Reservoir characterization involves the estimation elastic parameters from well-log data and seismic data, nonlinear elastic parameters estimation based on machine learning is a popular method for reservoir characterization. The common process is to construct a comprehensive mapping relationship between seismic data including multi attributes and well data, but the prediction accuracy may be low when using one prediction model in different sedimentary environments. In order to solve this problem, we put forward the new method of multi-model prediction based on machine learning to estimate elastic parameters directly, in other words, we use corresponding deep neural network to estimate elastic parameters in different sedimentary environments. In this paper, our work mainly includes three parts: firstly, waveform classification based on Selforganization map, secondly, training and optimizing multi prediction models based on Long Short Term Memoryrecurrent neural networks, thirdly, elastic parameters estimation using multi-model. This method is successfully applied in model and real data, it shows the estimation results are more reasonable and effective.
Presentation Date: Monday, October 12, 2020
Session Start Time: 1:50 PM
Presentation Time: 4:45 PM
Location: 351F
Presentation Type: Oral
