Research On Method Of Fracturing Well Selection Based On Two - Type Fuzzy Logic System

Hydraulic fracturing is the main technology of increasing oil as well as gas production and enhancing oil recovery. The selection of candidate wells and fracturing intervals is the first step to the implementation of hydraulic fracturing, and it is also the key to the whole process of hydraulic fracturing, so in order to make hydraulic fracturing efficient, candidate-well selection becomes particularly important. The accurate selection can not only reduce fracturing risk and save time, but also provide more technological choices of reservoir stimulation.The long practice of hydraulic fracturing shows that the procedure of candidate-well selection is complicated, and there is no universal method to solve it. The main reason is that selection of candidate wells is the complex and uncertain system of high dimension, nonlinear and strong coupling. Complicated geology and strong heterogeneity characteristics of reservoirs also determine its complexity. So it is difficult to use mathematical models to precisely describe it. The Fuzzy Logic System (FLS), dealing with complex and uncertain systems effectively, are widely used in nonlinear system identification.Conventional FLS, fully described by the type-1 fuzzy set, is the type-1 FLS. Because membership degree of type-1 fuzzy set is a precise value, it can’t deal directly with the uncertainty of rules. Type-2 FLS, described by at least one type-2 fuzzy set, is an extension of type-1 FLS. The membership degree is fuzzy, so type-2 FLS can remedy type-1 FLS’s defects. When it is difficulte to confirm the accurate membership function of fuzzy sets, type-2 FLS is very useful, because it can be used to deal with the uncertainty of rules, and even the uncertainty of measurement. So the thesis, based on Hechuan gas field in Sichuan, builds type-2 FLS for the research of candidate-well selection. The main research contens are as follows:(1)For the establishment of index system of candidate-well selection, qualitative analysis is used to find all the factors of influencing the fracturing effect at first. Then logical analysis is used to eliminate redundant factors. Quantitative analysis is applied to the rest of factors. Similarity analysis, based on grey clustering algorithm, is used in geologic factors and fracturing treatment factors, respectively. Subsequently, factor with the biggest correlation degree in each category is selected by grey correlation analysis. After these steps, index system of candidate-well selection is constructed.(2)Singleton type-1 Mamdani FLS and interval singleton type-2 Mamdani FLS are used in the selection of candidate wells. For parameter identification, Singular Value Decomposition-QR Decomposition (SVD-QR) algorithm and improved SVD-QR algorithm determine the number of fuzzy rules of type-1 and type-2 Mamdani FLS, respectively. The antecedent and consequent parameters of type-1 and type-2 FLS are tuned by back propagation method. The designed type-1 Mamdani FLS with the best performance is applied to parameter initialization of type-2 Mamdani FLS.(3)Singleton type-1 TSK FLS and interval singleton type-2 TSK FLS are used in the selection of candidate wells. For parameter identification, Fuzzy C-Means clustering (FCM) algorithm and Fuzzy Subtractive Clustering (FSC) algorithm determine the number of fuzzy rules of type-1 and type-2 TSK FLS, as well as some initial parameters of type-1 TSK FLS. The antecedent and consequent parameters of type-1 FLS are tuned by least square method and back propagation method. Those of type-2 FLS are tuned only by back propagation method. The designed type-1 TSK FLS with the best performance is applied to parameter initialization of type-2 TSK FLS.(4)The BP artificial neural network is used in the candidate-well selection. According to the criterion for reservoir’s stimulation potential evaluation, gained by mathematical statistic analysis and expert opinions, support vector machine and fuzzy neural network based on genetic algorithm are applied to the reservoir’s stimulation potential classification.