An intelligent control system for submersible electric pumps
By combining data cleaning and machine learning models with pressure sensor data, the error problem of submersible pump blockage detection was solved, accurate control and prediction of submersible pumps were achieved, and equipment damage was reduced.
Patent Information
- Application Number
- CN202410598613.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-05-15
- Publication Date
- 2025-09-09
- Estimated Expiration
- 2044-05-15
AI Technical Summary
Existing methods for detecting blockage of submersible oil pumps have large errors, which may cause damage to the submersible oil pumps before they become blocked, and existing control methods are not accurate enough.
Data cleaning and machine learning models (such as SVR and Transformer) are combined with pressure sensor data to predict the operating rate and viscosity of submersible oil pumps, and blockages are accurately identified through multiple judgment modules.
It achieves accurate detection of submersible pump blockage, reduces equipment damage, and improves detection accuracy and predictive capabilities.
Smart Images
Figure CN118257735B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of intelligent control technology, and in particular to an intelligent control system for a submersible electric pump. Background Art
[0002] A submersible pump is an electric centrifugal pump that pumps crude oil from an oil well to the surface. It uses the centrifugal force generated by the rotating impeller to transport liquid and is a non-variable displacement pump. A blockage in a submersible pump increases the pump's workload and, in severe cases, can cause generator burnout, bearing damage, and other failures. One existing method for detecting blockage in a submersible pump is to monitor the pump's inlet and outlet pressures. When a blockage occurs, the inlet and outlet pressures fluctuate. A key factor influencing these pressures is the viscosity of the crude oil. However, current methods for testing crude oil viscosity use crude oil samples collected from outside the submersible pump, which can have a certain degree of error compared to the actual value. This method does not account for the different pump pressure differentials associated with changes in crude oil viscosity, resulting in inaccurate detection and control systems for stopping submersible pumps. Furthermore, existing methods for stopping submersible pumps utilize overheat protection devices, but this can cause damage to the pump before the protection device takes effect. Summary of the Invention
[0003] In order to overcome the shortcomings of detecting problems encountered by submersible electric pumps, the present invention provides an intelligent control system for submersible electric pumps.
[0004] The technical solution is: an intelligent control system for a submersible electric pump, comprising the following steps:
[0005] The first data acquisition and calculation module is used to collect inlet pressure data and outlet pressure data of the submersible electric pump during normal operation, and calculate the pump pressure difference based on the inlet pressure data and outlet pressure data;
[0006] a blockage detection model training module, configured to clean the pump pressure difference data and viscosity data to obtain a fitting data set, input the fitting data set into an SVR model, train the SVR model, and finally obtain a trained blockage detection model;
[0007] The second data acquisition preprocessing module is used to acquire historical downhole data and perform data cleaning on it to obtain a submersible electric pump operation rate data set;
[0008] A prediction model training module is used to input the submersible pump operating rate data set into the Transformer model for training, and finally obtain a submersible pump operating rate prediction model;
[0009] The monitoring model training module is used to collect historical submersible pump operating rates and liquid production, perform linear regression fitting on them, and obtain a liquid production monitoring model;
[0010] The first judgment module is used to obtain real-time downhole data from the downhole monitoring system, input the data into the submersible pump operation rate prediction model to obtain the submersible pump predicted operation rate, and make a judgment based on the submersible pump predicted operation rate;
[0011] A second judgment module is configured to input the predicted operating rate of the submersible electric pump into a production monitoring model to obtain a liquid production, and make a judgment based on the liquid production;
[0012] The third judgment module is used to obtain relevant real-time viscosity according to the downhole monitoring system, average the relevant real-time viscosity and input it into the blockage detection model, and judge whether the submersible oil pump is blocked according to the result and perform relevant processing.
[0013] Preferably, the collecting of inlet pressure data and outlet pressure data during normal operation of the submersible electric pump and the calculating of the pump pressure difference based on the inlet pressure data and outlet pressure data include: obtaining the inlet pressure and outlet pressure through a dedicated pressure sensor installed inside or near the electric pump system, and obtaining the pump pressure difference by taking the difference between the inlet pressure and the outlet pressure.
