An adaptive intelligent butterfly valve control method with flow prediction model

Abstract

This invention discloses an adaptive intelligent butterfly valve control method with a flow prediction model, comprising the following steps: S1, collecting various parameters during butterfly valve operation and obtaining a standardized operating condition dataset after preprocessing; S2, constructing an improved flow prediction model incorporating an attention mechanism, establishing an improved loss function incorporating butterfly valve opening weights, and training the model; S3, constructing an adaptive fuzzy PID controller, calculating the adaptive fuzzy PID control parameters and butterfly valve opening adjustment amount to obtain the target butterfly valve opening; S4, converting the target butterfly valve opening into an actuator control signal to drive the butterfly valve to adjust its opening, while simultaneously collecting the actual flow rate and operating condition parameters after adjustment in real time; S5, calculating the flow deviation, correcting the flow prediction model parameters based on the deviation, and dynamically updating the adaptive fuzzy PID control parameters. This invention achieves accurate prediction and dynamic adaptive adjustment of butterfly valve flow, as well as forward-looking control.

Description

Technical Field

[0001] This invention relates to the field of butterfly valve control technology, and in particular to an adaptive intelligent butterfly valve control method with a flow prediction model. Background Technology

[0002] Butterfly valves, characterized by their simple structure, high flow capacity, and convenient adjustment, are widely used in industrial fluid transport systems. Their core function is to control fluid flow by adjusting the valve plate angle to change the fluid flow area. As industrial production moves towards intelligence and precision, higher demands are being placed on the control accuracy, response speed, adaptability, and energy efficiency of butterfly valves.

[0003] Currently, existing butterfly valve control methods suffer from the following technical shortcomings, making it difficult to meet the high-precision control requirements under complex operating conditions: (1) The control mode has lag: Most existing butterfly valve control adopts the "feedback control" mode, that is, by detecting the deviation between the current actual flow rate and the set flow rate, the butterfly valve opening is adjusted. This mode has obvious adjustment lag. When the fluid conditions (pressure, temperature, viscosity) change suddenly, it is easy to cause excessive flow fluctuation and adjustment oscillation, and it is impossible to achieve forward control of flow rate. The lag problem is more prominent, especially in large inertial fluid systems.

[0004] (2) Separation of flow prediction and control: Although some existing technologies have introduced flow prediction, the prediction model and control strategy are independent of each other. The prediction results are not effectively integrated into the adjustment process of control parameters. The prediction value is only used as a reference and does not play the core role of "prediction guidance". It cannot fundamentally solve the problem of feedback lag. At the same time, the prediction model mostly adopts a single time series prediction method and does not integrate the butterfly valve's own operating parameters (valve plate angle, pipe size) with the real-time fluid parameters. The prediction accuracy is low, especially in the specific opening range of the butterfly valve (15°~20°, 80°~90°). Due to factors such as nonlinear adjustment, cavitation, and vibration, the prediction error increases significantly.

[0005] (3) Lack of iterative optimization of model and control parameters: After the existing prediction model is trained, the parameters remain fixed. During long-term operation, the prediction accuracy will gradually decrease as the equipment ages and the operating conditions drift. The control parameters have not established a long-term iterative mechanism based on changes in operating conditions, which leads to the control performance decaying with the increase of operating time and failing to meet the operating requirements of long-term complex operating conditions. Summary of the Invention

[0006] The purpose of this invention is to provide an adaptive intelligent butterfly valve control method with a flow prediction model, which can solve the technical problems of existing butterfly valve control such as lag, low accuracy and poor adaptability, and realize accurate, adaptive and forward-looking control of butterfly valve flow.

[0007] To achieve the above objectives, the present invention provides an adaptive intelligent butterfly valve control method with a flow prediction model, comprising the following steps: S1. Collect fluid parameters, valve parameters, pipeline and flow parameters and environmental parameters during the operation of the butterfly valve. Obtain a standardized operating condition dataset through outlier removal, normalization and feature filtering. S2. Based on the preprocessed dataset, construct an improved ILSTM-Attention flow prediction model that integrates the attention mechanism, establish an improved loss function that integrates the butterfly valve opening weights, and train the model. S3. Based on the total deviation and the rate of change of the deviation between the predicted flow rate and the set flow rate, and combined with the real-time operating parameters, construct an adaptive fuzzy PID controller, calculate the adaptive fuzzy PID control parameters and the butterfly valve opening adjustment amount, and obtain the target opening of the butterfly valve. S4. Convert the calculated target opening degree of the butterfly valve into a control signal for the actuator to drive the butterfly valve to adjust the opening degree, and at the same time collect the actual flow rate and operating parameters after adjustment in real time. S5. Calculate the deviation between the predicted flow rate and the actual flow rate, and correct the flow prediction model parameters based on the deviation. At the same time, dynamically update the adaptive fuzzy PID control parameters according to real-time operating conditions until the flow rate stabilizes within the set threshold range.

