Continuous sand-removing and sticking-preventing control system of sand-preventing valve for oil extraction of oil well

Abstract

The invention relates to the technical field of sand prevention valves, in particular to a continuous sand removal and drill jamming prevention control system of a sand prevention valve for oil extraction of an oil well. The method comprises the following steps: a data acquisition unit acquires and normalizes parameters such as sand setting thickness, sand concentration, torque and hanging load; the risk prediction unit adopts a gradient lifting tree model to calculate the drilling tool jamming risk probability; the anti-jamming intelligent control unit generates a target opening instruction of the sand control valve through a fuzzy logic control algorithm; the desanding optimization control unit generates a target rotating speed instruction of the spiral separator through a PID control algorithm; and the execution unit drives the sand control valve and the spiral separator to execute corresponding actions. Risk prediction is carried out by fusing multi-dimensional features through a gradient boosting tree model, and fuzzy logic and PID cooperative control are combined, so that the problem that continuous self-adaptive sand removal and anti-jamming control are difficult to realize due to dependence on a single threshold value or artificial experience in the prior art is solved.

Description

Technical Field

[0001] This invention relates to the field of sand control valve technology, and more specifically, to a continuous sand removal and anti-stuck drill control system for oil well production using a sand control valve. Background Technology

[0002] During oil well production, formation sand particles enter the wellbore along with crude oil and are prone to deposit in valves, pipelines and other parts, leading to increased sand thickness and sand concentration, which in turn causes serious malfunctions such as valve jamming and drill string jamming.

[0003] Existing sand control systems mostly rely on manual experience or single parameter threshold control, lacking the ability to intelligently predict and coordinate the risks of multi-parameter coupling, making it difficult to achieve continuous and adaptive sand removal and anti-sticking control. Therefore, a continuous sand removal and anti-sticking control system for oil well sand control valves is designed. Summary of the Invention

[0004] The purpose of this invention is to provide a continuous sand removal and anti-sticking control system for oil well sand control valves, in order to solve the problems mentioned in the background art, that existing sand control systems mostly rely on manual experience or single parameter threshold control, lack the ability to intelligently predict and coordinate the risks of multi-parameter coupling, and are difficult to achieve continuous and adaptive sand removal and anti-sticking control.

[0005] To achieve the above objectives, the present invention aims to provide a continuous sand removal and anti-sticking control system for oil well sand control valves, including... The data acquisition unit collects parameters such as sediment thickness, sediment concentration, torque, and suspended weight through a sensor array, and normalizes the data to obtain valve parameter data. The risk prediction unit calculates the probability of stuck drill bit risk based on valve parameter data using a gradient boosting tree model. The anti-jamming intelligent control unit generates a target opening command for the anti-sand valve based on the probability of the drill jamming risk and the sand concentration using a fuzzy logic control algorithm. The sand removal optimization control unit uses the real-time collected sand concentration as a process variable and generates the target rotation speed command of the spiral separator through a PID control algorithm. An execution unit is used to receive the target opening degree command and the target speed command, and drive the sand valve and the spiral separator to perform corresponding actions.

[0006] As a further improvement to this technical solution, the data acquisition unit includes a data acquisition module and a normalization processing module; The data acquisition module uses a sensor group consisting of a sediment thickness sensor, a sand concentration sensor, a torque sensor, and a suspended weight sensor to collect sediment thickness, sand concentration, torque, and suspended weight parameters in real time. The normalization processing module uses the min-max normalization algorithm to map the collected data to the [0,1] interval and outputs the normalized valve parameter data.

[0007] As a further improvement to this technical solution, the risk prediction unit includes a model building module and a risk prediction module; The model building module constructs historical feature vectors based on historical sediment thickness, sediment concentration, torque, suspended weight and their rate of change as input features, and uses whether a stuck drill event occurred at the same historical moment as a label to form a training dataset for building a gradient boosting tree model. The risk prediction module constructs a real-time feature vector based on valve parameter data as input features, and uses a gradient boosting tree model to output the probability of stuck drill bit risk.

