Journal de géologie et géophysique
Libre accès

Abstrait

Use of Probabilistic Neural Network and Post Stack Inversion to Predict Reservoir Characterization in the Mediterranean Sea, Sapphire Field, Egypt

Ahmed Abosalama*

Geophysical parameters of probable reserves are interpreted using seismic inversion. It is essential for estimating porosity, saturation, and shale content. This article discusses the use of model-based Geophysical parameters of potential reserves are interpreted using seismic inversion. It is essential for determining porosity, saturation, and shale content. This article explores the use of model-based seismic inversion and probabilistic neural networks to characterize reservoirs. To make this assignment easier, the paper is divided into two portions. From 3D seismic data gathered in the research area (Sapphire Deep Seismic-2010), model-based inversion is used to generate acoustic impedance values. Seismic data is used to analyse five well logs. The average correlation coefficient between synthetic and seismic data is 0.997, with a 7% error, indicating the utility of model-based inversion. Second, a Probabilistic Neural Network (PNN) is trained and verified using estimated effective porosity, water saturation, and shale volume. The 3D variations in effective porosity, water saturation, and shale volume are obtained using the validated probabilistic neural network.

Our research revealed an undrilled section in the Sapphir-80 channel with favourable petro physical parameters, indicating a large volume of gas and condensate.

Seismic inversion connects observed seismic data to interpreted elastic physical parameters of probable reserves. Post-stack seismic inversion is used to estimate reservoir parameters such as porosity, saturation, shale content, etc. An application of model-based seismic inversion and probabilistic neural network to post-stack seismic data for reservoir characterization is described. The paper is divided into two pieces for this assignment. Initial post-stack seismic inversion approximating the Acoustic Impedance (AI) values using 3D seismic data recorded in the research area (Sapphire Deep Seismic-2010) in the time domain. Seismic data from five wells was gathered. As shown by 0.997 average correlation coefficient and 7% error between synthetic and seismic data, model-based inversion is effective. Second, a Probabilistic Neural Network (PNN) is trained and validated using data from the well sites. On the seismic volume, the probabilistic neural network calculates effective porosity, water saturation, and shale volume fluctuation in 3D.

The current analysis projected an undrilled area in the Sapphir-80 channel with good petro physical parameters, indicating a large volume of gas and condensate.