Adjusting valve opening / flow linear correction system and adjusting valve control method

Abstract

The invention relates to a regulating valve opening / flow linear correction system and a control method, and the system comprises a flow deviation unit which is used for calculating a real-time deviation value between an actual control flow and a dynamic ideal flow of a regulating valve and comprises a flow prediction module based on an LSTM network, and the flow prediction module outputs a working condition self-adaptive ideal flow reference value; the manual / automatic undisturbed switching unit is used for tracking an ideal flow reference value in a manual mode, keeping a deviation value at the switching moment in an automatic mode and pre-judging the switching opportunity; the deviation correction unit is used for dynamically generating a nonlinear compensation amount by adopting Gaussian process regression, and solving an optimal opening correction value through an online convex optimization algorithm; the correction condition limiting unit is internally provided with a digital twinborn simulation module and a safety boundary prediction model, virtual verification is carried out before opening correction is executed, and the maximum opening limit value is dynamically adjusted; and the digital valve fingerprint database is used for storing valve mechanical wear characteristics, seasonal characteristic drift data and a medium influence matrix.

Description

Technical Field

[0001] The present invention relates to a regulating valve opening / flow linear correction system and a regulating valve control method, belonging to the technical field of industrial control automation. Background Art

[0002] With the continuous development of industrial automation control technology, as a key actuator in fluid control systems, the control accuracy and response characteristics of regulating valves directly affect the operating efficiency and product quality of the entire system. Ideally, the opening of the regulating valve and the flow rate should have a linear relationship. However, in actual industrial applications, due to factors such as mechanical wear, medium property changes, and pressure fluctuations, the opening-flow characteristics of regulating valves often show a non-linear relationship, resulting in a decrease in control accuracy, an increase in energy consumption, and fluctuations in product quality.

[0003] Currently, the commonly used regulating valve flow control methods in the industrial field mainly include traditional methods based on PID control and methods based on special valve structure design. Although the traditional PID control method is simple to implement, it is difficult to adapt to dynamic changes under complex working conditions. For example, a natural gas station intelligent subtransmission method based on the Internet of Things disclosed in CN119395982A uses PID to control the valve opening, combines the historical operation data and current state of the valve for adaptive dead zone compensation, eliminates valve response hysteresis, and improves the regulation accuracy. However, when facing complex situations such as medium property changes and valve wear, this method is still difficult to maintain high-precision flow control.

[0004] To improve the control accuracy of regulating valves, CN114637343A discloses a uranium hexafluoride conversion material flow control system and method. This method uses a feedback correction module to calculate the steady-state opening according to the current actual flow rate of the material using a valve fitting equation, and calculates the corrected flow rate set value according to the pressure difference change rate at both ends of the valve and a non-linear equation. Although this method can improve the control accuracy under specific working conditions, its fitting equation is established based on static working conditions and is difficult to adapt to dynamic working condition changes.

[0005] Based on the above disclosed solutions, the main problems in regulating valve flow control are as follows:

[0006] 1. Traditional linearization methods mainly rely on the mechanical characteristics design of the valve (such as equal percentage valve cores). However, under actual working conditions, due to factors such as medium property changes, pressure fluctuations, and valve wear, the characteristic curve will drift, making it difficult to guarantee control accuracy;

[0007] 2. Existing adaptive control methods are mostly based on static models or simplified models, and it is difficult to accurately describe the dynamic characteristics of valves under complex working conditions. Especially under the conditions of large medium property changes and pressure fluctuations, the control accuracy significantly decreases;

[0008] 3. Although high-precision special valves can solve the linearization problem to a certain extent, their price is 3-5 times that of ordinary valves, increasing the equipment investment cost of enterprises, and still cannot completely solve the problem of characteristic drift during long-term use;

[0009] 4. The existing technology lacks an effective identification and compensation mechanism for the mechanical wear characteristics of valves during long-term use, resulting in an annual average energy consumption increase of 7.2% in some industrial installations (such as ethylene plants) due to valve nonlinearity;

