基于褶积模型的复杂地质下的煤层厚度预测技术

Prediction Technology of Coal Seam Thickness under Complex Geology Based on Convolution model

传统预测方法在煤层分级设置约束钻孔数过少时,在较为复杂的地质煤层中无法进行精细勘查,对煤层厚度的预测精度较低,不能保证煤炭资源的可靠开采。褶积模型能够利用反演方法进行脉冲和地质推进预测,在抗噪声和分辨能力上均有一定优势。在划分复杂地质下煤层厚度主控因素的基础上,基于褶积模型反演地质推断地层反射系数,设置线性堆叠网络层结构,来预测煤层厚度,完成基于褶积模型的复杂地质下煤层厚度预测技术设计。以某省三煤层矿区为测试对象,选择回归参数和神经网络两组传统方法,与基于褶积模型预测方法进行对比,预测8组井口坐标下的煤层厚度。结果表明,基于褶积模型预测方法的预测结果与实际数据较为一致,绝对误差小于0.02,而两组传统方法的绝对误差值均大于1.10,说明本方法的预测精度更高,具有实际应用效果。

When the number of constrained boreholes in coal seam classification is too small,the traditional prediction method can not carry out fine exploration in more complex geological coal seams,and the prediction accuracy of coal seam thickness is low,which can not ensure the reliable mining of coal resources.Convolution model can use inversion method to predict pulse and geological propulsion,and has certain advantages in anti-noise and resolution.Based on the division of the main controlling factors of coal seam thickness under complex geology,the reflection coefficient of the stratum is inversed based on the convolution model,the linear stacking network layer structure is set to predict the coal seam thickness,and the technical design of coal seam thickness prediction under complex geology based on the convolution model is completed.Taking the third coal seam mining area of a province as the test object,two groups of traditional methods of regression parameters and neural network are selected to predict the coal seam thickness under 8 groups of wellhead coordinates by comparing with the prediction method based on convolution model.The results show that the prediction results of the prediction method based on convolution model are consistent with the actual data,and the absolute error is less than 0.02,while the absolute error of the two groups of traditional methods is greater than 1.10.It shows that this method has higher prediction accuracy and practical application effect.