[0014] Preferably, the data cleaning of the pump pressure difference data and viscosity data to obtain a fitting data set, inputting the fitting data set into the SVR model, training the SVR model, and finally obtaining a trained blockage detection model includes: data cleaning of the pump pressure difference data and viscosity data to remove duplicate values, fill missing values, and eliminate outliers to obtain a fitting data set, and inputting the fitting data set into the SVR model, wherein the fitting data set uses viscosity as the independent variable and the pump pressure difference as the dependent variable to obtain a viscosity-pump pressure difference nonlinear equation.
[0015] Preferably, the historical downhole data is obtained and cleaned to obtain a submersible pump operating rate data set, and the submersible pump operating rate data set is input into a Transformer model for training, and finally a submersible pump operating rate prediction model is obtained, including: historical downhole data such as historical observation values of power, current, voltage, and liquid production are input into the Transformer for training, and the hidden time dependency between different time steps is captured through its self-attention mechanism, and a prediction result of the submersible pump operating rate is generated based on this, and the model is optimized using the mean square error, and finally a submersible pump operating rate prediction model is obtained.
[0016] Preferably, the collecting of historical submersible pump operating speed and liquid production, performing linear regression fitting on the collected data, and obtaining a liquid production monitoring model comprises: forming a scatter plot based on the historical data with the submersible pump operating speed as the independent variable and the liquid production as the dependent variable, performing linear fitting on the scatter plot using the least squares method, and obtaining a linear equation of the submersible pump operating speed-liquid production, wherein the first normal threshold is the sum of the standard deviation and the mean, wherein the mean is calculated as follows: , the standard deviation is calculated as: in The result obtained by the least squares method value.
[0017] Preferably, the real-time downhole data is obtained from the downhole monitoring system and input into the submersible pump operation rate prediction model to obtain the predicted operation rate of the submersible pump, and a judgment is made based on the predicted operation rate of the submersible pump, including: if the predicted operation rate of the submersible pump is lower than the preset rate threshold, then notifying relevant personnel to check the submersible pump.
[0018] Preferably, the predicted operating rate of the submersible electric pump is input into the production monitoring model to obtain the liquid production, and a judgment is made based on the liquid production, including: after the predicted operating rate of the submersible electric pump is input into the production monitoring model, if the obtained liquid production is lower than the first normal threshold value, a blockage detection model is used to judge whether it is blocked. If it is blocked, relevant personnel are notified to inspect and repair the submersible electric pump. If it is not blocked, it is considered that there is insufficient crude oil in the oil well, and the submersible electric pump is automatically stopped.
[0019] Preferably, the blockage detection model is used to determine the blockage, including: obtaining relevant real-time viscosity according to the downhole monitoring system, averaging the relevant real-time viscosity and inputting it into the blockage detection model, and judging whether the submersible electric pump is blocked according to the result and performing relevant processing.
[0020] Preferably, the relevant real-time viscosity is obtained according to the downhole monitoring system, and the relevant real-time viscosity is averaged and then input into the blockage detection model. According to the result, it is judged whether the submersible electric pump is blocked and relevant processing is performed, including: obtaining the current viscosity and the viscosity before the first preset time, taking the average of the two viscosities as the input of the blockage detection model, and then changing the viscosity at the current time to the viscosity after taking the average of the two viscosities, and finally obtaining the result of the blockage detection model.
[0021] Preferably, the viscosity at the current time is also changed to the viscosity after taking the average of the two viscosities, and finally the result of the blockage detection model is obtained, including: if the pump pressure difference obtained based on the viscosity after taking the average of the two viscosities is greater than or equal to the first correct threshold, it is considered that blockage has occurred, otherwise it is considered that no blockage has occurred. Beneficial effects
[0022] The present invention uses a transformer model to predict the operating rate of a submersible electric pump, thereby discovering power changes of the submersible electric pump in advance, detecting whether there is a problem with the submersible electric pump itself, obtaining liquid production through the predicted operating rate of the submersible electric pump, and judging whether the reduction in liquid production is due to insufficient crude oil or a blockage. Finally, by taking the average of the real-time viscosity of the crude oil and the viscosity at a preset time, the viscosity of the crude oil inside the submersible electric pump can be obtained more accurately without adding new equipment to the submersible electric pump, thereby making a more accurate judgment on the blockage phenomenon. BRIEF DESCRIPTION OF THE DRAWINGS
[0023] Figure 1 This is a flow chart of an intelligent control system for a submersible electric pump according to the present invention. DETAILED DESCRIPTION
[0024] The technical solution of the present invention will be further described below with reference to the accompanying drawings. Example 1
[0025] An intelligent control system for a submersible electric pump, such as Figure 1 As shown, the following steps are included:
[0026] The first data acquisition and calculation module is used to collect inlet pressure data and outlet pressure data of the submersible electric pump during normal operation, and calculate the pump pressure difference based on the inlet pressure data and outlet pressure data;
[0027] The inlet pressure and outlet pressure are obtained by installing a special pressure sensor inside or near the electric pump system, and the pump pressure difference is obtained by making the inlet pressure and outlet pressure difference.
