Logging curve prediction method based on modal decomposition reconstruction and depth LSTM-RNN model

Abstract

The invention discloses a logging curve prediction method based on modal decomposition reconstruction and a depth LSTM-RNN model. The logging curve prediction method comprises the steps: utilizing CEEMD to decompose logging curve data into limited intrinsic mode function IMF components and residual RES components which have local characteristics and are different in frequency; carrying out fluctuation degree detection on each component by a run-length detection method, and reconstructing CEEMD decomposition items with similar fluctuation frequencies into three new components with typical characteristics, namely a high-frequency item, a low-frequency item and a trend item; performing data normalization processing on the reconstructed new component and dividing training and test data; respectively establishing a depth LSTM-RNN model for each component and carrying out network training by utilizing the training data; and using the trained model to predict a missing or distorted logging curve, and finally reconstructing and reversely normalizing prediction results of the three components to obtain a logging curve prediction result. According to the logging curve prediction method, thenumber of prediction component modeling is reduced, and the prediction precision and speed are improved, and the simplicity and practicability are high, and the missing or distorted logging curve canbe predicted more accurately and effectively.

Description

technical field [0001] The invention relates to a method for predicting well logging curves based on modal decomposition and reconstruction and a deep LSTM-RNN model, belonging to the fields of deep learning methods and petroleum geophysical prospecting technologies. Background technique [0002] Well logging data, as a bridge and bond connecting seismicity and geology, play a vital role in oil and gas exploration. In practical applications, due to factors such as borehole diameter expansion, borehole wall collapse, and instrument failure, some logging data are often distorted or missing, which brings certain difficulties to subsequent interpretation work. And relogging is not only expensive, but also impossible to relog for some wellbores that have been completed. For this reason, it is of great significance to explore and develop logging curve prediction methods to correct or predict the logging data of distorted or missing well intervals to increase the accuracy of loggi...

AI Technical Summary

Problems solved by technology

[0006] To sum up, there have been a large number of researches on the prediction of well logging curves at home and abroad, but the above-mentioned models have not fully utilized the characteristics of time series, non-stationarity and strong nonlinearity of the well logging curve data, which has great impact on improving the well logging curve data. Prediction accuracy and robustness are of great significance

Images