论文部分内容阅读
为有效分析煤矿瓦斯监测数据以实现较准确的瓦斯浓度预测,研究应用希尔伯特-黄变换(HHT)方法进行瓦斯浓度时间序列分析与预测的方法。应用经验模态分解(EMD)方法将瓦斯浓度时间序列分解成不同频率的固有模态函数(IMF)分量的叠加,以获取瓦斯浓度时间序列的瞬时特征;通过Hilbert变换求得各IMF分量的瞬时频率,依据各IMF分量瞬时频率的均值将分解得到的IMF划分成较高频和低频2类新的分量,选取适合于各分量特征的预测模型分别进行预测,以消除局部随机性对预测精度的影响,结合自回归(AR)、径向基函数(RBF)神经网络和支持向量机(SVM)3种预测模型实现瓦斯浓度预测。实例分析表明:应用该方法所得预测结果比较准确,降低了预测复杂度,提高了预测精度。
In order to effectively analyze gas monitoring data of coal mine to achieve more accurate gas concentration prediction, a method of time series analysis and prediction of gas concentration using Hilbert-Huang transform (HHT) is studied. Empirical mode decomposition (EMD) method is used to decompose gas concentration time series into superposition of IMFs with different frequencies to obtain instantaneous characteristics of gas concentration time series. The instantaneous moment of each IMF component is obtained by Hilbert transform According to the average value of the instantaneous frequencies of each IMF component, the IMFs decomposed are divided into two categories of higher frequency and low frequency components, and the prediction models suitable for each component feature are selected for prediction respectively, so as to eliminate the influence of local randomness on the prediction accuracy The prediction of gas concentration was carried out by combining three prediction models: autoregressive (AR), radial basis function (RBF) neural network and support vector machine (SVM). The case study shows that the prediction results obtained by this method are accurate, the prediction complexity is reduced, and the prediction accuracy is improved.