[0028] a blockage detection model training module, configured to clean the pump pressure difference data and viscosity data to obtain a fitting data set, input the fitting data set into an SVR model, train the SVR model, and finally obtain a trained blockage detection model;
[0029] The pump pressure difference data and viscosity data were cleaned by removing duplicate values, filling missing values, and eliminating outliers to obtain a fitting data set, which was then input into the SVR model. The fitting data set used viscosity as the independent variable and pump pressure difference as the dependent variable to obtain a viscosity-pump pressure difference nonlinear equation.
[0030] It should be noted that when the fluid passes through the submersible electric pump, the fluid with higher viscosity will produce greater internal resistance, which will cause the submersible electric pump to provide a larger pressure differential to maintain the same flow rate or delivery volume. Therefore, the pressure differential is proportional to the viscosity of the fluid, but this relationship is usually nonlinear, especially under high viscosity and complex flow conditions. Therefore, the SVR model is used to fit the viscosity and pump pressure differential to obtain the viscosity-pump pressure differential nonlinear equation.
[0031] The second data acquisition preprocessing module is used to acquire historical downhole data and perform data cleaning on it to obtain a submersible electric pump operation rate data set;
[0032] A prediction model training module is used to input the submersible pump operating rate data set into the Transformer model for training, and finally obtain a submersible pump operating rate prediction model;
[0033] Historical downhole data such as power, current, voltage, and fluid production are input into the Transformer for training. The self-attention mechanism is used to capture the hidden time dependencies between different time steps, and the prediction results of the submersible pump operating rate are generated based on this. The model is optimized using the mean square error, and finally a submersible pump operating rate prediction model is obtained.
[0034] It should be noted that by inputting historical operating data of submersible oil pumps for training, it is possible to predict the future operating rate of submersible oil pumps and detect in advance whether there are any problems with the submersible oil pumps. By using the self-attention mechanism, time dependence can be better captured, and subsequent transformations are related to previous ones.
[0035] The monitoring model training module is used to collect historical submersible pump operating rates and liquid production, perform linear regression fitting on them, and obtain a liquid production monitoring model;
[0036] The first judgment module is used to obtain real-time downhole data from the downhole monitoring system, input the data into the submersible pump operation rate prediction model to obtain the submersible pump predicted operation rate, and make a judgment based on the submersible pump predicted operation rate;
[0037] Based on historical data, a scatter plot was formed with the submersible pump operating speed as the independent variable and the liquid production as the dependent variable. The scatter plot was linearly fitted using the least squares method to obtain the submersible pump operating speed-liquid production linear equation. The first normal threshold is the sum of the standard deviation and the mean, where the mean is calculated as follows: , the standard deviation is calculated as: in The result obtained by the least squares method value.
[0038] If the predicted operating rate of the submersible oil pump is lower than the preset rate threshold, relevant personnel are notified to check the submersible oil pump.
[0039] It should be noted that after fitting the linear equation of the submersible pump operating rate-liquid production, if the predicted operating rate of the submersible pump is lower than the preset rate threshold, it means that a fault may have occurred and someone can be sent to check it.
[0040] A second judgment module is configured to input the predicted operating rate of the submersible electric pump into a production monitoring model to obtain a liquid production, and make a judgment based on the liquid production;
[0041] After the predicted operating rate of the submersible pump is input into the production monitoring model, if the obtained liquid production is lower than the first normal threshold, the blockage detection model is used to determine whether it is blocked. If it is blocked, the relevant personnel are notified to inspect and repair the submersible pump. If it is not blocked, it is considered that there is insufficient crude oil in the oil well, and the submersible pump is automatically stopped.
[0042] It should be noted that when the predicted operating rate of the submersible pump is input into the production monitoring model, the predicted liquid production is obtained. When the liquid production that should be obtained by the operating rate of the submersible pump is lower than the first normal range, there are two possibilities. One is that the submersible pump is blocked and the amount of extraction is reduced. The other is that the amount of crude oil in the well is insufficient, resulting in less extraction. This method can effectively determine whether to repair the submersible pump or replace the working location of the submersible pump.
