CN-121997213-A - Intelligent linkage method for oil and gas exploitation process

CN121997213ACN 121997213 ACN121997213 ACN 121997213ACN-121997213-A

Abstract

The invention relates to the technical field of oil and gas exploitation, in particular to an intelligent linkage method of an oil and gas exploitation process, which comprises the following steps of S1, data acquisition and monitoring, wherein temperature, ‌ pressure and ‌ flow data in an oil refining process are acquired in real time through data acquisition equipment, S2, model optimization and prediction, and S3, intelligent control and adjustment are carried out, the production flow is optimized through data analysis, ‌, exploitation efficiency is improved, the model is utilized, ‌ to know the running state of the oil refining process, ‌ and make adjustment according to the need, high-efficiency decision making is realized, automatic control of each unit in the oil refining process is realized, ‌ enables the operation process to be more accurate and stable, the correlation between the data can be accurately analyzed through model prediction, future trend can be accurately predicted, the influence of subjective awareness and personal experience is avoided, the prediction result is relatively reliable, subjective interference in the human input and the prediction process is reduced, excessive bandwidth and computing capacity are not consumed, and seamless butting and sharing of the data can be realized.

Inventors

HE ZHANYOU
LI PEI
REN XIAORONG
LI ZHIGUANG
He Zhuopu
ZHAO XIAOCHUN
TONG JUNFENG
CHEN PENG

Assignees

中国石油天然气股份有限公司

Dates

Publication Date: 20260508
Application Date: 20241108

Claims (8)

