Study On Fracturing Optimization Model Of Shale Gas Well Based On Machine Learning

The fracturing mechanism of shale gas wells is complex,and there are many factors affecting the production of a well.When the difference of reservoir geological characteristics between wells is small,different fracturing parameters often lead to large differences in production between wells.This paper establishes a sample of study data based on geological,engineering and production data from 137 wells in the Weiyuan Block.Pearson correlation coefficient method was used to analyze the correlation between the 11 influencing factors and the production,and between influencing factors;the principal component analysis method was used to calculate the weight of each factor;and the average impurity reduction method was used to calculate the degree of influence of each factor on production.The less influential factors were excluded from the data sample.Six production prediction models based on six machine learning algorithms such as Random Forest,BP Neural Network and XGBoost are built,and four metrics are used to evaluate the model performance,and the results show that the XGBoost model performs best.A set of 14,805 block fracturing plans was created,single-well fracturing plans were screened based on six metrics,the XGBoost model was used to predict the production of these plans,and a combination of production and input-output ratios were used to select the final fracturing plans.In order to validate the fracturing parameter optimization model,17 wells with similar fracture section length and physical properties were selected for the model application test,and the production of the model-optimized matched wells ranked second,with an average cost of 4.3% lower and 34.7% higher than the standard wells in other tests;9.7% lower than the cost of the most productive wells,and only 1.4% lower than their gas production.It is proved that the fracturing parameter optimization model established in this paper is well applied and has practical value in the oil field.In the end of this paper,the fracturing scheme of a well is optimized by using the established fracturing parameter optimization model,and the gas production is expected to increase by 21.7%.