采矿与安全工程学报
採礦與安全工程學報
채광여안전공정학보
JOURNAL OF MINING AND SAFETY ENGINEERING
2013年
6期
946-952
,共7页
瓦斯涌出量%预测%SVM-LMD%采煤工作面
瓦斯湧齣量%預測%SVM-LMD%採煤工作麵
와사용출량%예측%SVM-LMD%채매공작면
gas emission volume%forecasting%SVM-LMD%coalface
提出利用 LMD(Local Mean Decomposition)方法获取生产函数分量(PF 分量)进行 SVM (Support Vector Machine)建模，用此方法对采煤工作面瓦斯涌出量进行预测。通过 LMD对瓦斯涌出量的历史数据进行分解得到其PF分量，然后，对应于每个PF分量各利用SVM函数拟合方法进行外推预测，再把不同 PF 分量的预测结果进行叠加重构合成，进而获得瓦斯涌出量预测的理论结果值。通过对某煤矿监测历史数据进行实例分析，可见此方法预测效果比常规 SVM 方法预测精度高，LMD 的引入可大幅度提高瓦斯涌出量的预测精度，表明此方法建立的采煤工作面瓦斯涌出量预测模型具有较好的合理性和可靠性。PF分量的获取和SVM方法小样本预测的结合，能够充分发掘数据本身所蕴含的物理机制和物理规律，这也十分符合利用数据自身驱动来获取其影响因素相互间的物理机制，从而为瓦斯涌出量预测精度的提高奠定较好基础。
提齣利用 LMD(Local Mean Decomposition)方法穫取生產函數分量(PF 分量)進行 SVM (Support Vector Machine)建模，用此方法對採煤工作麵瓦斯湧齣量進行預測。通過 LMD對瓦斯湧齣量的歷史數據進行分解得到其PF分量，然後，對應于每箇PF分量各利用SVM函數擬閤方法進行外推預測，再把不同 PF 分量的預測結果進行疊加重構閤成，進而穫得瓦斯湧齣量預測的理論結果值。通過對某煤礦鑑測歷史數據進行實例分析，可見此方法預測效果比常規 SVM 方法預測精度高，LMD 的引入可大幅度提高瓦斯湧齣量的預測精度，錶明此方法建立的採煤工作麵瓦斯湧齣量預測模型具有較好的閤理性和可靠性。PF分量的穫取和SVM方法小樣本預測的結閤，能夠充分髮掘數據本身所蘊含的物理機製和物理規律，這也十分符閤利用數據自身驅動來穫取其影響因素相互間的物理機製，從而為瓦斯湧齣量預測精度的提高奠定較好基礎。
제출이용 LMD(Local Mean Decomposition)방법획취생산함수분량(PF 분량)진행 SVM (Support Vector Machine)건모，용차방법대채매공작면와사용출량진행예측。통과 LMD대와사용출량적역사수거진행분해득도기PF분량，연후，대응우매개PF분량각이용SVM함수의합방법진행외추예측，재파불동 PF 분량적예측결과진행첩가중구합성，진이획득와사용출량예측적이론결과치。통과대모매광감측역사수거진행실례분석，가견차방법예측효과비상규 SVM 방법예측정도고，LMD 적인입가대폭도제고와사용출량적예측정도，표명차방법건립적채매공작면와사용출량예측모형구유교호적합이성화가고성。PF분량적획취화SVM방법소양본예측적결합，능구충분발굴수거본신소온함적물리궤제화물리규률，저야십분부합이용수거자신구동래획취기영향인소상호간적물리궤제，종이위와사용출량예측정도적제고전정교호기출。
In this paper, the method that using LMD (Local Mean Decomposition) to obtain production function components for SVM (Support Vector Machine) modeling was proposed, which was applied to forecast the gas emission volume in coalface. First, the historical data of gas emission volume were re-solved by LMD to get the production function components, i.e. PF components. Then, extrapolation fo-recasting of each PF component was carried out by using SVM function fitting method,respectively. In addition, the forecasting results were reconstructed, and the forecasted theoretical values of gas emis-sion volume were finally obtained. From the case study in one mine, the forecasting accuracy of the method proposed in this paper is higher than conventional SVM methods, and the established forecast-ing model of coalface gas emission based on this method has better rationality and reliability. Therefore, with the acquisition of production function components and small sample forecasting by SVM, the physical mechanisms and laws in data can be fully exploited, which accords well with the physical me-chanism that using data themselves to get their interaction. This method provides a basis for improving the forecasting accuracy of gas emission volume.
