Characterization of deep water turbidite channels and submarine fan lobes using artificial intelligence; Case study of Frem Field deep offshore Niger Delta

https://doi.org/10.1016/j.jafrearsci.2020.103852Get rights and content

Highlights

  • Artificial neural networks is very useful and effective in dep water reservoir characterization in the Niger Delta.

  • Unsupervised artificial neural network using competitive learning algorithm (CLA) can improve and better predict lithofacies as well as architectural patterns of deep water turbidites in the Niger Delta.

  • Submarine fan lobes and submarine channels complexes trending northwest southeast were imaged using CLA.

  • The conceptual model built can be used to constrain reservoir static models and could serve as analogues for turbidite sand within the same geologic setting in the deep-water Niger Delta.

Abstract

Two lithofacies and fluid discriminating seismic attributes are integrated using Artificial Intelligence (AI) via Unsupervised Artificial Neural Network (UANN) to characterize the architecture of deep water turbidite channels and submarine fan lobes across a hydrocarbon bearing reservoir within the Frem Field deep-water Niger Delta. A data-based approach including reservoir identification, environment of deposition prediction, seismic attribute analysis and finally UANN using the competitive learning algorithm (CLA) was used to match patterns from the two seismic attributes in order to reduce and capture uncertainties inherent with characterization of turbidite sands within the stratigraphic and structurally complex deep-water Niger Delta. One hydrocarbon bearing reservoir (Sand R001) with excellent reservoir quality was identified from the wireline logs interpretation after which gamma ray logs motifs as well as root mean square (RMS) amplitude and sweetness attributes imaging revealed the environment of deposition of the sand as an inner fan channel within a complex system of several channels and submarine fan lobes. Discreet facies map generated from the CLA enabled a better definition of the architecture, orientation and trend of the sand and lobate nature of the submarine fans lobes associated with the reservoir. The resulting output led to an enhanced characterization of the architectural patterns of the reservoir as well as associated deep-water facies in terms of reservoir architecture and orientation. The discreet facies map also revealed both northeast southwest and northwest southeast orientation of turbidite channels and submarine fan lobes and indicates the channels serves as feeders to the lobate submarine fan systems. The study has shown the efficacy of AI in enhancing deep water architectural patterns via pattern matching of facies and fluid related seismic attributes using CLA and thereby shows the method is effective in reducing uncertainties inherent with deep water reservoir characterization in the Niger Delta.

Introduction

Artificial Intelligence (AI) has been defined as the study of ideas that enable computers to do the things that make people seem intelligent (Schlumberger's Oilfield Glossary, 2011). In recent years, AI has shown an ever-increasing trend in development and affected a wide variety of research and commercial fields in the oil and gas industry (McCoy and Auret, 2019). Aside from the oil and gas industry, AI has found its applications still in in different areas, such as significant improvements to translation of different languages, self-driving automobiles, voice and facial recognition, online shopping, smart phone and furniture, video streaming and social media platform (Sutton and Barto, 2018).

Machine leaning (ML) is a branch of AI based on the biological learning process. It is an effective approach for both regression and/or classification of nonlinear systems involving a few or literally thousands of variables (Lary et al., 2016). Unsupervised Artificial Neural Network (UANN) is a type of classification-based Artificial Neural Network (ANN) capable of Machine Learning (ML) as well as pattern recognition. It can be used as a means of integrating different sets of inputs for facies classifications (Johann, 2001).

