Application of artificial intelligence in predicting the dynamics of bottom hole pressure for under-balanced drilling: Extra tree compared with feed forward neural network model

This study used six fields data alongside correlation heat map to evaluate the field parameters that affect accuracy of bottom hole pressure (BHP) estimation. The oil were acquired using measurement while drilling device collect surface measurements downhole drilling. For two case studies, measured wellbore filled with gasified mud system was utilized, and wellbores drilled rotary jointed drill strings. Extremely Randomized Tree feed forward neural network algorithms develop models can predict high accuracy, BHP from data. modeling purpose, an extensive used, proposed model further validated new fields. gathered encompasses a variety well data, general information/data, depths, size, depths. developed compared obtained based on its capability, stability accuracy. result model’s performance error analysis revealed Extra Feed Forward replicate R2 greater than 0.9. values for suggest relative reliability modelling techniques. magnitudes mean squared absolute percentage predicted BHPs both range 0.33 0.34 2.02%–2.14%, tree 0.40–0.41 3.90%–3.99% respectively; least errors recorded model. Also, (9.13–10.39 psi) are lower (10.98–11 psi), thus showing higher precision Literature has shown underbalanced operation does not guarantee improvement horizontal well’s extension ability, because it mainly depends relationship between bottomhole corresponding critical point. Thus, application this predicting trends.