A gas-bearing prediction model based on Ensemble Empirical Mode Decomposition and Deep Neural Network

To address the problem of weak seismic response and identification of hydrocarbon in deep reservoirs, a gas-bearing detection model based on the Ensemble Empirical Mode Decomposition (EEMD) and Deep Neural Network (DNN) is proposed. First, in order to avoid the problems of mode aliasing in the conventional empirical mode decomposition (EMD), the seismic signals are decomposed by the EEMD method. Then, a series of intrinsic mode functions (IMF) are obtained by EEMD, and the first two orders of IMF are chosen to reconstruct the seismic sub-signals. Finally, a deep neural network model is built to extract the weak fluid information straight from the reconstructed seismic data. The experimental results show that the DNN model trained on reconstructed seismic sub-signals generalizes better than the DNN model trained on raw seismic data, and can effectively indicate gas-bearing information while decreasing reliance on expert knowledge.