基于PCA_Fuzzy_RF模型的煤层底板突水预测

针对煤层底板突水问题,提出了基于主成分分析、模糊数学和随机森林的一种新预测模型。首先通过主成分分析将6个影响因素(水压、采高、隔水层厚度、断层落差、煤层倾角、断层距工作面距离)进行降维,提取4个主成分因子,其次对主成分因子进行模糊化,作为随机森林模型的输入变量,建立基于PCA_Fuzzy_RF的煤层底板突水预测模型。利用华北矿区实测资料的50组数据作为PCA_Fuzzy_RF模型的训练数据,10组数据作为测试数据,并将预测结果与BP神经网络及Fisher模型进行对比分析,结果表明,PCA_Fuzzy_RF模型的误判率为0,适用于解决煤层底板突水问题。

Prediction of the water inrush from the coal seam based on PCA_Fuzzy_ RF model

The paper attempts to propose a renovated model for forecasting the water inrush from the coal seam in hoping to reduce such risks in the coal production and increase the human and equipment safety.For the said purpose,we have proposed an entirely new prediction model based on the principal component analysis,the fuzzy mathematics and random forests.In doing so,we have first of all managed to reduce the dimension and extract the four principal factors through the Principal Component Analysis of the following six factors,that is,the water pressure,the mining height,the aquiclude thickness,the fault throw,the dip angle of the coal seam,and the distance from the fault site to the working face.All the four principal factors from the aforementioned six factors have been found including 88.717％ of the total information of the original factors without duplication in factors.And,then,it would be possible to establish a prediction model of the water inrush from the coal seam floor on the basis of the PCA _Fuzzy _ RF blurring the principal factors by eliminating the difference of the magnitude,so as to get some standardized data as the input variables of the random forests and by setting the model variable as m_try =2,whereas the number of trees of RF equals n_tree--150.By taking the 50 groups of data,we have gained from the data accumulated through investigation and measurement of the mining area in China North as the training data of PCA Fuzzy RF and using the 10 groups of data from the above said 50 groups as the investigation and testing data,in addition to comparing the prediction results with BP neural network and Fisher model,the results indicate that the misjudgment rate of PCA Fuzzy RF model is very low with its testing and measuring accuracy rate being 100％ confidential.Since the accuracy rate of BP neural network and the Fisher model proves to be at about 90％ and 80％,respectively,it can be regarded as being valuable and suitable to solve the problem of water inrush from the coal seam floor and provide a new method for the nonlinear classification problem.