CN-122020294-A - Lithology while drilling identification method, system, equipment and medium

CN122020294ACN 122020294 ACN122020294 ACN 122020294ACN-122020294-A

Abstract

The invention provides a lithology while drilling recognition method, a system, equipment and a medium, which belong to the field of petroleum and natural gas drilling engineering, and the method comprises the steps of performing feature extraction and fusion on time series data while drilling based on a multi-scale convolutional neural network of a lithology recognition model to obtain multi-scale fusion features; based on a bidirectional gating circulation unit in the lithology recognition model, global time sequence features containing context dependency relations are extracted according to front-back association of a multi-scale fusion feature sequence in the forward sequence direction and the backward sequence direction, the multi-scale fusion feature sequence comprises a plurality of multi-scale fusion features arranged in time sequence, and a lithology category recognition result is obtained based on the global time sequence features. The method accurately and stably carries out lithology identification while drilling in real time.

Inventors

HEI CHUANG
ZHANG XUHAO
LUO MINGZHANG
LI SHUNLI
ZHOU LUOYU
LI XIUQUAN
SUN CHANGHE
WANG XIONGWEI
LIU GENYUAN
WANG JIACHENG

Assignees

长江大学

Dates

Publication Date: 20260512
Application Date: 20260123

Claims (10)

1. A lithology while drilling recognition method, comprising: Performing feature extraction and fusion on the time sequence data while drilling to obtain multi-scale fusion features based on a multi-scale convolutional neural network of the lithology recognition model; Based on a bidirectional gating circulation unit in the lithology recognition model, extracting global time sequence features containing context dependency relations according to front-back association of a multi-scale fusion feature sequence in a forward sequence direction and a reverse sequence direction; and obtaining a lithology category identification result based on the global time sequence characteristic.
2. The lithology while drilling recognition method according to claim 1, wherein before the feature extraction and fusion of the time series while drilling data to obtain the multi-scale fusion feature, the multi-scale convolutional neural network based on the lithology recognition model comprises: Acquiring original while-drilling data, and calculating local statistics of the original while-drilling data based on a sliding window; and identifying and correcting abnormal data in the original while-drilling data according to the local statistic.
3. The lithology while drilling recognition method of claim 2, wherein the identifying and correcting anomalous data in the raw data while drilling from the local statistics comprises: if the continuous quantity of the abnormal data in the sliding window is smaller than a first threshold value, correcting the abnormal data by adopting a linear interpolation method; and if the continuous number of the abnormal data in the sliding window is larger than or equal to the first threshold value, correcting the abnormal data by adopting a cubic spline interpolation method.
4. The lithology while drilling recognition method according to claim 1, wherein the multi-scale convolutional neural network comprises at least three groups of one-dimensional convolutional layers connected in parallel, each group of one-dimensional convolutional layers adopts a convolutional kernel with different time dimensions, the multi-scale convolutional neural network based on the lithology recognition model performs feature extraction and fusion on time-series while drilling data to obtain multi-scale fusion features, and the method comprises the following steps: inputting the time sequence data while drilling into the at least three groups of one-dimensional convolution layers in parallel for convolution operation, and extracting local features of different time windows to obtain a corresponding feature map; Sequentially carrying out nonlinear activation and global average pooling treatment on the feature images output by each group of one-dimensional convolution layers; and splicing all the processed features to form the multi-scale fusion feature.
5. The lithology while drilling recognition method according to claim 1, wherein the bidirectional gating cycle unit comprises a forward gating cycle unit and a reverse gating cycle unit which are arranged in parallel and have opposite processing directions, wherein the extracting global time sequence features containing context dependency relations based on the bidirectional gating cycle unit in the lithology recognition model according to the front-back association of a multi-scale fusion feature sequence in the forward sequence direction and the reverse sequence direction comprises the following steps: The multi-scale fusion feature sequence is input into the forward gating circulation unit and the reverse gating circulation unit at the same time, wherein the forward gating circulation unit processes the multi-scale fusion feature sequence along the time increment direction, and the reverse gating circulation unit processes the multi-scale fusion feature sequence along the time decrement direction; Respectively acquiring a forward hidden state sequence obtained by processing the forward gating circulating unit and a reverse hidden state sequence obtained by processing the reverse gating circulating unit; Splicing the forward hidden state sequence and the reverse hidden state sequence to obtain a depth time sequence characteristic; and carrying out global average pooling treatment on the depth time sequence characteristics to obtain global time sequence characteristics.
6. The lithology while drilling recognition method according to claim 1, wherein the bi-directional gating circulation unit is connected with a full connection layer and an output layer, the obtaining the recognition result of the lithology category based on the global time sequence feature comprises: based on the full connection layer, performing linear change, nonlinear activation and random inactivation on the global time sequence feature to obtain a processed time sequence feature; and predicting and generating probability distribution of lithology categories based on the output layer according to the processed time sequence characteristics to obtain the identification result.
7. The lithology while drilling recognition method of claim 1, further comprising the training step of the lithology recognition model: calculating a difference value between the identification result and the real lithology tag by using a cross entropy loss function; and iteratively updating model parameters of the lithology recognition model by using the self-adaptive moment estimation algorithm with the minimum difference value as a target, monitoring the recognition accuracy of the verification set in the training process, and terminating the training if the accuracy is not improved in a continuous preset round.
8. A lithology while drilling recognition system, comprising: The multi-scale feature extraction module is used for carrying out feature extraction and fusion on the time sequence data while drilling on the basis of a multi-scale convolutional neural network of the lithology recognition model to obtain multi-scale fusion features; The global time sequence extraction module is used for extracting global time sequence features containing context dependency relations based on a bidirectional gating circulation unit in the lithology recognition model according to front-back association of a multi-scale fusion feature sequence in a front sequence direction and a back sequence direction; and the classification output module is used for obtaining the identifying result of the lithology category based on the global time sequence characteristic.
9. An electronic device comprising a memory and a processor, wherein, The memory is used for storing programs; The processor, coupled to the memory, for executing the program stored in the memory to implement the steps of the lithology while drilling recognition method of any one of the preceding claims 1 to 7.
10. A computer readable storage medium storing a computer readable program or instructions which when executed by a processor is capable of carrying out the steps of the lithology while drilling recognition method of any one of the preceding claims 1 to 7.

