Full-domain and full-process closed-loop control method for machine mining time rate based on big data analysis

Abstract

The invention relates to a big data analysis-based full-domain full-flow closed-loop control method for a mechanical extraction time rate, belonging to the technical field of oil and gas field big data application and industrial automation control, and comprising the following steps: S1, collecting data by a well stopping alarm instrument; s2, performing data management; s3, carrying out data classification; s4, carrying out influence factor identification; and S5, carrying out platform management and control. Comprising the following steps of: 1, acquiring a motion signal of the oil pumping unit by adopting a well stopping alarm instrument, uploading the motion signal to a server through a special wireless channel meeting the requirements of a network, and entering a background database; secondly, the server removes bad data through a data cleaning program independently developed by a background, data management is completed, and effective data capable of truly reflecting the field operation state is formed; and step 3, carrying out classification statistical analysis on the characteristics of the data by utilizing a clustering algorithm. According to the method, the oil extraction time rate of the oil pumping unit is scientifically and accurately calculated and is used for guiding production, and the whole-process closed-loop control of the oil extraction time rate is realized.

Description

Technical Field

[0001] This invention belongs to the field of oil and gas field big data application and industrial automation control technology, specifically involving a closed-loop control method for the whole-domain and whole-process machine extraction rate based on big data analysis. Background Technology

[0002] Given the current conditions of significant oil price fluctuations, lack of new production capacity replenishment, and declining output in older oilfields, improving oil recovery rate (ERR) and thereby enhancing work efficiency, maintenance quality, and management level, ultimately controlling natural decline, is a crucial issue that must be addressed. ERR refers to the ratio of a well's cumulative actual monthly production time to its cumulative monthly calendar production time. The construction of smart oilfields places higher demands on the statistical and analytical methods for ERR. Traditional ERR statistical analysis relies on manual inspections to identify well shutdowns, verify the reasons for shutdowns, and record them in paper ledgers or Excel spreadsheets for management and statistical analysis. This approach cannot guarantee timeliness, accuracy, and systematicness, and suffers from the following five problems:

[0003] 1. Insufficient timeliness. Manual inspection and recording methods are often affected by various factors such as manpower, weather, and the frequency of well inspections, resulting in delays in the discovery, verification, and recording of well shutdowns, and failing to reflect the actual operating status of oil wells in a timely manner.

[0004] 2. Accuracy is difficult to guarantee. Manual recording may contain errors, such as data entry errors or omissions. At the same time, due to the influence of human factors, the judgment and recording of the cause of well shutdown may also be subjective and inconsistent, leading to inaccurate statistical analysis results.

[0005] 3. Low work efficiency. Manual inspection and recording require a lot of manpower, resources and time, and it is difficult to process and analyze large amounts of data quickly and accurately, resulting in low work efficiency.

[0006] 4. Difficulty in data management and analysis. Data management methods using paper ledgers or Excel spreadsheets are not conducive to data storage, retrieval, and sharing. At the same time, inconsistencies in data format, source, and maintenance also increase the difficulty of data analysis and processing.

[0007] 5. Lack of systematicity and intelligence. Traditional technologies lack systematicity and intelligence, making it impossible to monitor and provide early warnings of the operating status of oil wells in real time, or to conduct in-depth analysis and investigation of the causes of well shutdowns, thus failing to provide strong support for production decisions. Summary of the Invention

[0008] To address the aforementioned problems, this invention proposes a closed-loop control method for machine-collected data rate across the entire process, based on big data analysis, comprising the following steps:

[0009] S1. Data collected by the well stoppage alarm device;

[0010] S2. Conduct data governance;

[0011] S3. Classify the data;

[0012] S4. Identify influencing factors;

[0013] S5. Conduct platform management.

[0014] The process includes the following:

[0015] The first step is to use a well stop alarm device to collect the motion signals of the pumping unit and upload them to the server through a dedicated wireless channel that meets network requirements, so that they can be entered into the background database.

[0016] The second step is for the server to use a data cleaning program developed in the background to remove bad data, complete data governance, and form effective data that can truly reflect the on-site operating status.

[0017] The third step is to use clustering algorithms to classify and statistically analyze the characteristics of the data.

[0018] Step S3 includes the following steps:

[0019] S31. Problem definition, data collection, and preparation for data classification;

[0020] S32. Separate the categorical data and perform cross-validation to improve data robustness;

[0021] S33. Form a feature stack to transform the data into the format required by the model;

[0022] S34. Form the overall feature stack;

[0023] S35, Decomposition and Refinement;

[0024] S36, Feature Selection;

[0025] S37. Algorithm training to obtain the optimal model;

[0026] S38. Solutions were found based on the analysis and mining.

