Hydraulic Optimization Predication Using Acritical Neural Networks (ANNs) In an Iraq Oil Field

Independent factors in this study include RPM, torque, differential pressure, hook load, and mud characteristics,whereas the dependent parameters Q, TFA, SPM, and formation variables like depth. An accurate projection of pumppressure alerts the driller to potential issues like circulation issues, washouts, subterranean blowouts, and kicks,allowing for the safe implementation of corrective measures. Presents an in-depth summary of drilling optimizationmethods. The MATLAB fitting tool was used to create an ANN model for hydraulics prediction. This modelsuccessfully predicts pump pressure vs depth in similar formations. Therefore, five input variables and a single outputvariable (pump pressure) were included in a three-layer feed-forward neural network that was trained using theLevenberg-Marquardt back-propagation technique. The ANN technique's simulated results showed a high degree ofagreement between the observed pump pressure and projected values for both the training and test data sets. Based onthe findings, the variables with the greatest influence on this model were: depth, effective diameter, and total flowarea, followed by flow rate, differential pressure, mud characteristics, etc. However, this model should be utilizedwith care owing to the small sample size. Since the information about inclination angle and dog leg severity wasdisregarded, this model may be used for vertical wells.

کلیدواژه ها:

Pump Pressures ، artificial neural network ، backpropagation neural network ، radial basis functional networks ، Hydraulic Fracturing ،