Sensitivity analysis and prediction of vertical multistacked well IPR -tight oil reservoir using artificial intelligence technique

The inflow performance relationship (IPR) can be defined as the relation between the well production rate (Q) and the flowing bottom hole pressure (BHP). There are so many factors that can affect the IPR of a well such as skin factor (S), formation permeability (K), fluid viscosity (μ), reservoir thickness (h) and wellbore radius. Unlike simple vertical wells, multilateral wells drilled through tight reservoirs impose some challenges when trying to model their IPR. This may come as a result of the fact that there is an interchange between the pressure drop behavior through the wellbore and the reservoir performance itself. The main goal of this paper is to generate a model capable of predicting the IPR of the vertical multistacked wells drilled through a low permeability oil reservoir Using Self Adaptive- Artificial Neural Network (SA-ANN). A set of simulated data generated by commercial software to represent our case is used for building our model. Different well and reservoir cases were implemented by changing the well configuration and reservoir properties. The effect of some factors such as horizontal permeability to vertical permeability ratio (kh/kv), number of well laterals (N), distance between laterals (D), length of the laterals (L) and reservoir porosity (Phi) on the dimensionless IPR curves was investigated. SA- ANN is used to create an empirical correlation that can predict the IPR for future well/reservoir cases without the need for the AI model. 70% of the generated data was used for model training while the rest 30% was used for testing the model. The model showed a very good match between the real (Q/Q_max) values and the predicted values. The coefficients of determination (R²) for both the training and testing stages were 0.989 and 0.988 respectively while the mean squared errors for the same stages were 4.08 × 10⁻⁵ and 1.6 × 10⁻⁴ respectively. This strongly shows how efficient the proposed SA-ANN model and its high reliability in predicting the IPR for vertical multistacked wells drilled through low permeability oil reservoirs. For the first time, an empirical correlation was developed based on the optimized SA-ANN model that can be used to predict the IPR for vertical multistacked wells with high accuracy. This technology will help the reservoir and production engineers to better design the well completion and manage the well productivity.