CN-122021224-A - Modeling method of intelligent pipeline data model machine

CN122021224ACN 122021224 ACN122021224 ACN 122021224ACN-122021224-A

Abstract

The invention relates to the field of pipeline data management, in particular to an intelligent pipeline data model machine modeling method; compared with the prior art, the intelligent pipeline data model machine modeling method is difficult to accurately grasp the pipeline running state in real time and has low gas transmission efficiency because the pipeline management method is often dependent on manual experience and regular inspection, and the technical scheme is that the intelligent pipeline data model machine modeling method comprises the specific related business process, the minimum movable unit of data is determined, the dimension and the fact in a data model are designed, a data warehouse is built, the liquid accumulation pre-judging model is built based on historical data and real-time data by using a machine learning and statistics method, the liquid accumulation situation, the abnormal gas flow speed and the like in the pipeline can be monitored and predicted in real time through the intelligent judging model and the liquid accumulation pre-judging analysis model, an automatic switching mechanism is triggered timely, and a compressor control strategy is optimized, so that high-efficiency gradient relay is realized, and the gas transmission efficiency is improved.

Inventors

WANG JIAN
XU WENLONG
REN XIAORONG
LI ERYANG
GUO SHENGZHONG
GAO LEI
XU BIN
LI ZHIGUANG
LIAO JIAN
HUANG ZHAN
XIA XING
XU ZIQIANG

Assignees

中国石油天然气股份有限公司

Dates

Publication Date: 20260512
Application Date: 20241111

Claims (10)

