CN-121997040-A - Drilling data generation method and system based on conditional diffusion model

CN121997040ACN 121997040 ACN121997040 ACN 121997040ACN-121997040-A

Abstract

The invention provides a drilling data generation method and system based on a conditional diffusion model, and relates to the technical field of oil and gas well drilling. The method comprises the steps of obtaining a first data set, preprocessing a plurality of drilling data, determining a working condition type corresponding to each drilling data, performing iterative training on a neural network model according to the drilling data and the working condition type corresponding to each drilling data to obtain a trained neural network model, receiving an input drilling data generation instruction, wherein the drilling data generation instruction carries a target working condition type, inputting the target working condition type into the trained neural network model, and outputting the drilling data corresponding to the target working condition type. According to the method and the device, drilling data corresponding to different working condition categories can be obtained rapidly and accurately based on the trained neural network model, and then the risk prediction model is trained through the drilling data corresponding to different working conditions, so that the prediction capability of the model is improved.

Inventors

YANG XINYI
WU DONG
CUI MENG
MEI YUNYI
GUO WEIHONG
TIAN YUMENG
JING LINGZHI
SHI XIAOYAN

Assignees

中国石油天然气集团有限公司
中国石油集团工程技术研究院有限公司

Dates

Publication Date: 20260508
Application Date: 20241107

Claims (10)

1. A method of generating well data based on a conditional diffusion model, the method comprising: Acquiring a first data set, wherein the first data set comprises a plurality of drilling data, each drilling data comprises a plurality of drilling characteristic parameters corresponding to different moments, and the drilling characteristic parameters comprise torque, total pool volume, weight on bit, inlet flow, rotary table rotating speed, outlet flow and delay well depth; Preprocessing the plurality of drilling data, wherein the preprocessing comprises normalization processing, denoising processing and missing value supplementing; determining the working condition category corresponding to each drilling data; Performing iterative training on the neural network model according to the drilling data and the working condition category corresponding to each drilling data to obtain a trained neural network model; receiving an input drilling data generation instruction, wherein the drilling data generation instruction carries a target working condition class; And inputting the target working condition category into the trained neural network model, and outputting a plurality of drilling data corresponding to the target working condition category.
2. The method of claim 1, wherein the determining the class of conditions for each drilling data comprises: And determining the working condition category corresponding to each drilling data through a voting classifier based on a preset classification algorithm, wherein the preset classification algorithm comprises one or more of a support vector machine algorithm, a logistic regression algorithm, a random forest algorithm, an adaptive lifting algorithm, a multi-layer perceptron algorithm and a decision tree algorithm.
3. The method of claim 2, wherein the neural network model is a diffusion model based on a transform structure, the diffusion model including a time encoding module, an input encoding module and a transform module, the time encoding module being configured to convert a time corresponding to each drilling data into an embedded vector, the input encoding module being configured to encode each drilling data into a feature vector through linear projection, and the transform module being configured to denoise the embedded vector and the feature vector to obtain the drilling data.
4. The method of claim 3, wherein the fransformer module comprises a plurality of fransformer layers arranged in a stacked manner, each fransformer layer comprises a self-attention layer, a cross-attention layer and a feedforward neural network layer, each fransformer layer is configured with a residual link and a layer normalization mechanism, the self-attention layer is used for determining context information included in a feature vector of each drilling data, the cross-attention layer is used for determining an association relationship between the feature vector of each expert data and a working condition category, and the feedforward neural network layer is used for processing according to the context information included in the feature vector and the association relationship between the feature vector and the working condition category to obtain a corresponding classification result.
5. The method of claim 4, wherein the iteratively training the neural network model according to the plurality of drilling data and the operating condition category corresponding to each drilling data to obtain a trained neural network model comprises: constructing an objective function, wherein the objective function is a mean square error function; and based on the objective function, performing iterative training on the neural network model according to the drilling data and the working condition category corresponding to each drilling data to obtain a trained neural network model.
6. A system for generating well data based on a conditional diffusion model, the system comprising: an acquisition unit for acquiring a first data set, wherein the first data set comprises a plurality of drilling data, each drilling data comprises a plurality of drilling characteristic parameters corresponding to different moments, and the drilling characteristic parameters comprise torque, total pool volume, weight on bit, inlet flow, rotary table rotating speed, outlet flow and delay well depth; The processing unit is used for preprocessing the plurality of drilling data, wherein the preprocessing comprises normalization processing, denoising processing and missing value supplementing; the determining unit is used for determining the working condition category corresponding to each drilling data; The training unit is used for carrying out iterative training on the neural network model according to the drilling data and the working condition category corresponding to each drilling data to obtain a trained neural network model; The receiving unit is used for receiving an input drilling data generation instruction, wherein the drilling data generation instruction carries a target working condition class; The generating unit is used for inputting the target working condition category into the trained neural network model and outputting a plurality of drilling data corresponding to the target working condition category.
7. The system according to claim 6, characterized in that the determining unit is specifically configured to: And determining the working condition category corresponding to each drilling data through a voting classifier based on a preset classification algorithm, wherein the preset classification algorithm comprises one or more of a support vector machine algorithm, a logistic regression algorithm, a random forest algorithm, an adaptive lifting algorithm, a multi-layer perceptron algorithm and a decision tree algorithm.
8. The system of claim 7, wherein the neural network model is a diffusion model based on a transform structure, the diffusion model including a time encoding module, an input encoding module and a transform module, the time encoding module being configured to convert a time corresponding to each drilling data into an embedded vector, the input encoding module being configured to encode each drilling data into a feature vector through linear projection, and the transform module being configured to denoise the embedded vector and the feature vector to obtain the drilling data.
9. The system of claim 8, wherein the fransformer module comprises a plurality of fransformer layers arranged in a stacked manner, each fransformer layer comprises a self-attention layer, a cross-attention layer and a feedforward neural network layer, each fransformer layer is configured with a residual link and a layer normalization mechanism, the self-attention layer is used for determining context information included in a feature vector of each drilling data, the cross-attention layer is used for determining an association relationship between the feature vector of each expert data and a working condition category, and the feedforward neural network layer is used for processing according to the context information included in the feature vector and the association relationship between the feature vector and the working condition category to obtain a corresponding classification result.
10. An electronic device, comprising: A memory for storing instructions executable by the processor; Wherein the processor is configured to execute the instructions to implement the conditional diffusion model based drilling data generation method of any one of claims 1-5.

