Module 15: Machine Learning for Integrating Well Data and Seismic Attributes

SubsurfaceAI 2024 proudly introduces Module 15, a revolutionary advancement incorporating machine learning to harmonize well data with seismic attributes seamlessly. This module is specifically engineered to bolster reservoir characterization and predictive analysis, employing cutting-edge data analysis and machine learning algorithms.

Key Features:

  1. Training Data Preparation: Simplify the preparation of training data for supervised classification and reservoir property prediction. Utilize upscaled well data, customized for specific well types or based on expert knowledge. Automatically label training data with facies logs or defined polygons, streamlining the model training process.
  2. Data Analysis Tools: Equip yourself with a comprehensive analytical toolkit to delve into the intricacies of your training data, featuring:
    • Histogram distributions for visual data variability insights.
    • Correlation tables to uncover variable interrelationships.
    • Cross plots for comparative analysis between upscaled well data and seismic attributes.
    • Heatmaps to easily identify data anomalies and patterns.
  3. Attribute Dimension Reduction and Selection: Enhance model performance by pinpointing the most predictive seismic attributes using sophisticated algorithms like Principal Component Analysis (PCA), Neighborhood Component Analysis (NCA), and Step-wise Regression. This optimizes attribute selection for heightened prediction accuracy.
  4. Unsupervised Classification: Craft detailed seismic facies maps, grids, or volumes employing advanced algorithms such as Self-Organized Maps (SOM) and Hierarchical Classification. This feature enables the discovery of unique geological features without the need for predefined labels, supporting exploratory data analysis and interpretation.
  5. Supervised Classification: Apply supervised classification algorithms for precise mapping and volumetric analysis of facies logs, lithofacies, and potential hydrocarbon zones. Supported algorithms include Waveform Correlation Maps, Deep Learning Neural Networks, and Bayes Classification, ensuring effective integration of well facies logs with seismic data.
  6. Reservoir Property Prediction: Leverage optimally selected attributes to predict reservoir properties using advanced algorithms like Multivariate Linear Regression, Deep Learning Neural Networks, XGBoost, and Random Forest. Generate pseudo-well log volumes to deepen reservoir understanding and inform decision-making.
  7. Machine Learning Performance Analysis: Access a robust toolkit for machine learning model evaluation. The Performance Window reveals the influence of each input attribute on the target variables and their relative importance. Visualize individual wells’ contributions and verify facies classification accuracy across different classes.

Module 15 of SubsurfaceAI 2024 marks a pivotal step forward in the synthesis of well data and seismic attributes, offering unparalleled tools for superior reservoir characterization and predictive accuracy. Embrace the future of subsurface analysis with Module 15.