Adicionando informação a priori ao formalismo do FWI através da entropia relativa

Abstract

Full waveform inversion is a technique used to obtain the subsurface velocity model with high accuracy. In the oil industry (mainly in reservoir characterization), it is common to use well information as priori information to reduce the ambiguity of the results obtained. In this sense, this work proposes to add priori information in the formalism of full waveform inversion through relative entropy. The application of entropy in inverse problems generally formulates the problem to allow the use of statistical concepts. However, there is an equivalence between the statistical and “deterministic’ formulation. Therefore, in this work, the proposal is to use relative entropy in a deterministic way. Some aspects of relative entropy will be discussed and three different ways of using it to add information to priori in the inverse problem. In regions where salt is present in the BP 2004 model, significant improvements were obtained by adding priori information to through relative entropy for synthetic data. It will be shown that adding information to priori through relative entropy, in the full waveform inversion formalism, can help to avoid local minima.