概率神经网络在地震岩性反演中的应用

卧龙湖煤矿北二采区岩浆岩侵入8煤层的现象较为严重,同时该区煤层中构造煤比较发育,瓦斯富集问题较为突出。利用三维地震资料、测井曲线进行约束反演得到的波阻抗作为外部属性,并使用step-wise属性选择法确定合适数目的地震属性,利用概率神经网络技术(PNN)对该区进行孔隙度预测反演。孔隙度反演结果与波阻抗反演结果的对比表明:孔隙度较波阻抗对于识别瓦斯富集带具有更高的分辨能力;概率神经网络具有高稳定性、计算精度高等特点,可作为研究构造煤发育和瓦斯赋存的有效手段。 

Application of probabilistic neural network in seismic lithological inversion

Magmatic invasion into No.8 coal seam is a relatively severe problem in the north mining district no.2 of Wolonghu mine,bringing out significant methane enrichment.Acoustic impedance obtained by restrained inversion using both seismic data and well log,along with other seismic attributes picked out by step-wise regression,is considered as an external attribute to predict porosity using the technology of probabilistic neural network(PNN).PNN has higher definition in identifying gas enrichment zone than acoustic impedance,while they both indicate the same lithological interpretation.In conclusion,PNN is a highly reliable and effective method of lithological inversion to study tectonic coal development and methane gas occurrence with a high stability and computational precision.