Quality Control Method of Exploration and Development Data Based on Machine Learning

Abstract

Because of increased exploration depth and increasingly complex geological environment survey area, makes the actual acquired seismic exploration data contains a lot of noise, the noise composition is serious interference signal effectively, affect the signal to noise ratio and resolution of the seismic data, reduce the quality of the seismic exploration data, to the subsequent inversion and interpretation, and finally brought difficulties such as oil and gas exploration work. This paper mainly studies the quality control method of exploration and development data based on machine learning. In this paper, the classification and source of desert noise are analyzed first, and a convolutional neural network with branch structure (BCDNet) is proposed to enhance the ability of extracting effective signal features from desert seismic exploration data, so as to better recover the seismic in-phase axis polluted by desert seismic random noise.