1. The intelligent linkage method for the oil and gas exploitation process is characterized by comprising the following steps of: s1, data acquisition and monitoring, namely acquiring temperature, ‌ pressure and ‌ flow data in the oil refining process in real time through data acquisition equipment, transmitting ‌ to a central control room through a network for data processing, knowing the running state of the oil refining process, ‌ and adjusting according to requirements; S11, data analysis, namely mining and integrating all acquired data, wherein the data analysis specifically comprises data preprocessing and data prediction analysis; S111, preprocessing data, namely, processing missing values, detecting and processing abnormal values, normalizing/normalizing the data, coding classification variables, selecting characteristics and reducing dimensions, transforming the data, dividing the data, processing unbalanced data, preprocessing text data and carrying out time sequence data and mathematical; S112, data predictive analysis is established by using statistical analysis and a model, and data are subjected to deep analysis and mining; S12, displaying the processed data in a graph or curve mode, so that a producer can more intuitively know the current production state, including but not limited to yield, productivity, temperature, humidity and pressure information; s2, model optimization and prediction ‌ Optimizing and predicting by establishing a mathematical model of the oil refining process, ‌ optimizing the model ‌ to realize the optimal operation of the oil refining process, ‌ improving the product quality and the production efficiency, ‌ predicting the possible problems in the oil refining process, ‌ taking corresponding measures in advance, ‌ reducing risks; The main method of the model optimization algorithm comprises a gradient descent algorithm, a random gradient descent algorithm and a dynamic gradient descent algorithm, wherein the gradient descent algorithm gradually updates parameters from randomly initialized parameters to minimize a loss function; a random gradient descent algorithm, starting from the randomly initialized parameters, gradually updating the parameters to minimize the loss function; a dynamic gradient descent algorithm, starting from the randomly initialized parameters, gradually updating the parameters to minimize the loss function; The algorithm of the model prediction comprises linear regression, a support vector machine, a decision tree and a random forest, wherein the formula of the linear regression mathematical model is as follows , Is a dependent variable, Is an independent variable, Is of the intercept, Is a slope, Is an error term, The number of the samples is that the mathematical model formula of the support vector machine is , Representing a decision function, Is a parameter of a support vector machine, Is a sample label, Is a kernel function, b is an offset, x is an input vector, and the decision tree mathematical model formula is if Wherein Is an input variable, Is an output variable, Representing category, c representing category number, and a random forest mathematical model formula is , Representing the predicted result of the random forest on the input data x, Representing the number of decision trees in a random forest, Represent the first A prediction result of the decision tree; And S3, intelligent control and regulation, namely ‌, realizing automatic control of each unit in the oil refining process through an automatic control system and an artificial intelligent algorithm, wherein ‌ ensures that the operation process is more accurate and stable.
2. The intelligent linkage method of the oil and gas exploitation process according to claim 1, wherein the data acquisition equipment comprises a pressure sensor, a temperature sensor and a flow sensor, and the pressure sensor, the temperature sensor and the flow sensor are arranged on the production equipment and the pipeline.
3. The intelligent linkage method for the oil and gas exploitation process according to claim 1, wherein the missing value processing comprises deleting records containing missing values; filling the missing values with mean, median or mode statistics; predicting a missing value by using a K-nearest neighbor and decision tree algorithm; outlier detection and processing includes detecting outliers using statistical methods, determining whether to delete, replace or preserve outliers based on business requirements, data normalization/normalization includes normalization and normalization, where normalization converts data to a distribution with a mean of 0 and standard deviation of 1, normalization scales data to a range of [0,1] or [ -1,1], encoding classification variables including one-hot encoding, tag encoding and sequential encoding, where one-hot encoding converts classification variables to binary columns, tag encoding converts classification variables to integers, sequential encoding converts ordered classification variables to integers, preserving sequential information, feature selection and dimension reduction includes selecting important features using statistical testing, model weighting methods, reducing complexity of features using PCA, t-SNE methods, data transformation includes logarithmic transformation, box-Cox transformation, polynomial feature generation, data partitioning divides data sets into training sets, validation sets and test sets to evaluate model performance and generalization capability, processing unbalanced data includes a majority class, preprocessing includes preprocessing of data includes de-word preprocessing, de-emphasis, feature selection and model weighting includes using statistical testing, feature selection and model weighting methods, reducing complexity using PCA, t-SNE method, data transformation includes a majority class, data sampling includes de-mining, and fuzzy processing, fuzzy pattern processing, and fuzzy pattern processing, and fuzzy pattern processing, fuzzy pattern, and fuzzy pattern, and, fuzzy pattern, fuzzy, and, fuzzy, and, fuzzy, to fuzzy, and, to the fuzzy, and the fuzzy, to the fuzzy, and the, punctuation and special characters, stem extraction or morphological reduction, text vectorization, time series data preprocessing including date and time feature extraction, time series stabilization processing, seasonal decomposition and trend decomposition.
4. The intelligent linkage method for the oil and gas exploitation process according to claim 1, wherein the statistical analysis method comprises description statistical analysis, hypothesis testing, correlation analysis and regression analysis, wherein the description statistical analysis is mainly used for describing basic characteristics of data, including central trend, discrete degree and distribution form of the data, the hypothesis testing is used for judging whether a certain hypothesis is met, the overall property is inferred according to sample data, the common hypothesis testing methods comprise t testing, variance analysis and chi-square testing, the correlation analysis is used for analyzing the relation strength and direction between two or more variables, the common methods comprise Pearson correlation analysis and Spearman correlation analysis, and the regression analysis is used for researching the influence degree and direction of independent variables on dependent variables and mainly comprises two methods of linear regression and nonlinear regression.
5. The intelligent linkage method for the oil and gas exploitation process according to claim 1, wherein the method for establishing the model comprises a linear model, a nonlinear model, a time sequence model and a machine learning model, wherein the linear model is commonly used for establishing a linear relation model between independent variables and dependent variables, such as a linear regression model, the nonlinear model is used for establishing a nonlinear relation model between the independent variables and the dependent variables, such as a polynomial regression model and a Logistic regression model, the time sequence model is used for analyzing and predicting trend and periodicity of time sequence data, such as an ARIMA model and a GARCH model, and the machine learning model is used for automatically learning data models, such as decision trees, support vector machines and neural networks, by means of algorithms and pattern recognition.
6. The intelligent linkage method for the oil and gas exploitation process is characterized by comprising the specific steps of loading a model, initializing parameters, defining a loss function, selecting an optimization declaration, training the model and evaluating the model, wherein a mathematical model formula of the model optimization algorithm is as follows: Wherein Representing the updated weight, Representing the old weight, Represent learning rate, Representing a loss function, Representing the gradient of the loss function.
7. The intelligent linkage method for oil and gas exploitation processes according to claim 1, wherein the intelligent control and adjustment is realized by adjusting the prediction time domain, the control time domain, the sampling period and the constraint problem.
8. The intelligent linkage method of the oil and gas exploitation process according to claim 7, wherein the control time domain is selected to be 10% -20% of the prediction time domain.

