论文部分内容阅读
煤矿井下瓦斯浓度时间序列预测是一个非线性数据处理问题。该时间序列含有丰富地质信息和物理信息等,挖掘其隐含信息对研究井下瓦斯浓度变化规律具有重要意义。为此利用经验模态分解技术(EMD)提取了隐含在瓦斯浓度序列中的非线性高频波动项和低频趋势项,再对各分量构造不同的极限学习机(ELM)进行跟踪预测,最后叠加各分量的预测值得到瓦斯浓度后序的预测值。算例的分析结果表明,EMD-ELM算法具有较强的预测能力,预测精度高于已有一些预测方法,对有效控制和预防瓦斯浓度超标,保证煤矿安全生产具有参考意义。
Prediction of coal mine gas concentration time series is a nonlinear data processing problem. The time series contains abundant geological information and physical information, and mining its implicit information is of great significance for studying the variation law of gas concentration in underground wells. Therefore, EMD is used to extract the non-linear high-frequency fluctuation term and the low-frequency trend term hidden in the gas concentration sequence, and then to trace and predict the ELM with different components. Finally, Superimposing the predicted value of each component gives the predictive value of the gas concentration sequence. The results of numerical examples show that the EMD-ELM algorithm has strong predictive ability and the prediction accuracy is higher than some existing prediction methods, which has reference significance for effective control and prevention of excessive gas concentration and ensuring coal mine safety production.