Valve actuator scheduling control method and system based on AI intelligence

Abstract

The invention provides a valve actuator dispatching control method and system based on AI intelligence, and relates to the technical field of industrial automation control, and the method comprises the steps: collecting the running state parameters of a steam turbine in real time, determining the relation between the steam inlet amount and the valve opening degree under the current working condition to obtain a preliminary valve opening degree estimated value, and carrying out dynamic response compensation; the method comprises the following steps: carrying out comprehensive coupling analysis on a theoretical valve opening value and mechanical parameters such as response time, inertial parameters and an opening limit threshold value of a valve actuator, calculating a dynamic adjustment weight and carrying out boundary constraint processing to obtain a processed parameter, and calling a preset valve opening-silicon controlled trigger angle nonlinear mapping model in a single chip microcomputer based on the parameter; and calculating a final silicon controlled trigger angle by inquiring adjacent data points and a linear interpolation method. By fusing intelligent prediction, multi-dimensional adjustment and a closed-loop feedback mechanism, the valve control precision and response speed are improved, and efficient and intelligent operation of industrial automatic control is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial automation control, and particularly to a valve actuator scheduling control method and system based on AI intelligence. Background Art

[0002] In the field of industrial process control, although traditional valve actuator scheduling control technologies can meet basic control requirements, they also have some limitations when facing complex working conditions, dynamically changing scenarios, and high-precision control requirements. For example, some traditional technologies rely on fixed control logics and lack the ability of adaptive learning for the dynamic coupling relationships of multiple variables such as grid load, fuel quality, and steam turbine operating status, resulting in insufficient prediction accuracy of steam inlet volume and lag in valve opening adjustment. For example, in a thermal power plant, when the grid load fluctuates frequently due to the access of new energy, traditional control methods need to rely on the load-steam inlet volume linear relationship preset by humans. If the fuel quality changes synchronously at this time, the non-linear coupling relationship between multiple variables cannot be captured in real time, resulting in an enlarged deviation in the prediction of the steam inlet volume of the steam turbine.

[0003] In addition, the compensation for the mechanical clearance of valves in traditional technologies relies on fixed compensation and lacks a real-time dynamic detection and adaptive adjustment mechanism. After long-term operation, compensation errors are likely to accumulate due to mechanical component wear and environmental temperature changes. For example, in a thermal power plant, traditional control only measures the mechanical clearance once at the initial stage of equipment commissioning and sets a fixed compensation value. As the usage time increases, the clearance of the gear transmission pair expands from the initial 0.05 mm to 0.15 mm due to wear, and the traditional technology does not trigger the compensation mechanism, resulting in a gradually enlarged deviation between the actual valve opening and the command value. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a valve actuator scheduling control method and system based on AI intelligence, which can adaptively adjust the valve opening and improve the control response speed and adjustment accuracy.

[0005] To solve the above technical problems, the technical solution of the present invention is as follows:

[0006] In the first aspect, a valve actuator scheduling control method based on AI intelligence, the method includes:

[0007] Step 1, collect grid load demand data, fuel quality parameters, and real-time energy consumption data of the steam turbine, and use a time series reinforcement prediction model to dynamically predict the final steam inlet volume of the steam turbine;

[0008] Step 2, collect steam turbine operating state parameters in real time, determine the relationship between the steam inlet volume and the valve opening under the current working condition to obtain a preliminary predicted value of the valve opening and perform dynamic response compensation. When the load changes rapidly, start a dynamic feedforward compensation mechanism and perform phase compensation in combination with the response delay characteristics of the valve actuator to obtain the theoretical valve opening value;

[0009] Step 3: Comprehensively and couplingly analyze the theoretical valve opening value with the valve actuator response time, inertia parameters, and mechanical parameters of the opening limit threshold, calculate the dynamic adjustment weight and perform boundary constraint processing to obtain the processed parameters. Then, based on these parameters, call the non-linear mapping model of valve opening - thyristor firing angle preset in the single-chip microcomputer, calculate the final thyristor firing angle through querying adjacent data points and the linear interpolation method. Finally, generate a pulse width modulation signal accordingly, add dead-time compensation for electrical delay, and adjust the amplitude to adapt to the drive power requirement to form an adjustment instruction that conforms to the electrical characteristics of the drive module;

[0010] Step 4: The valve actuator receives the adjustment instruction, drives the valve to perform corresponding actions, and feeds back the valve position signal to the time series enhanced prediction model in real time to form a closed-loop control;

[0011] Step 5: During the closed-loop control process, detect the set threshold of the valve mechanical clearance and automatically trigger the clearance compensation mechanism to correct the opening instruction.

[0012] Furthermore, collect the grid load demand data, fuel quality parameters, and real-time energy consumption data of the steam turbine, and use the time series enhanced prediction model to dynamically predict the final steam inlet volume of the steam turbine, including:

[0013] Collect the grid load demand data, fuel quality parameters, and real-time operation data of the steam turbine in real time, and perform time series alignment to obtain a multi-dimensional input vector;

[0014] Based on the multi-dimensional input vector, establish a time series enhanced prediction model, collect historical operation data as training samples, and input them into the time series enhanced prediction model for training;

[0015] During the training process, the time series enhanced prediction model continuously adjusts its own parameters to learn the collaborative relationship between the grid load, fuel quality, and steam turbine operation status, and obtains the trained time series enhanced prediction model;

[0016] Input the multi-dimensional input vector into the trained time series enhanced prediction model, and dynamically predict the final steam inlet volume of the steam turbine according to the learned load - fuel - steam turbine collaborative relationship.

[0017] Furthermore, collect the steam turbine operation status parameters in real time, determine the relationship between the steam inlet volume and the valve opening under the current working condition to obtain a preliminary estimated valve opening value and perform dynamic response compensation. When the load changes rapidly, start the dynamic feedforward compensation mechanism and perform phase compensation in combination with the response delay characteristics of the valve actuator to obtain the theoretical valve opening value, including:

[0018] Collect the operating state parameters of the steam turbine in real time, including steam temperature, steam pressure, steam turbine speed, power output, ambient temperature, humidity, and atmospheric pressure;

[0019] According to the operating state parameters of the steam turbine, determine the relationship between the steam inlet volume and the valve opening degree that matches the current working condition from the operating data, obtain the preliminary estimated value of the valve opening degree corresponding to the final steam inlet volume, and perform dynamic response compensation on the preliminary estimated value of the valve opening degree;

[0020] When a rapid load change is detected, start the dynamic feedforward compensation mechanism, and perform phase compensation through the response delay characteristic of the valve actuator to obtain the theoretical opening value of the valve.

[0021] Furthermore, when a rapid load change is detected, start the dynamic feedforward compensation mechanism, and perform phase compensation through the response delay characteristic of the valve actuator to obtain the theoretical opening value of the valve, including:

[0022] Determine the immediate adjustment component, the cumulative adjustment component, and the dynamic suppression component according to the current valve opening deviation value to obtain the basic adjustment amount;

[0023] According to the basic adjustment amount, monitor the change rate of the load demand in real time, and generate a feedforward compensation amount according to the change rate at a preset ratio;

[0024] Perform phase compensation using the dynamic lag compensation amount of the valve actuator according to the feedforward compensation amount;

[0025] Fuse the basic adjustment amount, the feedforward compensation amount, and the phase compensation to obtain the theoretical opening value of the valve.

[0026] Furthermore, the determination process of the adjustment instruction includes:

[0027] Perform comprehensive coupling analysis on the theoretical valve opening value and the mechanical parameters of the valve actuator, where the mechanical parameters include the response time, inertia parameters, and opening limit threshold of the valve actuator. According to the response time and inertia parameters, calculate the dynamic adjustment weight of the theoretical opening value, and perform boundary constraint processing on the theoretical opening value in combination with the opening limit threshold to obtain the processed parameters;

[0028] Based on the processed parameters, call the non-linear mapping model of valve opening - thyristor firing angle preset in the single-chip microcomputer. The mapping model takes the theoretical opening value as the input, locates the adjacent data points of the current opening value in the pre-stored discretized mapping relationship table by querying, and uses linear interpolation to calculate the final thyristor firing angle matching the current opening value in real time;

[0029] Generate a corresponding pulse width modulation signal according to the final thyristor firing angle, add dead-time compensation to the pulse signal for the electrical delay in the transmission of the pulse width modulation signal, and at the same time, adjust the amplitude of the pulse signal for power adaptation according to the driving power requirement of the valve actuator to form an adjustment instruction that conforms to the electrical characteristics of the driving module.

[0030] Further, the valve actuator receives the adjustment instruction, drives the valve to perform corresponding actions, and feeds back the valve position signal to the time-series enhanced prediction model in real time to form a closed-loop control, including:

[0031] The valve actuator receives the adjustment instruction, parses the instruction, and transmits the information recognition to the control unit of the actuator;

[0032] The control unit starts the driving device according to the adjustment instruction, sends corresponding electrical signals to the stepping motor, and makes the stepping motor operate;

[0033] During the movement of the valve, use the position sensor on the valve to detect the position of the valve in real time and convert the position information into an electrical signal;

[0034] Transmit the valve position signal to the time-series enhanced prediction model, and compare it with the theoretical opening value of the valve to evaluate whether the current state meets the target;

[0035] The time-series enhanced prediction model dynamically adjusts the instruction to control the thyristor firing angle according to the difference between the valve position and the theoretical opening value of the valve, and continuously optimizes the valve opening to match the target value to form a closed-loop control.

[0036] Further, during the closed-loop control process, detect the set threshold of the valve mechanical clearance and automatically trigger the clearance compensation mechanism to correct the opening instruction, including:

[0037] During the operation of the closed-loop control, collect the data of the valve mechanical state in real time through the sensors of the valve actuator, including the displacement change and the force condition during the opening and closing processes of the valve;

[0038] Judge the size of the mechanical clearance according to the data of the valve mechanical state, based on the displacement difference and the change characteristics of the movement resistance when the valve moves forward and backward;

[0039] Compare the mechanical clearance value with the preset threshold. If the mechanical clearance value ≥ the preset threshold, automatically trigger the clearance compensation mechanism and stop the current conventional control process;

[0040] In the clearance compensation, according to the preset rules, combine the size of the mechanical clearance, the valve type, and the current operating condition factors to correct the original opening instruction.