[0010] 5. The disturbance problem during the manual / automatic mode switching process has not been effectively solved, affecting the system stability and control accuracy. Summary of the Invention

[0011] In order to solve the problems existing in the above-mentioned prior art, the present invention proposes a regulating valve opening / flow linear correction system and a regulating valve control method, which solve the dynamic nonlinear problems caused by mechanical wear and working condition changes of existing regulating valves, and achieve the technical effects of an opening-flow response linearity error ≤ ±1% FS, automatic adaptation to medium / pressure changes, and enabling ordinary valves to reach the control accuracy of special valves.

[0012] The technical solution of the present invention is as follows:

[0013] On the one hand, the present invention provides a regulating valve opening / flow linear correction system, including:

[0014] Flow deviation unit: used to calculate the real-time deviation value between the actual control flow of the regulating valve and the dynamic ideal flow, including a flow prediction module based on the LSTM network. The flow prediction module receives valve opening, medium pressure, temperature, and historical flow data, and outputs an ideal flow reference value adapted to the working conditions;

[0015] Manual / automatic seamless switching unit: equipped with a switching strategy module optimized by reinforcement learning, tracking the ideal flow reference value output by the flow deviation unit in the manual mode, maintaining the deviation value at the moment of switching in the automatic mode, and using an operation mode predictor integrating the XGBoost model to predict the switching timing;

[0016] Deviation correction unit: dynamically generates a nonlinear compensation amount by Gaussian process regression, solves the optimal opening correction value through an online convex optimization algorithm, and outputs a corrected valve opening command;

[0017] Correction condition limiting unit: built-in digital twin simulation module and safety boundary prediction model, performs virtual verification before executing the opening correction, and dynamically adjusts the minimum opening limit value B and the maximum opening limit value C;

[0018] Digital valve fingerprint database: Stores valve mechanical wear characteristics, seasonal characteristic drift data, and medium influence matrix, providing device-specific parameters for each unit.

[0019] As a preferred embodiment, the flow deviation unit further includes:

[0020] A Kalman filter fusion module for integrating measurement data from electromagnetic flowmeters, ultrasonic flowmeters, and multi-parameter transmitters;

[0021] An online learning engine for automatically collecting data when the system is stable and triggering model fine-tuning at a set time.

[0022] As a preferred embodiment, the reinforcement learning strategy in the manual / automatic seamless switching unit adopts the DQN algorithm, and its reward function R is defined as:

[0023]

[0024] Where α and β are respectively the set first weight coefficient and the first weight coefficient, e(t) is the control error at time t, t0 and t1 are respectively the two start times and end times of the sampling time period, and N switch Is the number of switchings.

[0025] As a preferred embodiment, the Gaussian process regression model of the deviation correction unit uses a squared exponential kernel function:

[0026]

[0027] Where x is the input vector [opening, pressure, temperature], and σ f 、σ n And l are all hyperparameters.

[0028] As a preferred embodiment, the safety boundary prediction model of the correction condition restriction unit is a Transformer architecture, and the input includes:

[0029] The cumulative valve operation times, the medium particulate matter concentration, and the vibration spectrum characteristics in the most recent 30 days. The output is the dynamic safe opening interval [B', C'], where B' and C' are respectively the corrected minimum opening limit value and the maximum opening limit value.

[0030] As a preferred embodiment, it further includes a human-machine collaborative decision-making module, including:

[0031] A fine-tuned industrial large language model for generating natural language maintenance suggestions;

[0032] A knowledge graph reasoning engine for associating valve fault characteristics with the maintenance record library.

[0033] On the other hand, the present invention also provides a control method for a regulating valve, which is applied to the regulating valve opening / flow linear correction system described in any embodiment of the present invention, and includes the steps:

[0034] S1. Predict the ideal flow value under the current working condition through the LSTM network;

[0035] S2. Compare the actual flow with the ideal flow to generate a deviation signal;

[0036] S3. Calculate the opening compensation amount using Gaussian process regression and output it after being verified by digital twin;

[0037] S4. When the wear characteristics of the valve guide sleeve are detected, automatically relax the minimum opening limit value B.

