Research on Reservoir Development Indexes Prediction Model Based on CNN-BiLSTM
ZHU Bilei
ZHANG Qiang
ZHU Liutao
HAN Liting
Abstract：In view of the problems of insufficient feature extraction and low prediction accuracy in traditional oil production and water content prediction methods,a reservoir development indexes prediction model based on CNN-BiLSTM is proposed.First-ly,the model applies InceptionV1R and ResidualBlockX modules to increase the diversity of feature extraction structures,while us-ing shortcut to suppress network degradation and gradient dispersion.Secondly,the activation function of Gaussian Error Linear Unit(GELU)is introduced into BiLSTM to accelerate the convergence of the model.Finally,methods such as L2 regularization,Dropout and custom learning rate scheduler are used to suppress over-fitting and improve the generalization ability of the model.In order to verify the performance of the proposed model,it is compared with six models including RNN,LSTM,GRU,ConvLSTM,BiLSTM and CNN+LSTM.The results show that the average of mean absolute error of oil production and water content prediction results is lower than that of the comparison model,it indicates that the prediction model is better.
Keywords：BiLSTMInceptionV1R moduleResidualBlockX moduleprediction
Publication Date：2025-08-20
Online Publishing Date：2025-12-12(First online date of this platform, not the publication date of the document)
Pages：8( 2167-2173,2210 )