改进Transformer在产油量预测中的应用研究

Research on the Application of Improved Transformer in Oil Production Forecast

产油量预测有利于制定合理的采油策略。本文提出一种包含卷积神经网络、门控循环单元和Transformer的组合模型CNN-GRU-Transformer,可用于产油量预测。该模型应用CNN提取部分深层空间特征,GRU提取产油量数据的时序特征,并根据油井数据的特点,改进了Transformer原有结构。通过改进的Transformer,将提取到的特征与预测相结合。实验的结果表明,CNN-GRU-Transformer模型在预测产油量各项指标中均为最优值,在适应产油量基本趋势方面表现最佳。

Predicting oil production is beneficial for developing reasonable oil recovery strategies.This article proposes a combined model(CNN GRU Transformer)that includes convolutional neural networks,gated loop units,and transformers for oil production prediction.This model applies CNN to extract some deep spatial features,GRU to extract temporal features of oil production data,and improves the original structure of Transformer based on the characteristics of oil well data.By improving the Transformer,the extracted features are combined with predictions.The experimental results indicate that the CNN GRU Transformer model has the best performance in predicting various indicators of oil production and adapting to the basic trend of oil production.