第44卷第1期 2017年2月
矿业安金与环保
MINING SAFETY & ENVIRONMENTAL PROTECTION
张文娟，侯媛彬，张文涛，等，基于GA-LSSVR的煤矿瓦斯数据去噪研究[J].矿业安全与环保，2017,44(1)：45-48. 文章编号：10084495(2017)01004504
基于GA-LSSVR的煤矿瓦斯数据去噪研究
张文娟，侯媛彬"，张文涛”，刘迷，陈显
Vol. 44 No. 1 Feh.2017
（1.西安科技大学电气与控制工程学院，陕西西安710054；2.河南科技学院信息工程学院，河南新乡453003）摘要：针对煤矿瓦斯数据普遍含有噪声的间题，提出一种基于遗传算法优化的最小二乘支持向量
回归机(GA-LSSVR)的数据去噪算法。LSSVR通过求解只含一个等式约束的二次规划间题来求得最优解，从而改进了小波去噪局部最优的缺点。但LSSVR也存在收敛速度慢的缺点，通过遗传算法（GA)优化LSSVR，以提高算法的收敛速度。首先，对某煤矿的瓦斯浓度时间序列进行异常数据和缺失数据的处理，然后用GA-LSSVR建模训练。仿真实验结果表明，与小波去噪方法相比，GA-LSSVR 能有效去除噪声，并且能够避免数据失真，把有效信号分高出来，经过计算，GA-LSSVR能将输入输出均方根误差降低0.00294，相对降低了34.59%，去噪效果较好：与LSSVR方法相比，GA-LSSVR能明显缩短程序运行时间，可提高运行效率。
关键词：瓦斯浓度；数据去噪；LSSVR；遗传算法；小波去噪
中图分类号：TD712*.52文献标志码：A
网络出版时间：2017-02-0709：53
网络出版地址：http：//www.cnki.net/kcms/detail/50.1062.TD.20170207.0953.002.html
ResearchonNoiseEliminationof CoalMineGasDataBasedonGA-LSSVR
ZHANGWenjuan',HOUYuanbin,ZHANGWentao",LIUMi',CHENXian
(1. School of Electric and Control Engineering,Xi' an Unisersity of Science and Technology,Xi' an 710054, China; 2. School of Information Engineering,Henan Institute of Science and Technology,Xinxiang 453003, China)
Abstract: Aiming at the problem that the coal mine gas data generally contain noise,a data denoising algorithm based on od se (s re aaa eado s o e ossaoeodns sarens sea LSSVR,the optimal solution was obtained by solving the quadratie programming problem with only one equality constraint, and jo eapep s o sasaao sap ae ndo je uoo prr s s slow convergence rate,the genetic algorithms ( GA) was used to optimize the LSSVR so as to improve the convergence rate of the algorithm. First,the abnormal data and the missing data processing was made on the the time series of gas concentrations in a coal mine,then GALSSVR was used for modeling training, The simulation experiment results showed that GALSSVR can effectively eliminate the noise, avoid the occurrence of data distortion and separate the effective signals as compared with the wavelet denoising method. Through the calculation, GALSSVR can reduce the root mean square error of the input and output by 0. 002 94,relatively reduced by 34. 59% ,the noise eliminating effect was better; GALSSVR could obviously shorten the running time of the program and improve the running efficiency as compared with LSSVR method.
Keywords : gas concentration; noise elimination of data; LSSVR ; genetic algorithm;wavelet denoising
煤矿井下环境非常复杂，条件十分恶劣，布置在井下的瓦斯传感器经常受到各种干扰的影响，如烟尘、高温、水气等，并且还会受到电磁干扰的影响，致使采集到的瓦斯数据普遍含有噪声]。如果用含有
收稿日期：2016-06-22;2016-1203修订
作者简介：张文娟（1990一），女，河南周口人，硕士研究生，主要研究方向为控制理论与控制工程。E-mail：
1139467304qq.com 万方数据
噪声的瓦斯数据直接进行预测，不仅不能准确预测瓦斯涌出量，及时预警危险，而且浪费时间，做大量无用工作。因此，对瓦斯数据必须进行去噪处理[2]，还原其真实的发展变化趋势。董丁稳等(3)采用小波软阀值法对矿井瓦斯监测数据进行了去噪处理，为避免数据严重失真，选用了三层的小波分解，滤除了高频的噪声信号，然后通过小波逆变换将经阅值处理过的小波系数和尺度系数重构，得到消噪后的瓦
· 45 ·