ALINE: A Computational System Based on Seismic Data and Machine Learning for Gas Reservoir Detection
Palavras-chave:
Machine Learning, Seismic Acquisitions, Gas reservoirs, ALINEResumo
Reflection seismic is one of the most used geophysical methods by the O&G industry for subsurface
imaging. Through the processing and interpretation of seismic data, geoscientists infer the positioning and geome-
try of potential hydrocarbon accumulations. However, the reflection seismic method can produce ambiguous data
owing to similar signatures in natural bodies with different physical properties. Moreover, onshore seismic data
have in general less quality when compared to offshore seismic data, making the interpretation process even more
difficult. Artificial intelligence (AI) techniques have been adopted in several applications, particularly for the inter-
pretation of salt bodies and geological faults. However, for identification of hydrocarbon reservoirs, AI techniques
are still under development, particularly owing to the great amount of seismic data to be processed. Recently,
Eneva and Tecgraf/PUC-Rio developed the computational system ALINE (Automated Learning Intelligence for
Exploration) based on Machine Learning techniques and seismic data to generate indicators of potential gas accu-
mulations in on-shore fields. In this study, we focus on the description of ALINE’s system, its current capabilities
and methods, advantage and limitations, and future developments. The current methodology uses modern neural
network architectures through the analysis 1D of seimic traces to identify specific signatures of gas accumulation.
Several onshore seismic sections from Parque dos Gavioes at Parna ̃ ́ıba’s Basin were provided by Eneva during the
first validation tests. The results obtained for that region showed an accuracy of 75 - 80% of the gas class and 90
- 95% of the non-gas class. Although other approaches for similar applications are not available in the literature
for comparisons, the global average of success shows that the system has a significant potential for exploratory
purposes. Moreover, ALINE’s system also can be adopted for predictions considering offshore 3D seismic data, as
performed on the Block F3 in the North Sea that enabled better accuracy rates owing the best quality of data. Those
results highlight the potential of ALINE as a computational tool for the interpretation of 2D or 3D seismic data,
onshore or offshore, boosting the value of seismic data and minimizing uncertainties, representing an effective
technological advance in the sector O&G.