Lithology Recognition Research Based on Wavelet Transform and Artificial Intelligence

Lithology identification is one of the main application directions of deep learning in oil and gas field development. Artificial intelligence models can effectively improve the efficiency of oil and gas field development and on-site construction. By using wavelet denoising technology, 15 764 logging data from a block are denoised with 7 features to improve the signal-to-noise ratio of logs, and then random forest, XGBoost, and support vector machine models are constructed. The model is used to compare and evaluate the prediction effect of the model by using the precision, recall rate and F1 score. The research results show that the wavelet noise reduction technology can effectively improve the signal-to-noise ratio of the logs and highlight the lithological characteristics. After noise reduction, the XGBoost model performed the best, with test set accuracy, recall, and F1 score all reaching 0.998. The support vector machine adopts the data training after noise reduction, and the prediction effect is improved most obviously, and the total score is increased by 9.2%.