Study On Impedance Inversion Method Of Longitudinal And Shear Waves Based On BP Neural Network And Well Constraints

Inversion is the process of converting the seismic data obtained from the field into more concrete geological rock information through the geophysicists' hand, we can transform the mathematic-physical model into the field-source model that can reflect the rock information by inversion. The conventional linear inversion tends to have a high degree of dependence on the initial model, and it is prone to local extremum. Based on the above reasons, Some scholars began to introduce nonlinear ideas such as artificial neural network technology into inversion, BP neural network is the algorithm that can give an expression of the essence of artificial neural network technology.The BP neural network algorithm is introduced into the wave impedance inversion, and the logging data is taken as the output learning sample and a series of mathematical transformation processing is carried out with the seismic data to extract a variety of seismic attributes. The method of selecting attribute which is most helpful to the inversion is used to find the attributes as the input learning sample, and the mapping between the input and output is established through the training process.Then, the non-linear mapping relationship between the finite samples is extended to the section or slice to achieve the propagation of the wave impedance.In this paper, we did the joint inversion of P-wave and S-wave after the design of the BP neural network algorithm. Taken the longitudinal wave inversion as the main and the shear wave inversion as the auxiliary constraint instead of a single longitudinal wave data inversion, we tried to make use of the difference of wave impedance in the aspect of oil and gas reservoir to carry on cluster analysis to realize reservoir identification. In this paper, this work all through the program and the esults showed that with logging data as constraints, impedance inversion of wave data with BP neural network algorithm, and the longitudinal and transverse wave impedance are analyzed by cluster analysis, the prediction results obtained with the actual drilling data have a high degree of agreement and achieved an desired effect.