论文部分内容阅读
测井岩性识别是油气勘探中十分重要的基础工作,可为测井解释选择正确解释方法和解释参数提供依据。本文利用过程神经元网络建立了复杂的非线性岩性辨识模型;同时,为了提高对实际问题求解的适应性和算法执行效率,开发了一种基于样条函数拟合的过程神经元网络学习算法。最后,结合海拉尔盆地贝16区块的实际测井资料进行岩性识别。实验结果表明,基于样条过程神经元网络的岩性辨识方法避免了采用传统BP神经网络预先建立复杂的数学或物理模型来提取小层测井曲线形态模式特征的过程,有效改善了网络的运算速度和对实际数据的抗扰性,具有较好的稳定性和泛化能力。
Logging lithology identification is a very important basic work in oil and gas exploration, which can provide the basis for selecting the correct interpretation method and explaining the parameters for logging interpretation. In this paper, a complex nonlinear lithologic identification model is established by using process neural network. In the meantime, in order to improve the adaptability of practical problems and the efficiency of algorithm execution, a neural network learning algorithm based on spline function fitting . Finally, the lithology identification is carried out based on the actual logging data of the Bay 16 block in the Hailar Basin. The experimental results show that the lithology identification method based on spline process neural network can avoid the process of extracting complex mathematical or physical models by traditional BP neural network in order to extract morphological patterns of small-layer log curves and effectively improve the operation of network Speed and immunity to the actual data, with good stability and generalization ability.