[0008] As a further improvement to this technical solution, the specific steps involved in constructing the gradient boosting tree model in the model construction module are as follows: Time series data of core parameters are extracted from historical databases, and the rate of change characteristics of each core parameter are calculated. The core parameters and their rate of change characteristics are then concatenated column by column to form a historical feature vector. At the same time, historical feature vectors Each data entry is labeled to indicate whether a stuck drill event has occurred, thus obtaining event tags. Among them, the core parameters include the thickness of the sediment. Sand concentration Torque and suspension weight ; Set the initial number of shallow decision trees to be... The logarithmic loss function is used as the objective for optimizing the gradient boosting tree model; the complexity of a single shallow decision tree is limited to not exceeding the maximum depth of the tree. At the same time, each decision conclusion is restricted to be based on a sample size exceeding the minimum sample size of the leaf node. ; exist In each iteration, the log loss function is calculated based on the historical feature vector and event label, and the negative gradient of the current gradient boosting tree model with respect to each sample is calculated based on the log loss function. A new shallow decision tree is trained based on this negative gradient to learn the correction direction. For each leaf node generated by the new shallow decision tree, the optimal output value that can minimize the overall loss is calculated. Calculate the loss on the validation set after each iteration. If continuous If the wheel's losses no longer decrease, training should be stopped prematurely. The trained shallow decision trees are used with a learning rate Combined into the final gradient boosting tree model .

[0009] As a further improvement to this technical solution, the risk prediction module involves the following specific steps: constructing a real-time feature vector based on valve parameter data as input features, and using a gradient boosting tree model to output the probability of stuck drill bit risk. Based on the valve parameter data at the current moment and the previous moment, the normalized parameter value change rate is calculated; among which, the parameter value change rate includes the change rate of sediment thickness, the change rate of sediment concentration, the change rate of torque, and the change rate of suspended weight; Valve parameter data and their rate of change are concatenated sequentially to construct a real-time feature vector. ; Real-time feature vectors Input is a gradient boosting tree model In the middle, output the original predicted value ; The original predicted value is obtained by using the sigmoid function. Conversion to stuck drill risk probability .

[0010] As a further improvement to this technical solution, the specific steps involved in generating the target opening degree command for the sand-proof valve through the fuzzy logic control algorithm in the anti-jamming intelligent control unit are as follows: The probability of a stuck drill and real-time sand concentration Each value is mapped to a fuzzy linguistic value through a preset triangular membership function; among them, the fuzzy linguistic value for the probability of stuck drill bit is low, medium and high, and the fuzzy linguistic value for sand concentration is low, normal and high. Based on a predefined fuzzy rule base, the Mamdani inference method is used to obtain the fuzzy linguistic value of the aperture change from the fuzzy linguistic value of the stuck drill risk probability and the fuzzy linguistic value of the sand concentration. The centroid method is used to defuzzify the fuzzy linguistic values ​​of aperture change, and the precise aperture change is calculated. ; Current opening Change in opening Add them together to obtain the target opening command. And the target opening command. Amplitude limiting measures are implemented to ensure that the amplitude does not exceed the upper or lower limits of the effective opening of the sand control valve.

[0011] As a further improvement to this technical solution, the sand removal optimization control unit includes a concentration deviation calculation module and a PID parameter adjustment module; Among them, the concentration deviation calculation module compares the real-time sand concentration with the set target concentration to calculate the concentration deviation value and the deviation change rate. The PID parameter adjustment module dynamically adjusts the target rotation speed command of the spiral separator based on the concentration deviation value and the rate of change of deviation, using a PID control algorithm.

[0012] As a further improvement to this technical solution, the specific steps involved in calculating the concentration deviation value and the rate of change of deviation in the concentration deviation calculation module are as follows: The concentration deviation at the current moment is obtained by subtracting the real-time sand concentration from the preset target concentration value. ; The rate of change of concentration deviation is obtained by differential calculation of the concentration deviation values ​​within two consecutive sampling periods.