[0038] As a preferred embodiment, the digital twin verification in step S3 includes:

[0039] Based on CFD flow field simulation, predict the corrected flow characteristics;

[0040] The stress analysis module checks whether the force on the valve core exceeds the limit.

[0041] On the other hand, the present invention also proposes an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements the control method for the regulating valve described in any embodiment of the present invention.

[0042] On the other hand, the present invention also proposes a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the control method for the regulating valve described in any embodiment of the present invention.

[0043] The beneficial effects of the present invention are as follows:

[0044] 1. The control accuracy is significantly improved: Through the combined application of the LSTM network and Gaussian process regression, the linearity error of the opening-flow response is realized to be ≤±1% FS. Application examples in a refinery show that the PID set value tracking error is reduced from ±2.1% to ±0.7%, reaching the control accuracy level of special valves.

[0045] 2. The adaptability is enhanced: The system can automatically adapt to changes in the medium and pressure. By fusing multi-sensor data through Kalman filtering, combined with the seasonal characteristic drift data and the medium influence matrix in the digital valve fingerprint library, ordinary regulating valves have the working condition adaptability of special valves.

[0046] 3. Equipment life extension: By warning of mechanical wear through a digital fingerprint library, unplanned shutdowns are reduced by 63%. When the wear characteristics of the valve guide sleeve are detected, the system automatically relaxes the minimum opening limit value B to avoid excessive wear of the valve in the low-opening region.

[0047] 4. Significant energy-saving benefits: In a certain power plant's feed water system, the actual annual power savings reached 280,000 kWh. Compared with traditional linearization methods, the present invention effectively solves the problem of increased energy consumption caused by valve non-linearity. For example, in a certain ethylene plant, the adoption of this system avoided the 7.2% annual average energy consumption increase originally caused by valve non-linearity.

[0048] Additional aspects and advantages of the present invention will be set forth in the following description, and some of them will be obvious from the description, or can be learned by practicing the present invention. In addition, the various aspects and advantages of the present invention can be realized and obtained by the method steps and combinations specifically pointed out in the appended claims. Brief Description of the Drawings

[0049] Figure 1 It is a schematic diagram of the system framework of the first embodiment of the present invention;

[0050] Figure 2 It is a schematic diagram of the method flow of the second embodiment of the present invention. Detailed Embodiments

[0051] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0052] It should be understood that the step numbers used in the text are only for convenient description and do not limit the execution order of the steps.

[0053] It should be understood that the terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include the plural forms.

[0054] The terms "comprising" and "including" indicate the presence of the described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or their combinations.

[0055] The term "and / or" refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0056] Example 1:

[0057] See Figure 1 , this embodiment provides a linear correction system for the opening degree / flow rate of a control valve, including: a flow deviation unit, a manual / automatic seamless switching unit, a deviation correction unit, a correction condition limiting unit, and a digital valve fingerprint library, where:

[0058] The flow deviation unit is used to calculate the real-time deviation value between the actual control flow rate of the control valve and the dynamic ideal flow rate. This unit includes a flow prediction module based on the LSTM network, which receives valve opening degree, medium pressure, temperature, and historical flow rate data, and outputs an ideal flow rate reference value adapted to the working conditions. The LSTM network adopts a four-layer structure, including an input layer, two hidden layers, and an output layer. The number of neurons in the hidden layers is 128 and 64 respectively, and the ReLU activation function is used. The input layer receives valve opening degree data, medium pressure data, temperature data, and historical flow rate data for the past 24 hours, and these data are collected at a sampling interval of 5 minutes. The LSTM network is pre-trained and can predict the ideal flow rate value according to the current working conditions, and the prediction accuracy reaches ±1.5% of the actual flow rate.