[0008] Preferably, in S1, various sensors installed on the butterfly valve and conveying pipeline collect various operating condition data in real time, with a collection cycle of... Set to 0.1-1s; The fluid parameters include: fluid pressure, fluid temperature, fluid viscosity, and fluid density. Valve parameters include: the current angle of the butterfly valve plate and the feedback signal from the butterfly valve actuator; Pipeline and flow parameters include: pipeline inner diameter, real-time flow rate, and historical flow rate; Environmental parameters include: ambient temperature.

[0009] Preferably, in S2, the improved ILSTM-Attention traffic prediction model consists of an input layer, an ILSTM layer, an attention mechanism layer, a fully connected layer, and an output layer. The input layer takes in preprocessed feature parameters; the ILSTM layer replaces the traditional sigmoid function with the Leaky-ReLU activation function, improving the activation functions of the forget gate, input gate, and output gate; the attention mechanism layer introduces a channel attention mechanism to assign weights to the feature maps output by the ILSTM layer; the fully connected layer fuses and reduces the dimensionality of the features output by the attention mechanism layer, transforming them into a one-dimensional feature vector; and the output layer outputs the predicted traffic flow for future timeframes. The output of the improved ILSTM-Attention traffic prediction model is combined with the prediction correction coefficient to make real-time corrections to the predicted traffic. The calculation formula is as follows: ; in, for t Time prediction Real-time traffic, For the predicted duration, 1 ≤ ≤30min for t Historical traffic sequences up to and before time. They are respectively t The fluid pressure, temperature, viscosity, butterfly valve angle, and fluid density at any given time. D The inner diameter of the pipe. For prediction correction factors; Determined by both real-time prediction deviation and fluid pressure, its calculation formula is as follows: ; in, for t Predicted flow at any time for t Actual flow at any given moment This is a correction factor, with a value ranging from 0.5 to 1.5. This is the pressure influence coefficient, with a value ranging from 0.9 to 1.1. P avg This represents the historical average pressure.

[0010] Preferably, the improved loss function in S2 adds prediction bias weights for two weak intervals to the traditional mean squared error, and the calculation formula is as follows: ; in, N The number of training samples. Q p,i , Q a,i The first i The predicted flow and actual flow for each sample For the first i The butterfly valve angle corresponding to each sample I (·) is an indicator function. The loss weights are for the opening ranges of 15°-20° and 80°-90°, respectively, with values ​​ranging from 1.5 to 2.5. When training the model, the preprocessed training set is input into the improved ILSTM-Attention traffic prediction model, using an improved loss function.L With the goal of minimizing errors, the Adam optimizer was used for model training, with a learning rate of 0.001-0.01 and 100-500 iterations. During training, model parameters were adjusted in real-time using a validation set to avoid overfitting. After training, the model performance was validated using a test set, and the prediction error was minimized. When the model training is complete, if the prediction error exceeds the threshold, the model structure and parameters are adjusted, and the model is retrained until the requirements are met; among these, This is the prediction deviation threshold.

[0011] Preferably, in S3, the input to the adaptive fuzzy PID controller is the total deviation. and the rate of change of deviation Total deviation It incorporates the deviation between predicted and actual traffic and sets a weight for the prediction deviation. Weight greater than actual deviation This emphasizes the guiding role of predicted flow and achieves a prediction-first control mode. The specific formula is as follows: ; in, To set the flow rate, These are the weights of the prediction deviation and the actual deviation, respectively. ; By incorporating the influence of fluid viscosity, the characterization accuracy of the deviation change rate is corrected. The specific formula is as follows: ; in, for t The total deviation at time -1 To control the cycle, This represents the historical average fluid viscosity.

[0012] Preferably, the adaptive fuzzy PID control parameter calculation process integrates fuzzy inference and operating condition compensation, and the specific calculation formula is as follows: ; ; ; in, k p ( t ), k i ( t ), k d ( t The proportional coefficient, integral coefficient, and differential coefficient at time t are respectively. k p0 ,k i0 , k d0 These are the initial control parameters. For parameter adjustment coefficients, This represents the maximum threshold for the rate of change of deviation. Proportion system k p ( t The integral coefficient increases as pressure increases and the butterfly valve angle approaches 45°, thus improving the system's response speed. k i ( t The differential coefficient decreases as the rate of change of deviation and viscosity increase, thus avoiding integral saturation and reducing adjustment oscillations; k d ( t The rate of change of deviation decreases as the absolute value of the deviation increases, and increases as the butterfly valve angle increases, thereby improving the stability of the system and avoiding differential impact.

