基于优选地震属性与PSO-BP模型的煤层含气量预测

Prediction of coalbed methane content based on preferred seismic attributes and PSO-BP model

常规煤层含气量预测方法多基于测井约束的地震属性反演以及线性映射模型,造成预测结果的精度难以控制,严重限制了方法的普适性。本文从地震属性优选与BP神经网络预测模型改进两方面入手开展研究。利用Q型聚类分析方法,对提取的目标储层地震属性进行分类优选,得到了与地质目标相关性好且相互独立的4种地震属性;进一步利用粒子群寻优算法对BP神经网络算法的输入层与隐含层的连接权值和隐含层的阈值进行了优化,构建PSO-BP预测模型,并利用井位置的优选地震属性和含气量数据训练PSO-BP模型。基于训练好的PSO-BP模型,以整个工区的优选地震属性作为输入,进行研究区内煤层含气量预测。井位置含气量预测结果与实测结果对比表明,该预测方法准确率高。因此,可认为PSO-BP预测模型以及相应的预测方法流程能有效适用于煤储层含气量的预测。

Conventional coal seam gas content prediction methods are mostly based on logging constrained seismic attribute inversion and linear mapping model,which makes the prediction accuracy difficult to control,and severely limits the universality of the method.This paper starts with two aspects:seismic attribute optimization and BP neural network prediction model improvement.Using the Q-type clustering analysis method,the seismic attributes of the extracted target reservoirs are classified and optimized,and four kinds of seismic attributes with good correlation with the geological targets are obtained.The particle swarm optimization algorithm is further used to BP neural network algorithm.The connection weights of the input layer and the hidden layer and the threshold of the hidden layer are optimized,and the PSO-BP prediction model is constructed.The PSO-BP model is trained by using the preferred seismic attributes and gas content data of the well location.Based on the trained PSO-BP model,the coal seam gas content prediction in the study area is carried out with the preferred seismic attributes of the entire work area as input.The comparison between the predicted gas content of the well position and the measured results shows that the prediction method has high accuracy.Therefore,it can be considered that the PSO-BP prediction model and the corresponding prediction method flow can be effectively applied to the prediction of coal gas content in coal reservoirs.