[0059] The flow deviation unit also includes a Kalman filter fusion module and an online learning engine. The Kalman filter fusion module is used to integrate the measurement data of electromagnetic flowmeters, ultrasonic flowmeters, and multi-parameter transmitters. This module adopts an extended Kalman filter algorithm to perform weighted fusion on the data of different sensors, and the weight coefficients are dynamically adjusted according to the historical accuracy of each sensor. The measurement range of the electromagnetic flowmeter is 0 - 500 m 3 / h, with an accuracy of ±0.5%, the measurement range of the ultrasonic flowmeter is 0 - 600 m 3 / h, with an accuracy of ±0.8%, and the multi-parameter transmitter provides measurement data with a pressure range of 0 - 10 MPa and a temperature range of -20°C to 150°C. The Kalman filter fusion module performs data fusion calculations every 100 milliseconds and outputs a highly reliable actual flow rate value.

[0060] The online learning engine automatically collects data and triggers nightly model fine-tuning when the system is stable. The criterion for determining system stability is: the flow rate fluctuation amplitude is less than 3% of the set value and the duration exceeds 30 minutes. The online learning engine adopts an incremental learning strategy. During the period from 2:00 - 4:00 every night, it uses the steady-state data collected on the same day to fine-tune the LSTM network, with the learning rate set to 0.001 and the number of training iterations set to 200 times to adapt to the slow changes in valve characteristics.

[0061] The hand / automatic seamless switching unit is equipped with a switching strategy module optimized by reinforcement learning. In the manual mode, it tracks the ideal flow reference value output by the flow deviation unit. In the automatic mode, it maintains the deviation value at the moment of switching and uses an operation mode predictor integrating the XGBoost model to predict the switching timing. The operation mode predictor integrating the XGBoost model analyzes the historical behavior patterns of the operator to predict the switching timing and gets ready 3 - 5 seconds in advance. This predictor uses 100 decision trees, with a maximum depth of 6, a learning rate of 0.1, a regularization parameter λ = 1.0, and γ = 0.5. The input features include 20 features such as time features (time period, shift), process parameters (upstream and downstream pressures, temperature change rate), and operator ID. The output is the switching probability within the next 10 seconds. When the predicted probability exceeds 75%, the system enters the standby state to ensure the smoothness of the switching process.

[0062] Among them, the reinforcement learning strategy adopts the DQN algorithm, and its reward function R is defined as:

[0063]

[0064] Among them, α and β are respectively the set first weight coefficient and the first weight coefficient. In this embodiment, α and β are 0.8 and 0.2 respectively, e(t) is the control error at time t, t0 and t1 are respectively the two start times and end times of the sampling time period, and N switch is the number of switchings. In this embodiment, the network structure of the DQN algorithm includes three fully connected layers, with the number of neurons being 64, 32, and 16 respectively. The size of the experience replay buffer is 10,000, and the target network update frequency is 100 steps.

[0065] The deviation correction unit uses Gaussian process regression to dynamically generate a non - linear compensation amount, solves the optimal opening correction value through the online convex optimization algorithm, and outputs the corrected valve opening command. Among them, the Gaussian process regression model uses the squared exponential kernel function:

[0066]

[0067] where x is the input vector [opening, pressure, temperature], σ f 、σ n and l are all hyperparameters. In this embodiment, σ f = 1.5, σ N = 0.1, l = 0.8. This model is updated once every 30 minutes using the latest 500 sets of historical data to adapt to the dynamic changes of the valve characteristics.

[0068] Specifically, the online convex optimization algorithm adopts the interior - point method to solve the objective function:

[0069] min||Δu||2+λ||e||2;

[0070] Where Δu is the opening correction amount, e is the predicted flow error, and λ = 2.5 is the balance parameter. The maximum number of iterations is set to 50 and the convergence threshold is 10^-4 in the optimization process to ensure that the calculation is completed within 100 milliseconds. The range of the opening correction value output by the algorithm is limited within ±15% to avoid system instability caused by excessive adjustment.

