Logging Reservoir Evaluation Based On Gramian Angular Field Transformation

Logging reservoir evaluation is a process of comprehensive evaluation and analysis of reservoir in target block by combining logging data with geological information,development data and logging data.This process can solve the predetermined geological tasks,such as lithology stratification,oil and gas reservoir evaluation and other geological engineering problems.Traditional logging reservoir evaluation generally requires physical modeling.The modeling process is usually affected by the level of logging workers,and the workload is also very complex.As the core of interdisciplinary and artificial intelligence,machine learning can use a large number of scientifically processed logging data to establish a machine learning model that meets the production and application conditions.The logging reservoir evaluation process using machine learning algorithm can divide the knowledge structure by simulating the interpretation mode of logging workers,so as to improve the production efficiency and greatly reduce the workload of logging experts.Machine learning is widely used in the field of pattern recognition,especially in computer vision.However,when encountering loggings with rich formation information,there are still many difficulties in establishing machine learning model for classification.When the traditional machine learning method carries out lithologic stratification and oil-water identification,it usually manually labels multiple one-dimensional data of the target interval as the input characteristics.The obtained lithologic stratification and oil-water identification results are greatly affected by the original data,personnel level and other factors,and the accuracy of model identification is also poor.In this paper,when using logging and other data for lithologic stratification and oil-water identification,using gramian angle field(GAF)transformation,the logging similar to time series can be converted into pictures,and the one-dimensional sequence information can be upgraded into two-dimensional picture information.From the results,this method will not lose the information of the logging itself,but also retain the relevant information between the logging data,so as to make full use of the current advantages of machine vision,mine the existing logging data to the greatest extent,and establish the lithology stratification and oil-water identification model combined with gramian angle field and convolution neural network.The accuracy of this method is more than 80% in lithologic stratification and more than 75% in oil-water identification.