[0013] As a further improvement to this technical solution, the specific steps involved in dynamically adjusting the target speed command of the spiral separator based on the PID control algorithm in the PID parameter adjustment module are as follows: Receive real-time concentration deviation values ​​and deviation change rates, and adjust them according to a preset proportional coefficient. Multiplying by the real-time concentration deviation yields the proportional term. ; The historical concentration deviation values ​​are summed and multiplied by a preset integral coefficient. Obtain the integral term ; Based on preset differential coefficients The differential term is obtained by multiplying the concentration deviation rate by the concentration deviation rate. ; Adding the proportional, integral, and derivative terms yields the output control quantity of the PID controller at the current moment. ; Output control quantity Current speed of the screw separator Add them together to get the target speed command. Subsequently, a target speed command was issued. A speed limit is applied to ensure that the speed does not exceed the upper or lower limits of the allowable safe speed of the spiral separator.

[0014] As a further improvement to this technical solution, the specific steps involved in driving the sand control valve and the spiral separator to perform corresponding actions in the execution unit are as follows: Receive the target opening command Converted into control voltage signal for sand control valve actuator ; The received target speed command Converted into control current signal for the screw separator drive motor ; Control voltage signal The electric actuator output to the sand control valve drives the valve to adjust to the target opening degree; the control current signal is then transmitted. The variable frequency drive outputs to the screw separator, driving the motor to reach the target speed; Real-time reading of the actual opening feedback value of the sand control valve And the actual speed feedback value of the screw separator It is then compared with the target opening command and the target speed command. If any deviation exceeds the maximum allowable deviation, the command retransmission mechanism is triggered. When the actual opening degree matches the target opening degree and the actual speed matches the target speed, the judgment action is completed, the system enters steady-state operation mode, continuously monitors and waits for new control commands.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. In this continuous sand removal and anti-sticking control system for oil well production, the risk prediction unit adopts a gradient lifting tree model, integrating multi-dimensional features such as sand thickness, sand concentration, torque, suspended weight, and their rate of change for joint prediction, enabling accurate calculation of the probability of sticking. Compared to traditional methods relying on a single threshold or human experience, this system achieves early, quantitative warning of sticking risk, significantly improving the safety and reliability of downhole operations.

[0016] 2. In this continuous sand removal and anti-sticking control system for oil well production, the anti-sticking intelligent control unit and the sand removal optimization control unit work together, combining fuzzy logic and PID control algorithms, to dynamically coordinate the opening degree of the sand removal valve and the rotation speed of the spiral separator based on real-time sticking risk and sand concentration. Compared with existing independently controlled sand removal systems, this system achieves a dynamic balance between sand removal efficiency and anti-sticking control, effectively suppressing sand deposition during continuous oil production, reducing the sticking rate, and improving oil well production efficiency and continuity. Attached Figure Description

[0017] Figure 1 This is an overall flowchart of the present invention; The meanings of the labels in the diagram are as follows: 1. Data acquisition unit; 11. Data acquisition module; 12. Normalization processing module; 2. Risk prediction unit; 21. Model building module; 22. Risk prediction module; 3. Anti-jamming intelligent control unit; 4. Sand removal optimization control unit; 41. Concentration deviation calculation module; 42. PID parameter adjustment module; 5. Execution unit. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example Please see Figure 1 As shown, a continuous sand removal and anti-sticking control system for oil well sand control valves is provided, including... Data acquisition unit 1 collects parameters such as sediment thickness, sediment concentration, torque, and suspended weight through a sensor group, and normalizes the data to obtain valve parameter data. In this embodiment, the data acquisition unit 1 includes a data acquisition module 11 and a normalization processing module 12; Among them, the data acquisition module 11 uses a sensor group consisting of a sediment thickness sensor, a sand concentration sensor, a torque sensor, and a suspended weight sensor to collect sediment thickness, sand concentration, torque, and suspended weight parameters in real time. The normalization processing module 12 uses the min-max normalization algorithm to map the collected data to the [0,1] interval and outputs the normalized valve parameter data.

