Preferred Fracturing Wells Were Studied

Candidate selection is a very important part of the fracturing engineering. There are many factors that affect choosing the right candidate wells. The relationship between them is very complicated and they have different effects on the fracturing result. The traditional methods of choosing the wells are based on experience, which possess definite subjectivity. Several candidate selection methods have been researched by using the present-day mathematics and computer technology for the ZhongYuan oilfield. The main achievements acquired in this paper lie in the following:(1) The main factors that affect the fracturing effect of the ZhongYuan oilfield have been found by using the grey correlation analysis method and fuzzy ranking method. We get the correlation of the input parameters and the output by using the genetic neural network and backward elimination method, which proves that the selected factors are appropriate.(2) The data records may be incomplete in the candidate selection database, which can not reflect the reality of the formations. This paper found the "good" dataset by using fuzzy clustering and fuzzy neuro-cluster data classification system, which can be used in the candidate selection.(3) This paper deeply researches the fuzzy models for the candidate selection including multi stage comprehensive evaluation model, fuzzy decision-making model, fuzzy analysis model, fuzzy comprehensive decision-making model and grey correlation analysis model. The weights of the factors have been found by using level analysis principle of systematic engineering. This paper also researches the factors that affect the fracturing success and their weights.(4) This paper improves the training algorithm of the neural network and researches the methods to improve the generalization ability of the neural network. The neural network configuration and initialization parameters have been decided at last. The fracturing candidate wells are selected by using the neural network.(5) The traditional fuzzy neural network has been improved and we get the improved Takagi—Sugeno fuzzy neural network , which has high accuracy, good global convergence, good generalization ability. The fracturing candidate wells has been selected by using the improved Takagi—Sugeno fuzzy neural network. The network can get the results in a short time and has high accuracy.(6) This paper improves the genetic operators and the fitness function and confirms the main parameters of the genetic algorithm to optimize the neural network configuration and weights. The fracturing candidate wells have been selected by using the optimized genetic neural network. The fracturing treatment proves that the candidate wells are right and the network has good generalization ability and high accuracy.(7) We contrast the candidate selection methods with each other and get theadvantages and disadvantages of every method, so we can use different methods in different conditions. The results shows that the fuzzy method can be easily used but the weights of the factors can not be decided easily. The genetic neural network has the best generalization ability and accuracy, which is the best candidate selection method.