[0013] Preferably, in the calculation of the butterfly valve opening adjustment, a specific range compensation amount is added to the traditional PID adjustment amount. The specific calculation formula is as follows: ; in, This refers to the butterfly valve opening adjustment amount. This is the opening compensation amount; The calculation formula is: ; in, All are compensation coefficients. for t The current actual opening degree of the butterfly valve plate is measured at all times. 17.5° and 85° are the midpoint angles of the two weak zones, respectively. Butterfly valve target opening The formula for calculation is: ; Set 0°≤ The constraint of ≤90° ensures that the target opening is within the effective adjustment range of the butterfly valve.

[0014] Preferably, S4 specifically includes the following steps: S41. Control signal conversion: Convert the calculated target opening degree... This is converted into a control signal for the butterfly valve actuator, ensuring precise matching between the control signal and the target opening degree. The specific formula is as follows: ; in, U( t () represents the actuator control signal at time t. U min , U max These are the minimum and maximum control signals for the actuator, respectively; S42, Opening Execution: The actuator receives the control signal. U ( t After that, drive the butterfly valve plate to rotate to the target opening degree. The flow rate is regulated by adjusting the fluid flow area; simultaneously, the actuator provides real-time feedback on the actual opening degree. To determine if the butterfly valve is operating normally; S43. Feedback Data Acquisition: After adjustment, the actual flow rate, fluid parameters, and actual butterfly valve opening data are collected in real time by sensors. These data are compared with the predicted flow rate and target butterfly valve opening to calculate the deviation and provide data support for subsequent iterative optimization.

[0015] Preferably, S5 specifically includes the following steps: S51. Prediction Model Parameter Correction: Calculate Prediction Error ,when At that time, the weight parameters of the improved ILSTM-Attention traffic prediction model are corrected based on the gradient descent method, and the step size is also corrected. The calculation formula is as follows: ; in, The initial correction step size, This represents the largest prediction deviation in history. S52. Iterative Update of Control Parameters: Update initial control parameters based on the rate of change of real-time operating parameters. k p0 , k i0 , k d0 This allows the initial control parameters to be dynamically adjusted as operating conditions drift over a long period, improving the long-term adaptability of the control strategy. Simultaneously, based on the actual application effects of the fuzzy inference rules, the fuzzy rule base is periodically optimized to ensure the rationality of control parameter adjustments. The specific formula is as follows: ; ; ; in, To update the coefficients, ΔP ( t )= P ( t )- P (t -1) is t Time and t Pressure difference at time -1.

[0016] Preferably, S5 also includes a fault diagnosis and emergency control mechanism. When butterfly valve jamming, sensor malfunction, or flow deviation continuously exceeding the threshold is detected, it automatically switches to emergency control mode. The formula for calculating the emergency opening adjustment is: ; in, For emergency opening adjustment, This is the emergency adjustment coefficient.

[0017] Therefore, the present invention adopts the above-mentioned adaptive intelligent butterfly valve control method with a flow prediction model, which solves the technical defects of existing butterfly valve control such as flow regulation lag, low regulation accuracy in specific opening range, and poor parameter adaptive capability. It realizes accurate prediction and dynamic adaptive regulation of butterfly valve flow, improves the regulation accuracy and stability of butterfly valve under complex working conditions, and is suitable for intelligent control of butterfly valves in various industrial fluid transportation systems.

[0018] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0019] Figure 1 This is a schematic diagram illustrating the steps of an embodiment of an adaptive intelligent butterfly valve control method with a flow prediction model according to the present invention. Detailed Implementation

[0020] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0021] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0022] Example 1: like Figure 1As shown, this embodiment provides an adaptive intelligent butterfly valve control method with a flow prediction model, including the following steps: S1. Collect fluid parameters, valve parameters, pipeline and flow parameters, and environmental parameters during the operation of the butterfly valve. Through outlier removal, normalization, and feature filtering, a standardized operating condition dataset is obtained.

[0023] The specific implementation process is as follows: Data Acquisition: Various sensors installed on butterfly valves and delivery pipelines collect real-time data on various operating conditions, with a collection cycle of [missing information]. Set to 0.1-1s (can be adjusted according to actual working conditions): Fluid parameters include: fluid pressure, fluid temperature, fluid viscosity, and fluid density; Valve parameters include: the current angle of the butterfly valve plate and the feedback signal from the butterfly valve actuator; Pipeline and flow parameters include: pipeline inner diameter, real-time flow rate, and historical flow rate; Environmental parameters include: ambient temperature, used to correct for the temperature effects on fluid parameters.

[0024] Outlier Removal: Due to electromagnetic interference and sensor malfunctions in industrial settings, outliers may exist in the collected data. An outlier removal method is employed, based on the 3σ principle combined with a butterfly valve operating condition compensation formula. Compared to the traditional 3σ principle, this embodiment incorporates a butterfly valve angle compensation term in the formula to avoid misclassifying normal flow fluctuations as outliers in weak adjustment ranges such as 15°-20° and 80°-90°, thus ensuring data validity. The specific formula is as follows: in, This refers to traffic data collected in a single instance. Let σ be the mean of historical traffic data, and σ be the standard deviation of historical traffic data. This is the butterfly valve angle compensation coefficient (range 0.1-0.3), and 45° is the optimal angle for butterfly valve adjustment sensitivity.

