A new carbonate reservoir prediction method and its application in the South Atlantic Ocean

Abstract:

With the development of technology, offshore oilfield exploration targets have gradually shifted from shallow to the deep. The characterization of carbonate reservoirs with strong heterogeneity, especially the prediction of fluids in deep-sea pre-salt carbonate reservoirs, has become an urgent problem to be solved in the field of oil and gas exploration. In view of the difficulty of pre-salt carbonate reservoir exploration, this paper proposes a multi-parameter pre-stack inversion method based on rock-physics analysis, Bayesian theory and neural network porosity prediction, which improves the fluid prediction accuracy of complex pre-salt carbonate reservoirs.

We obtained the Lame coefficient inversion results using the pre-stack multi-parameter inversion method, the pre-stack multi-parameter inversion and reservoir quantitative prediction are realized. Then wells with P-wave velocity, S-wave velocity, density, and porosity data were selected, the degree of regression between the logging Lame coefficient and the inversion Lame coefficient is calculated respectively. Extract the Lame coefficient from the seismic inversion and analyse its error with the Lame coefficient calculated by the wells. The parameters errors within 10% are included in the confidence interval, finally the correlation regression between the two is 92%, which indicates that the Lame coefficient from seismic inversion is reliable. It shows that the inversion Lame coefficient can be used for porosity prediction. Wells without shear wave velocity data were added together for analysis, the changing trend of point group was observed, the control parameter of the fitting formula was adjusted, and the neural network training was conducted again to obtain a larger agreement rate. We optimized the control parameters from 10 to 13.5. According to the error analysis of these wells it can be seen that the overall agreement rate increased from 91.2553% to 96.4739% after adjustment, and the new neural network is more suitable for the porosity prediction. This method provides exploration guidance and targets for fluid prediction of carbonate reservoirs, and has been well applied in the Santos Basin of the South Atlantic Ocean.