Study on Fracturing Sweet Spot Prediction Method Based on Machine Learning

Fracability evaluation of reservoirs is very important for the optimization of fractured well segments, the optimization of development plans and the prediction of economic benefits. Based on machine learning, this paper summarizes the geological engineering related parameters of 36 horizontal Wells in a block in Mahu area, and obtains nine types of parameters, including Young's modulus, permeability (PERM), porosity, stress difference, minimum horizontal principal stress, pore pressure, Poisson's ratio, total amount of fluid added in a single stage, total amount of sand added in a single stage, and analyzes their correlation with the stimulated reservoir volume (SRV). Based on the prediction effect of SRV and production, the fracturing sweet spot model was reoptimized in the study area. The fracturing sweet spot model in the study area was optimized based on the prediction effect. Multiple linear regression and Pearson coefficient method were used to further provide a comprehensive sweet spot calculation method, and a set of evaluation methods for fracturing sweet spot in the Mahu area was formed. The results showed that the method is reliable and adaptable, which is instructive and useful for clarifying the dominant layers and fracturing sections.