[0027] In step S31, the problem of the analysis business is clearly defined, and it is determined whether it belongs to classification, regression or other problems.

[0028] In step S32, the data is separated using cross-validation.

[0029] In step S33, a feature stack of members, trade union organizations, and inclusive activities is formed, and the data is transformed into the form required by the model.

[0030] In step S34, the feature stacks are combined to form the total feature stack.

[0031] In step S35, the model is further optimized by decomposition.

[0032] In step S36, feature selection is performed to avoid the "curse of dimensionality" by eliminating redundant feature data.

[0033] In step S37, an algorithm is selected for training, parameters are optimized, and the model is evaluated to obtain the optimal model.

[0034] In step S38, the data to be analyzed is analyzed and mined according to the optimal model, and an analysis result report is generated.

[0035] The beneficial effects of this invention are as follows: This invention provides a method for statistical analysis of pumping unit start-up and shutdown status data, scientifically and accurately calculating the pumping unit's oil production rate, which is then used to guide production. Oilfields applying this method can achieve closed-loop management of the entire oil production rate process. Company-level, oilfield-level, and operator-level operations can monitor well production rate changes 24 / 7 through an IoT management platform. Operators can control the start-up and shutdown of oil wells through the platform, automatically collecting start-up and shutdown information and compiling and organizing the reasons for well shutdowns. Company-level production management departments can utilize approximately 190 well start-up and shutdown data collected monthly. The system generates tens of thousands of records, constructs a database of well start-ups and shutdowns across the entire oilfield, and automatically calculates the monthly performance of time-rate indicators for different oil production plants. It automatically calculates the company's average time-rate and performs year-on-year (comparison with the same period last year) and month-on-month (comparison with adjacent months) analyses. It also calculates the percentage of each factor affecting the time-rate, identifies the main influencing factors, and formulates production management measures. Furthermore, it calculates the company's average daily well shutdown time, the average daily number of well starts per well (start-up frequency), and compares the number of remote well start-ups and shutdowns. Finally, it ranks oil production plants based on their time-rate indicators, compares changes in ranking, and distributes the rankings to the oil production plants for performance evaluation and management. Attached Figure Description

[0036] Figure 1 This is a flowchart illustrating the full-domain, full-process closed-loop control of the oil production rate of the pumping unit in this invention.

[0037] Figure 2 This is an external view of the well stop alarm device;

[0038] Figure 3 This is a circuit board diagram for a well stop alarm device.

[0039] Figure 4 This is a framework diagram of the oil production rate algorithm for the pumping unit of the present invention;

[0040] Figure 5 This is a summary diagram of the shutdown factors in oil production plants (non-intermittent pumping wells) according to the present invention;

[0041] Figure 6 This is a single-well analysis example diagram of the present invention;

[0042] Figure 7 This is a single-well analysis example diagram of the present invention;

[0043] Figure 8 This is a diagram of the well shutdown data after preprocessing according to the present invention;

[0044] Figure 9 This is a correlation analysis diagram of the time rate and well group of the non-intermittent pumping well in Team 2 of Oil Production Plant No. 1 in this invention.

[0045] Figure 10 This is a correlation analysis chart showing the time rate of non-intermittent pumping wells and the number of well shutdowns in the No. 2 Oil Production Team of Oilfield No. 1 of this invention.

[0046] Figure 11 This is a correlation analysis chart of the time rate and shutdown duration of the non-intermittent pumping well in Team 2 of Oil Production Plant No. 1 in this invention.

[0047] Figure 12 This is a correlation analysis diagram of the time rate and operating time of the non-intermittent pumping well in Team 2 of Oil Production Plant No. 1 in this invention.

[0048] Figure 13 This is a correlation analysis diagram of the pumping rate and well group between the No. 2 oil production team of the No. 1 oil production plant in this invention.

[0049] Figure 14 This is a correlation analysis chart showing the time rate of well pumping and the number of well shutdowns in the No. 2 oil production team of the No. 1 oil production plant according to the present invention.

[0050] Figure 15 This is a correlation analysis diagram of the pumping rate and shutdown time of the No. 1 oil production plant No. 2 team in this invention.

