【摘要】 概率神经网络是一种基于概率统计思想的神经网络,利用概率神经网络进行储层的流体密度反演,通过它的非线性扩展进行多个属性的优选组合,完成神经网络的训练学习和概率估算,有效地剔除个别数据的不利影响,使反演过程更加稳定,减少反演结果的多解性。川西某气藏的概率神经网络储层流体密度反演结果表明,该反演方法准确性很高,反演结果与实际试气结果对应良好,能解决一些常规地震勘探油气检测方法不能解决的问题,并可对储层的含气性进行定量分析,为储层预测、气水识别、油气藏描述提供重要的数据支持。更多还原

【Abstract】 The fluid-density in reservoir rock porosity can reflect the reservoir characteristics directly.This is one of the key parameters of today′s seismic exploration for hydrocarbon.Probabilistic neural network(PNN) is a neural net work based on probability and statistics idea.The probabilistic neural network inversion reservoir fluid-density is used for the preferred combination of multi ple attributes to complete the training of neural network learning and probabili ty estimation through its non-linear expansion.This can effectively eliminate t he adverse effects of individual data,stabilize the inversion process and reduc e the inversion multi-solvability.The gas reservoir fluid-density inversion res ult of a gas field in western Sichuan shows that the PNN inversion method,with its high accuracy and correspondence with the actual test gas well,can solve so me problems that some conventional seismic hydrocarbon detection methods are una ble to solve.It can conduct a quantitative analysis for reservoir gas-bearing, and provide an important data support for reservoir prediction,gas-water identi fication and gas reservoir description.更多还原