Abstract: In order to predict the reservoir with strong heterogeneity,a probabilistic neural network technique was developed on the theory of probabilistic density function and trained using known information of sandstone-mudstone reservoir.The probabilistic neural network was used to predict the reservoirs of a lithologic oilfield in the southwest China.The targets in the study area are the beach sandstone and channel sandstone of Sha-1 Member where the reservoir thickness is small and the lateral change of lithology is big.The correlations among the features of logging,seismic attributes and geological lithology were developed to train the probabilistic neural network,and the seismic attributes were transformed by the trained network to identify the lithologic information of reservoirs.The results show that the probabilistic neural network technique has good application effectiveness in the actual data processing.

Key words: probabilistic neural network technique, reservoir prediction, lithology inversion, correlation, seismic data, stratum feature

摘要： 提出了一种由多测井和多地震属性参数组成的概率神经网络方法,来进行非均质性较强的油气储层的预测.介绍了该方法的网络模型构建和地层岩性预测的过程.利用该概率神经网络方法,研究了我国西南某一岩性油气田沙一段湖滩砂及河道砂体.运用测井响应特征、地震属性特征与地质岩性特征的相关性对概率神经网络进行了培训,从而对地层特征进行了预测和识别,并取得了较好的应用效果.

关键词: 概率神经网络技术, 储层预测, 岩性反演, 相关性, 地震参数, 地层特征