[0051] Figure 16 This is a correlation analysis diagram of the time rate and operating time of the pumping well in Team 2 of Oil Production Plant No. 1 in this invention.

[0052] Figure 17 The spectrum of well shutdowns in Oil Production Plant No. 1 of this invention Figure 1 ;

[0053] Figure 18 This invention provides a spectral analysis of well shutdown duration and frequency at Oil Production Plant No. 1. Figure 2 ;

[0054] Figure 19 This invention relates to the shutdown duration spectrum of a certain well group in the No. 1 oil production plant. Figure 3 ;

[0055] Figure 20 This is a statistical chart of well shutdown time after the non-intermittent pumping well data treatment according to the present invention;

[0056] Figure 21This is the interface of the statistical chart of the reasons for well shutdown in Oil Production Plants No. 1, No. 2, and No. 3 of this invention;

[0057] Figure 22 This is the interface for analyzing the pumping rate between oil production plants in this invention.

[0058] Figure 23 This is the interface for counting the number of wells pumped between oil production plants in this invention.

[0059] Figure 24 This is the interface for analyzing the pumping rate between Team 2 and Team 1 of Oil Production Plant in this invention;

[0060] Figure 25 This is a statistical chart of the pumping rate between Team 2 and Team 1 of Oil Production Plant No. 1 in this invention;

[0061] Figure 26 This is a graph showing the analysis results of the well-pumping clustering algorithm for Oil Production Plant No. 1 with an operating time of >9000 hours according to the present invention.

[0062] Figure 27 This is a chart showing the manual statistical analysis of the original data on the number of wells pumped during the operating time of Oil Production Plant No. 1 for more than 9000 hours in this invention.

[0063] Figure 28 This is a diagram of the MATLAB tool software program editing interface of the present invention;

[0064] Figure 29 This is a diagram of the .dat data interface of the present invention;

[0065] Figure 30 This is a line graph showing the loss error of the present invention;

[0066] Figure 31 This is a line graph showing the convergence results of this invention;

[0067] Figure 32 This is the homepage interface of the platform of this invention. Figure 1 ;

[0068] Figure 33 This is the homepage interface of the platform of this invention. Figure 2 ;

[0069] Figure 34 This is a screenshot of the plant-level time rate data list interface of the present invention;

[0070] Figure 35 This is an interface diagram showing the plant-level time rate variation trend of the present invention;

[0071] Figure 36 This is a diagram of the plant-level time-rate comparison interface of the present invention;

[0072] Figure 37 This is a diagram of the team-level user information query interface of the present invention;

[0073] Figure 38 This is a percentage chart showing the reasons for well shutdowns in this invention.

[0074] Figure 39 This is a comparative chart showing the percentage of causes of well shutdown in this invention;

[0075] Figure 40 This is a summary diagram of the factors contributing to well shutdown in this invention.

[0076] Figure 41 This is a block diagram showing the main features of the method of the present invention;

[0077] Figure 42 This is a block diagram illustrating the scope of the method of the present invention. Detailed Implementation

[0078] To make the technical means and objectives of this invention easier to understand, the invention is further described below with reference to specific embodiments, such as a closed-loop control method for machine sampling rate across the entire domain and process based on big data analysis. Figure 1 As shown, the process includes the following steps: S1, data collection by the well shutdown alarm device; S2, data processing; S3, data classification; S4, identification of influencing factors; S5, platform management. The process includes the following steps: First, the well shutdown alarm device collects the motion signals of the pumping unit and uploads them to the server via a dedicated wireless channel that meets network requirements, entering the backend database; Second, the server uses a self-developed data cleaning program to remove bad data, completing data processing and forming effective data that accurately reflects the on-site operating status; Third, clustering algorithms are used to classify and statistically analyze the characteristics of the data.

[0079] Specifically as follows:

[0080] 1. Develop an automatic acquisition device for the start and stop signals of an oil pumping unit.

[0081] In 2015, a pumping unit shutdown alarm device was invented, and in 2018, a new type of patent for the "Pumping Unit Shutdown Alarm Device" was applied for. This product is battery-powered and relies on encrypted communication via a supplier's 2G or 4G private network. It uses an accelerometer to determine the well's start-up and shutdown status. Because it is battery-powered, it can accurately upload the start-up and shutdown status to the database even in the event of a power outage at the well site. The oilfield has achieved full coverage installation of the shutdown alarm device, realizing automatic data collection of well start-up and shutdown status, achieving high efficiency and accuracy in front-end data collection, and providing a reliable data foundation for big data analysis. Figures 2-3 As shown.

