构造性神经网络在测井岩性识别中的应用

Constructive Neural Networks for Well-Log Lithology Identification

测井岩性识别是石油勘探中十分重要的基础工作,准确的岩性识别结果可以为勘探和开发提供可靠的依据。人工神经网络方法可对复杂的高维数据进行非线性映射,在模式识别领域得到越来越广泛的应用。针对传统的BP神经网络算法存在收敛速度慢、隐层数以及隐层节点数难以确定等缺点,研究了一种构造性神经网络学习算法———级联算法(Cascade Correlation Algorithm,简称CC算法)及其在石油工程中的应用。采用该算法对甘肃镇原-泾川地区进行岩性识别研究,通过与BP神经网络的识别结果进行比较,体现出构造性神经网络的优越性。

Well-Log lithology identification is a very important basis tasks in petroleum exploration and the correct lithology identification may supply a decision criterion for oil exploration and exploitation. Artificial neural network can implement a non-linear mapping for high-dimension complex data and it has been widely used in the field of pattern recognition. To overcome some limitations of BP algorithm, such as fixed topology and hidden units to be predefined, this paper shows the study of a constructive neural network algorithm, i, e. , Cascade Correlation Algorithm （CCA） and its application in petroleum engineering. The CCA is applied to identify well-log lithology identification of Gan-Su Zhenyuan-Jingchuan oil area. Compared with BP algorithm, the identification precision and convergence have been greatly improved by using CCA.