AI Method for Development Potential Prediction with High Water Cut Reservoir

Medium to high permeability reservoirs have generally entered the high water cut development stage, and maintaining stable production requires continuous increase in adjustment measures. Quickly and accurately identifying potential development areas is a prerequisite for implementing measures. At present, the evaluation of potential areas mainly relies on numerical simulation, which have a large computational workload and time consuming. Meanwhile, traditional simulation methods have insufficient ability to characterize the dynamic heterogeneity during the high water cut stage. The numerical simulation agent model based on artificial intelligence has realized the acceleration of optimization calculation by a single model, but the general generalization ability is significantly insufficient and does not have the potential evaluation ability. This article improves the numerical simulation algorithm for high water cut oil reservoirs, establishes saturation samples for typical medium high permeability oil reservoirs, and comprehensively considers of oil enrichment and flow capacity. The whole reservoir is divided into three potential areas: water consumption area, affected area, and inactive area, achieving the establishment of learning samples and effectively avoiding problems such as large data volume, complex rules, and difficult training in directly learning saturation parameter fields; Build a ConvLSTM model, effectively mining the correlation between parameters such as porosity, permeability, and pressure. Through training, establish a fast prediction model from dynamic data, reservoir boundary and physical property parameters to development potential zoning. The results show that the prediction speed and accuracy of development potential zones are significantly improved