Description

Intelligent linkage method for oil and gas exploitation process Technical Field The invention relates to the technical field of oil and gas exploitation, in particular to an intelligent linkage method for an oil and gas exploitation process. Background Petroleum and natural gas are indispensable energy sources in modern society, ‌ is used for power generation, ‌ transportation, ‌ chemical industry and other industries, ‌ can meet the energy demands of countries or regions by exploiting the resources, ‌, ‌ supports economic development and improvement of the living standard of people, ‌ can drive development of related industries, ‌ is used for petrochemical industry, ‌ transportation and the like, ‌ promotes economic growth and employment of countries or regions. In the existing digital oilfield architecture in the oil and gas industry, after operation data is collected by an internet of things sensor, the data is usually transmitted back to a data center, so that excessive bandwidth and computing capacity are consumed, seamless connection and sharing of the data cannot be realized, and the running state cannot be automatically adjusted. Therefore, the intelligent linkage method of the oil and gas exploitation process can be designed aiming at the problem that the running state cannot be automatically adjusted due to excessive consumption of bandwidth and computing capacity. Disclosure of Invention In order to overcome the problem that the running state cannot be automatically adjusted. The technical scheme of the invention is that the intelligent linkage method of the oil and gas exploitation process comprises the following steps: s1, data acquisition and monitoring, namely acquiring temperature, ‌ pressure and ‌ flow data in the oil refining process in real time through data acquisition equipment, transmitting ‌ to a central control room through a network for data processing, knowing the running state of the oil refining process, ‌ and adjusting according to requirements; S11, data analysis, namely mining and integrating all acquired data, wherein the data analysis specifically comprises data preprocessing and data prediction analysis; S111, preprocessing data, namely, processing missing values, detecting and processing abnormal values, normalizing/normalizing the data, coding classification variables, selecting characteristics and reducing dimensions, transforming the data, dividing the data, processing unbalanced data, preprocessing text data and carrying out time sequence data and mathematical; S112, data predictive analysis is established by using statistical analysis and a model, and data are subjected to deep analysis and mining; and S12, displaying the processed data in a graph or curve mode, so that a producer can more intuitively know the current production state, including but not limited to yield, productivity, temperature, humidity and pressure information. S2, model optimization and prediction ‌ Optimizing and predicting by establishing a mathematical model of the oil refining process, ‌ optimizing the model ‌ to realize the optimal operation of the oil refining process, ‌ improving the product quality and the production efficiency, ‌ predicting the possible problems in the oil refining process, ‌ taking corresponding measures in advance, ‌ reducing risks; The main method of the model optimization algorithm comprises a gradient descent algorithm, a random gradient descent algorithm and a dynamic gradient descent algorithm, wherein the gradient descent algorithm gradually updates parameters from randomly initialized parameters to minimize a loss function; a random gradient descent algorithm, starting from the randomly initialized parameters, gradually updating the parameters to minimize the loss function; a dynamic gradient descent algorithm, starting from the randomly initialized parameters, gradually updating the parameters to minimize the loss function; The algorithm of the model prediction comprises linear regression, a support vector machine, a decision tree and a random forest, wherein the formula of the linear regression mathematical model is as follows ,Is a dependent variable,Is an independent variable,Is of the intercept,Is a slope,Is an error term,The number of the samples is that the mathematical model formula of the support vector machine is,Representing a decision function,Is a parameter of a support vector machine,Is a sample label,Is a kernel function, b is an offset, x is an input vector, and the decision tree mathematical model formula is ifWhereinIs an input variable,Is an output variable,Representing category, c representing category number, and a random forest mathematical model formula is,Representing the predicted result of the random forest on the input data x,Representing the number of decision trees in a random forest,Represent the firstA prediction result of the decision tree; And S3, intelligent control and regulation, namely ‌, realizing automatic control of each u