[0041] In a second aspect, a valve actuator scheduling and control system based on AI intelligence includes:

[0042] A multi-source acquisition module, configured to collect power grid load demand data, fuel quality parameters, and real-time steam turbine energy consumption data in real time, and obtain a predicted value of the final steam inlet volume of the steam turbine;

[0043] A valve mapping module, configured to calculate a theoretical valve opening value according to the predicted final steam inlet volume, the frictional torque of the mechanical actuator, and the gear clearance parameters;

[0044] A firing angle adjustment module, configured to generate a thyristor firing angle adjustment command in real time according to the mapping relationship between the theoretical valve opening value and the valve actuator;

[0045] A closed-loop feedback module, configured to detect the mechanical clearance threshold of the valve according to the thyristor firing angle adjustment command,

[0046] to correct the opening command;

[0047] A clearance compensation module, configured to automatically trigger a clearance compensation mechanism according to the mechanical clearance threshold of the valve.

[0048] In a third aspect, a computing device includes:

[0049] One or more processors;

[0050] A storage device, configured to store one or more programs, and when the one or more programs are executed by the one or more processors, enable the one or more processors to implement the method.

[0051] In a fourth aspect, a computer-readable storage medium stores a program, and when the program is executed by a processor, the method is implemented.

[0052] The above solutions of the present invention at least include the following beneficial effects:

[0053] By analyzing and processing the power grid load demand data, fuel quality parameters, and real-time steam turbine energy consumption data, it is possible to dynamically predict the final steam inlet volume of the steam turbine, adapt to complex and changeable working conditions, and provide a basis for valve opening adjustment. According to the predicted result, the theoretical valve opening value is determined, and combined with the real-time adjustment of the thyristor firing angle by the single-chip microcomputer, a precise valve opening adjustment command is generated, enabling the valve to quickly and accurately respond to the control demand, effectively improving the adjustment efficiency and reducing the adjustment time. While the valve actuator drives the valve to act upon receiving the adjustment command, it simultaneously feeds back the valve position signal to the time series reinforcement prediction model in real time to form a closed-loop control. It can continuously adjust the command dynamically according to the difference between the actual valve position and the theoretical opening value to optimize the valve opening, thereby improving the stability.

[0054] During the closed-loop control process, the set threshold of the valve mechanical clearance can be detected, and the clearance compensation mechanism can be automatically triggered. By real-time monitoring of the valve mechanical state data, including the displacement changes and force conditions during the valve opening and closing processes, the size of the mechanical clearance is judged, and then the original opening command is corrected according to the preset rules combined with various factors. Ensure the accuracy of the valve opening, avoid control errors caused by mechanical clearances, and improve the control accuracy. When determining the theoretical valve opening value, according to various operating state parameters of the steam turbine, including steam temperature, pressure, rotational speed, power output, as well as environmental temperature, humidity, and atmospheric pressure. By matching the relationship between the steam intake and the valve opening under the current operating conditions from the operating data, and starting the dynamic feedforward compensation mechanism and phase compensation during rapid load changes, it can better adapt to complex and changeable operating conditions. Through precise control and timely clearance compensation, unnecessary actions and wear of the valve actuator are reduced, the maintenance cost of the equipment is lowered, and the service life of the equipment is extended. At the same time, the efficient regulation performance contributes to energy utilization efficiency and reduces energy consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 is a schematic flow chart of a valve actuator scheduling control method based on AI intelligence provided by an embodiment of the present invention.

[0056] Figure 2 is a schematic diagram of a valve actuator scheduling control system based on AI intelligence provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0057] Hereinafter, exemplary embodiments of the present disclosure will be described in more detail with reference to the drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be completely conveyed to those skilled in the art.

[0058] As Figure 1 shown, an embodiment of the present invention proposes a valve actuator scheduling control method based on AI intelligence, and the method includes the following steps:

[0059] Step 1, collect grid load demand data, fuel quality parameters, and real-time energy consumption data of the steam turbine, and use a time series reinforcement prediction model to dynamically predict the final steam intake of the steam turbine;

[0060] Step 2, collect the operating state parameters of the steam turbine in real time, determine the relationship between the steam intake and the valve opening under the current operating conditions to obtain a preliminary predicted valve opening value and dynamically respond to compensation, and start the dynamic feedforward compensation mechanism and perform phase compensation in combination with the response delay characteristics of the valve actuator during rapid load changes, so as to obtain the theoretical valve opening value;

[0061] Step 3: Comprehensively and couplingly analyze the theoretical valve opening value with the response time of the valve actuator, the inertia parameter, and the mechanical parameters such as the opening limit threshold, calculate the dynamic adjustment weight and perform boundary constraint processing to obtain the processed parameters, then based on these parameters, call the non-linear mapping model of valve opening - thyristor firing angle preset in the single-chip microcomputer, calculate the final thyristor firing angle through querying adjacent data points and the linear interpolation method, and finally generate a pulse width modulation signal accordingly, add dead-time compensation for electrical delay, and adjust the amplitude to adapt to the drive power requirement to form an adjustment instruction that conforms to the electrical characteristics of the drive module;

[0062] Step 4: The valve actuator receives the adjustment instruction, drives the valve to perform corresponding actions, and feeds back the valve position signal to the time series enhanced prediction model in real time to form a closed-loop control;

[0063] Step 5: During the closed-loop control process, detect the set threshold of the valve mechanical clearance and automatically trigger the clearance compensation mechanism to correct the opening instruction.

[0064] In the embodiment of the present invention, by collecting the power grid load demand data, fuel quality parameters, and real-time steam turbine energy consumption data, the key factors affecting the steam inlet volume of the steam turbine can be comprehensively captured. Combined with the time series enhanced prediction model, the complex relationships between factors can be dynamically learned to adapt to the changes under different working conditions. The time series enhanced prediction model has the ability of dynamic prediction and can quickly adjust the predicted value of the steam inlet volume according to the change of real-time data. When the load suddenly changes, it can quickly respond and output a new prediction result, accelerating the entire control response speed and reducing the fluctuations and energy waste caused by untimely adjustment of the steam inlet volume.

[0065] Determine the theoretical valve opening value according to the final steam inlet volume and working characteristics of the steam turbine, so that the valve opening closely matches the actual operating state of the steam turbine. Different steam inlet volumes and working characteristics require different valve openings, ensuring that the valve adjustment can meet the operating requirements of the steam turbine under various working conditions, improving the operating efficiency and stability. The accurate theoretical valve opening value helps to optimize the steam flow and energy conversion process inside the steam turbine. Make the steam enter the steam turbine with a determined flow rate and pressure, reduce steam leakage and energy loss, improve the energy conversion efficiency of the steam turbine, and thus reduce energy consumption and improve economy.

[0066] The single-chip microcomputer adjusts the thyristor trigger angle in real time according to the theoretical opening value of the valve and the mechanical parameters of the valve actuator to generate an adjustment instruction. This method can control the opening change of the valve, realize the fine adjustment of the valve. More accurately, the valve opening is adjusted to the theoretical value, improving the control accuracy and reducing the control error. The single-chip microcomputer has programmability and real-time processing ability, and can quickly adjust the thyristor trigger angle according to different theoretical opening values and mechanical parameters of the valve. This makes it have stronger flexibility and adaptability when facing different types of valve actuators and complex working conditions changes, and can carry out parameter adjustment and function expansion.

[0067] The valve actuator receives the adjustment instruction to drive the valve to act, and in real time feeds back the valve position signal to form a closed-loop control. By continuously comparing the actual position of the valve and the theoretical opening value, the deviation can be detected and adjusted in time, so that the valve opening is always maintained in an ideal state, improving the stability and reliability. The closed-loop control mechanism can automatically adjust the control strategy according to the actual operation situation of the valve. During the valve execution process, if there are changes in resistance and mechanical wear, the adjustment instruction can be sensed and adjusted in time through the feedback signal, realizing adaptive adjustment to ensure that the valve can always accurately respond to the control requirements.

[0068] During the closed-loop control process, the set threshold of the valve mechanical clearance is detected, and the clearance compensation mechanism is automatically triggered, which can effectively solve the control error problem caused by the mechanical clearance. As the equipment operation time increases, the mechanical clearance may gradually increase, affecting the accuracy of the valve opening. Through the clearance compensation mechanism, the opening instruction can be corrected in time to ensure that the valve control accuracy remains stable during the long-term operation process. Timely clearance compensation can reduce the impact and wear during the valve execution process, reduce the failure rate of the equipment. By correcting the opening instruction, the valve operates in a reasonable state, avoiding abnormal vibration and uneven force caused by excessive clearance, thereby prolonging the service life of the valve actuator and related equipment and reducing the equipment maintenance cost.

[0069] In a preferred embodiment of the present invention, step 1 above, collecting the power grid load demand data, fuel quality parameters and real-time energy consumption data of the steam turbine, and using the time series reinforcement prediction model to dynamically predict the final steam inlet volume of the steam turbine, may include:

[0070] Step 110, collecting the power grid load demand data, fuel quality parameters and real-time operation data of the steam turbine in real time, and performing time series alignment to obtain a multi-dimensional input vector;

[0071] Step 111, according to the multi-dimensional input vector, establish a time series reinforcement prediction model, and collect historical operation data as training samples, and input them into the time series reinforcement prediction model for training;

[0072] Step 112, during the training process, the time-series reinforcement prediction model learns the synergy relationship among the grid load, fuel quality, and steam turbine operating status by continuously adjusting its own parameters, and obtains the trained time-series reinforcement prediction model;

[0073] Step 113, input the multi-dimensional input vector into the trained time-series reinforcement prediction model, and dynamically predict the final steam admission volume of the steam turbine according to the learned load-fuel-steam turbine synergy relationship.

