A logging curve prediction method and system based on deep learning

Abstract

The present invention provides a method and system for predicting well logging curves based on deep learning. The method includes steps: S10: use the logging curve data of one of the data wells as label data, and use the logging curve data of the remaining data wells as label data. Input data, and calculate the eigenvalue of the input data; S20, obtain the aggregation value and the maximum eigenvalue in the eigenvalue, and calculate the ratio of the aggregation value to the maximum eigenvalue; S30, obtain according to the ratio Deep study the initial parameters and floating range of the DNN model, and adjust the parameters within the floating range until the training model of the best deep learning DNN model is obtained; S40: Send the well logging curves of all data wells in the work area to the A trained model of the best deep learning DNN model to predict well log data from other virtual wells. The invention has higher calculation efficiency and higher precision, and can provide more accurate curves for subsequent seismic data processing.

Description

technical field [0001] The present invention relates to the technical field of seismic data processing in the field of exploration, in particular to a logging curve prediction method and system based on deep learning. Background technique [0002] The prediction or reconstruction of well log curves has become an essential work in the early stage of combined well-seismic inversion technology. However, in actual exploration, well logging cannot be very dense, because it involves the cost of logging. The prediction of well logging curve plays an important role in the later oil and gas prediction and the inversion of elastic parameters. [0003] At present, empirical methods are used to predict well logging curves in exploration. This method can only be used for mathematical conversion of one or more curves. The speed is fast and the whole operation process is easy, but the disadvantage is that the accuracy of the results is not high. There is also the seismic attribute analys...

AI Technical Summary

Problems solved by technology

[0003] At present, empirical methods are used in exploration to predict well logging curves. This method can only be used for mathematical conversion of one or more curves. The speed is fast and the whole operation process is easy, but the disadvantage is that the accuracy of the results is not high.

There is also the seismic attribute analysis method to predict the logging curve. This method mainly simulates the seismic attribute curve through the linear data in the horizontal and vertical directions. The accuracy of this method is higher than that of the empirical method. 3D seismic data volume, and there is an overtraining problem

Images