A Study on Production Forecasting Method for Offshore Complex Reservoirs Based on Clustering Analysis and Deep Learning

With the increasing proportion of complex reservoirs in the development of offshore oil and gas fields, traditional production forecasting methods are facing more and more difficulties. In order to achieve accurate production forecasting of oil and gas wells and optimize production planning, this study adopts a data-driven strategy and uses clustering classification methods to analyze the data characteristics of injection and production parameters of complex reservoirs. Based on the factors affecting oil well production, including effective thickness, porosity, permeability, saturation, injection volume, oil production volume, etc., clustering features are selected and a production forecasting data system is constructed to form corresponding analytical results. On this basis, relying on the results of clustering analysis, the production laws of various types of complex reservoir oil wells are further explored, and each type of oil well is abstracted as a typical well. Subsequently, a deep learning neural network is used to build a production forecasting model, which not only effectively reduces the difficulty and workload of labeling complex reservoir oil well data as training samples, but also significantly increases the number of samples and greatly improves forecasting efficiency.