[0074] In the embodiment of the present invention, real-time data collection is carried out according to various sensors distributed on the power grid, fuel supply system, and steam turbine. At the power grid end, a load sensor is used to collect the power demand values at different times as the power grid load demand data. For fuel, key indicators such as the composition and calorific value of the fuel are obtained through quality detection equipment to form fuel quality parameters. For the steam turbine, temperature sensors, pressure sensors, and speed sensors are used to collect temperature, pressure, and speed operation data, which constitute the real-time operation data of the steam turbine. Since the collection frequencies and time points of different types of data may vary, in order to ensure the time consistency of the data, a time series alignment operation is required. First, a unified time reference is determined, such as a time stamp in seconds. Then, for each collected data point, it is mapped to the unified time series according to the corresponding time stamp. For missing data points, linear interpolation and spline interpolation methods can be used for filling; for duplicate data points, duplicate removal processing is performed. After processing, all the data is integrated into a multi-dimensional input vector, where each dimension represents a specific type of data.

[0075] According to the characteristics of the multi-dimensional input vector and the requirements of the problem, a time-series reinforcement prediction model is constructed, which can determine an appropriate neural network architecture, including recurrent neural networks (RNN) and their variants (long short-term memory network LSTM, gated recurrent unit GRU). At the same time, combining the idea of reinforcement learning, a reward mechanism is introduced to encourage the time-series reinforcement prediction model to optimize in the direction of more accurate prediction. A large amount of historical operation data is collected as training samples, and these historical data should cover the grid load demand, fuel quality, and steam turbine operation status under different working conditions to ensure that the time-series reinforcement prediction model can learn the laws in various situations. The collected historical data is normalized to improve the training effect and convergence speed of the time-series reinforcement prediction model. The normalized training samples are input into the time-series reinforcement prediction model for training. During the training process, the time-series reinforcement prediction model makes predictions based on the input data, and compares the prediction results with the actual steam turbine steam admission volume to calculate the error. Based on the reward mechanism of reinforcement learning, the time-series reinforcement prediction model is given corresponding rewards or punishments according to the size of the error. The time-series reinforcement prediction model continuously adjusts its own parameters through optimization algorithms (stochastic gradient descent, Adam) to reduce the error and maximize the reward. In each training iteration, the time-series reinforcement prediction model attempts to find the synergistic relationship between the grid load, fuel quality, and steam turbine operation status, that is, how they interact with each other and ultimately determine the steam admission volume of the steam turbine. As the training progresses, the parameters of the model gradually converge, and the learning of this synergistic relationship becomes more and more accurate, and finally the trained time-series reinforcement prediction model is obtained.

[0076] The multi-dimensional input vector obtained by time-series alignment is input into the trained time-series reinforcement prediction model. The time-series reinforcement prediction model analyzes and processes the input data according to the previously learned load-fuel-steam turbine synergistic relationship. The time-series reinforcement prediction model uses the internal neural network structure and learned parameters to dynamically predict the final steam admission volume of the steam turbine. The prediction results will be updated with the real-time changes of the input data to adapt to the dynamic changes of the grid load, fuel quality, and steam turbine operation status, so as to provide an accurate basis for the valve actuator scheduling control.

[0077] Suppose there is a thermal power plant supplying electricity to a specific area. Intelligent load sensors are deployed on the grid side to use AI algorithms to filter abnormal data in real time and collect grid load demand data every 15 minutes. Through analysis, it is found that the electricity consumption peak is from 8 am to 10 am every day, with an increase in load demand, and it enters the low valley period at night. In terms of fuel supply, coal is sampled and analyzed every hour. With the help of AI image recognition and spectral analysis technologies, key parameters such as the calorific value and sulfur content of coal are quickly obtained. When it is detected that a certain batch of coal has a high calorific value, AI automatically records and analyzes the potential improvement effect on combustion efficiency. In the steam turbine area, sensors collect steam pressure and temperature operation data at a frequency of once per second, eliminate invalid data caused by interference in real time, and according to the characteristics of the power generation period, intelligently mark the change patterns of steam pressure and temperature during the power generation peak.

[0078] Facing data collected at different frequencies, AI automatically selects the final time series alignment strategy. For missing data points, instead of being limited to traditional interpolation methods, a deep learning-based Seq2Seq model is used for filling. This model can generate more realistic values based on the semantic association of the front and back data; for duplicate data, the AI clustering algorithm quickly identifies and de-duplicates them, and finally integrates them into a multi-dimensional input vector. When constructing a time series reinforcement prediction model, through AI automated architecture search technology, from a large number of neural network structure combinations, the GRU network variant that best fits the data characteristics of this thermal power plant is accurately selected. During training, AI dynamically adjusts the rewards of reinforcement learning. When the complex relationship between high load, high calorific value coal, and the steam inlet volume of the steam turbine is accurately captured, high rewards are given; if there is a prediction deviation, combined with Shapley value analysis, the key influencing factors of the decision are located, and the parameters are optimized accordingly. Using the AdamW optimization algorithm, AI continuously adjusts the parameters. After hundreds of iterations, it masters the coordination law among grid load, coal quality, and the operating state of the steam turbine. At 9 am the next day, after the multi-dimensional data collected in real time is processed, it is input into the trained time series reinforcement prediction model. Considering the current peak electricity consumption load, the combustion advantage of high calorific value coal, and the real-time operating state of the steam turbine, it quickly predicts that the final steam inlet volume of the steam turbine is 50 cubic meters per minute. Subsequently, based on the prediction results, the AI decision-making system combines the response data and mechanical characteristic parameters of the valve actuator, and uses the reinforcement learning algorithm to formulate the final adjustment strategy - first quickly adjust the valve opening with a driving power of 70%, and then fine-tune with a power of 30% when the steam inlet volume is close to the target value to ensure that the steam inlet volume reaches the target value smoothly, which not only meets the grid load demand but also achieves the balance of energy consumption and stable operation of the equipment.

[0079] By dynamically predicting the final steam inlet volume of the steam turbine, the steam inlet volume can be adjusted in real time according to the grid load demand and fuel quality, enabling the steam turbine to operate efficiently under different working conditions, reducing energy waste, and improving power generation efficiency. The load-fuel-steam turbine coordination relationship learned by the time-series reinforcement prediction model can accurately predict the steam inlet volume, avoid the unstable operation of the steam turbine caused by unreasonable steam inlet volume, and enhance the stability and reliability of the power generation system. Dynamically adjusting the steam inlet volume according to the fuel quality parameters can make full use of fuels with different qualities, improve the fuel utilization rate, and reduce the power generation cost. Real-time data collection and dynamic prediction can quickly respond to changes in grid load demand and fuel quality, timely adjust the steam inlet volume, and improve flexibility and adaptability.

[0080] In a preferred embodiment of the present invention, in step 2 above, the operating state parameters of the steam turbine are collected in real time, the relationship between the steam inlet volume and the valve opening degree under the current working condition is determined to obtain a preliminary predicted value of the valve opening degree and a dynamic response compensation is performed. When the load changes rapidly, a dynamic feedforward compensation mechanism is started and phase compensation is performed in combination with the response delay characteristics of the valve actuator, so as to obtain the theoretical valve opening degree value, which may include:

[0081] Step 220, collect the operating state parameters of the steam turbine in real time, including steam temperature, steam pressure, steam turbine speed, power output, ambient temperature, humidity, and atmospheric pressure;

[0082] Step 221, according to the operating state parameters of the steam turbine, determine the relationship between the steam inlet volume and the valve opening degree that matches the current working condition from the operating data, obtain a preliminary predicted value of the valve opening degree corresponding to the final steam inlet volume, and perform a dynamic response compensation on the preliminary predicted value of the valve opening degree;

[0083] Step 222, when it is detected that the load changes rapidly, start a dynamic feedforward compensation mechanism, and perform phase compensation through the response delay characteristics of the valve actuator to obtain the theoretical valve opening degree value.

[0084] In an embodiment of the present invention, intelligent sensors equipped with edge AI chips are respectively deployed at the steam inlet pipeline, the rotating shaft, and the power generation output end of the steam turbine. A temperature sensor and a pressure sensor are installed on the steam inlet pipeline of the steam turbine to measure the steam temperature and steam pressure in real time. These sensors should have the characteristics of high sensitivity and fast response to accurately capture the instantaneous changes of steam parameters. A speed sensor is installed on the shaft of the steam turbine to monitor the speed of the steam turbine in real time through the principle of magnetoelectric induction or photoelectric induction. At the same time, a power sensor is installed at the output end of the generator to measure the power output of the steam turbine to reflect the actual work capacity.

[0085] Install temperature and humidity sensors and atmospheric pressure sensors in the steam turbine engine room to measure the ambient temperature, humidity, and atmospheric pressure respectively. These environmental parameters will affect the operating efficiency and performance of the steam turbine, so real-time monitoring is required. Use a data acquisition card or a programmable logic controller (PLC) as the core acquisition device. The data acquisition card features a high sampling rate and multi-channel input, suitable for high-speed data acquisition; the PLC has the advantages of high reliability and strong anti-interference ability, suitable for the industrial field environment. Connect the output signals of each sensor to the corresponding input channels of the data acquisition device. Write a data acquisition program using programming languages (Python, LabVIEW). The main functions of the program include sensor signal acquisition, filtering, calibration, and storage. Filter the collected raw signals to remove noise interference; calibrate the signals according to the calibration parameters of the sensors and convert them into actual physical quantity values; finally, store the processed data in the database.

[0086] Clean the data stored in the database to remove outliers and missing values. Outliers may be caused by sensor failures or interference, and statistical analysis methods ( criteria) can be used for identification and elimination; for missing values, interpolation methods (linear interpolation, spline interpolation) can be used for filling. Normalize the data by mapping the data of different parameters to the same value range ([0, 1]) to eliminate the influence of data dimensions and scales. Extract characteristic parameters that can reflect the operating conditions from the normalized operating data, including steam temperature, pressure, and steam turbine speed. Use a clustering analysis method (such as K-means clustering) to classify the stored data according to the operating condition characteristics to obtain different operating condition categories. Calculate the similarity with each operating condition category based on the real-time collected steam turbine operating state parameters. Euclidean distance and cosine similarity methods can be used for calculation, and determine the operating condition category with the highest similarity as the current operating condition.