[0025] Normalization: Due to the significant differences in the dimensions of parameters under different operating conditions (e.g., pressure is measured in MPa, temperature in ℃, and angle in °), directly inputting them into the model would lead to training bias. Therefore, an improved linear normalization formula incorporating operating condition weights is adopted to normalize all parameters to the [0, 1] interval. Simultaneously, the butterfly valve angle weight is incorporated to enhance the representational ability of operating condition parameters within a specific opening range, addressing the problem of traditional normalization neglecting the influence of the butterfly valve's own state. The specific formula is as follows: in, These are the normalized parameter values. x The original value of the parameter.x min , x max These are the historical minimum and maximum values ​​of the parameter, respectively.

[0026] Feature selection: Redundant information may exist in the collected multi-dimensional parameters (e.g., the correlation between ambient temperature and fluid temperature is too high). A formula based on mutual information entropy combined with operating condition correlation is used to select parameters with strong correlation to flow rate as input features for subsequent models. This eliminates redundant information, reduces model training complexity, and improves model training efficiency and prediction accuracy. The mutual information entropy threshold is used in this process. I ( X , Q The value ≥0.6 can be finely adjusted according to actual working conditions. The specific formula is as follows: in, Operating parameters X With traffic Q Mutual information entropy, P ( x , q ) is a parameter X With traffic Q The joint probability density, P ( x ), P ( q ) are parameters X With traffic Q Marginal probability density, The relevant weights for operating conditions (range 0.8-1.2).

[0027] After preprocessing, a standardized operating condition dataset is obtained, which is divided into a training set (70%~80%), a validation set (10%~15%), and a test set (10%~15%), which are used for the training and validation of the traffic prediction model.

[0028] S2. Based on the preprocessed dataset, construct an improved ILSTM-Attention flow prediction model that incorporates the attention mechanism, establish an improved loss function that incorporates the butterfly valve opening weights, and train the model.

[0029] The specific implementation process is as follows: Model Structure Design: The improved ILSTM-Attention traffic prediction model consists of an input layer, an ILSTM layer, an attention mechanism layer, a fully connected layer, and an output layer. Input layer: Inputs preprocessed feature parameters, including historical flow sequences. Real-time fluid parameters ( ), butterfly valve angle Pipe inner diameter D ; ILSTM layer: Based on the traditional LSTM layer, the Leaky-ReLU activation function is used to replace the traditional Sigmoid function, and the activation functions of the forget gate, input gate and output gate are improved to solve the gradient vanishing problem, enhance the model's ability to fit long sequence data, and effectively capture the long-term dependence between historical traffic and real-time operating parameters. Attention Mechanism Layer: Introducing a channel attention mechanism, weighting is applied to the feature map output by the ILSTM layer, focusing on features that are strongly correlated with flow rate changes (such as butterfly valve angle and fluid pressure), improving the model's sensitivity to key operating parameters, and solving the problem of existing prediction models not responding to changes in operating conditions in a timely manner; Fully connected layer: fuses and reduces the dimensionality of the features output by the attention mechanism layer, transforming them into a one-dimensional feature vector; Output layer: Outputs the predicted flow at future time points.

[0030] The output of the improved ILSTM-Attention traffic prediction model is combined with the prediction correction coefficient to make real-time corrections to the predicted traffic. The calculation formula is as follows: in, for t Time prediction Real-time traffic, For the predicted duration, 1 ≤ ≤30min for t Historical traffic sequences up to and before time. They are respectively t The fluid pressure, temperature, viscosity, butterfly valve angle, and fluid density at any given time. D The inner diameter of the pipe. This is the prediction correction factor.

[0031] Determined by both real-time prediction deviation and fluid pressure, it can correct prediction results in real time, reduce the impact of pressure fluctuations on prediction accuracy, and ensure the adaptability of predicted flow rate to actual operating conditions. Its calculation formula is as follows: in, for t Predicted flow at any time for t Actual flow at any given moment This is a correction factor, with a value ranging from 0.5 to 1.5. This is the pressure influence coefficient, with a value ranging from 0.9 to 1.1. P avgThis represents the historical average pressure.