The Frem Field is located in the deep-water Niger Delta where both stratigraphic and structural complexities are known to make the morphology and architecture of channel sands and submarine fan lobes which are key elements in reservoir characterization and deep-water hydrocarbon exploration and exploitation difficult to define (Nyantakyi1 et al., 2015). These uncertainties however, are not randomly generated and it can be reduced by better use of integration of different geologic information, technology, scientific concepts and constraints (Reza et al., 2013) and one of such ways to capture and reduce these uncertainties has been through the applications of artificial intelligence. It is important to know though that several authors such as Aminu and Oloruniwo (2008), Njoku et al. (2013), Oyeyemi et al. (2018), Fajana et al. (2019), and Fajana (2020) have worked on using multilayer feedforward neural network algorithms (MLFFNN) or multi-layer perceptron neural network (MLPNN) to characterize reservoirs in the Niger Delta, however, few authors have exploited and highlighted the importance of the use of competitive learning algorithm (CLA) to capture risk and uncertainties associated with stratigraphic changes in the deep offshore Niger Delta.

CLA in unsupervised neural network is highly effective in accentuating salient stratigraphic features based on its competitive feed forward processes (Johann, 2001) and thereby making it appropriate in capturing uncertainties in complex geologic settings where heterogeneity and anisotropic is difficult to predict and capture (Al Moqbel and Wang, 2011). Since, the deep-water Niger Delta is a complex geologic setting as explained earlier, the use of CLA approach was necessitated in this study. Consequently, this study aims at using CLA to integrate and match patterns from seismic attributes to improve the characterization of the architectural patterns of turbidite channels and submarine fan lobe systems within the field. The specific objectives are to; identify a seismically resolvable hydrocarbon bearing reservoir, predict environment of deposition of the identified reservoir, use seismic attributes to predict lithofacies and fluid distribution across the reservoirs and finally use as inputs seismic attributes into the CLA to improve the imaging of the architectural patterns of the reservoir.

Section snippets

Location of the study area

The study area is located on the upper continental slope downward the transition between the extensional and compressional belts, offshore western Niger Delta (Fig. 1). It covers an area of about 638 km2 at a distance of approximately 120 km from the present day 124 coastline.

(Fig. 1) with water depths ranging between 810 and 3240 m. The area is predominantly characterized by mobile shales, mass transport deposits and channelized submarine fan systems and lobes.

The occurrences and architecture

Theoretical background

Artificial Neural Networks (ANN) uses information processing algorithms that try to mimic the human brain. Unlike conventional computation algorithms that always follow the same programmed steps independently of the input data, neural networks learn by trial, using a set of inputs, and so are not programmed to perform a specific task (Azevedo, 2009).

ANN works like Biological neural networks which are made up of billions of neurons interconnected, building extremely complex information networks.

Dataset

The dataset used for the study includes a 3D seismic volume, five wells (Frem-001, Frem-002ST3, Frem-003ST1, Frem-004ST1 and Frem-005) with composites suites of well logs including gamma ray log, resistivity, sonic, density and neutron logs respectively. The seismic volume includes a 1x3D post stacked time migrated seismic data with frequency range of 10–125 Hz covering an area of 1814 km2. The horizontal resolution of the data is characterized by a stacking bin spacing of 25 m by 25 m (in-line

Reservoir identification and well logs E.O.D characterization

Amalgamated submarine turbidite channels, submarine fan lobes as well as submarine turbidite channel margin deposits were identified form the well logs across the field (Fig. 9). This generalized E.O.D characterization from the GR logs motifs from all the wells indicated the predominant deep-water facies within the field. The general stratigraphy of the field however is that of sands encased in deep water shale.

GR log signatures for sand R001 also shows sharp erosive base and a blocky signature

Conclusions

One hydrocarbon bearing sand was identified and analyzed. Gamma ray logs motifs and seismic attributes analysis suggests the E.O.D of the sand to be an inner fan channel within a complex system of turbidite channels and submarine fan lobes. The UANN using CLA analysis enabled the generation of discreet facies map which is a better predictor of the distribution and architectural patterns of the sand and associated submarine channels and fan lobes systems across the reservoir. The method has

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements

This is to acknowledge Shell Petroleum Development Company of Nigeria for providing the data used for the study and the University of Lagos Geosciences Department for the use of the Workstation lab.

References (22)

  • H. Doust et al.

    American “Niger delta,”

  • Cited by (0)

    View full text