[0020] The oil well production sand control system for continuous sand removal and anti-sticking control also includes a risk prediction unit 2, which calculates the probability of sticking risk based on valve parameter data using a gradient boosting tree model. In this embodiment, the risk prediction unit 2 includes a model building module 21 and a risk prediction module 22; Among them, the model building module 21 constructs historical feature vectors based on historical sediment thickness, sediment concentration, torque, suspended weight and their rate of change as input features, and uses whether a stuck drill event occurs at the same historical moment as a label to form training data for constructing a gradient boosting tree model. Specifically, time-series data of core parameters are extracted from historical databases, and the rate of change characteristics of each core parameter are calculated. The core parameters and their rate of change characteristics are then concatenated column-wise to form a historical feature vector. At the same time, historical feature vectors Each data entry is labeled to indicate whether a stuck drill event has occurred, thus obtaining event tags. Among them, the core parameters include the thickness of the sediment. Sand concentration Torque and suspension weight ; Historical feature vectors: ; in, For the thickness of the sediment, The sand concentration, For torque, For suspension weight, The rate of change of the sediment thickness. The rate of change of sand concentration, The rate of change of torque, The rate of change of suspended weight. For historical feature vectors; This is the sampling time interval, typically 1 second; Event Tags (1 indicates that the drill bit is stuck, and 0 indicates that the drill bit is not stuck); Set the initial number of shallow decision trees to be... The logarithmic loss function is used as the objective for optimizing the gradient boosting tree model; the complexity of a single shallow decision tree is limited to not exceeding the maximum depth of the tree. At the same time, each decision conclusion is restricted to be based on a sample size exceeding the minimum sample size of the leaf node. ; Specifically, This represents the initial number of shallow decision trees (i.e., weak learners). If the value is too small, the model will underfit (failing to capture complex patterns), while if it is too large, it may overfit (remembering noise in the training data). In the oil well scenario, it is usually set to 50~200 (which needs to be dynamically adjusted in conjunction with the subsequent early stop mechanism). This represents the maximum depth of the tree, typically set to 3-5, to avoid overfitting to local data due to excessively deep trees. This is the minimum number of samples for a leaf node, typically between 5 and 10, to avoid decision bias caused by a small amount of abnormal data (such as instantaneous sensor errors). Logarithmic loss function: ; in, For real tags (i.e. event tags) ), This represents the current prediction value of the gradient boosting tree model. The sigmoid function (which converts the predicted value) Mapping to the [0,1] interval to achieve probability transformation); exist In each iteration, the log loss function is calculated based on the historical feature vector and event label, and the negative gradient of the current gradient boosting tree model with respect to each sample is calculated based on the log loss function. A new shallow decision tree is trained based on this negative gradient to learn the correction direction. For each leaf node generated by the new shallow decision tree, the optimal output value that can minimize the overall loss is calculated. For each iteration Calculate the negative gradient: ; in, Indicates the current iteration number (from 1 to ...). ), For sample index; For the first Round model for the first The negative gradient of each sample. The first of the historical feature vectors One sample, For the first The true label of each sample For the model number Predicted values ​​for each sample; This is the output of the previous model. Calculate the loss on the validation set after each iteration. If continuous If the wheel's losses no longer decrease, training should be stopped prematurely. The trained shallow decision trees are used with a learning rate Combined into the final gradient boosting tree model ; Among them, the gradient boosting tree model can receive real-time normalized feature data and output a continuous value between 0 and 1 to represent the probability of stuck pipe in the current downhole working condition. The learning rate, typically set to 0.01 to 0.1, is used to control the contribution weight of each shallow decision tree, essentially limiting the optimization range of a single iteration. The risk prediction module 22 constructs a real-time feature vector based on valve parameter data as input features, and uses a gradient boosting tree model to output the probability of stuck drill bit risk. Specifically, based on the valve parameter data at the current moment and the previous moment, the normalized parameter value change rate is calculated; among which, the parameter value change rate includes the change rate of sediment thickness, the change rate of sand concentration, the change rate of torque, and the change rate of suspended weight; Valve parameter data and their rate of change are concatenated sequentially to construct a real-time feature vector. ; ; in, For real-time sediment thickness, For real-time sand concentration, For real-time torque, For real-time suspension; This represents the real-time rate of change in sediment thickness. This represents the real-time rate of change in sand concentration. This represents the real-time torque change rate. This represents the real-time rate of change of suspended weight. The sampling time interval; For real-time feature vectors; Real-time feature vectors Input is a gradient boosting tree model In the middle, output the original predicted value ; The original predicted value is obtained by using the sigmoid function. Conversion to stuck drill risk probability ; The formula is: ; in, This represents the probability of a stuck drill event occurring at the current moment, with a value range of [0,1]. This is the sigmoid function.