1. The modeling method of the intelligent pipeline data model machine is characterized by comprising the following steps of: s1, collecting operation data of a single well, a gas collecting station and a pipeline, and cleaning and standardizing the data; S2, determining a modeling target, and performing model design and verification; S3, determining the minimum activity unit of data, designing the dimension and the fact in the data model and constructing a data warehouse in the explicitly related business process; S4, based on historical data and real-time data, building a hydrops prejudging model by utilizing a machine learning and statistics method, training the model by utilizing the historical data, adjusting parameters to improve the prediction accuracy, and verifying the accuracy and reliability of the model by comparing a prediction result with actual data; S5, formulating a compressor control strategy according to the accumulated liquid pre-judging result and the pipeline running state; s6, integrating the intelligent judging and identifying model, the automatic switching mechanism, the hydrops pre-judging and analyzing model and the compressor control strategy into a unified intelligent pipeline data model; And S7, collecting new data in the running process of the system for model optimization and strategy adjustment.
2. The modeling method of the intelligent pipeline data modeling machine according to claim 1, wherein when the operation data of a single well, a gas gathering station and a pipeline are collected, the data are acquired in real time through devices such as a sensor and an instrument, and the acquired data are transmitted to a cloud storage platform in a wireless transmission mode, and the method comprises the following specific steps of: S101, determining data sources including single well data, gas gathering station data, pipeline data and external data; S102, acquiring data in real time through equipment such as a sensor, an instrument and the like, and acquiring the real-time data; S103, acquiring historical data from a database, a file or a third party data source; s104, inputting data which cannot be automatically acquired into a system in a manual mode; s105, transmitting the acquired data to a cloud storage platform in a wireless transmission mode; And S106, storing data by using a NoSQL database, and ensuring the safety and expandability of the data.
3. The modeling method of intelligent pipeline data modeling machine of claim 2, wherein the data is cleaned and standardized, the data from different sources is converted into a unified format, and the data from different data sources is combined into a unified data set for subsequent analysis, and the specific steps are as follows: s201, checking missing values in data, and filling or deleting according to business logic; s202, identifying and processing abnormal values, and identifying, deleting or correcting abnormal data by means of setting a threshold value, a box diagram and the like; s203, deleting the completely repeated data records, and merging or reserving partial repeated records according to service requirements; S204, converting data from different sources into a unified format, converting text data into numerical data, and converting data of different units into a unified unit; S205, carrying out normalization or standardization processing on the numerical data so as to eliminate the influence of dimension on data analysis; s206, merging the data from different data sources into a unified data set; S207, linking the data in different forms or data sets through a main key, an external key and the like to form a complete data view; s208, checking whether the data are complete or not, and checking whether the data are consistent or not; s209, setting data quality evaluation indexes such as integrity, accuracy, consistency and timeliness; S210, feeding back and optimizing the data collection and processing flow according to the evaluation result; s211, defining a data model according to service requirements and data characteristics, wherein the data model comprises a data table structure, field types and constraint conditions; s212, establishing a data dictionary, and recording detailed description, data sources, update frequency and other information of each field; And S213, formulating data standards and constraints to ensure the consistency and accuracy of the data.
4. The modeling method of the intelligent pipeline data modeling machine according to claim 3, wherein the modeling target is determined, and the model design and verification are performed, and the specific steps are as follows: s301, defining main targets of the project and collecting specific target requirements; s302, classifying requirements according to functional modules, such as data acquisition and monitoring, intelligent judgment, automatic switching, liquid accumulation pre-judgment and compressor control; s303, according to the importance and urgency of the service demands, the demands are prioritized, and the key demands are ensured to be met first; S304, analyzing whether conflict exists among all the demands, and making a solution to ensure that all the demands can be reasonably met.
5. The modeling method of intelligent pipeline data modeling machine of claim 4, wherein determining modeling targets and performing model design and verification, further comprises the steps of: s305, designing the whole architecture of a system according to service requirements, wherein the whole architecture comprises a data acquisition layer, a data processing layer, a model calculation layer and a decision support layer; s306, designing an algorithm and rules for automatically identifying the pipeline running state; s307, designing logic rules, and automatically triggering switching operation under specific conditions; S308, establishing a hydrops prejudging model by using a machine learning and statistical method, and predicting the hydrops condition of a pipeline; S309, designing a compressor control strategy to realize high-efficiency step relay and optimize the gas flow rate; s310, designing a flow path and processing logic of data among all modules, and ensuring the accuracy and timeliness of the data; s311, defining interface specifications among the modules, including data formats, transmission protocols, calling modes and the like, and ensuring seamless butt joint among the modules; S312, primarily evaluating the system performance according to the system architecture and the function module design, and performing necessary optimization according to the evaluation result; S313, verifying the validity and accuracy of the model through simulation test or small-scale test run.
6. The modeling method of intelligent pipeline data modeling machine of claim 5, wherein the business process is explicitly involved, the minimum activity unit of the data is determined, the dimensions and facts in the data model are designed, and a data warehouse is constructed; the method comprises the following specific steps: s401, identifying and defining a core business process related to an intelligent pipeline data model, such as gas delivery, compressor control and effusion treatment; s402, determining a granularity level of stored data in a data warehouse, namely a minimum activity unit of the data, wherein in an intelligent pipeline data model, the granularity level comprises data such as hourly, daily and monthly flow, pressure and temperature; S403, defining dimensions in the data model, wherein the dimensions are used for describing the context of a business process, such as time, place, equipment, gas composition and the like; s404, designing attributes for each dimension; S405, defining facts in a data model, namely service indexes such as flow, pressure, energy consumption and accumulated liquid amount to be measured; s406, determining a calculation mode and a storage format of the fact; s407, constructing a logic model of the data warehouse by using a star pattern or snowflake pattern dimension modeling technology; S408, designing a relation between a fact table and a dimension table, and ensuring effective association and quick query of data; s409, designing a physical storage structure of a data warehouse according to a logic model, wherein the physical storage structure comprises a database table structure, an index and a partition; s410, developing an ETL process, extracting data from a source system, cleaning, converting and loading the data into a data warehouse; s411, executing an ETL process, and loading the cleaned and converted data into a data warehouse; And S412, verifying whether the data loaded into the data warehouse is accurate, complete and consistent.