[0087] For each operating condition category, determine a suitable data fitting method according to the corresponding relationship between the steam inlet volume and the valve opening in the operating data. Polynomial regression, linear regression, and neural network methods can be used. Polynomial regression is suitable for cases where the data shows a non-linear relationship; linear regression is simple and easy to understand with high calculation efficiency; the neural network has a powerful non-linear mapping ability and can handle complex relationships. Divide the operating data of each operating condition category into a training set and a validation set, use the training set to train the selected fitting method and adjust the parameters to accurately fit the relationship between the steam inlet volume and the valve opening. Use the validation set to verify the fitting method and evaluate the accuracy and generalization ability of the model.

[0088] Substitute the final steam intake into the steam intake-valve opening relationship under the current operating conditions to calculate the preliminary estimated value of the valve opening. Analyze the dynamic characteristics of the steam turbine system, including the influence of inertia and delay factors on the valve opening response. Use a proportional-integral-derivative (PID) controller for dynamic response compensation. The PID controller calculates the compensation amount based on the current valve opening error, the integral of the error, and the derivative of the error, and optimizes the compensation effect by adjusting the proportional, integral, and derivative coefficients. Add the calculated compensation amount to the preliminary estimated value of the valve opening to obtain the valve opening value after dynamic response compensation. During the actual operation process, continuously monitor the actual response of the valve opening, and adjust the compensation parameters according to the feedback information to ensure the finalization of the compensation effect. Calculate the change rate of the steam turbine power output in real time, that is, the change amount of power per unit time. The difference method can be used to calculate the power change rate. For example, calculate the difference between the power outputs at the current moment and the previous moment, and divide it by the time interval. Set the threshold value of the load change rate according to the design parameters, operating experience, and safety requirements of the steam turbine. When the calculated load change rate > this threshold value, it is determined that the load changes rapidly, and the dynamic feedforward compensation mechanism is immediately triggered.

[0089] Through the analysis of the operation data, establish a relationship model between the load change and the feedforward compensation amount of the valve opening. The transfer function and state space model methods can be used for modeling to describe the influence law of the load change on the valve opening. When a rapid load change is detected, calculate the feedforward compensation amount according to the established feedforward compensation model and the current load change amount. The calculation of the feedforward compensation amount should consider the factors of the direction, amplitude, and speed of the load change to ensure that the influence of the load change can be compensated in a timely and accurate manner. Superimpose the calculated feedforward compensation amount on the valve opening value after dynamic response compensation to obtain the preliminary compensated opening value. By testing the response delay characteristics of the valve actuator, record the time delay and phase shift from the moment the valve receives the control signal to the actual opening change. The step response test and frequency response test methods can be used for testing to obtain the dynamic characteristic parameters of the valve actuator. According to the response delay characteristics of the valve actuator, design a phase compensation algorithm. The lead-lag compensator and Smith predictor methods can be used for phase compensation to adjust the phase of the valve opening signal to eliminate the influence of the response delay. Input the preliminary compensated opening value into the phase compensation algorithm to calculate the valve opening value after phase compensation, and use this value as the theoretical opening value of the valve. During the actual operation process, continuously monitor the operating state of the actual opening of the valve, and adjust the phase compensation parameters according to the feedback information to ensure the effect of the phase compensation.

[0090] Suppose that during the operation of a thermal power plant, the steam turbine supplies power and heat to the surrounding area simultaneously. Sensors collect operation data in real time and transmit it to the data processing center, including parameters such as steam temperature (500 °C), steam pressure (10 MPa), steam turbine speed (3000 revolutions per minute), power output (50 MW), ambient temperature (25 °C), humidity (60%), and atmospheric pressure (101 kPa). The data processing center uses AI-driven operating condition matching to quickly identify similar operating conditions. The AI system discovers that the current operating condition is highly similar to a certain historical operating state - in this historical condition, when the steam inlet flow rate is 100 t / h, the valve opening is 70%. Combining with the predicted value of the current steam inlet flow rate (105 t / h), the AI calculates the preliminary predicted value of the valve opening to be 73% based on the reinforcement learning model. Subsequently, by analyzing the change trend of the steam flow rate, the compensation coefficient is dynamically adjusted to optimize the valve opening to 74% to more accurately match the current operating condition. Suddenly, the electricity demand in the surrounding area surges, and the power output of the steam turbine quickly rises from 50 MW to 60 MW, and the power change rate > the set threshold (5 MW / s). The AI determines it as a rapid load change event. Based on the reinforcement learning strategy, the dynamic feedforward compensation mechanism is immediately activated, and it is obtained that an additional 3% valve opening needs to be increased, so that the current valve opening is adjusted to 77%.

[0091] Due to a 2-second response delay in the valve actuator, the AI calculates the power change trend within the next 2 seconds based on the predictive control algorithm (MPC model predictive control) and introduces a phase compensation mechanism to increase the valve opening by an additional 1% to ensure the stability of the steam turbine output. Finally, the theoretical valve opening value of 78% is output and sent to the actuator for adjustment.

[0092] By collecting various operating state parameters of the steam turbine in real time, performing operating condition matching and dynamic response compensation, the relationship between the valve opening and the steam inlet flow rate can be determined more accurately, thereby improving the accuracy of valve opening control and enabling the steam turbine to better meet the demand for the final steam inlet flow rate. The dynamic feedforward compensation mechanism and phase compensation enable a reaction to be made when a rapid load change is detected, adjust the valve opening in advance, reduce the adjustment time, and improve the response speed and stability. According to the various operating state parameters and dynamic characteristics of the steam turbine, the steam turbine can maintain good operating performance under different operating conditions, improve the power generation efficiency, reduce energy consumption, and extend the service life of the equipment. By adjusting the valve opening in real time according to different operating conditions and environmental conditions, it has stronger adaptability and can cope with various complex actual situations.

[0093] In another preferred embodiment of the present invention, when a rapid load change is detected, the dynamic feedforward compensation mechanism is activated, and phase compensation is performed through the response delay characteristic of the valve actuator to obtain the theoretical opening value of the valve, which may include:

[0094] Determine the immediate adjustment component, the cumulative adjustment component, and the dynamic suppression component according to the current valve opening deviation value to obtain the basic adjustment amount;

[0095] According to the basic adjustment amount, monitor the change rate of the load demand in real time, and generate a feed-forward compensation amount according to the change rate at a preset ratio;

[0096] According to the feed-forward compensation amount, perform phase compensation using the dynamic lag compensation amount of the valve actuator;

[0097] Fuse the basic adjustment amount, the feed-forward compensation amount, and the phase compensation to obtain the theoretical opening value of the valve.

[0098] In the embodiment of the present invention, the actual opening of the current valve is obtained in real time through a sensor, compared with the target opening, and the valve opening deviation value is calculated . This deviation value reflects the gap between the current valve opening and the expected opening. is the immediate adjustment component, which is obtained by multiplying the deviation value at the current moment by the proportionality coefficient . The magnitude of determines the response speed to the deviation, The larger it is, the more rapid the response to the deviation, but it may lead to a decrease in stability; on the contrary, The smaller it is, the slower the response, but the stability may be better. Let = 0.8. When the valve opening deviation value at a certain moment = 5, then the immediate adjustment component × is 4. According to this calculation result, the valve opening will be adjusted quickly. Since = 0.8 is relatively moderate, it can not only ensure a relatively fast response speed to the deviation, but also will not cause instability due to too large a coefficient. is the cumulative adjustment component, which is the integral of the valve opening deviation value over a period of time, reflecting the cumulative effect of the deviation. is the integration coefficient. The role of integration is to eliminate the steady-state error. As time goes by, the cumulative adjustment component will be continuously adjusted until the deviation is completely eliminated. Suppose = 0.1. If within a period of time (set to 10 minutes), the integral result of the valve opening deviation value = 20 (the specific calculation of the integral will be determined according to changing with time), then the cumulative adjustment component = 2. As time goes by, this cumulative adjustment component will continue to act, gradually eliminating the steady-state error and helping the valve opening to finally stabilize at the target value.

[0099] is the dynamic suppression component, which calculates the change rate of the deviation value and then multiplies it by the differential coefficient to obtain it. The larger the [[value]], the stronger the suppression effect on the deviation change, which can effectively reduce the overshoot and reach the stable state faster. Assume = 0.3. When the change rate of the valve opening deviation value at a certain moment = 3, the dynamic suppression component is 0.9. This component will dynamically suppress the valve opening adjustment according to the speed of the deviation change. For example, when the valve opening changes too fast, it will play a buffering role, reduce the overshoot, and reach the stable state faster. By calculating and superimposing these three components, the basic adjustment amount is obtained. The load demand data of the power grid is monitored in real time, and the change rate of the load demand is calculated. This can be achieved by analyzing the load demand data at different times and calculating the ratio of the difference to the time interval. The feedforward compensation amount is . Among them is a preset proportional coefficient, which is determined according to the characteristics and actual operation experience. When the change rate of the load demand is large, the feedforward compensation amount will increase accordingly, and the valve opening will be adjusted in advance to quickly respond to the change of the load and reduce the response delay. Assume = 0.2. When the change rate of the power grid load demand = 10 (assuming the load demand unit is kilowatt / minute) is monitored in real time, then the feedforward compensation amount is 2. When the change rate of the load demand is large, the feedforward compensation amount will increase accordingly, and the valve opening will be adjusted in advance to respond to the load change in time and reduce the response delay.