[0043] The third judgment module is used to obtain relevant real-time viscosity according to the downhole monitoring system, average the relevant real-time viscosity and input it into the blockage detection model, and judge whether the submersible oil pump is blocked according to the result and perform relevant processing.
[0044] Relevant real-time viscosity is obtained from the downhole monitoring system, and the relevant real-time viscosity is averaged and then input into the blockage detection model. According to the result, it is determined whether the submersible electric pump is blocked and relevant processing is performed.
[0045] The current viscosity and the viscosity before the first preset time are obtained, and the average of the two viscosities is used as the input of the blockage detection model. Then the viscosity at the current time is also changed to the viscosity after the average of the two viscosities, and finally the result of the blockage detection model is obtained.
[0046] If the pump pressure difference obtained by taking the average of the two viscosities is greater than or equal to the first correct threshold, it is considered that blockage has occurred; otherwise, it is considered that no blockage has occurred.
[0047] It should be noted that when obtaining the current crude oil viscosity, because the viscosity of the downhole crude oil is collected rather than the viscosity inside the submersible pump, the detection of the pump pressure difference of the submersible pump at this time may cause some errors, especially when the submersible pump first extracts crude oil with lower viscosity and then extracts crude oil with higher viscosity. At this time, the viscosity of the crude oil inside the submersible pump will change. This error can be effectively reduced by averaging the viscosity of the crude oil collected in real time with the viscosity before the first preset time. The viscosity at the current time is then changed to the viscosity after averaging the two viscosities, so that it is convenient to use the viscosity after averaging the viscosity at the current time. This method can reduce errors without adding an additional viscosity detection device to the submersible pump device. Finally, the blockage detection model is used to detect whether the pump pressure difference exceeds the range, thereby determining whether the submersible pump is blocked.
[0048] The technical principles of the embodiments of the present invention have been described above in conjunction with specific embodiments. These descriptions are intended solely to explain the principles of the embodiments of the present invention and should not be construed in any way as limiting the scope of protection of the embodiments of the present invention. Based on the explanations herein, those skilled in the art will be able to conceive of other specific implementations of the embodiments of the present invention without inventive effort, and such implementations will fall within the scope of protection of the embodiments of the present invention.
Claims
1. An intelligent control system for a submersible electric pump, characterized in that: The following steps are involved: The first data acquisition and calculation module is used to collect inlet pressure data and outlet pressure data of the submersible electric pump during normal operation, and calculate pump pressure difference data based on the inlet pressure data and outlet pressure data; a blockage detection model training module, configured to clean the pump pressure difference data and viscosity data to obtain a fitting data set, input the fitting data set into an SVR model, train the SVR model, and finally obtain a trained blockage detection model; The second data acquisition preprocessing module is used to acquire historical downhole data and perform data cleaning on it to obtain a submersible electric pump operation rate data set; A prediction model training module is used to input the submersible pump operating rate data set into the Transformer model for training, and finally obtain a submersible pump operating rate prediction model; The monitoring model training module is used to collect historical submersible pump operating rates and liquid production, perform linear regression fitting on them, and obtain a liquid production monitoring model; The first judgment module is used to obtain real-time downhole data from the downhole monitoring system, input the data into the submersible pump operation rate prediction model to obtain the submersible pump predicted operation rate, and make a judgment based on the submersible pump predicted operation rate; A second judgment module is configured to input the predicted operating rate of the submersible electric pump into a production monitoring model to obtain a liquid production, and make a judgment based on the liquid production; The third judgment module is used to obtain relevant real-time viscosity according to the downhole monitoring system, average the relevant real-time viscosity and input it into the blockage detection model, and judge whether the submersible oil pump is blocked according to the result and perform relevant processing.
2. The intelligent control system for a submersible electric pump according to claim 1, characterized in that: The method of collecting inlet pressure data and outlet pressure data during normal operation of the submersible electric pump and calculating the pump pressure difference based on the inlet pressure data and outlet pressure data includes: obtaining the inlet pressure and outlet pressure through a dedicated pressure sensor installed inside or near the electric pump system, and obtaining the pump pressure difference by taking the difference between the inlet pressure and the outlet pressure.