[0021] The oil well production sand prevention valve continuous sand removal and anti-sticking control system also includes an anti-sticking intelligent control unit 3, which generates the target opening degree command of the sand prevention valve according to the sticking risk probability and sand concentration through a fuzzy logic control algorithm. In this embodiment, the probability of drill bit jamming is... and real-time sand concentration Each value is mapped to a fuzzy linguistic value through a preset triangular membership function; among them, the fuzzy linguistic value for the probability of stuck drill bit is low, medium and high, and the fuzzy linguistic value for sand concentration is low, normal and high. Each fuzzy linguistic value corresponds to a triangular membership function, which is used to convert continuous numerical values ​​into membership degrees in a fuzzy set; Based on a predefined fuzzy rule base, the Mamdani inference method is used to obtain the fuzzy linguistic value of the aperture change from the fuzzy linguistic value of the stuck drill risk probability and the fuzzy linguistic value of the sand concentration. The rule format of the fuzzy rule base is as follows: IF is AND is THEN Opening change is ; in, This is a fuzzy linguistic value representing the probability of a stuck drill bit. This is a fuzzy linguistic value for sand concentration. For fuzzy linguistic values ​​representing changes in opening degree; Example of a rule: IF is high AND The opening degree change is at its maximum. The centroid method is used to defuzzify the linguistic values ​​of opening changes, and the precise amount of opening change is calculated. ; The formula is: ; in, For the first Membership degree corresponding to a fuzzy linguistic value with varying openness. For the first Typical output values ​​corresponding to fuzzy linguistic values ​​with varying degrees of openness; This represents the change in opening degree; Current opening Change in opening Add them together to obtain the target opening command. And the target opening command. Amplitude limiting measures are implemented to ensure that the amplitude does not exceed the upper or lower limits of the effective opening of the sand control valve.

