Well inflow performance relationship (IPR) is defined as he relationship between the volumetric production rate and lowing bottomhole pressure. Determination of well IPR is ssential for well performance evaluation and optimization. ell IPR depends on pay zone thickness, rock permeability, luid viscosity, wellbore radius, drainage area, and skin factor. ithout knowing all values of these parameters, well IPR urves can be established using various empirical models. A inear IPR can be used for the production from undersaturated il reservoir. In this form the liquid production rate is directly roportional to the pressure drawdown defined as average eservoir pressure minus flowing bottomhole pressure. The roportionality factor is called productivity index. For aturated oil reservoirs, Vogel's equation can be used as an PR model to account for multiphase flow. The IPR of gas wells an be modeled using Forchheimer equation or Backpressure quation. Predicting well inflow performance relationship ccurately is very important for production engineers. Well nflow performance can be predicted using neural networks. In his study, a neural network models used simulation result to onstruct the IPR for oil wells considering all the important actors. The new models give better match than Vogel model.
The inflow performance relationship (IPR) describesthe relationship between well flow rate and bottomhole ressure. IPR helps petroleum engineers to optimize roduction, identify the optimal design for artificial lift, nd predict future production following a stimulation reatment.
Vogel1 built a simulation model using Weller's2 ssumptions for solution-gas-drive reservoirs. Vogel1 roduced IPR curves for different cases, and then plotted imensionless IPR curves for these cases. Noticing that ost of these curves exhibited the same behavior, he used egression to derive an equation that related flow rate to lowing bottomhole pressure in an equation for a curve. rtificial neural networks have been widely used3 and re gaining attention in petroleum engineering because of heir ability to solve problems that previously were difficult or even impossible to solve. Neural networks ave particularly proved their ability to solve complex roblems with nonlinear relationships.
Shippen4 developed a neural network model for rediction of liquid holdup in two-phase horizontal flowand the results exhibited better overall performance than ther existing methods.
Zambrano5 developed a neural network model to redict dewpoint pressure for retrograde gases; the model howed better estimation than all existing correlations. An artificial neural network is a network that imulates the learning processes in the human brain. It uilds its own model on the basis of given information, nd then estimates an output from new input. The etwork consists of neurons and connections between hem. In neural network language, we call the onnections weights. Specific values are stored in those eights to simulate the human brain. The main advantage of the neural network is that it can learn from given input nd output and establish its own model and relationships etween output and input to estimate future values.
I used the neural network to build three models. Two odels are able to establish the IPR curve with an overall ccuracy better than Vogel's equation and with a easonable accuracy for the third model.