A Prediction Model for Water Absorption Profile Based on IDW-DTW-RNN Method

Abstract

At present, many low-permeability oil fields are entering the later stages of development, and there will be problems of difficulty in water injection and poor water injection effects. Obtaining water absorption profiles and their changes can more effectively help formulate layered water injection strategies. Generally, the isotope water absorption profile method is often used to obtain the water absorption profile. This method is accurate, but the test cost is higher and the actual oil field measurement data is less. Based on the measured water absorption profile data, this paper selects 8 factors including effective thickness of oil layer, effective permeability, measure coefficient, crude oil viscosity, crude oil volume coefficient, injection-production pressure difference, well spacing, and connection coefficient between oil and water wells as the main influencing factors of the profile prediction model. Taking water injection wells data with 8 factors as input, a water injection profile prediction neural network model based on inverse distance weighting method and dynamic time warping, namely called IDW-DTW-RNN, is trained in this study. This model is applied in oil field, and the accuracy is obviously improved compared with the conventional method, which is in line with the actual development situation. It provides a scientific basis for the adjustment and optimization of the later injection-production structure of the oilfield.