[0022] The continuous sand removal and anti-sticking control system for oil well sand control valves also includes a sand removal optimization control unit 4. The sand removal optimization control unit 4 uses the real-time collected sand concentration as a process variable and generates the target rotation speed command of the spiral separator through a PID control algorithm. In this embodiment, the sand removal optimization control unit 4 includes a concentration deviation calculation module 41 and a PID parameter adjustment module 42; Among them, the concentration deviation calculation module 41 compares the real-time sand concentration with the set target concentration and calculates the concentration deviation value and the deviation change rate. Specifically, the concentration deviation value at the current moment is obtained by subtracting the real-time sand concentration from the preset target concentration value. ; ; in, The target concentration value is preset by the system, representing the maximum allowable sand concentration threshold. This is the concentration deviation value, which reflects the degree of deviation between the current sand concentration and the target concentration value. A positive deviation indicates that the concentration is too high (sand removal needs to be strengthened), while a negative deviation indicates that the concentration is too low (sand removal can be appropriately reduced). The rate of change of concentration deviation is obtained by calculating the difference between the concentration deviation values ​​in two consecutive sampling periods; ; in, This is the deviation value at the previous sampling time; The concentration deviation rate reflects the trend of sand concentration deviation. A positive value indicates that the deviation is widening (the sand concentration is deviating from the target value at an accelerated rate), while a negative value indicates that the deviation is narrowing (the sand concentration is approaching the target value). The PID parameter adjustment module 42 dynamically adjusts the target speed command of the spiral separator based on the concentration deviation value and the deviation change rate, according to the PID control algorithm. The proportional term responds quickly to the deviation, the integral term eliminates steady-state error, and the derivative term suppresses overshoot, thereby achieving stable control of the sand concentration. Specifically, it receives real-time concentration deviation values ​​and deviation change rates, and uses a preset proportional coefficient. Multiplying by the real-time concentration deviation yields the proportional term. ; ; in, This is a proportionality coefficient, dynamically set according to the sand concentration of the oil well, and is usually taken as 0.5 to 5.0 (the specific value needs to be adjusted on-site in conjunction with the sensitivity of the spiral separator; it can be appropriately reduced in scenarios with large fluctuations in sand concentration). This is a proportional term used for rapid response to current concentration deviations; The historical concentration deviation values ​​are summed and multiplied by a preset integral coefficient. Obtain the integral term ; ; in, This is the integration coefficient, typically taken as 0.01–0.5 (corresponding to the sampling time interval). (For related applications, the sampling time interval can be appropriately increased when it is large to avoid the integral effect being too weak.) This is the integral term, used to eliminate steady-state error; Based on preset differential coefficients The differential term is obtained by multiplying the concentration deviation rate by the concentration deviation rate. ; ; in, This is the differential coefficient, usually taken as 0.1 to 2.0 (it can be appropriately increased in scenarios with frequent fluctuations in sand concentration to enhance anti-interference ability). This is the differential term, used to predict the trend of deviation changes and suppress system overshoot; Adding the proportional, integral, and derivative terms yields the output control quantity of the PID controller at the current moment. ; ; in, The output control quantity represents the amount of adjustment required to the speed of the spiral separator to eliminate concentration deviation; Output control quantity Current speed of the screw separator Add them together to get the target speed command. Subsequently, a target speed command was issued. A speed limit is applied to ensure that the speed does not exceed the upper or lower limits of the allowable safe speed of the spiral separator.

[0023] The continuous sand removal and anti-stuck drill control system for sand control valves used in oil well production also includes an execution unit 5, which is used to receive the target opening degree command and the target rotation speed command, and drive the sand control valve and the spiral separator to perform corresponding actions. In this embodiment, the received target opening command will be... Converted into control voltage signal for sand control valve actuator ; The received target speed command Converted into control current signal for the screw separator drive motor ; Control voltage signal The electric actuator output to the sand control valve drives the valve to adjust to the target opening degree; the control current signal is then transmitted. The variable frequency drive outputs to the screw separator, driving the motor to reach the target speed; Real-time reading of the actual opening feedback value of the sand control valve And the actual speed feedback value of the screw separator It is then compared with the target opening command and the target speed command. If any deviation exceeds the maximum allowable deviation, the command retransmission mechanism is triggered. The instruction retransmission mechanism is triggered if any of the following conditions are met: ; ; in, This represents the maximum permissible deviation in opening. This represents the maximum permissible deviation in rotational speed. When the actual opening degree matches the target opening degree and the actual speed matches the target speed, the judgment action is completed, the system enters steady-state operation mode, continuously monitors and waits for new control commands.

