DEVELOPMENT OF AN ARTIFICIAL NEURAL NETWORK FOR PREDICTION OF FLOWING BOTTOMHOLE PRESSURE IN VERTICAL OIL WELLS
Open Access
- Author:
- Rodge, Mohnish
- Area of Honors:
- Petroleum and Natural Gas Engineering
- Degree:
- Bachelor of Science
- Document Type:
- Thesis
- Thesis Supervisors:
- Turgay Ertekin, Thesis Supervisor
Turgay Ertekin, Thesis Honors Advisor
Hamid Emami-Meybodi, Faculty Reader - Keywords:
- Pressure gradient curves
Artificial Neural Network
Well Test Analysis
Petroleum and Natural Gas Engineering
Multirate Well Test
Wellhead Pressure
Bottomhole Pressure
Wellbore Pressure
Oil
Natural Gas
artificial lift
multiphase flow - Abstract:
- Pressure gradient curves are working multiphase flow correlations used by a production engineer in the oil and gas industry. These curves take into account various practical ranges of physical parameters like flow rates, tubing sizes, gas-liquid ratios and fluid properties. The prediction of flowing pressure changes in vertical multiphase wellbores becomes imperative to choose optimum production strings, predict accurately flowing bottom hole pressure and designing of efficient artificial lift systems. An artificial neural network (ANN) is an efficient and an adaptive computer model that resembles the working of a human nervous system. A neural network is able to apply the knowledge gained through training to predict the output within a given training range. After training the neural network by inputting data from the flowing pressure traverse curves, the resulting ANN model can predict the pressure changes within a vertical wellbore. For more accurate results, various ANN architectures with a different number of neurons, hidden layers, learning rate and minimum ranges of error were used for obtaining a better-trained network. The vertical wellbore flowing pressures will be the predicted output of the designed artificial neural network. A well-test analysis that utilizes the values of wellbore flowing pressure to measure the permeability is performed to test the applicability of the ANN model. The success of this model will eliminate the need of wireline equipment to be lowered into the wellbore to measure the wellbore flowing pressure.