Abstract:
This study integrates seismic attributes using a machine learning algorithm for hydrocarbon zone prediction over the study area. It attempts to resolve the problem of making accurate and efficient interpretations from a large database of derived seismic attributes. The central objectives were to delineate hydrocarbon reservoirs, analyze some seismic attributes, develop and evaluate the predictive models, and validate the developed predictive models. The research utilized post-stack 3D seismic volume for the delineation of structures and generation of seismic attributes. The well logs were used to identify and correlate the tops of hydrocarbon-bearing reservoirs. A computer programme (in Python language) was developed for the generation and performance assessment of predictive models. The binary cross-entropy classification metric was used to evaluate the performance of the developed predictive models. An individual seismic attribute (Maximum Amplitude and Extract value) map was used to validate the predictive models. Three hydrocarbon-bearing reservoirs were delineated with their thickness variations from 39ft (in R4) to 89.1ft in R1. Six faults were mapped across the horizons. The Multi-Layer Perceptron (MLP) model for horizons 1, 2, and 3 predicted hydrocarbon zones with a training accuracy variation of between 0.98 – 0.99 as well as a testing accuracy variation range of 0.96 - 0.989 over their individual epoch intervals. This MLP model map gave higher precision of the predicted hydrocarbon zones over the Self-Organising Map (SOM) predictive model, thus reinforcing the confidence level of the former. The developed computer programme was able to combine seismic attributes that predicted six conformable hydrocarbon zones. Also, enhanced the visibility of the other five hydrocarbon zone predictions as compared to the image interpretation of the individual seismic attribute in the study area.