7. The modeling method of intelligent pipeline data modeling machine according to claim 6, wherein the model is built by machine learning and statistics based on historical data and real-time data, the model is trained by using the historical data, parameters are adjusted to improve prediction accuracy, and the accuracy and reliability of the model are verified by comparing the prediction result with actual data, and the method comprises the following steps: s501, screening out data related to hydrops prejudgement from a data warehouse, wherein the data comprise historical pressure, flow, temperature, gas components and the like; S502, dividing a data set into a training set, a verification set and a test set; s503, selecting a characteristic with larger influence on the hydrops prejudgment based on data correlation analysis; s504, further processing the selected features, and improving the learning efficiency and the prediction accuracy of the model through feature scaling, feature coding and feature combination; S505, according to the characteristics of the problems and the data characteristics, a logistic regression algorithm is selected to construct a hydrops prejudgement model; s506, adjusting model parameters to optimize model performance; S507, training the model by using training set data, and adjusting model parameters by an iterative optimization algorithm to minimize the prediction error of the model on the training set; S508, controlling the generalization capability of the model by adjusting the complexity of the model, introducing regularization terms and using a cross-validation method; s509, verifying the trained model by using verification set data, and evaluating the prediction performance of the model on unseen data; s510, evaluating the performance of the model by calculating indexes such as accuracy, recall rate, F1 fraction, ROC curve, AUC value and the like of the model; s511, final testing is carried out on the optimized model by using test set data, the tested model is deployed into an intelligent pipeline data model, and the intelligent pipeline data model is integrated with modules such as an intelligent judgment model and an automatic switching mechanism to jointly realize the pre-judgment and automatic processing of the effusion.
8. The modeling method of intelligent pipeline data modeling machine according to claim 7, wherein the compressor control strategy is formulated according to the liquid accumulation pre-judging result and the pipeline running state, and the specific steps are as follows: s601, selecting proper compressor control technology and algorithm, such as PID control, fuzzy control and neural network control, according to actual requirements and technical feasibility; S602, according to the hydrops prejudgement result, the pipeline running state and the gas transmission efficiency requirement, defining the target and constraint condition of the compressor control strategy; s603, establishing a specific compressor control strategy, including selection of control variables and formulation of control rules; s604, optimizing key parameters in a control strategy, such as proportional, integral and differential coefficients of a PID controller, through theoretical calculation or simulation; S605, establishing a simulation model of equipment such as a pipeline, a compressor and the like by using simulation software PipeSim, and ensuring that the model can accurately reflect the actual running condition; s606, implementing a formulated compressor control strategy in a simulation model to simulate the pipeline running states under different working conditions; S607, collecting data in the simulation process, and evaluating whether the effect of the control strategy reaches an expected target; s608, identifying problems and defects in the control strategy according to the simulation test result; S609, adjusting and optimizing the control strategy according to the problems, and performing iterative test on the adjusted control strategy to ensure that the control strategy can stably and efficiently run under different working conditions; S610, deploying the optimized compressor control strategy into an actual pipeline system to replace the original control strategy; s611, utilizing the monitoring function of the intelligent pipeline data model machine to monitor the running state of the compressor and the running state of the pipeline in real time; and S612, collecting actual operation data and evaluating the actual effect of the control strategy.
9. The modeling method of the intelligent pipeline data model machine according to claim 8, wherein the intelligent judging and identifying model, the automatic switching mechanism, the hydrops prejudging and analyzing model and the compressor control strategy are integrated into a unified intelligent pipeline data model, and the specific steps are as follows: s701, defining the functions to be realized by the system, including data integration, model calling, real-time monitoring, early warning notification and control instruction issuing; S702, establishing an overall architecture of a system, wherein the overall architecture comprises a front-end display layer, a business logic layer, a data access layer and a data storage layer; s703, defining interface protocols among the modules to ensure smooth data circulation and instruction transmission; S704, packaging the intelligent judgment model, the automatic switching mechanism, the accumulated liquid pre-judgment analysis model, the compressor control strategy and other modules to provide a unified calling interface; s705, butting the packaged model interface with a business logic layer to ensure that the system can call each model according to the need; And S706, carrying out integrated test on each model, and verifying the accuracy of the collaborative capability and the data interaction between the models.
10. The modeling method of an intelligent pipeline data model machine according to claim 9, wherein the intelligent judgment model, the automatic switching mechanism, the hydrops prejudgment analysis model and the compressor control strategy are integrated into a unified intelligent pipeline data model, and further comprising the steps of: S707, developing a user-friendly front-end interface based on Web or a mobile platform to realize the functions of visual display of data, issuing of operation instructions and the like; s708, writing a business logic code, processing a front-end request, calling a model interface and executing a control instruction; S709, designing a database structure, and storing various data generated in the running process of the system, wherein the various data comprise real-time data, historical data and model parameters; S710, integrating the front end, the rear end, the database and each model module to form a complete intelligent pipeline management system; S711, testing various functions of the system to ensure that the service requirements are met; s712, testing performance indexes such as response time, throughput, number of concurrent users and the like of the system, and ensuring that the system can still stably run under high load; s713, checking the security hole of the system, and ensuring the security of data transmission and storage; S714, preparing hardware resources such as a server and network equipment, and installing software environments such as an operating system, a database and middleware; and S715, deploying the developed system into a production environment.

