BP神经网络算法在陆域天然气水合物成藏预测中的应用

The application of BP neural network algorithm to the prediction of terrestrial gas hydrate accumulation

勘察陆域天然气水合物的过程中,会产生大量的地质、地球物理和地球化学资料,将这些数据资料与数学方法有机结合,建立起综合信息预测模型对寻找水合物具有重要意义。笔者选取木里地区为主要研究区域,综合区内勘查已获得的地质、地球物理和地球化学数据,分析和提取了对水合物成藏有利的特征,给出了相应的预测变量转化规则。采用BP神经网络这种非线性预测方法进行成藏预测研究,并对结果进行对比评估。结果显示,钻遇水合物的钻井与预测得到的高有利度区吻合,未遇水合物的钻井基本落于低有利度区,算法有效实用,建立的转换规则切实可行。

During the exploration of terrestrial gas hydrates,Large quantities of geological, geophysical and geochemical data will be produced. The search for hydrate is of significance for effective combination of multi-source information with mathematical methods so as to establish a comprehensive information forecasting model. In this paper, the features which are favorable for gas hydrate accumulation were extracted from geological, geophysical and geochemical data in Muli area, and the corresponding transformation regularity was proposed. BP artificial neural network was used to do the study of gas hydrate prediction, and the effects of these two methods are compared and assessed. The results show that the prediction area is highly correlated with existing drilling result, suggesting that the methods are effective and the transformation regularity is feasible.