Design And Implementation Of Lithology Recognition System Based On Neural Network

In the work of logging lithology interpretation and evaluation,owing to the complex data sources involved in the logging interpretation business and the inconsistent data standards,the efficiency of relying on manual experience to explain lithology is low.Consequently it is crucial to realize the integration of logging interpretation data and the intelligence of lithology identification.Aiming at the above problems,this paper analyzes the business process and requirements of logging lithology interpretation,proposes a lithology identification method based on neural network,and designs and implements a lithology identification system.Firstly,it establishes logging raw data standard and logging comprehensive map data standard according to the actual situation of logging data interpretation and evaluation center,introduces the concept of metadata to construct the basic metadata based on logging interpretation business,and establishes the mapping relationship between source data and target data through the design of mapping rules,so as to design a set of logging data integration method based on metadata and data mapping.Second,the sample features affecting lithology are analyzed,the data features are preprocessed,and the feature screening is completed using correlation analysis and MIC;for the problem of uncertainty in the structure selection of BP networks,this paper uses genetic algorithm to evolve the BP network structure and selects the best optimization algorithm for BP networks through experimental comparison,uses real historical data to train and predict the model,and compares it with the empirical formula-based method The model is trained and predicted using real historical data,and compared with the empirical formula-based network structure selection method.The experimental results show that the method based on genetic algorithm optimized BP neural network structure improves the accuracy of lithology identification by about 10%.The model was applied to the lithology identification in the actual production well data of the same block and achieved the expected results.Finally,the design and implementation of the lithology identification system including data integration module,data management module,data maintenance module and lithology identification model were realized from the system requirements using C/S architecture,in which the lithology identification module used the BP neural network model optimized by genetic algorithm.At present,the neural network-based lithology identification system designed in this paper has been applied to the lithology interpretation work in the actual production of logging,which verifies the reasonableness of the lithology identification model proposed in this topic,and the lithology interpretation efficiency of logging has been improved to a certain extent,and the system operates stably.