Estimation of Petrophysical Properties Using Linear Programming Sparse Spike Inversion and Deep Feed-Forward Neural Network Techniques Over F3 Block, Netherlands: A Case Study

Abstract

In this study, acoustic impedance (P-impedance) distribution in the subsurface of the F3 block, Netherlands is determined using the linear programming (l₁-norm) sparse spike inversion (LPSSI) method. The objectives of the study are to characterize the sand channel and extract high-resolution subsurface rock features from the low-resolution seismic data. To estimate rock properties from seismic data, a variety of conventional post-stack seismic inversion techniques are available. However, the LPSSI technique is a reasonably quick and easy-to-compute subsurface model that can be employed for both quantitative and qualitative interpretation. The method is employed in two steps: first, composite traces close to well locations are retrieved and inverted for acoustic P-impedance, and then optimization of the LPSSI parameters is done using comparison with well log impedance. According to the analysis of the composite traces, the algorithm performs well and has a high average correlation (0.98). The F3 block seismic data are utilized in the second stage to estimate the distribution of acoustic impedance in the subsurface by using the LPSSI method. A sand channel-like low impedance anomaly with a range of 3800–7400 m/s g/cc is evident in the inverted acoustic impedance analysis at the 1380–1400 ms time interval. Then, using a deep feed-forward neural network (DFNN), many other crucial rock parameters, including porosity, density, and P-wave velocity, were estimated in the inter-well region to corroborate the sand channel. Following the analysis of these petrophysical properties, a high porosity zone (24–40%), low-density zone (1.9–2.02 g/cc), and low P-wave velocity zone (1700–2300 m/s) are present in the 1380–1400 ms time interval, which aligns with the low impedance zone and validates the presence of the sand channel.