[0071] The correction condition limiting unit is built-in with a digital twin simulation module and a safety boundary prediction model. It conducts virtual verification before performing the opening correction and dynamically adjusts the minimum opening limit value B and the maximum opening limit value C; The digital twin simulation module constructs a high-precision model of the internal flow field of the control valve based on the computational fluid dynamics principle, and the simulation accuracy reaches ±2% of the actual flow rate. This module uses finite element mesh division, with the number of meshes being 500,000 and the computational time step being 0.01 seconds, and it can complete the simulation calculation within 1 second. The safety boundary prediction model is an architecture of Transformer. The inputs include: the cumulative valve action times, the medium particle concentration, and the vibration spectrum characteristics in the last 30 days. The output is the dynamic safety opening range [B', C'], where B' and C' are the corrected minimum and maximum opening limit values respectively. The Transformer model adopts a 4-layer encoder structure, with 8 attention heads, a hidden layer dimension of 256, and a feed-forward network dimension of 512. The upper limit of the cumulative valve action times received by the model is 1 million times, the medium particle concentration range is 0 - 500mg / L, and the vibration spectrum characteristics include 128-point spectrum data in the range of 0 - 1000Hz. The initial value of the output dynamic safety opening range is [5%, 95%], and it is dynamically adjusted with the increase of the valve usage time and the change of working conditions.

[0072] The digital valve fingerprint database stores the valve mechanical wear characteristics, seasonal characteristic drift data, and medium influence matrix, providing device-specific parameters for each unit. Among them, the mechanical wear characteristics include the geometric dimension change data of the valve core and valve seat, with an accuracy of 0.01mm; The seasonal characteristic drift data records the changes in the valve characteristic curve under different environmental temperatures (-30°C to 60°C); The medium influence matrix contains the influence coefficients of different media (such as water, oil, gas, etc.) on the valve characteristics. The database adopts a time-series database structure, supporting a write frequency of up to 10kHz and a query response time of sub-milliseconds.

[0073] As a preferred implementation manner of this embodiment, the system further includes a human-machine collaborative decision-making module, including:

[0074] The fine-tuned industrial large language model is based on the pre-trained Transformer architecture with a parameter scale of 700 million. It has been fine-tuned with 500,000 valve maintenance records and is used to generate natural language maintenance suggestions. The model can understand professional terms and generate specific maintenance suggestions according to the current valve status, such as "It is recommended to replace the valve core during the next planned shutdown, and the remaining service life is expected to be 45 days."

[0075] The knowledge graph reasoning engine is used to associate valve fault features with the maintenance record library. The engine contains more than 100,000 entity nodes and 500,000 relationship edges, covering 500 common valve fault modes and their solutions. The reasoning engine adopts a path-based reasoning algorithm with a maximum reasoning depth of 5 layers and can complete the reasoning process from fault features to maintenance suggestions within 200 milliseconds.

[0076] Embodiment 2:

[0077] This embodiment provides a control method for a control valve, which is applied to the control valve opening / flow linear correction system as described in Embodiment 1 above, and includes the steps:

[0078] S1. Predict the ideal flow value under the current working condition through the LSTM network;

[0079] The LSTM network receives the valve opening data, medium pressure data, temperature data, and historical flow data in the past 24 hours as inputs. The valve opening data ranges from 0 - 100%, with a sampling interval of 5 minutes; the medium pressure data ranges from 0 - 10 MPa, with a sampling interval of 5 minutes; the temperature data ranges from -20°C to 150°C, with a sampling interval of 5 minutes; the historical flow data ranges from 0 - 500 m 3 / h, with a sampling interval of 5 minutes. The LSTM network processes these input data through a four-layer structure, including an input layer, two hidden layers, and an output layer. The number of neurons in the hidden layers is 128 and 64 respectively. The network uses the Adam optimizer with a learning rate of 0.001 and a batch size of 32. The network outputs the ideal flow value under the current working condition, and the prediction accuracy reaches ±1.5% of the actual flow.

