Prediction of Shale Gas Production by Hydraulic Fracturing in Changning Area Using Machine Learning Algorithms

Machine learning has been widely used for the production forecasting of oil and gas fields due to its low computational cost. This paper studies the productivity prediction of shale gas wells with hydraulic fracturing in the Changning area, Sichuan Basin. Four different methods, including multiple linear regression (MLR), support vector machine (SVM), random forest (RF) and artificial neural network (ANN) are used, and their performances are compared by the value of the mean absolute percentage error to determine the best method of all. The training and validation results show that the MLR and SVM methods exhibit poor performances with relatively high errors (> 15%), while the ANN and RF methods show obviously better results, where the RF has a median error (~12%) and the ANN has the smallest error (<10%). After the production forecasting, the particle swarm optimization is implemented as a parameter optimization approach to improve the gas production, which can be increased by around two times after optimization. This study provides a guideline for the shale gas production via hydraulic fracturing in the Changning area.