[0100] There is a certain delay, that is, dynamic hysteresis, when the valve actuator responds to the control signal. Through the characteristic analysis and experimental testing of the valve actuator, the dynamic hysteresis compensation amount is determined. This involves the research on the action time and response speed parameters of the valve actuator. The phase compensation amount is obtained by using the dynamic hysteresis compensation amount of the valve actuator for phase compensation, and it is . Among them is the coefficient related to the dynamic hysteresis of the valve actuator, is a time constant, which reflects the time characteristic of the dynamic hysteresis. This phase compensation amount is used to adjust the valve opening deviation caused by the delay of the valve actuator. Let = 0.15, = 2 (the time unit is seconds). At a certain moment = 3 seconds, the phase compensation amount is 0.03345. The phase compensation amount is used to correct the opening deviation caused by the dynamic lag of the valve actuator. Combining with specific and values, the valve opening can be adjusted. The calculated basic adjustment amount, feedforward compensation amount and phase compensation amount are fused to finally obtain the theoretical opening value of the valve .

[0101] The immediate adjustment component in the formula can quickly respond according to the current deviation, so that the valve opening can be adjusted towards the target value in time; the cumulative adjustment component eliminates the steady-state error and ensures that the valve opening is finally stable; the dynamic suppression component reduces overshoot. The three work together to make the valve opening adjustment accurate and fast. The feedforward compensation amount acts in advance according to the change rate of the load demand, pre-adjusts the valve opening when the load changes, without waiting for the deviation to appear, shortens the response time, effectively reduces the fluctuation caused by the load mutation, and improves the stability. The phase compensation term corrects the dynamic lag of the valve actuator to ensure that the actual valve opening is consistent with the theoretical expectation, avoids the accumulation of control errors caused by delay, and improves the control accuracy. The theoretical opening value of the valve is calculated by fusing the basic adjustment amount, feedforward compensation amount and phase compensation amount, comprehensively considering the valve state, load change and actuator characteristics, and forming an adaptive control system. Whether in the steady state or dynamic working conditions, the final control of the valve opening can be realized, the energy consumption is reduced, the equipment life is extended, and the stable and efficient operation of industrial production is guaranteed.

[0102] In a preferred embodiment of the present invention, in the above step 3, the theoretical opening value of the valve is comprehensively coupled and analyzed with the response time of the valve actuator, the inertia parameter and the opening limit threshold mechanical parameter, the dynamic adjustment weight is calculated and boundary constraint processing is performed to obtain the processed parameter, and then based on this parameter, the non-linear mapping model of valve opening - thyristor trigger angle preset in the single-chip microcomputer is called, and the final thyristor trigger angle is calculated by querying adjacent data points and the linear interpolation method. Finally, a pulse width modulation signal is generated accordingly, adding dead-time compensation for electrical delay and adjusting the amplitude to adapt to the drive power demand, forming an adjustment instruction that conforms to the electrical characteristics of the drive module, which may include:

[0103] In an embodiment of the present invention, during the operation of the valve control system, a series of key data are acquired. Through the scheduling instructions of the control system, the theoretical opening value of the valve is determined. Using a time measurement device, the time from when the valve actuator receives the instruction to when it starts to move is recorded multiple times. Through statistics and analysis, the response time of the valve actuator is obtained. The inertia parameters of the valve are found from the valve design documents and product manuals, which are related to the mass, size, and structural characteristics of the valve. The opening limit threshold is preset to ensure that the valve operates within a safe and reasonable opening range and to avoid over-opening or over-closing. The collected mechanical parameters, namely the theoretical opening value of the valve, the response time of the valve actuator, the inertia parameters, and the opening limit threshold, are considered comprehensively. Different weights are assigned to them according to the importance of each parameter to valve control. For example, since the theoretical opening value of the valve plays a key role in the final control target, a relatively high weight is assigned; the response time of the valve actuator reflects the action speed of the actuator and affects the timeliness of control, so a certain weight is also given; the inertia parameters reflect the inertia magnitude of the valve during movement and affect the adjustment accuracy of the valve, and corresponding weights are assigned; the opening limit threshold is an important parameter to ensure system safety, and an appropriate weight is also assigned. These parameters and their corresponding weights are combined to obtain a dynamically adjusted weight. To ensure that the dynamically adjusted weight is within a reasonable range, boundary constraint processing is required. According to the actual operation conditions and safety requirements of the valve control system, a maximum weight value and a minimum weight value are set. If the result of the dynamically adjusted weight > the maximum weight value, it is adjusted to the maximum weight value; if < the minimum weight value, it is adjusted to the minimum weight value. After such processing, a processed parameter is obtained, which takes into account the influences of various mechanical parameters and ensures the rationality of the value.

[0104] A non - linear mapping model of valve opening - thyristor firing angle is pre - established in the single - chip microcomputer. At different valve openings, the corresponding thyristor firing angles are recorded. During the valve opening process, various possible working conditions should be simulated, including different pressure, temperature, and flow conditions, to ensure that the acquired data is representative. Then, analyze and process these data to find the internal relationship between the valve opening and the thyristor firing angle. The method of data fitting can be used to determine a suitable mathematical function to approximately describe this relationship, thus establishing a non - linear mapping model of valve opening - thyristor firing angle. Based on the obtained processed parameters, call the pre - established non - linear mapping model of valve opening - thyristor firing angle in the single - chip microcomputer. Since the processed parameters may not exactly correspond to a certain data point in the non - linear mapping model of valve opening - thyristor firing angle, it is necessary to find two adjacent data points to the processed parameters. In the data of the non - linear mapping model of valve opening - thyristor firing angle, find a data point that is smaller than and closest to the processed parameter, and a data point that is larger than and closest to the processed parameter. Then, use the method of linear interpolation. According to the valve openings and thyristor firing angles corresponding to these two adjacent data points, estimate the final thyristor firing angle corresponding to the processed parameter. Specifically, assume that between these two adjacent data points, the valve opening and the thyristor firing angle are linearly related, and calculate the approximate value of the final thyristor firing angle through proportional calculation.

[0105] According to the final thyristor firing angle, use the timer function of the single-chip microcomputer to generate a Pulse Width Modulation (PWM) signal. The timer counts time according to a certain period. Within each period, the duration of the high level is determined according to the final thyristor firing angle. The larger the firing angle, the longer the duration of the high level; the smaller the firing angle, the shorter the duration of the high level. By continuously repeating this process, a continuous PWM signal can be generated, and its duty cycle corresponds to the final thyristor firing angle. Due to the delay in the electrical system, in order to avoid short-circuit problems when the thyristor switches states, a dead time needs to be added to the PWM signal. In the timer setting of the single-chip microcomputer, when the PWM signal transitions from high level to low level, a period of time is delayed before outputting the low level; when transitioning from low level to high level, the same period of time is delayed before outputting the high level. This delayed time is the dead time, and its size is adjusted according to the actual electrical system characteristics. By adding the dead time, the electrical delay can be effectively compensated. According to the electrical characteristics and power requirements of the valve actuator drive module, the amplitude of the generated PWM signal is adjusted. If the drive module requires a larger signal amplitude while the PWM signal output by the single-chip microcomputer is smaller, an amplifier is needed to amplify the signal; if the drive module requires a smaller signal amplitude while the signal amplitude output by the single-chip microcomputer is larger, an attenuator or a level conversion circuit is needed to attenuate the signal. By adjusting the amplitude, the PWM signal can meet the power requirements of the drive module, ensuring that the valve actuator can work properly. After processing, the obtained PWM signal is the adjustment instruction that conforms to the electrical characteristics of the drive module. This adjustment instruction is transmitted to the drive module of the valve actuator through the electrical circuit. The drive module controls the conduction and cutoff of the thyristor according to the adjustment instruction, thereby precisely adjusting the opening of the valve and achieving precise control of the entire valve control system.

[0106] In another preferred embodiment of the present invention, in the above step 3, the determination process of the adjustment instruction includes:

[0107] Step 3330: Perform comprehensive coupling analysis on the theoretical valve opening value and the mechanical parameters of the valve actuator. The mechanical parameters include the response time, inertia parameter, and opening limit threshold of the valve actuator. According to the response time and inertia parameter, calculate the dynamic adjustment weight of the theoretical opening value, and perform boundary constraint processing on the theoretical opening value in combination with the opening limit threshold to obtain the processed parameters;

[0108] Step 3331: Based on the processed parameters, call the non-linear mapping model of valve opening - thyristor firing angle preset in the single-chip microcomputer. The mapping model takes the theoretical opening value as the input, locates the adjacent data points of the current opening value in the table by querying the pre-stored discretized mapping relationship table, and uses linear interpolation to calculate the final thyristor firing angle matching the current opening value in real time;

[0109] Step 3332: Generate a corresponding pulse width modulation signal according to the final thyristor firing angle. For the electrical delay in the transmission of the pulse width modulation signal, add dead time compensation to the pulse signal. At the same time, according to the driving power requirement of the valve actuator, adjust the amplitude of the pulse signal for power adaptation to form an adjustment command that conforms to the electrical characteristics of the driving module.

[0110] In the embodiment of the present invention, sensors are used to collect the theoretical opening value of the valve and the mechanical parameters of the valve actuator in real time. These parameters include the response time, inertia parameter, and opening limit threshold of the valve actuator. The AI technology monitors the working state of the sensors in real time and automatically identifies the outliers in the sensor data. For example, if the data of a certain sensor suddenly fluctuates greatly, the AI will immediately mark the data and cross-verify it with the data of other sensors to ensure the accuracy of the data. The collected data will be uniformly stored in a specially designed data structure. The AI will classify and store the data according to the importance and usage frequency of the data. For the frequently used key data, it will be stored in the cache for quick access; while for the historical data, it will be compressed and stored to save storage space. At the same time, the data will be backed up regularly to prevent data loss. Analyzing the influence of the response time and inertia parameter on the valve opening adjustment is a key step to achieve precise control. The response time reflects the time from when the valve actuator receives the command to when it starts to move, and the inertia parameter reflects the influence generated by the valve's own inertia during the movement. A large amount of operation data will be analyzed to find out the internal law between the response time, inertia parameter, and valve opening adjustment. For example, through learning thousands of valve adjustment data, it can be found that when the response time is long, the valve action is relatively slow. To speed up the adjustment speed, it is necessary to appropriately increase the dynamic adjustment weight; while when the inertia parameter is large, to avoid excessive adjustment of the valve, it is necessary to appropriately reduce the dynamic adjustment weight.

