Research on High-resolution Reconstruction of AI Sequence-constrained Phased Logging Curve

Abstract Due to factors such as the development period and the stability of logging instrument accuracy, the early old logging series had missing or low resolution logging curves, which affected the calculation of reservoir parameters and resulted in low accuracy and precision. Logging curve reconstruction is an important means to solve the problem of missing logging curves and improve the resolution of logging curves. The reconstruction of conventional logging curves mainly adopts physical model regression method, but it is difficult to establish suitable physical models between some curves, and the regression of model formula parameters is also difficult. The efficiency and resolution of curve reconstruction are also low. This article proposes a high-resolution reconstruction technique for sequence constrained sedimentary facies control logging curves based on artificial intelligence deep learning. Different wells are aligned according to the sequence to eliminate the effects of depth and thickness differences, solving the problem of traditional methods only considering the correspondence between curve values and trends without considering the influence of vertical depth on curve variation patterns; By discretizing sedimentary facies values into curves, the influence of phase transitions on logging curves is introduced, which solves the problems of traditional methods not considering horizontal variation patterns and insufficient utilization of adjacent well data; Using artificial intelligence methods to solve the problem of complex and difficult to model models with too many parameters. The correlation coefficient between the reconstructed curve and the target curve through acoustic and density tests has reached 0.85 or above, indicating that this technology effectively improves the resolution of logging curves and enhances the accuracy of reservoir parameter interpretation.