Main controlling factors of fracturing effect in low permeability heavy oil reservoir based on data mining

Due to strong heterogeneity and uneven fracturing conditions of each well, it is difficult to guarantee the fracturing effect. For this purpose, modeling parameters were selected by combining correlation coefficient and variance expansion factor on the basis of fracturing data statistics. Secondly, a full subset regression model of the output and its influencing factors was established, and the optimal fitting equation was optimized based on Akaike's information criterion. The results show that the prediction accuracy of the model is 87%. The regression coefficient and relative weight of flowback rate are 0.89 and 0.45 respectively, which are the main controlling factors of fracturing effect. The optimization of the fracturing parameters of 31~# well shows that the production can be increased to 10 t/d when the backflow rate is increased to 52%. Therefore, the mining historical fracturing data can realize the diagnosis of main controlling factors of fracturing effect, and provide quantitative guidance and basis for oil field fracturing operation.