Gas chimney and hydrocarbon detection using combined BBO and artificial neural network with hybrid seismic attributes

The exact interpretation of structure of the Earth is essential to avoid drilling of dry holes or locating hydrocarbon reservoirs in faulty regions. The economic approaches are required to locate the reservoirs without damaging the tectonic plates of the Earth. The identification of gas chimneys provides an impressive approach to indirectly interpret the hydrocarbon reservoirs as the chimneys form the migration pathway for gas from reservoirs, and it looks like a gas cloud in the seismic data. The proposed method using biogeography-based optimization (BBO) with supervised neural networks enables the effective interpretation of chimneys and hydrocarbons. Seismic attributes are measured on the pre-processed seismic data, where the attributes play an important role in providing qualitative information on structure of the Earth. Continuity, instantaneous and amplitude attributes are measured to provide the details on the areas of discontinuity, bright-spot regions with high amplitude and low frequency. The calculated attributes are picked from selected locations in the pre-processed seismic data, and the hybrid combination of attributes enables effective, economic chimney identification and provides indications to locate the reservoirs. The redundant and highly correlated features are eliminated using BBO which in turn improves the classification accuracy. BBO algorithm retains the best solution in each generation, and the fitness of the algorithm is verified with supervised neural networks. The proposed algorithm is applied on F3 block dataset, and BBO selects 20 predominant features among 75 features. The optimal solution after 50 generations enables the classification of chimneys through multilayer feed-forward supervised neural network. The results indicate that the chimney classification accuracy is improved by 90%, and the mean square error is minimized with BBO as a feature selection algorithm compared to other evolutionary algorithms.