Oil Well Drilling Parameters Estimation And Lithology Recognition Based On Genetic Algorithms And Neural Network

At present, more and more attention has been paid to the application of the intellectual technology such as genetic algorithms and neural network in petroleum exploration and production.The oil well drilling parameters real-time estimation and lithology recognition have become one of the important tasks for the researchers. While the key factors that influence the recognition ability and velocity lie in the recognition method used and the real-time ability and stability of the algorithms. For this reason, based on intellectual technology such as genetic algorithms and neural network, this dissertation presents a more systematic and deeper study on the oil well drilling parameter estimation and the lithology recognition method.By analyzing the working mechanisms and the basic method of genetic algorithms, a general genetic algorithm for nonlinear varying parameter estimation based on float encoding program is proposed.Aiming at the deficiency of the traditional genetic algorithms in dealing with oil drilling parameters real-time estimation such as the slow convergence velocity and the poor local searching ability, an improved genetic algorithm is proposed and realized by computer simulation. Combined the example of oil drilling parameters rval-time estimation,the control parameter of genetic algorithm influencing the properties of the algorithm is analyzed.On the basis of analyzing the working process and deficiency of the backpropagation neural networks ART-2 network and fuzzy-adaptive Hamming net, improved algorithms have been proposed respectively, the softwares for the algorithms have been performed and all these have been proved by experiments.A new idea for using the oil well drilling parameter curve trend variation to realize real-time lithology recognition, which is based on theanalysis of the oil well drilling parameter curve and the data about core lithology, has been proposed. The improved algorithms of BP network, ART-2 network and Fuzzy-adaptive Hamming net have been applied in the real-time lithology recognition and classification. And all these have been proved by the computer simulation experiments.