[0082] Establish a big data statistical analysis algorithm framework for oil production rate.

[0083] The first step is problem definition, which aims to prepare for data classification; such as the process. Figure 4 The stage where the number 1 in the text is located;

[0084] The second step is to classify and separate the data and perform cross-validation to improve data robustness; as shown in the process. Figure 4 The stage where number 2 is located;

[0085] The third step is to create a feature stack, transforming the data into the format required by the model; as shown in the process. Figure 4 The stage where the number 3 is located;

[0086] The fourth step is to combine the various feature stacks to form the final feature stack; as shown in the flowchart. Figure 4 The stage indicated by the number 4 in the text;

[0087] The fifth step is decomposition and refinement, with the aim of further optimizing the model; such as the process. Figure 4 The stage indicated by the number 5 in the text;

[0088] The sixth step, feature selection, aims to avoid the "curse of dimensionality" by eliminating redundant feature data; as shown in the process... Figure 6 The stage where the number 1 in the text is located;

[0089] The seventh step is algorithm training, the goal of which is to obtain the optimal model; as shown in the flowchart. Figure 4 The stage indicated by the number 7 in the text;

[0090] Step 8 involves analysis and data mining to arrive at a solution; as shown in the flowchart. Figure 4 The stage indicated by the number 8 in the text;

[0091] 3. Complete the development of big data statistical analysis and intelligent judgment programs.

[0092] (1) Development of preprocessing program for big data on oil production rate.

[0093] After obtaining the raw data, it was found that some data was structured while others were unstructured; some data was complete while others were missing; some data was accurate while others contained noise. Therefore, numerous preprocessing algorithms, including data cleaning, outlier handling, and median filtering, were implemented.

[0094] Taking the initial analysis of shutdown factors in three oil production plants (non-intermittent pumping wells) as an example, the summary results are as follows: Figure 5 As shown, a single-well analysis example (Well 1) is as follows: Figure 6 As shown, a single-well analysis example (Well 2) is as follows: Figure 7 As shown, there are a large number of outliers. The preprocessed well shutdown data is as follows: Figure 8 As shown. (2) Complete the development of a big data correlation analysis program for oil production rate.

[0095] Taking the "2018-1-1 to 2019-4-281# Oil Production Plant Team 2 - All Well Groups - Production Statistics" as an example, relevant algorithm analysis was performed, and the "non-intermittent pumping wells" were statistically analyzed. The analysis results are presented as a scatter plot.

[0096] First, a matrix is ​​formed using the algorithmic tool MATLAB. Variables include the number of well shutdowns, well groups, shutdown duration, and positive active energy. The mean and covariance (covariance is a method to measure the strength of the linear relationship between two variables) are calculated. The absolute value with the highest correlation is then assigned to the X and Y axes. A second image window is opened, the current image window is cleared, and the properties of the graphic object are specified. A two-dimensional line graph is then drawn. Through the calculation and analysis of the matrix data's covariance and correlation coefficient across multiple dimensions, a result is obtained and plotted graphically. Figure 9 Figure 10 The scatter plot in the upper right area indicates that the frequent well shutdowns, despite a relatively good oil production rate, suggest that the problem stems from frequent short-term well shutdowns. The underlying management issues should be identified and addressed.

[0097] The provided raw data was graphically analyzed using software tools, yielding the following graphical results. A comparison between the manually collected statistical results and the graphical analysis results proves that the algorithm used is correct. Figures 11-16 This provides the output results for the relevant algorithms. Figure 11 The correlation analysis chart of the non-intermittent pumping well of the No. 2 Oil Production Team of Oilfield No. 1 is a diagram showing the correlation between the pumping rate and the shutdown time of the present invention. Figure 12 The correlation analysis chart of the time rate and running time of the non-intermittent pumping well of the No. 2 team of Oil Production Plant 1 is shown in the figure. Figure 13 The graph showing the correlation between the pumping rate and the well group in the No. 2 oil production team of the No. 1 oilfield is a correlation analysis chart of the pumping rate in this invention. Figure 14 The correlation analysis chart of the pumping rate and the number of well shutdowns in the No. 2 pumping well of Oil Production Plant 1 is a diagram illustrating the relationship between the pumping rate and the number of well shutdowns in this invention. Figure 15 The graph shows the correlation between the pumping rate and the shutdown time of the No. 1 oilfield No. 2 pumping well. Figure 16 The graph showing the correlation between the pumping rate and the operating time of the No. 1 oilfield No. 2 pumping well is an example of the correlation analysis of the pumping rate and the operating time of this invention.

