Lithology identification method based on feature contribution degree and PSO-Catboost

• Main Authors: DENG SONG, PENG MINGGUO, LI CHAOWEI, XU SHOUKUN, LI QIU, WANG YIWEN, YAN XIAOPENG, WANG SONGBAI, PAN HAOYU, HAO HONGDA, WANG JIANGSHUAI, SHI LIN
• Format: Patent
• Language: Chinese; English
• Published: 26.09.2023
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; DETECTING MASSES OR OBJECTS; EARTH DRILLING; EARTH DRILLING, e.g. DEEP DRILLING; ELECTRIC DIGITAL DATA PROCESSING; FIXED CONSTRUCTIONS; GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; MEASURING; MINING; OBTAINING OIL, GAS, WATER, SOLUBLE OR MELTABLE MATERIALS OR ASLURRY OF MINERALS FROM WELLS; PHYSICS; TESTING

The invention discloses a lithology identification method based on a characteristic contribution degree and PSO-Catboost, and the method is characterized in that the method comprises the steps: collecting well logging and logging data of a drilled well; defining different digital tags for different lithologies, dividing sample data into a training set and a test set, and setting operation iteration times and particle quantity parameters of a PSO algorithm; optimizing a random forest algorithm by using a PSO algorithm; and inputting the optimized features into a Catboost model, and automatically optimizing hyper-parameters of the Catboost model through a PSO algorithm to obtain a model with an optimal lithology recognition effect. According to the method, the catboost model is applied to the lithology identification field, and compared with a traditional integrated model based on a decision-making tree, a sorting lifting strategy is provided to solve the problems of gradient deviation and prediction offset exi