Abstract
The comprehensive water cut of periodic waterflood reservoirs is random and volatile, and it is a non-stationary time series. In order to solve the problems of poor prediction results of conventional methods or heavy workload and long time-consuming in reservoir numerical simulation, an EMD-LSTM neural network model and a smoothing spline regression (Smoothing Spline) variable weight combined prediction model are proposed. This method introduces empirical mode decomposition (EMD) to process non-stationary and non-linear data, and combines the advantages of machine learning and curve regression to improve the prediction accuracy of the model. The variable weight combination model is used to predict the water content of the periodic water injection reservoir. The results show that the prediction accuracy of the variable weight combination model is significantly higher than that of the single model. The variable weight combination model can reduce the prediction error and can effectively and accurately predict the periodic water injection oil. The comprehensive water content of the reservoir can guide the adjustment of water injection parameters for periodic water injection reservoirs.
Export citation and abstract BibTeX RIS
Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.