Research And Construction Of Log Lithology Identification Integrated Model

With the advancement and development of artificial intelligence technology,more and more traditional disciplines have introduced algorithms such as machine learning and deep learning to research and analyze problems in traditional disciplines.In recent years,the booming development of geological big data has promoted the combination of geophysics and artificial intelligence.Lithology identification plays an important role both in formation evaluation and in geological surveys such as reservoir description and reservoir calculation.However,traditional lithology identification methods have problems such as long cycles,low efficiency,and large subjective factors,making it difficult for geological staff to identify lithology in a timely and accurate manner.At present,neural network algorithms are mostly used in the field of lithology recognition,but the network structure is not well designed,it is difficult to select appropriate parameters,and it is impossible to learn the wrong samples again.Therefore,based on the in-depth study of traditional machine learning algorithms,this paper establishes an integrated model based on a variety of machine learning algorithms and applies it to lithology recognition.In addition,this article also combines particle swarm optimization algorithms to find the problem of optimal hyperparameters is studied and improved.The main tasks completed in this article are as follows:(1)Aiming at the problems of low efficiency and redundant information of traditional logging lithology identification method based on logging response equation,this paper proposes a logging lithology identification method based on ensemble learning.An integrated model using three models of K-nearest neighbor,random forest,and support vector machine as the primary trainer,and a logistic regression model as the secondary trainer.(2)Aiming at the problem that the standard particle swarm optimization algorithm converges too fast,an improved method is proposed in this paper.This method improves the standard particle swarm algorithm by adding a compression factor and a dynamic selection learning factor to make the particle swarm search In the early stage,the best particles in the population can be found as soon as possible,and in the later stage of thesearch,it has a stronger ability to jump out of the local optimal value.(3)Aiming at the problem that the logging lithology identification integrated model has many hyperparameters and is difficult to select,this article uses an improved particle swarm optimization algorithm to optimize some parameters in the integrated model.Experiments show that the improved particle swarm algorithm is used.The performance of the optimized integrated model is better than the performance of the integrated model optimized by the standard particle swarm algorithm,and the accuracy of lithology recognition is higher.Experimental results show that the recognition rate of the integrated model proposed in this paper is higher than that of random forest,gradient boosting decision Tree,and artificial neural network models.The improved particle swarm algorithm has a stronger local search capability,and the recognition rate of the integrated model using the improved particle swarm optimization has been improved to a certain extent.The logging lithology identification model proposed in this paper provides a new method for lithology identification,and solves the problems of long lithology identification cycle and low efficiency through integrated learning.