Description

Lithology while drilling identification method, system, equipment and medium Technical Field The invention relates to the field of petroleum and natural gas drilling engineering, in particular to a lithology while drilling identification method, system, equipment and medium. Background In petroleum and natural gas drilling engineering, accurate and real-time lithology identification is a key link for grasping stratum geological features, optimizing drilling parameters and guaranteeing operation safety and efficiency. In the face of complex and changeable underground geological conditions, whether the lithology of drilling is encountered can be rapidly and reliably judged, and the scientificity and the risk control capability of drilling decisions are directly related. To address this need, two main classes of technologies, traditional and data-driven intelligent, have been developed in the industry. The traditional method is represented by drilling and coring, logging rock cuttings and traditional logging (cable/early logging while drilling), or has high cost, low efficiency or relies on manual experience and has insufficient real-time performance. In order to overcome the limitations, intelligent recognition technology based on real-time data while drilling has been developed, the early stage mainly depends on a traditional machine learning model (such as a support vector machine and a random forest), and the later stage turns to a deep learning method capable of automatically extracting features. However, existing smart identification technologies still have significant drawbacks. While the traditional machine learning method severely depends on artificial characteristic engineering, a complex high-dimensional nonlinear mode in time series data while drilling is difficult to capture, the current mainstream deep learning model (such as a common convolutional neural network or a unidirectional cyclic neural network) is often poor in robustness to drilling parameter fluctuation, and long sequence dependency of stratum in the depth direction is difficult to fully capture, particularly in lithology gradual change or unstable recognition precision of thin interbed sections. Therefore, a new method capable of deeply fusing multi-scale features and effectively modeling time sequence context dependence is needed to improve accuracy, instantaneity and stability of lithology while drilling identification under complex geological conditions. Disclosure of Invention In view of the foregoing, it is necessary to provide a lithology while drilling recognition method, system, device and medium for solving the technical problems of high artificial dependency, insufficient real-time performance, insufficient lithology recognition precision and insufficient stability in the prior art. In order to solve the technical problem, in a first aspect, the present invention provides a lithology while drilling identification method, including: Performing feature extraction and fusion on the time sequence data while drilling to obtain multi-scale fusion features based on a multi-scale convolutional neural network of the lithology recognition model; Based on a bidirectional gating circulation unit in the lithology recognition model, extracting global time sequence features containing context dependency relations according to front-back association of a multi-scale fusion feature sequence in a forward sequence direction and a reverse sequence direction; and obtaining a lithology category identification result based on the global time sequence characteristic. In one possible implementation manner, before the multi-scale convolutional neural network based on the lithology recognition model performs feature extraction and fusion on the time-series data while drilling to obtain multi-scale fusion features, the method includes: Acquiring original while-drilling data, and calculating local statistics of the original while-drilling data based on a sliding window; and identifying and correcting abnormal data in the original while-drilling data according to the local statistic. In one possible implementation manner, the identifying and correcting the abnormal data in the original while-drilling data according to the local statistic includes: if the continuous quantity of the abnormal data in the sliding window is smaller than a first threshold value, correcting the abnormal data by adopting a linear interpolation method; and if the continuous number of the abnormal data in the sliding window is larger than or equal to the first threshold value, correcting the abnormal data by adopting a cubic spline interpolation method. In one possible implementation manner, the multi-scale convolutional neural network includes at least three groups of one-dimensional convolutional layers connected in parallel, each group of one-dimensional convolutional layers adopts a convolutional kernel with different time dimensions, the multi-scale convolutional neural net