Description

Drilling data generation method and system based on conditional diffusion model Technical Field The invention relates to the technical field of oil and gas well drilling, in particular to a method and a system for generating drilling data based on a conditional diffusion model. Background With the rapid development of artificial intelligence technology, the application of the artificial intelligence technology in the oil and gas industry is increasingly wide, and particularly in the technical field of oil and gas well drilling, the introduction of the intelligent technology greatly improves the operation efficiency and the safety. Through a data-driven algorithm model, various complex problems in the drilling process, such as real-time monitoring of underground environment, automatic control of drilling equipment, risk prediction of underground accidents and the like, are effectively solved. However, one of the core challenges of intelligent drilling technology is the lack of data, especially drilling data involving high risk or special conditions. In reality, high quality, comprehensive drilling data that can be used to train the smart model is often inadequate, especially in the face of risky conditions (e.g., blowouts, sags, etc.), and it is almost impossible to obtain large amounts of such drilling data due to the low probability of occurrence and the high potential risk of these conditions. This lack of data directly limits the predictive power of the neural network model, resulting in the inability of existing intelligent systems to provide sufficiently accurate predictions and guidance when dealing with complex and high risk conditions. For example, during drilling, accidents such as blowout, while occurring at a lower rate, may have serious safety and economic consequences once they occur. In order to effectively predict and avoid such risks, the model must be trained with enough drilling data corresponding to risk conditions. However, because such data is difficult to accumulate through practical operations, models tend to perform poorly in coping with similar emergency conditions. In the related art, CN116304701a provides a method for generating an HRRP sample based on a conditional denoising diffusion probability model, and although a high-resolution radar pulse sample in the radar field can be obtained through the conditional denoising diffusion probability model, the probability model cannot be applied to the technical field of oil-gas well drilling because the drilling data respectively correspond to different working conditions, and further cannot obtain the drilling data corresponding to different working conditions. Therefore, a method and a system for generating drilling data based on a conditional diffusion model are needed, which can obtain drilling data corresponding to different working conditions based on the trained conditional diffusion model, further train a risk prediction model through the drilling data corresponding to different working conditions, and improve the prediction capability of the model. Disclosure of Invention The embodiment of the invention provides a drilling data generation method and a drilling data generation system based on a conditional diffusion model, which can quickly and accurately obtain drilling data corresponding to different working condition categories based on a trained neural network model, further train a risk prediction model through the drilling data corresponding to different working conditions, and improve the prediction capability of the model. In order to achieve the above purpose, the embodiment of the present invention adopts the following technical scheme: A first aspect provides a drilling data generation method based on a condition diffusion model, which comprises the steps of obtaining a first data set, wherein the first data set comprises a plurality of drilling data, each drilling data comprises a plurality of drilling characteristic parameters corresponding to different moments, the drilling characteristic parameters comprise torque, total pool volume, drilling pressure, inlet flow, rotating disc rotating speed, outlet flow and delay well depth, preprocessing the drilling data, including normalization processing, denoising processing and missing value supplementing, determining a working condition category corresponding to each drilling data, performing iterative training on a neural network model according to the drilling data and the working condition category corresponding to each drilling data to obtain a trained neural network model, receiving an input drilling data generation instruction, wherein the drilling data generation instruction carries a target working condition category, inputting the target working condition category into the trained neural network model, and outputting the drilling data corresponding to the target working condition category. In one possible implementation manner of the first aspect, determining the