[0111] In practical applications, the AI will automatically calculate the dynamic adjustment weight of the theoretical opening value based on the response time and inertia parameters collected in real time. This calculation process is real-time and adaptive, and can make timely adjustments according to different working conditions and parameter changes. The opening limit threshold stipulates the maximum and minimum values of the valve opening, ensuring that the valve operates within a safe and reasonable range. The theoretical opening value is compared with the opening limit threshold in real time. If the theoretical opening value > the maximum value, it will be automatically adjusted to the maximum value, and the relevant information of this adjustment will be recorded, including the adjustment time and the opening values before and after the adjustment. At the same time, the reason for the opening value > the maximum value will be analyzed, whether it is due to an incorrect input command, an actuator failure or other factors, and corresponding warning information will be given. If the theoretical opening value is lower than the minimum value, it will be adjusted to the minimum value, and the same records and analysis will be carried out. After such processing, the processed parameters take into account both the dynamic characteristics of the valve actuator and ensure that the valve opening is within a safe range.

[0112] Taking the processed parameters as the input, call the non-linear mapping relationship between valve opening and thyristor firing angle pre-set in the single-chip microcomputer. This mapping relationship describes the non-linear relationship between the valve opening and the thyristor firing angle. In the mapping relationship, a discretized mapping relationship table is pre-stored, which records the thyristor firing angles corresponding to different valve opening values. The AI will use an efficient search algorithm to quickly query this mapping relationship table and locate the adjacent data points of the current processed opening value in the table. Since the mapping relationship table is discrete, the current opening value may not exactly match a certain data point in the table, so the AI will find two adjacent data points. Next, using the linear interpolation method, based on the information of the adjacent data points, the final thyristor firing angle matching the current opening value is calculated in real time. In this process, the AI will take into account the errors and uncertainties of the data and optimize and correct the interpolation results to improve the calculation accuracy. According to the calculated final thyristor firing angle, a corresponding pulse width modulation (PWM) signal is generated. The PWM signal is a commonly used control signal, which controls the output power of the circuit by changing the width of the pulse. When generating the PWM signal, the duty cycle of the pulse is determined according to the final thyristor firing angle, that is, the ratio of the high-level time of the pulse to the period. At the same time, the electrical delay problem that may occur during the transmission of the PWM signal will be considered. To avoid the impact of this delay on valve control, dead-time compensation will be added to the pulse signal. The dead time refers to setting a short interval between the high level and the low level of the pulse signal to ensure that incorrect triggering does not occur during the signal conversion. The AI will dynamically adjust the size of the dead time according to the electrical parameters monitored in real time to achieve the final compensation effect.

[0113] Finally, the AI adjusts the amplitude of the pulse signal according to the driving power requirement of the valve actuator. Different valve actuators require different driving powers. By monitoring the working state and power requirement of the valve actuator in real time and automatically adjusting the amplitude of the pulse signal, it ensures that the valve actuator can work properly. The adjusted pulse signal forms a regulation command that conforms to the electrical characteristics of the driving module, and this command can accurately control the action of the valve actuator.

[0114] Monitor the working state of the sensor in real time and identify outliers, and ensure data accuracy through cross-validation to avoid incorrect data caused by sensor failures or interference. The classification storage and management strategy for data not only improves the access efficiency of key data but also saves storage space. At the same time, the regular backup mechanism effectively prevents data loss and ensures the security and integrity of data assets. Through in-depth analysis of a large amount of operation data, the AI discovers the internal laws between the response time, inertia parameters, and valve opening adjustment, and realizes the real-time adaptive calculation of dynamically adjusted weights. This enables the valve opening adjustment to accurately match the dynamic characteristics of the valve actuator under different working conditions, improves the response speed and adjustment accuracy of valve control, effectively avoids problems such as over-adjustment or slow adjustment, and enhances stability and reliability. Based on the real-time comparison and automatic adjustment mechanism of the opening limit threshold, the AI can ensure that the valve opening is always within a safe and reasonable range. Once an over-limit situation occurs, it not only makes timely adjustments but also deeply analyzes the reasons and issues early warnings, which helps the staff quickly locate and solve problems, reduces the risk of safety accidents caused by abnormal valve actions, and ensures the safe and stable operation of the equipment. During the calculation of the thyristor trigger angle, the AI optimizes the search algorithm and interpolation results, improving the calculation accuracy; when generating the PWM signal, it dynamically adjusts the duty cycle, dead time, and signal amplitude, fully considering the signal transmission delay and the actual driving requirements of the valve actuator, so that the generated regulation command can accurately control the action of the valve actuator, effectively improving the performance and efficiency of the entire control system and reducing energy waste.

[0115] In a preferred embodiment of the present invention, step 4, where the valve actuator receives the regulation command, drives the valve to perform corresponding actions, and feeds back the valve position signal to the time series reinforcement prediction model in real time to form a closed-loop control, may include:

[0116] Step 440, the valve actuator receives the regulation command, parses the command, and transmits the information identification to the control unit of the actuator;

[0117] Step 441, according to the regulation command, the control unit starts the driving device, sends corresponding electrical signals to the stepping motor, and makes the stepping motor operate;

[0118] Step 442: During the movement of the valve, use the position sensor on the valve to detect the position of the valve in real time and convert the position information into an electrical signal;

[0119] Step 443: Transmit the valve position signal to the time-series reinforcement prediction model and compare it with the theoretical opening value of the valve to evaluate whether the current state meets the target;

[0120] Step 444: The time-series reinforcement prediction model dynamically adjusts the command to control the thyristor trigger angle according to the difference between the valve position and the theoretical opening value of the valve, and continuously optimizes the valve opening to match the target value to form a closed-loop control.

[0121] In the embodiment of the present invention, AI serves as the intelligent center of the entire control process and plays a role in the instruction reception link. When the valve actuator receives an adjustment instruction, it quickly decodes the instruction according to the preset communication protocol rule library. This rule library is constructed by AI through learning a large amount of running communication data and can automatically identify different types of instruction formats and coding methods. For example, when receiving an instruction based on the Modbus protocol, AI can instantly determine the address code, function code, data area, and check code in the data frame and perform intelligent verification on the check code. If the verification fails, it will automatically request retransmission of the instruction to ensure the accuracy of the data. After decoding, the instruction information is repackaged in a format recognizable by the control unit and transmitted to the control unit of the actuator through an optimized path. This path selection is dynamically determined according to the real-time load condition of the internal data transmission network of the actuator to ensure the efficiency of information transmission. According to the target valve opening in the instruction and the dynamically learned characteristics of the valve actuator, the optimal control parameters of the drive device are obtained, and a series of precise electrical signal control strategies are generated, including the frequency, pulse width, and sequence of the signals, to ensure that the stepper motor operates at the most reasonable speed and manner. At the same time, continuously monitor the operating state data of the motor, including current, voltage, and temperature. These data are fed back to AI in real time through sensors. Once an abnormality is detected, such as a sudden increase in current that may indicate motor jamming, the protection mechanism will be immediately triggered to stop the motor from running, send a fault alarm to the control system, and perform a preliminary analysis of the fault cause based on the fault data.

[0122] During the movement of the valve, the AI ​​monitors the data acquisition process of the position sensor in real time. When high-frequency noise generated by electromagnetic interference is detected, the filtering parameters are automatically adjusted to specifically filter out the noise. The processed signal is converted into an electrical signal and normalized. In addition, the working state of the sensor is also diagnosed in real time. By comparing the data of multiple sensors, it is judged whether the sensor has failed or its performance has declined. Once an abnormality is found, measures are taken in a timely manner, such as switching to a standby sensor or issuing a sensor failure warning. The AI ​​is responsible for efficiently transmitting the valve position signal to the time-series reinforcement prediction model. During the transmission process, the final transmission protocol and path are dynamically selected according to the real-time status of the network. For example, when the network bandwidth is low, the data is automatically compressed and sent using a transmission protocol with strong low-bandwidth adaptability to ensure that the data can reach the time-series reinforcement prediction model in a timely and complete manner. A complex comparison algorithm is used to deeply compare the valve position signal and the theoretical opening value of the valve. Driven by the AI, the time-series reinforcement prediction model performs in-depth decision-making analysis based on the difference between the valve position and the theoretical opening value. Using the reinforcement learning algorithm and combining the current system state, it calculates the best solution for adjusting the thyristor trigger angle. In this process, the response of the system under different adjustment solutions is continuously simulated, the advantages and disadvantages of each solution are evaluated, and the solution that can make the valve opening match the target value fastest and most stably is selected to generate a new control instruction and optimize the instruction to ensure the accuracy and reliability of the instruction during transmission and execution. After the new instruction is sent to the valve actuator, it continuously tracks the execution effect of the instruction, continuously collects feedback data, and online updates and optimizes the time-series reinforcement prediction model to form an evolving closed-loop control system.

[0123] Suppose in the steam inlet valve control system of a thermal power plant, at a certain moment, due to a sudden increase in the grid load, through a series of calculations, it is determined that the opening of the steam inlet valve needs to be increased from the current 50% to 70%, and the corresponding adjustment instructions are generated and sent to the valve actuator. After receiving the instructions, the AI in the valve actuator quickly starts the parsing program. Based on the learned communication protocol knowledge, it accurately identifies that the instructions are based on the Profibus protocol. The AI decodes the instructions, extracts the key information that the target opening is 70%, and checks the checksum to ensure the data is correct. Then, according to the internal network topology and real-time load conditions of the actuator, the final path is selected to transmit the information to the control unit. After receiving the instructions, the control unit combines the actual state of the current valve (such as the initial opening of the valve, the wear degree of the valve, etc.) and the dynamic characteristics of the valve actuator to obtain the control parameters of the stepper motor, and generates a series of electrical signals to control the stepper motor to operate at an appropriate speed and acceleration, while monitoring the current and voltage data of the motor in real time. During the operation of the motor, it is found that the current has a slight fluctuation. Through analysis, it is judged that this is caused by a slight change in the internal friction of the valve. The AI automatically fine-tunes the electrical signals to keep the motor running stably.