[0024] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. A continuous sand removal and anti-sticking control system for sand control valves used in oil well production, characterized in that: include The data acquisition unit (1) collects the sediment thickness, sand concentration, torque and suspended weight parameters through the sensor group, and normalizes the data to obtain valve parameter data. Risk prediction unit (2), which calculates the probability of stuck drill bit risk based on valve parameter data using a gradient boosting tree model; Anti-jamming intelligent control unit (3), the anti-jamming intelligent control unit (3) generates the target opening degree command of the anti-sand valve through fuzzy logic control algorithm according to the jamming risk probability and sand concentration; The sand removal optimization control unit (4) uses the real-time collected sand concentration as a process variable and generates the target speed command of the spiral separator through a PID control algorithm. The execution unit (5) is used to receive the target opening command and the target speed command, and drive the sand valve and the spiral separator to perform corresponding actions.

2. The continuous sand removal and anti-sticking control system for oil well sand control valves according to claim 1, characterized in that: The data acquisition unit (1) includes a data acquisition module (11) and a normalization processing module (12). Among them, the data acquisition module (11) collects the sediment thickness, sand concentration, torque and suspension parameters in real time through a sensor group consisting of a sediment thickness sensor, a sand concentration sensor, a torque sensor and a suspension weight sensor. The normalization processing module (12) uses the min-max normalization algorithm to map the collected data to the [0,1] interval and outputs the normalized valve parameter data.

3. The continuous sand removal and anti-sticking control system for oil well sand control valves according to claim 1, characterized in that: The risk prediction unit (2) includes a model building module (21) and a risk prediction module (22). Among them, the model building module (21) constructs a historical feature vector based on the historical sediment thickness, sand concentration, torque, suspended weight and its rate of change as input features, and uses whether a stuck drill event occurs at the same historical moment as a label to form a training dataset for constructing a gradient boosting tree model. The risk prediction module (22) constructs a real-time feature vector based on valve parameter data as input features and uses a gradient boosting tree model to output the probability of stuck drill.

4. The continuous sand removal and anti-sticking control system for oil well sand control valves according to claim 3, characterized in that: The specific steps involved in constructing the gradient boosting tree model in the model construction module (21) are as follows: Time series data of core parameters are extracted from historical databases, and the rate of change characteristics of each core parameter are calculated. The core parameters and their rate of change characteristics are then concatenated column by column to form a historical feature vector. At the same time, historical feature vectors Each data entry is labeled to indicate whether a stuck drill event has occurred, thus obtaining event tags. Among them, the core parameters include the thickness of the sediment. Sand concentration Torque and suspension weight ; Set the initial number of shallow decision trees to be... The logarithmic loss function is used as the objective for optimizing the gradient boosting tree model; the complexity of a single shallow decision tree is limited to not exceeding the maximum depth of the tree. At the same time, each decision conclusion is restricted to be based on a sample size exceeding the minimum sample size of the leaf node. ; exist In each iteration, the log loss function is calculated based on the historical feature vector and event label, and the negative gradient of the current gradient boosting tree model with respect to each sample is calculated based on the log loss function. A new shallow decision tree is trained based on this negative gradient to learn the correction direction. For each leaf node generated by the new shallow decision tree, the optimal output value that can minimize the overall loss is calculated. Calculate the loss on the validation set after each iteration. If continuous If the wheel's losses no longer decrease, training should be stopped prematurely. The trained shallow decision trees are used with a learning rate Combined into the final gradient boosting tree model .

5. The continuous sand removal and anti-sticking control system for an oil well sand control valve according to claim 3, characterized in that: In the risk prediction module (22), the specific steps involved in constructing a real-time feature vector based on valve parameter data as input features and using a gradient boosting tree model to output the probability of stuck drill bit risk are as follows: Based on the valve parameter data at the current moment and the previous moment, the normalized parameter value change rate is calculated; among which, the parameter value change rate includes the change rate of sediment thickness, the change rate of sediment concentration, the change rate of torque, and the change rate of suspended weight; Valve parameter data and their rate of change are concatenated sequentially to construct a real-time feature vector. ; Real-time feature vectors Input is a gradient boosting tree model In the middle, output the original predicted value ; The original predicted value is obtained by using the sigmoid function. Conversion to stuck drill risk probability .