3. The intelligent control system for a submersible electric pump according to claim 1, characterized in that: The data cleaning of the pump pressure difference data and the viscosity data to obtain a fitting data set, inputting the fitting data set into the SVR model, training the SVR model, and finally obtaining a trained blockage detection model includes: data cleaning of the pump pressure difference data and the viscosity data to remove duplicate values, fill in missing values, and eliminate outliers to obtain a fitting data set, and inputting the fitting data set into the SVR model, wherein the fitting data set uses viscosity as an independent variable and the pump pressure difference as a dependent variable to obtain a viscosity-pump pressure difference nonlinear equation.
4. The intelligent control system for a submersible electric pump according to claim 1, characterized in that: The method acquires historical downhole data and performs data cleaning on the data to obtain a submersible pump operating rate data set, inputs the submersible pump operating rate data set into a Transformer model for training, and finally obtains a submersible pump operating rate prediction model, including: inputting historical downhole data including historical observation values of power, current, voltage, and liquid production into a Transformer for training, capturing the hidden time dependency between different time steps through its self-attention mechanism, and generating a prediction result of the submersible pump operating rate based on the result, and optimizing the model using mean square error, to finally obtain a submersible pump operating rate prediction model.
5. The intelligent control system for a submersible electric pump according to claim 1, characterized in that: The method collects historical operating speeds and liquid production rates of submersible electric pumps and performs linear regression fitting on the operating speeds and liquid production rates to obtain a liquid production monitoring model, including: forming a scatter plot based on the historical data with the operating speed of the submersible electric pumps as an independent variable and the liquid production rate as a dependent variable, performing linear fitting on the scatter plot using the least squares method to obtain a linear equation of the operating speed of the submersible electric pumps and the liquid production rate, wherein the first normal threshold is the sum of the standard deviation and the mean, wherein the mean is calculated as follows: , the standard deviation is calculated as: in The result obtained by the least squares method The value of .
6. The intelligent control system for a submersible electric pump according to claim 1, characterized in that: The real-time downhole data is obtained from the downhole monitoring system and input into the submersible pump operation rate prediction model to obtain the predicted operation rate of the submersible pump, and a judgment is made based on the predicted operation rate of the submersible pump, including: if the predicted operation rate of the submersible pump is lower than the preset rate threshold, the relevant personnel are notified to check the submersible pump.
7. The intelligent control system for a submersible electric pump according to claim 1, characterized in that: The predicted operating rate of the submersible electric pump is input into the production monitoring model to obtain the liquid production, and a judgment is made based on the liquid production, including: after the predicted operating rate of the submersible electric pump is input into the production monitoring model, if the obtained liquid production is lower than a first normal threshold value, a blockage detection model is used to judge whether it is blocked. If it is blocked, relevant personnel are notified to inspect and repair the submersible electric pump. If it is not blocked, it is considered that there is insufficient crude oil in the oil well, and the submersible electric pump is automatically stopped.
8. The intelligent control system for a submersible electric pump according to claim 7, characterized in that: The blockage detection model is used to determine the blockage, including: obtaining relevant real-time viscosity according to the downhole monitoring system, averaging the relevant real-time viscosity and inputting it into the blockage detection model, and judging whether the submersible electric pump is blocked based on the result and performing relevant processing.
9. The intelligent control system for a submersible electric pump according to claim 8, characterized in that: The method obtains relevant real-time viscosity based on the downhole monitoring system, averages the relevant real-time viscosity and inputs it into the blockage detection model, and judges whether the submersible oil pump is blocked based on the result and performs relevant processing, including: obtaining the current viscosity and the viscosity before a first preset time, taking the average of the two viscosities as the input of the blockage detection model, then changing the viscosity at the current time to the viscosity after taking the average of the two viscosities, and finally obtaining the result of the blockage detection model.
10. The intelligent control system for a submersible electric pump according to claim 9, characterized in that: After that, the viscosity at the current time is also changed to the viscosity after taking the average of the two viscosities, and finally the result of the blockage detection model is obtained, including: if the pump pressure difference obtained based on the viscosity after taking the average of the two viscosities is greater than or equal to the first correct threshold, it is considered that blockage has occurred, otherwise it is considered that no blockage has occurred.
Citation Information
Patent Citations
Centrifugal pump fault determination method and equipment
CN117846979A
Electrical submersible pump control
WO2020097301A1