Enhanced Oil and Gas Production Forecasting Through Stacked generalization Ensemble Learning Technique

Planning a strategy throughout the oil and gas sector depends on production forecasting. Precise projections aid in estimating future output rates, streamlining processes, and effectively allocating resources. Techniques like “ Decline Curve Analysis (DCA) and Numerical Reservoir Simulation (NRS) ” have been used in the past, but they have drawbacks such reliance on static models and time consumption. A stacked generalization ensemble learning method for predicting oil and gas production is presented in this work. Using Python and data from wells in the state of “New York State”, the model contains four machine learning techniques: “ Random Forest Regressor (RFR), Extremely Randomized Trees Regressor (ETR), K-Nearest Neighbors (KNN), and Gradient Boosting Regressor (GBR) ”. The stacked model works better than separate models, according to the results of experiments, via R2 scores of 0.9709 per oil and 0.9998 per gas.