Application of Neural Networks to Modeling Fluid Contacts in Prudhoe Bay

Abstract Modeling the dynamic behavior of fluid movement in an oil reservoir is complicated because of non-linear interactions between reservoir heterogeneity and fluid flow. Simple regression, geostatistical, and numerical simulation techniques have been used in the past to model fluid movement with various degrees of success. Most of these methods, however, suffer from common drawbacks that they are time consuming and difficult to automate. This paper presents a new method based on artificial neural networks (ANNs) to model dynamic gas-oil contacts in the Prudhoe Bay reservoir. This method is fast, efficient, and highly automated and requires minimum user intervention. The proposed method uses oil, gas, and water production, perforation history, permeability, sand and shale distribution, and surveillance data at surrounding wells as ANN input to predict the fluid distribution a t a target well. A two-step method is developed to design an ANN. The first step trains the network using previously measured data as input and output, and thus establishing the internal rules of the network. The second step uses the trained network to estimate the fluid distribution at target wells. Results show that the ANN method can predict fluid distribution at target wells more accurately and consistently than conventional methods.