[0032] An improved loss function integrating butterfly valve opening weights was established: To address the low prediction accuracy of butterfly valves in the 15°-20° and 80°-0° opening ranges, a multi-objective loss function integrating opening range compensation was developed. This function, based on the traditional mean squared error (MSE), incorporates prediction bias weights for two weak ranges. By increasing the loss weights of the weak ranges, the model is forced to focus on optimizing the prediction accuracy of these ranges during training, overcoming the technical bottleneck of poor adaptability of existing prediction models to weak ranges. The specific formula is as follows: in, N The number of training samples. Q p,i , Q a,i The first i The predicted flow and actual flow for each sample For the first i The butterfly valve angle corresponding to each sample I (·) is an indicator function (when the condition in parentheses is satisfied) I =1, otherwise I =0), The loss weights are for the opening ranges of 15°-20° and 80°-90°, respectively, with values ​​ranging from 1.5 to 2.5.

[0033] When training the model, the preprocessed training set is input into the improved ILSTM-Attention traffic prediction model, using an improved loss function. L With the goal of minimizing errors, the Adam optimizer was used for model training, with a learning rate of 0.001-0.01 and 100-500 iterations. During training, model parameters were adjusted in real-time using a validation set to avoid overfitting. After training, the model performance was validated using a test set, and the prediction error was minimized. When the model training is complete, it can be used; if the prediction error exceeds the threshold, the model structure and parameters are adjusted, and retraining is performed until the requirements are met. The prediction deviation threshold is set to a value ranging from 0.05 to 0.15m. 3 / h.

[0034] S3. Based on the total deviation and the rate of change of the deviation between the predicted flow rate and the set flow rate, and combined with the real-time operating parameters, construct an adaptive fuzzy PID controller, calculate the adaptive fuzzy PID control parameters and the butterfly valve opening adjustment amount, and obtain the target opening of the butterfly valve.

[0035] Specifically, the input to the adaptive fuzzy PID controller is the total deviation of the fused prediction bias. and the rate of change of deviation : Total Deviation It incorporates the deviation between predicted and actual traffic, setting a greater weight for the prediction deviation than the actual deviation. (0.55~0.75) greater than the actual deviation weight This emphasizes the guiding role of predicted flow, achieving a "prediction-first" control mode, fundamentally solving the lag problem of traditional feedback control. The specific formula is as follows: in, To set the flow rate, These are the weights of the prediction deviation and the actual deviation, respectively. ; Incorporating the influence of fluid viscosity, correcting the rate of change of deviation To improve characterization accuracy and avoid misjudgment of deviation rates due to viscosity changes, thus ensuring the rationality of control parameter adjustments, the specific formula is as follows: in, for t The total deviation at time -1 To control the cycle, This represents the historical average fluid viscosity.

[0036] Adaptive PID control parameter calculation: Since traditional PID control parameters are fixed, and conventional fuzzy PID control parameters only depend on the deviation and the rate of change of the deviation, this embodiment designs an adaptive control parameter calculation formula that integrates fuzzy inference and operating condition compensation, taking fluid pressure as an example. P ( t Viscosity m ( t ), butterfly valve angle i ( t Real-time operating parameters, such as those mentioned above, are incorporated into the parameter adjustment process to achieve adaptive control of the parameters based on operating conditions. The specific calculation formula is as follows: in, k p ( t ), k i ( t ), k d ( t ) are respectively t The proportional coefficient, integral coefficient, and differential coefficient at time intervals.k p0 , k i0 , k d0 These are the initial control parameters. For parameter adjustment coefficients, This is the maximum threshold for the rate of change of deviation.

[0037] Proportion system k p ( t The integral coefficient increases as pressure increases and the butterfly valve angle approaches 45°, thus improving the system's response speed. k i ( t The differential coefficient decreases as the rate of change of deviation and viscosity increase, thus avoiding integral saturation and reducing adjustment oscillations; k d ( t The rate of change of deviation decreases as the absolute value of the deviation increases, and increases as the butterfly valve angle increases, thereby improving the stability of the system and avoiding differential impact.

[0038] The above-mentioned adaptive PID control parameter calculation formula breaks through the limitations of traditional control parameter adjustment, realizes dynamic adaptation of control parameters to real-time operating conditions, and improves the versatility and adaptability of control strategies.

[0039] Butterfly valve opening adjustment calculation: In the calculation of the butterfly valve opening adjustment, a specific range compensation is added to the traditional PID control. This compensation is specifically applied to the two weak adjustment ranges of the butterfly valve: 15°-20° and 80°-90°. The specific calculation formula is as follows: in, This refers to the butterfly valve opening adjustment amount. This is the opening compensation amount; The calculation formula for the weak range of butterfly valve regulation is as follows: in, All are compensation coefficients. for t The current actual opening degree of the butterfly valve plate is 17.5° and 85°, which are the midpoint angles of the two weak zones, respectively.

[0040] In the 15°-20° range, the nonlinearity of the adjustment is corrected by compensation, which reduces the impact of cavitation and vibration on the adjustment accuracy. In the 80°-90° range, the flow rate is precisely fine-tuned by compensation, which solves the problem of the adjustment of flow rate changes that are not obvious in this range.

