Equalization Optimizer-Based LSTM Application in Reservoir Identification.

Area of Science:

• Artificial Intelligence
• Machine Learning
• Geoscience

Background:

• Reservoir identification is crucial yet challenging in well logging.
• Long Short-Term Memory (LSTM) networks show promise but have limitations.
• Optimizing LSTM parameters is key to enhancing classification accuracy.

Purpose of the Study:

• To improve the accuracy of LSTM-based reservoir identification in well logging.
• To introduce an improved equalization optimizer algorithm (TAFEO) for LSTM parameter optimization.
• To evaluate the effectiveness of the TAFEO-optimized LSTM model.

Main Methods:

• Developed the Tent Chaotic Mapping-based Equalization Optimizer Algorithm (TAFEO).
• Applied TAFEO to optimize LSTM neurons and parameters for reservoir identification.
• Validated TAFEO using benchmark functions and Wilcoxon rank-sum test.
• Evaluated the optimized LSTM model using Receiver Operating Characteristic (ROC) curves and UCI datasets.

Main Results:

• TAFEO demonstrated superior accuracy and convergence speed compared to other optimization algorithms.
• The TAFEO-optimized LSTM model achieved a maximum Area Under the ROC Curve (AUC) of 99.43% on UCI datasets.
• In practical well-logging applications, the TAFEO-optimized LSTM model reached a recognition accuracy of 95.01%.

Conclusions:

• The TAFEO algorithm effectively optimizes LSTM for enhanced reservoir identification.
• The proposed method significantly improves accuracy and robustness in well-logging applications.
• This approach offers a more effective solution for reservoir identification compared to existing methods.