Volume 51

An Integrated Method of Data-Driven and Water-Drive Mechanism for Oil Production Forecast in High Water Cut Reservoir Qingshuang Jin, Yongchao Xue, Xiangyu Ren, Aile Zheng, Xiaobiao Wang

Abstract

The Q oilfield is a medium-to-high permeability heavy oil reservoir, currently in the high water cut development stage, with continuously declining production. Accurate production forecasting can provide guidance for adjusting production strategies in the oilfield. In recent years, the rise of machine learning models has offered a better alternative for predicting well production. This paper is based on data-driven daily production dynamic forecasting of oil wells, integrating data-driven models with water drive characteristic curves to enhance the accuracy of dynamic production forecasting for oil wells.
Firstly, using exclusion discriminant analysis, time, water cut, and daily oil production were selected as machine learning feature data, and the optimal water drive characteristic curve was determined. Secondly, the optimization of long short-term memory neural network structure is carried out, and the optimal results are used to carry out the recursive multi-step prediction of long short-term memory neural network. Furthermore, the optimal data-driven model was further integrated with the optimal water drive characteristic curve to establish an oil well daily oil production forecasting model that combines water drive characteristic curves and LSTM models. Finally, using data from two wells in the Q oilfield, the predictive performance of the pure data-driven model, the integrated model, and the numerical simulation model was compared and evaluated. The results showed that the integrated model had the best predictive performance and the lowest error.
This paper establishes a dynamic production forecasting model for oil wells that deeply integrates water drive characteristics with data-driven models, improving traditional daily production forecasting methods and achieving superior predictive results.

Keywords oil production prediction, data driven, water-drive characteristics, integrated model