Prediction Method Of Missing Log Curves Based On Deep Mining Of Amplitude Correlation Relation

In the field of petroleum engineering,logging is an important measurement method in the geological exploration industry.Geophysical logging data is one of the most important information sources for obtaining resource description information of oil and gas reservoirs.Especially in the face of secondary and tertiary development of middle and late stage old oil fields,the role of large scale high precision logging curves is more prominent.Geologists can build accurate geological models based on the logging data to further design development strategies.However,the application of reservoir geological description using logging data often has the problem of distortion or missing of some logging curves,and the recovery of logging curves has been a research hotspot and difficulty in related research fields.To this end,this paper is dedicated to effectively improving the characterization of symbiotic relationships among logging curves based on signal processing,machine learning,and carries out a series and research on logging curve prediction methods based on deep mining of magnitude relationships.In response to the problems of inadequate representation and utilization of correlation information between different logging curves of the same well and poor adaptability of cross-well models in traditional signal recovery methods and recovery methods based on machine learning such as neural networks,a method for recovery of logging curves based on long short-term memory neural network(LSTM)multi-scale symbiotic relationship mining is proposed.Based on the neural network-based logging curve recovery method,this method can characterize the lateral correlation information between different logging curves by introducing multi-scale gray symbiosis to realize the comprehensive utilization of longitudinal and transverse semantic information of the logging curve set,and then achieve the recovery of missing logging curves.Through the verification of actual logging data,the method can effectively exploit the horizontal and vertical information of logging curves to achieve more accurate prediction of density logging curves.Due to the complexity and non-homogeneity of geological reservoirs,most of the current recovery methods recover only one kind of log curves.In order to further improve the accuracy of prediction results of multiple log curves in different wellheads,a log curve recovery method based on multi-scale spatio-temporal correlation map information mining is proposed.The method is based on the graph representation learning commonly used in the field of signal processing to describe the graph structure between different logging curves so that the logging curves can establish nonlinear mapping relationships.Then,the hierarchical structure of the depth forest is used to achieve the characterization of the longitudinal information of the logging curves,and finally,the multi-scale information is fused with the LSTM network to realize a new logging curve prediction method.The prediction results of multiple logging curves for different wellheads are obtained through experimental validation of the method and compared with other methods,and the results show that the method has higher accuracy and better robustness.