[0041] Butterfly valve target opening The formula for calculation is: Set 0°≤ The constraint of ≤90° ensures that the target opening is within the effective adjustment range of the butterfly valve.

[0042] Fuzzy Inference Rule Optimization: Combining the operating characteristics of butterfly valves with industrial control experience, the fuzzy inference rules are optimized by incorporating changes in operating parameters (pressure, viscosity) into the fuzzy rule base. For example: when... big, Small, P ( t > P avg At that time, the output of fuzzy inference is relatively large. k p ( t Adjust the amount to improve response speed; when Small, big, m ( t > m avg When the output is smaller k i ( t Adjusting the amount of data helps avoid oscillations, making the fuzzy inference rules more closely match actual working conditions and improving control accuracy.

[0043] S4. Convert the calculated target opening degree of the butterfly valve into a control signal for the actuator to drive the butterfly valve to adjust its opening degree, while simultaneously collecting the actual flow rate and operating parameters after adjustment in real time. Specifically, this includes the following steps: S41. Control signal conversion: Convert the calculated target opening degree... This is converted into a control signal (voltage or current) for the butterfly valve actuator, ensuring precise matching between the control signal and the target opening degree. The actuator employs a servo motor or electro-hydraulic servo mechanism, offering fast response and high adjustment accuracy. The specific formula is as follows: in, U ( t )for t The actuator control signal at any given time, U min , U max These are the minimum and maximum control signals for the actuator, respectively.

[0044] S42, Opening Execution: The actuator receives the control signal. U ( t After that, drive the butterfly valve plate to rotate to the target opening degree. The flow rate is regulated by adjusting the fluid flow area; simultaneously, the actuator provides real-time feedback on the actual opening degree. Determine if the butterfly valve is operating normally (e.g., whether there is any jamming).

[0045] S43. Feedback Data Acquisition: After adjustment, the actual flow rate, fluid parameters, and actual butterfly valve opening data are collected in real time by sensors. These data are compared with the predicted flow rate and target butterfly valve opening to calculate the deviation and provide data support for subsequent iterative optimization.

[0046] S5. Calculate the deviation between the predicted flow rate and the actual flow rate, and correct the flow prediction model parameters based on the deviation. At the same time, dynamically update the adaptive fuzzy PID control parameters according to real-time operating condition changes until the flow rate stabilizes within the set threshold range. This includes the following steps: S51. Prediction Model Parameter Correction: Calculate Prediction Error ,when At that time, the weight parameters of the improved ILSTM-Attention traffic prediction model are corrected based on the gradient descent method to achieve adaptive adjustment of the correction step size, avoid model oscillation, and retain the model's generalization ability; when At this time, the model parameters remain unchanged, reducing the amount of computation.

[0047] Correction step size The calculation formula is as follows: in, This is the initial correction step size (range 0.001-0.01). This represents the largest prediction error in history.

[0048] S52. Iterative update of control parameters: Update the initial control parameters based on the rate of change of real-time operating parameters (butterfly valve opening rate of change, pressure rate of change, viscosity rate of change). k p0 , k i0 , k d0 This allows the initial control parameters to be dynamically adjusted as operating conditions drift over a long period, improving the long-term adaptability of the control strategy. Simultaneously, based on the actual application effects of the fuzzy inference rules, the fuzzy rule base is periodically optimized to ensure the rationality of control parameter adjustments. The specific formula is as follows: in, This is for updating coefficients (value range 0.05-0.2). ΔP (t )= P ( t )- P ( t -1) is t Time and t Pressure difference at time -1.

[0049] S53. Emergency control mechanism triggered: When the following fault conditions are detected, the system will automatically switch to emergency control mode to ensure safe and stable operation: Butterfly valve jamming: Current actual opening degree relative to target opening The deviation lasts ≥5° (duration ≥3s); Sensor malfunction: Abnormal values ​​account for ≥10% of the collected data for a continuous period of ≥3 seconds; Flow deviation exceeds standard: Total deviation Continuous ≥0.2m 3 / h (duration ≥5s).

[0050] Calculate the emergency opening adjustment amount in emergency control mode. Simultaneously, an alarm signal is issued to remind staff to handle the fault promptly. After the fault is resolved, it automatically switches back to normal control mode. The specific formula is as follows: in, This is the emergency adjustment coefficient, with a value range of 0.8-1.2.

[0051] Based on the above method, repeat steps S1-S5 to achieve a closed-loop iteration of the entire process of "data acquisition - flow prediction - adaptive control - feedback optimization", ensuring that the butterfly valve flow is always stable within the set threshold range, improving control accuracy and stability, and reducing energy consumption.