6. The continuous sand removal and anti-sticking control system for an oil well sand control valve according to claim 1, characterized in that: In the anti-jamming intelligent control unit (3), the specific steps involved in generating the target opening degree command of the sand-proof valve through the fuzzy logic control algorithm are as follows: The probability of a stuck drill and real-time sand concentration Each value is mapped to a fuzzy linguistic value through a preset triangular membership function; among them, the fuzzy linguistic value for the probability of stuck drill bit is low, medium and high, and the fuzzy linguistic value for sand concentration is low, normal and high. Based on a predefined fuzzy rule base, the Mamdani inference method is used to obtain the fuzzy linguistic value of the aperture change from the fuzzy linguistic value of the stuck drill risk probability and the fuzzy linguistic value of the sand concentration. The centroid method is used to defuzzify the fuzzy linguistic values ​​of aperture change, and the precise aperture change is calculated. ; Current opening Change in opening Add them together to obtain the target opening command. And the target opening command. Amplitude limiting measures are implemented to ensure that the amplitude does not exceed the upper or lower limits of the effective opening of the sand control valve.

7. The continuous sand removal and anti-sticking control system for an oil well sand control valve according to claim 1, characterized in that: The sand removal optimization control unit (4) includes a concentration deviation calculation module (41) and a PID parameter adjustment module (42). Among them, the concentration deviation calculation module (41) compares the real-time sand concentration with the set target concentration and calculates the concentration deviation value and the deviation change rate. The PID parameter adjustment module (42) dynamically adjusts the target speed command of the spiral separator based on the concentration deviation value and the deviation change rate using the PID control algorithm.

8. The continuous sand removal and anti-sticking control system for an oil well sand control valve according to claim 7, characterized in that: The specific steps involved in calculating the concentration deviation value and the rate of change of deviation in the concentration deviation calculation module (41) are as follows: The concentration deviation at the current moment is obtained by subtracting the real-time sand concentration from the preset target concentration value. ; The rate of change of concentration deviation is obtained by differential calculation of the concentration deviation values ​​within two consecutive sampling periods.

9. The continuous sand removal and anti-sticking control system for an oil well sand control valve according to claim 1, characterized in that: In the PID parameter adjustment module (42), the specific steps involved in dynamically adjusting the target speed command of the spiral separator based on the PID control algorithm are as follows: Receive real-time concentration deviation values ​​and deviation change rates, and adjust them according to a preset proportional coefficient. Multiplying by the real-time concentration deviation yields the proportional term. ; The historical concentration deviation values ​​are summed and multiplied by a preset integral coefficient. Obtain the integral term ; Based on preset differential coefficients The differential term is obtained by multiplying the concentration deviation rate by the concentration deviation rate. ; Adding the proportional, integral, and derivative terms yields the output control quantity of the PID controller at the current moment. ; Output control quantity Current speed of the screw separator Add them together to get the target speed command. Subsequently, a target speed command was issued. A speed limit is applied to ensure that the speed does not exceed the upper or lower limits of the allowable safe speed of the spiral separator.

10. A continuous sand removal and anti-sticking control system for an oil well sand control valve according to claim 1, characterized in that: The specific steps involved in driving the sand control valve and the spiral separator to perform corresponding actions in the execution unit (5) are as follows: Receive the target opening command Converted into control voltage signal for sand control valve actuator ; The received target speed command Converted into control current signal for the screw separator drive motor ; Control voltage signal The electric actuator output to the sand control valve drives the valve to adjust to the target opening degree; Control current signal The variable frequency drive outputs to the screw separator, driving the motor to reach the target speed; Real-time reading of the actual opening feedback value of the sand control valve And the actual speed feedback value of the screw separator It is then compared with the target opening command and the target speed command. If any deviation exceeds the maximum allowable deviation, the command retransmission mechanism is triggered. When the actual opening degree matches the target opening degree and the actual speed matches the target speed, the judgment action is completed, the system enters steady-state operation mode, continuously monitors and waits for new control commands.

Images