Fuzzy logic and grey clustering analysis hybrid intelligence model applied to candidate-well selection for hydraulic fracturing in hydrocarbon reservoir

Candidate-well selection (CWS) aims to recognize wells that have potential for higher production after hydraulic fracturing stimulation in petroleum development process, which is natural nonlinear, strong-coupling, uncertain, multi-input, and single-output mathematical problem. CWS hybrid intelligence model is developed by integrating widely applied fuzzy logic systems (FLS), namely, type-2 Takagi-Sugeno-Kang (T2-TSK) FLS, with grey clustering analysis (GCA) for hydraulic fracturing in H gas field of Sichuan Basin, one of the large natural gas field in Southwest of China. The T2-TSK FLS is constructed based on field data involving 49 fractured wells, while the GCA is used to determine the dominant input variables data, and these dominant variables have great influence on post-fractured production. Then we use 39 fractured wells data to train the T1-TSK and T2-TSK FLS to predict post-fractured production. The accuracy of the trained models is validated by comparing predicted post-fractured production with real post-fractured production for the rest of the 10 fractured wells. The evaluation results for the gas field case demonstrate that the T2-TSK FLS is superior to the traditional T1-TSK FLS for CWS using the same input data. The T2-TSK FLS developed in this paper gives high accuracy predicted post-production in H gas field, which is very helpful in selecting the candidate well exactly for hydraulic fracturing.