[0098] (3) Complete the development of the big data spectrum analysis program for oil production rate

[0099] Based on big data analysis of well downtime and causes related to oil production rate, a spectral analysis was employed to explore the main reasons for low downtime rates in a particular team or plant. The analysis is presented in the form of a spectral chart of the number of downtimes for a specific well group in a continuously operating team of Oil Production Plant No. 1, a spectral chart analyzing both downtime and frequency for a specific well group in continuously operating operations, and a spectral chart of downtime for a specific well group in continuous operation. Figures 17-19 As shown.

[0100] Preliminary conclusions from the analysis of factors affecting oil production rate indicate that the main problems are concentrated in: motion sensor malfunction, wiring faults (most prominent in #2), platform power supply (no problem in #1), unrepaired areas (very prominent in #5), surface pipelines (no problem in #1, most prominent in #2), single-well power distribution (no problem in #1), oil well maintenance (no problem in #1), surface faults (no problem in #1), and maintenance (no problem in #1). The main problems in #1 oil production plant are significantly different from those in the other two plants, such as... Figures 20-21 As shown.

[0101] Taking "from January 1, 2018 to April 28, 2019, Oilfield No. 1 - all production teams - intermittent pumping wells" as an example, the analysis results are verified, such as... Figures 22-25 As shown in the figure. The results indicate that Team 1#2 does not manage the most wells, but its downtime is particularly long. Teams 1#6 and 1#7 manage a larger number of wells, while their downtime frequency and duration remain at the average. Further analysis of Team 2 reveals that the problem is concentrated in Well Group 100, with two wells in Well Group 100 experiencing excessively long downtimes. Validation results show that using big data clustering algorithms is a highly efficient algorithm for searching for the causes of downtime and identifying wells with low downtime rates.

[0102] The median operating time of all intermittent pumping wells in Oil Production Plant No. 1 was 9003.49 hours. The distribution pattern is largely consistent with the pattern summarized on the previous page. Using the (operating / shutdown time ratio) as a good indicator, wells with a value greater than 20 belong to Team 7, indicating excellent management within Team 7. The optimal ratio points to wells 6-8.2 in Well Group 19 of Team 2. The analysis results of the big data clustering algorithm are consistent with the summarized statistical results of the original data, such as... Figures 26-27 As shown.

[0103] (4) Complete the development of the BP algorithm analysis program for big data on oil production rate.

[0104] To better conduct research on well shutdown rates, this project utilized a big data algorithm based on a backpropagation (BP) neural network. Taking the overall data from Oil Production Plant No. 1 in 2019 as an example, the analysis was carried out, and the algorithm results are as follows: Figures 28-31 As shown. The core algorithm code performs the following functions: First, it reads the training data, normalizes the feature values, constructs the output matrix, creates and applies the BP neural network algorithm, sets the training parameters, number of iterations, loss error, applies the gradient descent algorithm, reads the test data, and starts testing; it normalizes the test data, statistically analyzes the recognition accuracy, and repeatedly tests and learns until it reaches the optimal solution.

[0105] (5) Complete the development of supporting display functions for the oil production rate data platform.

[0106] Platform Homepage: The homepage displays the overall oil production time rate for the most recent month. The left side shows the time rate data for 11 production units: Oil Production Plant #8, Oil Production Plant #6, Oil Production Plant #2, Oil Production Plant #1, Oil Production Plant #4, Gas Production Plant #9, Oil Production Plant #5, Oil Production Plant #3, Oil Production Plant #7, Oil Production Plant #10, and Oil Production Plant #11. The right side shows the frequency statistics of the main factors affecting the oil production time rate for Oil Production Plants #5, #3, and #2, such as... Figures 32-33 As shown.

[0107] Plant-level hourly rate statistics: Figure 34 It allows for comparison of historical data trends for any of the 11 production units. Specific values ​​are displayed in a floating window when the mouse hovers over them.

[0108] You can save the comparison trend curve as an image by clicking the icon in the upper right corner. Clicking the time option allows you to select historical data from the database by month and display the hourly rate data for the current month in a bar chart format, showing 11 units. Figure 35 , Figure 36 As shown.

[0109] Team-level hourly rate statistics: Clicking the time option and the factory-level unit drop-down menu allows you to select historical data from the database by month. The data will be displayed in a bar chart format, showing the current month's hourly rate data for all team-level units in a given factory. Figure 37 As shown.

