模糊神经网络及其在油气预测中的应用

Fuzzy neural network with application to hydrocarbon prediction

四川石油管理局地质调查处研究所,南充 637O0O

Geologic Survey Division, Sichuan Petroleum Administration, Nanchong 637000

常规模式识别方法进行油气预测存在样本数量大、参数非线性及已知样本数类型分类有明显的模糊性等特点。本文根据神经网络的自适应谐振理论(AdaptiveResonanceTheory), 结合模糊聚类算法、特征参数非线性映射的有效性校验, 构成了一个性能完善的地震油气预测系统。该系统对样本特征空间用模糊神经网络建立未知样本的预测度量, 更加符合客观地质特征和油气分布规律, 从而能判别油气富集的有利区域。数值分析和实践应用表明, 该方法数值稳定可靠, 在储层油气预测方面有较好的效果。

Hydrocarbon prediction using conventional pattem reqnition methods has drawbacks ofhuge sample quantity, parameter nonlinearity and obvious arnbiguity of known sample types. seismichydrocarbon prediction system with perfect performance is proposed in this paper. It is based on the neu-ral network adaptive resonance theory, fuzzy clustering algorithm and charactetistic parameter nonlinearrnapping. The system uses hey neural network to construct the predictive measure of unknown samplesin sample characteristic space, which more conforms to the objctive geologic characteristics and hydro-carbon distribution law. Numetical anaysis and actual application dernohstrate that this systern can serveas a stable, reiable and effective means for oil and gas prediction.