Artificial-neural-network (ANN) based proxy model for performances forecast and inverse project design of water huff-n-puff technology

Water huff-n-puff is an effective improved oil recovery (IOR) technology for tight or low-permeability reservoirs which are developed by fractured horizontal wells. This paper trains a forward-looking and an inverse project design ANN based models by using large amount of sample data generated from embedded discrete fracture model (EDFM). The models fully consider reservoir properties, fracture geometry, capillary pressure, relative permeability and other fluid properties, and engineering design parameters. The forward-looking model can accurately and efficiently predict the oil/water production rates according to input reservoir parameters and schedules, the reverse design model can determine the reasonable water huff-n-puff schedule according to the expected average daily oil production rate. Besides, the models are applied to fast parametric studies of capillary pressure, fracture related parameters to analyze the influencing factors of water huff-n-puff process, and uncertainty analysis of the inverse design ANN model is conducted by Monte Carlo Simulation. In all, the developed ANN models can help petroleum engineers make better use of this important IOR technology of water huff-n-puff for low permeability or tight reservoirs.