[0110] Well Shutdown Cause Statistics: Through big data analysis, based on three oil production units (No. 5, No. 3, and No. 2), the main causes of well shutdowns were identified as follows: well occupation due to measures, well occupation due to monitoring, routine maintenance, downhole malfunctions, surface malfunctions, electrical circuit malfunctions, other production methods, natural causes, long-term and short-term shutdowns, and blank categories (10 categories in total). This page allows for a comparison of the percentage of well shutdown causes across the three units. The 10 causes can also be selectively added or removed using the 10 options in the upper left corner. Figures 38-39 As shown.

[0111] Summary of Influencing Factors: This page uses time length as the horizontal axis to divide the main reasons for well shutdowns into 10 categories. It compares the proportion of causes affecting the overall downtime rate for each plant from both overall and individual perspectives. Figure 40 As shown.

[0112] like Figure 41 , Figure 42This control method can be widely applied to oil and gas field enterprises and has great promotional value. The front-end data acquisition equipment is installed using magnetic adsorption, eliminating the need for wiring. It is battery powered (replaceable once every 3 years), features wireless transmission, high integration, simple structure, and convenient maintenance, achieving zero on-site construction. The platform software has automatic data cleaning, classification, and identification functions, and the embedded algorithm has self-learning and self-iteration capabilities. The platform's human-machine interface is simple and intuitive. The platform has a B / S architecture, allows remote multi-user access without installing plugins, shares server hardware and software resources, and achieves zero backend development. Furthermore, this control method has been applied and iterated in oil fields for more than 5 years, and has been tested in actual applications in tens of thousands of oil wells, meeting the daily closed-loop control needs of 10 oil production plants.

[0113] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A closed-loop control method for machine sampling rate across the entire domain and process based on big data analysis, characterized in that, Includes the following steps: S1. Data collected by the well stoppage alarm device; S2. Conduct data governance; S3. Classify the data; S4. Identify influencing factors; S5. Conduct platform management.

2. The closed-loop control method for machine sampling rate based on big data analysis as described in claim 1 includes the following processes: The first step is to use a well stop alarm device to collect the motion signals of the pumping unit and upload them to the server through a dedicated wireless channel that meets network requirements, so that they can be entered into the background database. The second step is for the server to use a data cleaning program developed in the background to remove bad data, complete data governance, and form effective data that can truly reflect the on-site operating status. The third step is to use clustering algorithms to classify and statistically analyze the characteristics of the data.

3. The closed-loop control method for machine sampling rate based on big data analysis as described in claim 1, wherein step S3 includes the following steps: S31. Problem definition, data collection, and preparation for data classification; S32. Separate the categorical data and perform cross-validation to improve data robustness; S33. Form a feature stack to transform the data into the format required by the model; S34. Form the overall feature stack; S35, Decomposition and Refinement; S36, Feature Selection; S37. Algorithm training to obtain the optimal model; S38. Solutions were found based on the analysis and mining.

4. In the closed-loop control method for machine sampling rate based on big data analysis as described in claim 3, in step S31, the problem of the analysis business is clearly defined and it is determined whether it belongs to classification, regression or other problems.

5. The closed-loop control method for machine sampling rate based on big data analysis as described in claim 3, wherein in step S32, data is separated by cross-validation.

6. In the closed-loop control method for machine collection rate based on big data analysis as described in claim 3, in step S33, a feature stack of members, trade union organizations, and inclusive activities is formed, and the data is transformed into the form required by the model.

7. In the closed-loop control method for machine sampling rate based on big data analysis as described in claim 6, in step S34, the feature stacks are combined to form a total feature stack.

8. In the closed-loop control method for machine sampling rate based on big data analysis as described in claim 3, step S35 involves decomposing and further optimizing the model.

9. In the closed-loop control method for machine sampling rate based on big data analysis as described in claim 3, in step S36, feature selection is performed to avoid the "curse of dimensionality" by eliminating redundant feature data.

10. The closed-loop control method for machine sampling rate based on big data analysis as described in claim 3, wherein in step S37, an algorithm is selected for training, parameters are optimized, and the model is evaluated to obtain the optimal model.

11. The closed-loop control method for machine sampling rate based on big data analysis as described in claim 3, wherein in step S38, the data to be analyzed is analyzed and mined according to the optimal model to generate an analysis result report.

Images