Shale oil production predication based on an empirical model-constrained CNN-LSTM

Abstract

Accurately predicting the production rate and estimated ultimate recovery (EUR) of shale oil wells is vital for efficient shale oil development. Although numerical simulations provide accurate predictions, their high time, data, and labor demands call for a swifter, yet precise, method. This study introduces the Duong–CNN–LSTM (D-C-L) model, which integrates a convolutional neural network (CNN) with a long short-term memory (LSTM) network and is grounded on the empirical Duong model for physical constraints. Compared to traditional approaches, the D-C-L model demonstrates superior precision, efficiency, and cost-effectiveness in predicting shale oil production.