基于注意力机制的地层参数预测模型研究

Research on Formation Parameter Prediction Model Based on Attention Mechanism

随着油气资源的深度开发勘探难度不断增加,合理的参数解释帮助石油工程师和地质学家更好地理解储层特性,优化油田开发方案,深度学习为高精度地层参数预测提供了新思路,文章提出了一个基于Transfomer架构的预测网络模型,利用注意力机制构建输入与输出的全局依赖关系,该模型学习多点之间的映射关系通过并行计算实现测井数据特征对地层参数的预测。实际用例表明该方法预测计算得到的地层参数数据较为准确,与实测数据相比具有较高的相似性,具有较好的实用价值。

As the complexity of deep oil and gas exploration increases,the optimal parameter in-terprctation helps petroleum cnginccrs and gcologists to better understand the rescrvoir charac-teristics and optimize the solution.Deep learning provides a new way to predict the formation parameters accurately.In this paper,a prediction network model based on the Transfomer archi-tecture is proposed.The global dependence relationship between input and output is constructed using the attention mechanism.The model learns the mapping relationship between multiple points and realizes the prediction of formation parameters through parallel computation.The ac-tual application case shows that the formation parameter data obtained by this method is more accurate and has higher similarity with the measured data,and has better practical value.