Waterflooding Interwell Connectivity Characterization and Productivity Forecast with Physical Knowledge Fusion and Model Structure Transfer

Waterflooding reservoir interwell connectivity characterization is the fundamental work in oil development, aiming to inverse vital connecting channels between injectors and producers. In this paper, we endow an artificial neural network (ANN) with strong interpretability through ordinary differential equation (ODE) of material balance equation, proposing a physical knowledge fusion (PKFNN). addition, proposed model could inherit learned from different injector–producer pairs, fully improving training efficiency. way, PKFNN combines merits both machine learning approaches. Firstly, based on control law ODE highly transparent modular architectures framework feedforward network. our has high excellent approximation ability, combining The shows great performance productivity forecast reflection several experiments. provides novel way enhance robustness data-driven-based connectivity-analyzing models by integrating waterflooding reservoirs.