[0052] Therefore, the present invention adopts the above-mentioned adaptive intelligent butterfly valve control method with a flow prediction model, which solves the technical defects of existing butterfly valve control such as flow regulation lag, low regulation accuracy in specific opening range, and poor parameter adaptive capability. It realizes accurate prediction and dynamic adaptive regulation of butterfly valve flow, improves the regulation accuracy and stability of butterfly valve under complex working conditions, and is suitable for intelligent control of butterfly valves in various industrial fluid transportation systems.

[0053] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. An adaptive intelligent butterfly valve control method with a flow prediction model, characterized in that, Includes the following steps: S1. Collect fluid parameters, valve parameters, pipeline and flow parameters and environmental parameters during the operation of the butterfly valve. Obtain a standardized operating condition dataset through outlier removal, normalization and feature filtering. S2. Based on the preprocessed dataset, construct an improved ILSTM-Attention flow prediction model that integrates the attention mechanism, establish an improved loss function that integrates the butterfly valve opening weights, and train the model. S3. Based on the total deviation and the rate of change of the deviation between the predicted flow rate and the set flow rate, and combined with the real-time operating parameters, construct an adaptive fuzzy PID controller, calculate the adaptive fuzzy PID control parameters and the butterfly valve opening adjustment amount, and obtain the target opening of the butterfly valve. S4. Convert the calculated target opening degree of the butterfly valve into a control signal for the actuator to drive the butterfly valve to adjust the opening degree, and at the same time collect the actual flow rate and operating parameters after adjustment in real time. S5. Calculate the deviation between the predicted flow rate and the actual flow rate, and correct the flow prediction model parameters based on the deviation. At the same time, dynamically update the adaptive fuzzy PID control parameters according to real-time operating conditions until the flow rate stabilizes within the set threshold range.

2. The adaptive intelligent butterfly valve control method with a flow prediction model according to claim 1, characterized in that: In S1, various sensors installed on the butterfly valve and conveying pipeline collect various operating data in real time, with a collection cycle of [missing information]. Set to 0.1-1s; The fluid parameters include: fluid pressure, fluid temperature, fluid viscosity, and fluid density. Valve parameters include: the current angle of the butterfly valve plate and the feedback signal from the butterfly valve actuator; Pipeline and flow parameters include: pipeline inner diameter, real-time flow rate, and historical flow rate; Environmental parameters include: ambient temperature.

3. The adaptive intelligent butterfly valve control method with a flow prediction model according to claim 1, characterized in that: In S2, the improved ILSTM-Attention traffic prediction model consists of an input layer, an ILSTM layer, an attention mechanism layer, a fully connected layer, and an output layer. The input layer takes in preprocessed feature parameters; the ILSTM layer replaces the traditional sigmoid function with the Leaky-ReLU activation function, improving the activation functions of the forget gate, input gate, and output gate; the attention mechanism layer introduces a channel attention mechanism to assign weights to the feature maps output by the ILSTM layer; the fully connected layer fuses and reduces the dimensionality of the features output by the attention mechanism layer, transforming them into a one-dimensional feature vector; and the output layer outputs the predicted traffic flow for future timeframes. The output of the improved ILSTM-Attention traffic prediction model is combined with the prediction correction coefficient to make real-time corrections to the predicted traffic. The calculation formula is as follows: ； in, for t Time prediction Real-time traffic, For the predicted duration, 1 ≤ ≤30min for t Historical traffic sequences up to and before time. They are respectively t The fluid pressure, temperature, viscosity, butterfly valve angle, and fluid density at any given time. D The inner diameter of the pipe. For prediction correction factors; Determined by both real-time prediction deviation and fluid pressure, its calculation formula is as follows: ； in, for t Predicted flow at any time for t Actual flow at any given moment This is a correction factor, with a value ranging from 0.5 to 1.

5. This is the pressure influence coefficient, with a value ranging from 0.9 to 1.

1. P avg This represents the historical average pressure.

4. The adaptive intelligent butterfly valve control method with a flow prediction model according to claim 3, characterized in that: The improved loss function in S2 adds prediction bias weights for two weak intervals to the traditional mean squared error, and the calculation formula is as follows: ； in, N The number of training samples. Q p,i , Q a,i The first i The predicted flow and actual flow for each sample For the first i The butterfly valve angle corresponding to each sample I (·) is an indicator function. The loss weights are for the opening ranges of 15°-20° and 80°-90°, respectively, with values ​​ranging from 1.5 to 2.

5. When training the model, the preprocessed training set is input into the improved ILSTM-Attention traffic prediction model, using an improved loss function. L With the goal of minimizing errors, the Adam optimizer was used for model training, with a learning rate of 0.001-0.01 and 100-500 iterations. During training, model parameters were adjusted in real-time using a validation set to avoid overfitting. After training, the model performance was validated using a test set, and the prediction error was minimized. When the model training is complete, if the prediction error exceeds the threshold, the model structure and parameters are adjusted, and the model is retrained until the requirements are met; among these, This is the prediction deviation threshold.