[0124] The position sensor on the valve starts to collect the valve position information in real time, performs noise reduction processing on the original signal output by the sensor, and removes the noise generated by the on-site electromagnetic environment interference. Then, the processed signal is converted into a standard electrical signal and normalized. During this process, the AI monitors that the output data of one of the position sensors shows abnormal fluctuations. Immediately compare the data of other sensors, judge that the sensor may be faulty, switch to the standby sensor in time, and issue a fault warning. The processed valve position signal is transmitted to the time series reinforcement prediction model. Due to a certain degree of congestion in the current network, the data is automatically compressed, and the UDP protocol is selected for fast transmission. After receiving the data, the time series reinforcement prediction model conducts a comparative analysis. At this time, the actual opening of the valve is 55%, which is different from the target opening of 70%. Considering the urgency of the current load increase and the response characteristics of the valve, it is judged that the current state has not reached the target and further adjustment is needed. In the time series reinforcement prediction model, the reinforcement learning algorithm is used to simulate and evaluate various schemes for adjusting the thyristor firing angle. After calculation, it is determined that increasing the thyristor firing angle by 10° is the final scheme, and a new control instruction is generated and sent to the valve actuator. After the instruction is executed, continuously track the change of the valve opening, continuously collect feedback data, and optimize the time series reinforcement prediction model. As the valve opening gradually approaches the target value, fine-tune the control strategy according to the actual situation, and finally stabilize the valve opening within the range of 70%±1% to complete the closed-loop control.

[0125] AI is deeply involved in the whole process of instruction parsing, execution, and feedback adjustment. It can accurately understand and process adjustment instructions, and perform precise control in combination with the dynamic characteristics of the valve actuator. Through learning a large amount of data, high-precision adjustment of the valve opening can be achieved. Compared with traditional control methods, the control accuracy can be greatly improved, meeting the extremely high requirements for valve control accuracy in thermal power plants, ensuring the stable control of key parameters such as steam flow, and improving power generation efficiency. Automatically adjust the control strategy according to different working conditions and system states. Whether facing sudden changes in grid load, gradual decline in equipment performance, or interference from external environmental factors, final decisions can be made through real-time data analysis and model calculation. For example, when the performance of the valve changes due to wear after a period of use, AI can automatically adjust the control parameters to ensure that the valve can still accurately respond to instructions, with strong adaptive capabilities, reducing the cost of manual intervention and improving the intelligence level of the system. During the entire control process, continuously monitor the operation status data of the equipment. Through real-time analysis and anomaly detection of the data, potential fault hazards can be detected in a timely manner. Such as abnormal current of the motor and fault signals of the sensor, and conduct a preliminary analysis of the cause of the fault. This early fault diagnosis and warning function allows maintenance personnel to take measures in advance to avoid the occurrence and expansion of faults, reduce equipment downtime, lower maintenance costs, and improve the reliability and service life of the equipment.

[0126] In a preferred embodiment of the present invention, in step 5 above, during the closed-loop control process, detecting the set threshold of the mechanical clearance of the valve and automatically triggering the clearance compensation mechanism to correct the opening command may include:

[0127] Step 550, during the operation of the closed-loop control, use the sensor of the valve actuator to collect the data of the mechanical state of the valve in real time, including the displacement change during the opening and closing processes of the valve and the force condition;

[0128] Step 551, based on the data of the mechanical state of the valve, judge the size of the mechanical clearance according to the displacement difference and the change characteristics of the movement resistance when the valve moves forward and backward;

[0129] Step 552, compare the mechanical clearance value with the preset threshold. If the mechanical clearance value ≥ the preset threshold, automatically trigger the clearance compensation mechanism and stop the current conventional control process;

[0130] Step 553, during the clearance compensation, according to the preset rules, combine the size of the mechanical clearance, the valve type, and the current operating condition factors to correct the original opening command.

[0131] In the embodiments of the present invention, during the closed-loop control operation, AI monitors the sensor array on the valve actuator in real time. The displacement sensor continuously and highly frequently collects the position change data during the opening and closing processes of the valve, recording hundreds of displacement points per second; the pressure sensor is closely attached to the valve transmission components to capture the pressure values at each force-bearing point during movement in real time. The original data collected by the sensors is processed immediately. At the same time, the working states of the sensors are diagnosed in real time. By comparing the data of multiple sensors of the same type, it is judged whether the sensors have faults or data anomalies. Once a problem is found, the system immediately switches to the backup sensor and issues an alarm to ensure the continuity and accuracy of data collection. Analyze the collected valve mechanical state data, focusing on the displacement differences and the characteristics of the change in movement resistance during the forward and reverse movements of the valve. When the valve moves, AI calculates the displacement difference under the same stroke during the forward and reverse movement processes, and at the same time analyzes the change trend of the data of the pressure sensor to judge the change in movement resistance. For example, at the moment of valve commutation, if the displacement difference suddenly increases and the movement resistance significantly decreases, AI will initially judge that there is a mechanical clearance. It will also combine the valve operation data, compare the change amplitude of the current data, and exclude misjudgments caused by operating condition fluctuations. By comprehensively analyzing these data characteristics, the size of the mechanical clearance is estimated.

[0132] The mechanical clearance value is compared with a preset threshold in real time and dynamically adjusted according to factors such as the service life and maintenance status of the valve. Once the mechanical clearance value ≥ the preset threshold, the clearance compensation mechanism is immediately triggered. First, a pause command is sent to the closed-loop control system to interrupt the current conventional control process and prevent the valve control deviation from further expanding due to the mechanical clearance. At the same time, detailed warning information is generated, including the specific value of the mechanical clearance, the triggering time, and the possible affected content, and is pushed to the system administrator and maintenance personnel to remind them to pay attention to the valve status in a timely manner. During the clearance compensation stage, according to the built-in preset rules, considering multiple factors such as the size of the mechanical clearance, the valve type, and the current operating conditions, the original opening command is corrected. AI will call the corresponding compensation strategy template according to the valve type (gate valve, butterfly valve). For different mechanical clearance sizes, the opening command is adjusted according to the established proportional relationship. For example, the larger the clearance, the larger the opening increment for compensation. At the same time, it will also combine the current operating conditions, such as system pressure, flow demand, etc., to fine-tune the corrected command. During the adjustment process, the corrected opening command is simulated to predict the impact of the valve action on the system operation. If it is found that it may cause problems such as pressure fluctuations and flow instability, the correction plan will be re-optimized until the generated opening command can both compensate for the mechanical clearance and ensure stable operation. Finally, the corrected command is sent to the valve actuator.

[0133] Suppose in a thermal power plant, a butterfly valve is in a closed-loop control state and is responsible for regulating the material flow rate in the pipeline. The displacement sensor and pressure sensor controlled by AI continuously collect data of the butterfly valve. During a valve closing process, the displacement sensor collects 300 position data per second, and the pressure sensor synchronously records the force on the transmission components. The collected data is processed by dynamic filtering to effectively filter out the interference signals generated by pipeline vibration, ensuring that the data truly reflects the mechanical state of the butterfly valve. By analyzing the data during the forward opening and reverse closing processes of the butterfly valve, it is found that when the butterfly valve closes and changes direction, the displacement difference increases by 0.5 mm compared to the normal state, and the movement resistance during closing significantly decreases. It is judged that this is not caused by the change of working conditions, but there is a mechanical clearance. By further analyzing the data characteristics, the current mechanical clearance is estimated to be 0.7 mm. Comparing the obtained mechanical clearance value of 0.7 mm with the preset threshold of 0.5 mm, it is found that the threshold has been exceeded. The clearance compensation mechanism is quickly triggered, and a pause instruction is sent to the closed-loop control system to stop the current conventional opening adjustment operation. At the same time, a warning is pushed to the enterprise's equipment management system: "The mechanical clearance of the butterfly valve in Pipeline 3 reaches 0.7 mm, exceeding the threshold. It is recommended to check immediately!"

[0134] According to the preset rules, considering the type of this butterfly valve and the current material flow rate demand in the pipeline, it is decided to increase the original opening instruction by 8% for compensation. Before the adjustment, the AI simulates the execution of the corrected instruction and finds that it may cause a small fluctuation in the pipeline pressure. Therefore, the correction plan is optimized, and the increase amount of the opening instruction is adjusted to 6%. After re-simulating and verifying, it is confirmed that this plan can not only compensate for the mechanical clearance but also have no obvious impact on the operation. Finally, the corrected opening instruction is sent to the butterfly valve actuator to complete the clearance compensation and ensure the stable operation of the pipeline system.

[0135] Monitor in real time and accurately judge the mechanical clearance of the valve, promptly correct the opening command, effectively avoid the deviation between the actual opening and the target opening of the valve caused by mechanical clearance, and improve the accuracy of valve control. In fields such as chemical industry and power where strict requirements are imposed on flow and pressure control, stable operation can be ensured, product quality and production efficiency can be improved. The automatically triggered clearance compensation mechanism can intervene before the mechanical clearance affects the normal operation of the valve, reduce problems such as valve component wear and increased vibration caused by excessive clearance, reduce the degree of equipment loss, thereby extend the service life of the valve and related transmission components, reduce the frequency of equipment replacement and maintenance, and reduce the operation and maintenance costs of the enterprise. By promptly handling mechanical clearance problems, AI can prevent abnormal operations caused by valve failures, such as pipeline pressure fluctuations and unstable flow, ensure stable operation, effectively reduce the risks of production interruption and safety accidents caused by valve problems, and improve the continuity and safety of enterprise production. AI automatically completes the detection, judgment, and compensation of the mechanical clearance of the valve without the need for frequent manual inspections and interventions, realizing the intelligentization of equipment operation and maintenance. At the same time, the generated warning information and processing records provide clear reference bases for maintenance personnel, facilitating quick problem location and formulation of maintenance plans, and improving the efficiency and scientific nature of operation and maintenance management. When correcting the opening command, fully consider factors such as valve type and current operating conditions, and be able to formulate personalized compensation strategies for different usage scenarios, enabling the valve to maintain good control performance under complex and changeable operating conditions, enhancing the adaptability to different working environments, and improving the flexibility of the enterprise to meet various production requirements.