Description

Modeling method of intelligent pipeline data model machine Technical Field The invention relates to the field of pipeline data management, in particular to an intelligent pipeline data model machine modeling method. Background The pipeline management method is a method for ensuring normal operation, prolonging the service life and improving the management efficiency of the pipeline, and relates to multiple layers, including detection, maintenance, repair, informatization management and the like; firstly, in the aspect of pipeline detection, various means are provided in the prior art, such as visual detection, pressure test, ultrasonic detection, magnetic powder detection and the like, the problems of cracks, sediments, blockage, corrosion and the like in the pipeline can be found timely, accurate basis is provided for subsequent maintenance and repair work, secondly, in the aspect of pipeline maintenance, the pipeline maintenance mainly comprises measures of cleaning a pipeline, anti-corrosion treatment, periodic maintenance, external damage prevention and the like, the pipeline can be cleaned regularly, blocking and friction reduction can be prevented, corrosion can be effectively prevented by the anti-corrosion treatment, the service life of the pipeline is prolonged, in addition, normal operation of the pipeline can be ensured and worn or aged components can be replaced timely by periodically checking components such as connecting pieces, valves and accessories of the pipeline, a protective cover, a fixed pipeline and warning marks are also required to be arranged for preventing damage of the pipeline by external factors, but the pipeline management method in the prior art often depends on manual experience and periodic inspection, the pipeline operation state is difficult to master accurately in real time, and the gas transmission efficiency is low. Disclosure of Invention Compared with the pipeline management method in the prior art, the pipeline management method often depends on manual experience and regular inspection, and is difficult to accurately grasp the running state of the pipeline in real time, so that the gas transmission efficiency is low. The technical scheme of the invention is that the modeling method of the intelligent pipeline data model machine comprises the following steps: s1, collecting operation data of a single well, a gas collecting station and a pipeline, and cleaning and standardizing the data; S2, determining a modeling target, and performing model design and verification; S3, determining the minimum activity unit of data, designing the dimension and the fact in the data model and constructing a data warehouse in the explicitly related business process; S4, based on historical data and real-time data, building a hydrops prejudging model by utilizing a machine learning and statistics method, training the model by utilizing the historical data, adjusting parameters to improve the prediction accuracy, and verifying the accuracy and reliability of the model by comparing a prediction result with actual data; S5, formulating a compressor control strategy according to the accumulated liquid pre-judging result and the pipeline running state; s6, integrating the intelligent judging and identifying model, the automatic switching mechanism, the hydrops pre-judging and analyzing model and the compressor control strategy into a unified intelligent pipeline data model; And S7, collecting new data in the running process of the system for model optimization and strategy adjustment. When operation data of a single well, a gas gathering station and a pipeline are collected, the data are acquired in real time through equipment such as a sensor, an instrument and the like, and the acquired data are transmitted to a cloud storage platform in a wireless transmission mode, and the method comprises the following specific steps: S101, determining data sources including single well data, gas gathering station data, pipeline data and external data; S102, acquiring data in real time through equipment such as a sensor, an instrument and the like, and acquiring the real-time data; S103, acquiring historical data from a database, a file or a third party data source; s104, inputting data which cannot be automatically acquired into a system in a manual mode; s105, transmitting the acquired data to a cloud storage platform in a wireless transmission mode; And S106, storing data by using a NoSQL database, and ensuring the safety and expandability of the data. The method comprises the following steps of cleaning and standardizing data, converting the data from different sources into a unified format, merging the data from different data sources into a unified data set for subsequent analysis, and specifically comprises the following steps: s201, checking missing values in data, and filling or deleting according to business logic; s202, identifying and processing abnormal values, and identifying, deleting or c