5. The adaptive intelligent butterfly valve control method with a flow prediction model according to claim 4, characterized in that: In S3, the input to the adaptive fuzzy PID controller is the total deviation. and the rate of change of deviation Total deviation It incorporates the deviation between predicted and actual traffic and sets a weight for the prediction deviation. Weight greater than actual deviation This emphasizes the guiding role of predicted flow and achieves a prediction-first control mode. The specific formula is as follows: ； in, To set the flow rate, These are the weights of the prediction deviation and the actual deviation, respectively. ; By incorporating the influence of fluid viscosity, the characterization accuracy of the deviation change rate is corrected. The specific formula is as follows: ； in, for t The total deviation at time -1 To control the cycle, This represents the historical average fluid viscosity.

6. The adaptive intelligent butterfly valve control method with a flow prediction model according to claim 5, characterized in that: The calculation of adaptive fuzzy PID control parameters integrates fuzzy inference and operating condition compensation. The specific calculation formula is as follows: ； ； ； in, k p ( t ), k i ( t ), k d ( t ) are respectively t The proportional coefficient, integral coefficient, and differential coefficient at time intervals. k p0 , k i0 , k d0 These are the initial control parameters. For parameter adjustment coefficients, This represents the maximum threshold for the rate of change of deviation. Proportion system k p ( t The integral coefficient increases as pressure increases and the butterfly valve angle approaches 45°, thus improving the system's response speed. k i ( t The differential coefficient decreases as the rate of change of deviation and viscosity increase, thus avoiding integral saturation and reducing adjustment oscillations; k d ( t The rate of change of deviation decreases as the absolute value of the deviation increases, and increases as the butterfly valve angle increases, thereby improving the stability of the system and avoiding differential impact.

7. The adaptive intelligent butterfly valve control method with a flow prediction model according to claim 6, characterized in that: In the calculation of the butterfly valve opening adjustment, a specific range compensation is added to the traditional PID control. The specific calculation formula is as follows: ； in, This refers to the butterfly valve opening adjustment amount. This is the opening compensation amount; The calculation formula is: ； in, All are compensation coefficients. for t The current actual opening degree of the butterfly valve plate is measured at all times. 17.5° and 85° are the midpoint angles of the two weak zones, respectively. Butterfly valve target opening The formula for calculation is: ； Set 0°≤ The constraint of ≤90° ensures that the target opening is within the effective adjustment range of the butterfly valve.

8. The adaptive intelligent butterfly valve control method with a flow prediction model according to claim 7, characterized in that: S4 specifically includes the following steps: S41. Control signal conversion: Convert the calculated target opening degree... This is converted into a control signal for the butterfly valve actuator, ensuring precise matching between the control signal and the target opening degree. The specific formula is as follows: ； in, U ( t )for t The actuator control signal at any given time, U min , U max These are the minimum and maximum control signals for the actuator, respectively; S42, Opening Execution: The actuator receives the control signal. U ( t After that, drive the butterfly valve plate to rotate to the target opening degree. The flow rate is regulated by adjusting the fluid flow area; simultaneously, the actuator provides real-time feedback on the actual opening degree. To determine if the butterfly valve is operating normally; S43. Feedback Data Acquisition: After adjustment, the actual flow rate, fluid parameters, and actual butterfly valve opening data are collected in real time by sensors. These data are compared with the predicted flow rate and target butterfly valve opening to calculate the deviation and provide data support for subsequent iterative optimization.

9. The adaptive intelligent butterfly valve control method with a flow prediction model according to claim 8, characterized in that: S5 specifically includes the following steps: S51. Prediction Model Parameter Correction: Calculate Prediction Error ,when At that time, the weight parameters of the improved ILSTM-Attention traffic prediction model are corrected based on the gradient descent method, and the step size is also corrected. The calculation formula is as follows: ； in, The initial correction step size, This represents the largest prediction deviation in history. S52. Iterative Update of Control Parameters: Update initial control parameters based on the rate of change of real-time operating parameters. k p0 , k i0 , k d0 This allows the initial control parameters to be dynamically adjusted as operating conditions drift over a long period, improving the long-term adaptability of the control strategy. Simultaneously, based on the actual application effects of the fuzzy inference rules, the fuzzy rule base is periodically optimized to ensure the rationality of control parameter adjustments. The specific formula is as follows: ； ； ； in, To update the coefficients, ΔP ( t )= P ( t )- P ( t -1) is t Time and t Pressure difference at time -1.

10. The adaptive intelligent butterfly valve control method with a flow prediction model according to claim 9, characterized in that: S5 also includes a fault diagnosis and emergency control mechanism. When butterfly valve jamming, sensor malfunction, or flow deviation continuously exceeding the threshold is detected, it automatically switches to emergency control mode. The formula for calculating the emergency opening adjustment is: ； in, For emergency opening adjustment, This is the emergency adjustment coefficient.

Images