[0136] As Figure 2 shown, the embodiment of the present invention further provides an AI intelligent-based valve actuator scheduling control system, including:

[0137] A multi-source acquisition module, used to collect real-time grid load demand data, fuel quality parameters, and real-time steam turbine energy consumption data, and obtain the predicted value of the final steam admission volume of the steam turbine;

[0138] A valve mapping module, used to calculate the theoretical valve opening value according to the predicted final steam admission volume and the friction torque and gear clearance parameters of the mechanical actuator;

[0139] A firing angle adjustment module, used to generate a thyristor firing angle adjustment command in real time according to the mapping relationship between the theoretical valve opening value and the valve actuator;

[0140] A closed-loop feedback module, used to detect the mechanical clearance threshold of the valve according to the thyristor firing angle adjustment command,

[0141] to correct the opening command;

[0142] A clearance compensation module, used to automatically trigger the clearance compensation mechanism according to the mechanical clearance threshold of the valve.

[0143] It should be noted that this system corresponds to the above method, and all implementation manners in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0144] An embodiment of the present invention further provides a computing device, including: a processor and a memory storing a computer program. When the computer program is run by the processor, it executes the method as described above. All implementation manners in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0145] An embodiment of the present invention further provides a computer-readable storage medium storing instructions. When the instructions are run on a computer, the computer is caused to execute the method as described above. All implementation manners in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0146] The above is the preferred embodiment of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A valve actuator scheduling control method based on AI intelligence, characterized in that: The method comprises: Step 1: Collect grid load demand data, fuel quality parameters and real-time energy consumption data of steam turbines, and use the time series enhanced prediction model to dynamically predict the final steam inlet of steam turbines; Step 2: Collect the turbine operating status parameters in real time, determine the relationship between the steam inlet and the valve opening under the current working conditions to obtain a preliminary valve opening estimate and dynamically respond to compensation. When the load changes rapidly, start the dynamic feedforward compensation mechanism and perform phase compensation in combination with the valve actuator response delay characteristics, so as to obtain the valve theoretical opening value. Step 3, the valve theoretical opening value is comprehensively coupled and analyzed with the valve actuator response time, inertia parameters and opening limit threshold mechanical parameters, the dynamic adjustment weight is calculated and the boundary constraint processing is performed to obtain the processed parameters, and then the valve opening-thyristor trigger angle nonlinear mapping model preset in the single-chip microcomputer is called based on the parameters, and the final thyristor trigger angle is calculated by querying adjacent data points and linear interpolation method, and finally a pulse width modulation signal is generated based on this, and the dead time is added to compensate for the electrical delay and the amplitude is adjusted to adapt to the driving power demand, so as to form a regulation instruction that meets the electrical characteristics of the driving module; Step 4: The valve actuator receives the adjustment instruction, drives the valve to perform corresponding actions, and feeds back the valve position signal to the timing enhancement prediction model in real time to form a closed-loop control; Step 5: During the closed-loop control process, the set threshold of the valve mechanical clearance is detected and the clearance compensation mechanism is automatically triggered to correct the opening instruction.

2. The valve actuator scheduling control method based on AI intelligence according to claim 1 is characterized in that: Collect grid load demand data, fuel quality parameters and real-time energy consumption data of steam turbines, and use the time series enhanced prediction model to dynamically predict the final steam inlet volume of steam turbines, including: Collect grid load demand data, fuel quality parameters and real-time turbine operation data in real time, and align the time series to obtain a multi-dimensional input vector; According to the multi-dimensional input vector, a time series reinforcement prediction model is established, and historical operation data is collected as training samples, which are input into the time series reinforcement prediction model for training; During the training process, the time series reinforcement prediction model continuously adjusts its own parameters to learn the synergistic relationship between grid load, fuel quality and turbine operating status, and obtains the trained time series reinforcement prediction model; The multidimensional input vector is input into the trained time series intensive prediction model, and the final steam inlet of the turbine is dynamically predicted based on the learned load-fuel-turbine synergy relationship.

3. The valve actuator scheduling control method based on AI intelligence according to claim 2 is characterized in that: Collect turbine operating status parameters in real time, determine the relationship between steam inlet and valve opening under current operating conditions to obtain preliminary valve opening estimates and dynamically respond to compensation. When the load changes rapidly, start the dynamic feedforward compensation mechanism and perform phase compensation in combination with the valve actuator response delay characteristics to obtain the valve theoretical opening value, including: Collect steam turbine operating status parameters in real time, including steam temperature, steam pressure, turbine speed, power output, ambient temperature, humidity and atmospheric pressure; According to the operating state parameters of the steam turbine, the relationship between the steam inlet volume and the valve opening that matches the current operating conditions is determined from the operating data, a preliminary valve opening estimation value corresponding to the final steam inlet volume is obtained, and dynamic response compensation is performed on the preliminary valve opening estimation value; When a rapid load change is detected, the dynamic feedforward compensation mechanism is activated, and phase compensation is performed through the response delay characteristics of the valve actuator to obtain the theoretical opening value of the valve.

4. The valve actuator scheduling control method based on AI intelligence according to claim 3 is characterized in that: When a rapid load change is detected, the dynamic feedforward compensation mechanism is activated, and phase compensation is performed through the response delay characteristics of the valve actuator to obtain the theoretical opening value of the valve, including: According to the current valve opening deviation value, determine the immediate adjustment component, the cumulative adjustment component and the dynamic suppression component to obtain the basic adjustment amount; Based on the basic adjustment amount, the change rate of the load demand is monitored in real time, and the feedforward compensation amount is generated according to the preset ratio based on the change rate; According to the feedforward compensation amount, the dynamic hysteresis compensation amount of the valve actuator is used to perform phase compensation; The basic adjustment amount, feedforward compensation amount and phase compensation are integrated to obtain the theoretical opening value of the valve.

5. The valve actuator scheduling control method based on AI intelligence according to claim 4 is characterized in that: The process of determining the adjustment instructions includes: A comprehensive coupling analysis is performed on the theoretical valve opening value and the mechanical parameters of the valve actuator, wherein the mechanical parameters include the response time, inertia parameters and opening limit threshold of the valve actuator, and the dynamic adjustment weight of the theoretical opening value is calculated according to the response time and inertia parameters, and the theoretical opening value is subjected to boundary constraint processing in combination with the opening limit threshold to obtain the processed parameters; Based on the processed parameters, the valve opening-thyristor trigger angle nonlinear mapping model preset in the single-chip microcomputer is called. The mapping model takes the theoretical opening value as input, locates the adjacent data points of the current opening value in the table by querying the pre-stored discretization mapping relationship table, and uses the linear interpolation method to calculate the final thyristor trigger angle matching the current opening value in real time; According to the final thyristor trigger angle, the corresponding pulse width modulation signal is generated. The dead time compensation is added to the pulse signal to compensate for the electrical delay in the pulse width modulation signal transmission. At the same time, according to the driving power requirement of the valve actuator, the amplitude of the pulse signal is power-adapted and adjusted to form a regulation instruction that meets the electrical characteristics of the drive module.

6. The valve actuator scheduling control method based on AI intelligence according to claim 5 is characterized in that: The valve actuator receives the adjustment command, drives the valve to perform the corresponding action, and feeds back the valve position signal to the timing enhancement prediction model in real time to form a closed-loop control, including: The valve actuator receives the adjustment command, analyzes the command, and transmits the information identification to the control unit of the actuator; The control unit starts the driving device according to the adjustment instruction, sends a corresponding electrical signal to the stepper motor, and makes the stepper motor run; During the movement of the valve, the position sensor on the valve is used to detect the position of the valve in real time and convert the position information into an electrical signal; The valve position signal is transmitted to the time-series enhanced prediction model and compared with the theoretical valve opening value to evaluate whether the current state meets the target; The timing enhanced prediction model dynamically adjusts the command control thyristor trigger angle according to the difference between the valve position and the theoretical valve opening value, and continuously optimizes the valve opening to match the target value to form a closed-loop control.

7. The valve actuator scheduling control method based on AI intelligence according to claim 6 is characterized in that: During the closed-loop control process, the set threshold of the valve mechanical clearance is detected and the clearance compensation mechanism is automatically triggered to correct the opening command, including: During the closed-loop control operation, the valve actuator’s sensors are used to collect data on the valve’s mechanical status in real time, including displacement changes during valve opening and closing, as well as force conditions; According to the data of the valve mechanical state, the size of the mechanical clearance is determined by the displacement difference and movement resistance change characteristics of the valve during forward and reverse movement; Compare the mechanical clearance value with the preset threshold. If the mechanical clearance value is greater than or equal to the preset threshold, the clearance compensation mechanism is automatically triggered and the current routine control process is stopped. In clearance compensation, the original opening instruction is corrected according to preset rules, combined with the size of the mechanical clearance, valve type and current operating conditions.

8. A valve actuator scheduling control system based on AI intelligence, the system implements the method as described in any one of claims 1 to 7, characterized in that: include: Multi-source acquisition module, used to collect real-time grid load demand data, fuel quality parameters and real-time energy consumption data of steam turbines, and obtain the final steam inlet prediction value of steam turbines; The valve mapping module is used to calculate the theoretical valve opening value based on the final steam inlet prediction and the friction torque and gear clearance parameters of the mechanical actuator; The trigger angle adjustment module is used to generate a thyristor trigger angle adjustment instruction in real time according to the mapping relationship between the valve theoretical opening value and the valve actuator; The closed-loop feedback module is used to detect the valve mechanical clearance threshold according to the thyristor trigger angle adjustment instruction. To correct the opening instruction; The clearance compensation module is used to automatically trigger the clearance compensation mechanism according to the valve mechanical clearance threshold.

9. A computing device, characterized in that include: one or more processors; A storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the method as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program, which, when executed by a processor, implements the method according to any one of claims 1 to 7.

Images