer based time series prediction method for mine internal caused fire[J]. Journal of Mine Automation,2024,50(3):65-70,91.
[16] 李艳昌,张连超,陶华焕. 铁北煤矿褐煤自燃预测函数模型研究[J]. 工业安全与环保,2024,50(6):28-31. DOI: 10.3969/j.issn.1001-425X.2024.06.006 LI Yanchang,ZHANG Lianchao,TAO Huahuan. Study on prediction function model of lignite spontaneous combustion in Tiebei Coal Mine[J]. Industrial Safety and Environmental Protection,2024,50(6):28-31. DOI: 10.3969/j.issn.1001-425X.2024.06.006
[17] LI Pengfei,ZHANG Tong,JIN Yantao. A spatio-temporal graph convolutional network for air quality prediction[J]. Sustainability,2023,15(9):7624. DOI: 10.3390/su15097624
[18] 程子均,马六章,张翼翔. 基于LSTM−FC的瓦斯浓度时空分布预测[J]. 计算机工程与应用,2020,56(16):258-264. DOI: 10.3778/j.issn.1002-8331.2005-0250 CHENG Zijun,MA Liuzhang,ZHANG Yixiang. Prediction of spatiotemporal distribution of gas concentration based on LSTM-FC model[J]. Computer Engineering and Applications,2020,56(16):258-264. DOI: 10.3778/j.issn.1002-8331.2005-0250
[19] VELIČKOVIĆ P,CUCURULL G,CASANOVA A,et al. Graph attention networks[EB/OL]. [2024-07-22]. https://arxiv.org/abs/1710.10903v3.
[20] ZHOU Haoyi,ZHANG Shanghang,PENG Jieqi,et al. Informer:beyond efficient transformer for long sequence time-series forecasting[C]. AAAI Conference on Artificial Intelligence,Vancouve,2021:11106-11115.
[21] 王苏健,贾澎涛,金声尧. 基于随机森林回归的围岩应力插值方法[J]. 西安科技大学学报,2021,41(2):274-281. WANG Sujian,JIA Pengtao,JIN Shengyao. An interpolation method of surrounding rock stress based on random forest regression[J]. Journal of Xi'an University of Science and Technology,2021,41(2):274-281.
[22] 余琼芳,唐高峰,王联港. 基于改进Informer的综采工作面顶板压力多步长预测[J]. 矿业研究与开发,2023,43(11):61-67. YU Qiongfang,TANG Gaofeng,WANG Liangang. Multi-step prediction of roof pressure of comprehensive mining face based on the improved informer[J]. Mining Research and Development,2023,43(11):61-67.
-
期刊类型引用(8)
1. 于丹，颜伟. 煤矿井下避灾路径规划研究综述. 中国煤炭. 2022(02): 40-47 . 
百度学术
2. 马小陆，梅宏. 基于改进势场蚁群算法的移动机器人全局路径规划. 机械工程学报. 2021(01): 19-27 . 百度学术
3. 杨华玲. 基于混合算法的多停靠点物流路径规划仿真. 计算机仿真. 2021(04): 119-123 . 百度学术
4. 郭长恒，邢玉忠. 灾后巷道堆积体力学特征分析. 煤矿安全. 2020(05): 225-230 . 百度学术
5. 郝天轩，赵立桢. 跨平台矿井应急救援路径寻优方案研究. 工矿自动化. 2020(05): 108-112 . 本站查看
6. 胡人元. 深井被困事故井下救援处置程序体系构建. 武警学院学报. 2020(02): 35-38 . 百度学术
7. 侯远韶. 基于改进蚁群算法在机器人路径优化中的应用. 安阳工学院学报. 2020(06): 39-42 . 百度学术
8. 孙瑞，张文胜. 基于改进蚁群算法的移动机器人平滑路径规划. 图学学报. 2019(02): 344-350 . 百度学术
其他类型引用(5)
计量
- 文章访问数: 298
- HTML全文浏览量: 21
- PDF下载量: 31
- 被引次数: 13