[0080] S2. Compare the actual flow with the ideal flow to generate a deviation signal;

[0081] The system integrates the measurement data of the electromagnetic flowmeter, ultrasonic flowmeter, and multi-parameter transmitter through the Kalman filter fusion module to obtain a high-precision actual flow value. The measurement range of the electromagnetic flowmeter is 0 - 500 m 3 / h, with an accuracy of ±0.5%; the measurement range of the ultrasonic flowmeter is 0 - 600 m 3 / h, with an accuracy of ±0.8%; the multi-parameter transmitter provides measurement data for a pressure range of 0 - 10 MPa and a temperature range of -20°C to 150°C. The Kalman filter fusion module performs data fusion calculations every 100 milliseconds and outputs the actual flow value. The system compares the actual flow value with the predicted ideal flow value in step S1 and calculates the deviation value e = actual flow - ideal flow. The calculation frequency of the deviation signal is 10 Hz to ensure that the system can respond to flow changes in a timely manner.

[0082] S3. Calculate the opening compensation amount using Gaussian process regression, and output it after being verified by digital twin;

[0083] The Gaussian process regression model receives the deviation signal calculated in step S2, combines the current valve opening, medium pressure, and temperature data, and calculates the required opening compensation amount. The calculated opening compensation amount range is limited within ±15% to avoid system instability caused by excessive adjustment.

[0084] Digital twin verification includes a CFD-based flow field simulation and a stress analysis module. The CFD-based flow field simulation predicts the corrected flow characteristics. Using the finite element method, with 500,000 meshes and a calculation time step of 0.01 seconds, the simulation calculation can be completed within 1 second. The simulation results include parameters such as flow prediction values, pressure drops, and flow field distributions. The stress analysis module checks whether the force on the valve core exceeds the limit, calculates the stress distribution of the valve core at different openings, and ensures that the maximum stress does not exceed 70% of the material yield strength. When the simulation results show that the corrected opening can effectively improve the flow linearity and does not cause the valve mechanical stress to exceed the limit, the system outputs the corrected valve opening command.

[0085] S4. When the wear characteristics of the valve guide sleeve are detected, automatically relax the minimum opening limit value B;

[0086] The system detects the wear state of the valve guide sleeve by combining the valve mechanical wear characteristic data stored in the digital valve fingerprint library with the real-time monitored valve vibration data. The characteristics of guide sleeve wear include: the energy in the 80 - 120 Hz frequency band of the vibration spectrum increases by more than 30%, the valve stem displacement fluctuation increases by more than 0.2 mm, and abnormal sounds occur during the valve opening and closing process. When the system detects these characteristics, it determines that there is wear on the valve guide sleeve and automatically relaxes the minimum opening limit value B from the initial 5% to 8 - 12%, and the specific adjustment amount is dynamically calculated according to the wear degree. This can avoid the aggravation of guide sleeve wear under small opening conditions and extend the service life of the valve.

[0087] Embodiment 3:

[0088] This embodiment provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements the control method of the regulating valve as described in any embodiment of the present invention.

[0089] Embodiment 4:

[0090] This embodiment provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the control method of the regulating valve as described in any embodiment of the present invention.

[0091] In the embodiments of the present application, "at least one" means one or more, and "a plurality" means two or more. "And / or" describes the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent the cases of A existing alone, A and B existing simultaneously, and B existing alone. Where A and B can be singular or plural. The character " / " generally represents an "or" relationship between the associated objects before and after. "At least one of the following" and its similar expressions refer to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, and c can represent: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, and c can be single or multiple.

[0092] Those of ordinary skill in the art can realize that the units and algorithm steps described in the embodiments disclosed herein can be implemented by a combination of electronic hardware, computer software, and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0093] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0094] In several embodiments provided by the present application, if any function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks, or optical discs.

[0095] The above are only the embodiments of the present invention, and do not limit the patent scope of the present invention accordingly. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present invention.

