Research on dynamic monitoring and prediction method of water drive reservoir based on big data analysis

The development of water drive reservoir is affected by many factors, and the traditional monitoring methods have some problems, such as single data dimension, lagging analysis methods and insufficient prediction accuracy. With the development of big data and artificial intelligence, the fusion of multi-source heterogeneous data combined with machine learning and reservoir engineering theory provides a new path for the transformation from static description to dynamic prediction. In this study, a data-knowledge-driven dynamic monitoring and prediction method is proposed, and a multi-scale data set is constructed by fusing multi-source heterogeneous data such as geology, engineering and monitoring. Combining deep learning with reservoir engineering mechanism, key features are extracted and a prediction model with physical constraints is established. With the help of online learning and adaptive optimization algorithm, the model can be updated in real time and the injection-production strategy can be adjusted dynamically. Taking the N21 layer of Bohai BZ25-1 Oilfield as an example, the results show that the new method can closely track the actual trend, and the maximum deviation is only 4.2% in 18 months, which is 60% less than that of the traditional method. The single iteration of online sliding window updating method only takes 1.5 hours, and the fitting frequency is raised to the weekly level, which significantly improves the model updating efficiency and response speed.