Claims

1. A regulating valve opening / flow linear correction system, characterized in that: include: Flow deviation unit: used to calculate the real-time deviation between the actual control flow of the regulating valve and the dynamic ideal flow, including a flow prediction module based on the LSTM network. The flow prediction module receives valve opening, medium pressure, temperature and historical flow data, and outputs an ideal flow reference value that is adaptive to the working condition; Manual / automatic non-disturbance switching unit: It is equipped with a switching strategy module optimized by reinforcement learning. In manual mode, it tracks the ideal flow reference value output by the flow deviation unit, maintains the deviation value at the switching moment in automatic mode, and uses an operation mode predictor integrated with the XGBoost model to predict the switching timing. Deviation correction unit: Gaussian process regression is used to dynamically generate nonlinear compensation, the optimal opening correction value is solved through an online convex optimization algorithm, and the corrected valve opening instruction is output; Correction condition restriction unit: Built-in digital twin simulation module and safety boundary prediction model, virtual verification before executing opening correction, and dynamically adjust the minimum opening limit value B and the maximum opening limit value C; Digital valve fingerprint library: stores valve mechanical wear characteristics, seasonal characteristic drift data and medium impact matrix, and provides equipment-specific parameters for each unit.

2. The regulating valve opening / flow linearity correction system according to claim 1, characterized in that: The flow deviation unit also includes: Kalman filter fusion module for integrating measurement data from electromagnetic flowmeters, ultrasonic flowmeters and multi-parameter transmitters; The online learning engine is used to automatically collect data when the system is stable and trigger model fine-tuning at a set time.

3. The regulating valve opening / flow linearity correction system according to claim 1, characterized in that: The reinforcement learning strategy in the manual / automatic non-disturbance switching unit adopts the DQN algorithm, and its reward function R is defined as: Where α and β are the first weight coefficient and the second weight coefficient respectively, e(t) is the control error at time t, t0 and t1 are the two starting times and ending times of the sampling time period respectively, B switch is the number of switching times.

4. The regulating valve opening / flow linearity correction system according to claim 1, characterized in that: The Gaussian process regression model of the bias correction unit uses a squared exponential kernel function: Where x is the input vector [opening, pressure, temperature], σ f , σ n and l are hyperparameters.

5. The regulating valve opening / flow linearity correction system according to claim 1, characterized in that: The safety boundary prediction model of the correction condition restriction unit is a Transformer architecture, and the input includes: The cumulative number of valve operations, medium particle concentration, and vibration spectrum characteristics in the last 30 days are output as a dynamic safety opening range [B', C']. B' and C' are the corrected minimum opening limit value and maximum opening limit value, respectively.

6. The regulating valve opening / flow linearity correction system according to claim 1, characterized in that: It also includes a human-machine collaborative decision-making module, including: A fine-tuned industrial-scale language model for generating natural language maintenance recommendations; Knowledge graph reasoning engine, used to associate valve failure characteristics with maintenance record library.

7. A control method for a regulating valve, applied to the regulating valve opening / flow linearity correction system according to any one of claims 1 to 6, comprising the steps of: S1, predict the ideal flow value under the current working conditions through LSTM network; S2, compare the actual flow rate with the ideal flow rate to generate a deviation signal; S3, use Gaussian process regression to calculate the opening compensation amount, and output it after verification by digital twin; S4. When the wear characteristics of the valve guide sleeve are detected, the minimum opening limit value B is automatically relaxed.

8. The regulating valve control method according to claim 7, characterized in that: The digital twin verification in step S3 includes: Based on CFD flow field simulation, predict the corrected flow characteristics; The stress analysis module verifies whether the valve core force exceeds the limit.

9. A computer-readable storage medium storing program instructions, which implement the regulating valve control method according to any one of claims 7 to 8 when the instructions are executed by a processor.

10. An electronic device comprising a memory, a processor and a computer program stored in the memory, wherein the processor implements the regulating valve control method according to any one of claims 7 to 8 when executing the program.

Images