Intelligent control method for variable frequency motor

Abstract

The invention discloses an intelligent control method for a variable frequency motor, and belongs to the field of variable frequency motor control, and the method comprises the following steps: S1, constructing an original multi-source state data set, and obtaining a normalized state parameter and a data credibility weight through calibration noise reduction and normalized fusion; s2, constructing a control law-prediction model-fault constraint modular control parameter library; s3, constructing a state-control-effect annotation data set, training a PMP-ANN hybrid prediction model, and generating a multi-stage adaptive control scheme; and S4, constructing an energy consumption-control precision-part life-robustness multi-objective optimization model by combining data credibility weight based on the multi-stage adaptive control scheme, and determining an optimal control parameter scheme. By adopting the intelligent control method for the variable frequency motor, collaborative optimization of energy consumption, control precision, part service life and robustness under dynamic load is realized, and theoretical optimality and engineering practicability are both achieved.

Description

Technical Field

[0001] This invention relates to the field of variable frequency motor control technology, and in particular to an intelligent control method for variable frequency motors. Background Technology

[0002] In the field of variable frequency motor control technology, motors, as core power equipment in industrial production and intelligent manufacturing, directly affect production efficiency, energy costs, and operational safety through their control performance. However, existing technologies suffer from the following key shortcomings, making it difficult to meet the intelligent control requirements under complex operating conditions: 1. Traditional control methods (such as PID control) rely on fixed control parameters and lack the ability to adapt to the nonlinear dynamic characteristics of motor-load. When faced with dynamic load fluctuations (such as sudden changes in load torque and changes in operating pressure), the control accuracy drops significantly and it is difficult to balance response speed and stability. 2. Insufficient utilization of multi-source data: Existing technologies have not established a reliable weighted fusion mechanism for multi-source data (inverter built-in parameters, external sensor data, and industrial historical data). The data is susceptible to noise interference, and similar data lacks calibration, noise reduction, and weighted integration, resulting in insufficient reliability of the data source for control decisions. 3. Single optimization objective: Existing control schemes often focus on a single objective (such as reducing energy consumption or improving accuracy) without building a multi-objective collaborative optimization model that considers energy consumption, control accuracy, component lifespan, and robustness. This results in poor overall performance (such as reducing energy consumption while simultaneously causing increased component wear and decreased robustness). Summary of the Invention

[0003] The purpose of this invention is to provide an intelligent control method for variable frequency motors to solve the above-mentioned technical problems.

[0004] To achieve the above objectives, the present invention provides an intelligent control method for a variable frequency motor, comprising the following steps: S1. Combining the built-in monitoring data of the frequency converter and the historical industrial operation data of the motor operating parameters, an original multi-source state dataset is constructed. The credibility weight of the original multi-source state data is calculated based on the AHP-entropy weight method. The rotor slot harmonic features are extracted by sub-band decomposition. After calibration, noise reduction, normalization and fusion, normalized state parameters and data credibility weights are obtained. S2. Based on the normalized state parameters and data confidence weights output in step S1, and combined with the nonlinear dynamic characteristics of the motor-load, construct a modular control parameter library of control law-prediction model-fault constraint. S3. Based on the modular control parameter library built in step S2, and combined with dynamic load demand parameters, construct a state-control-effect labeled dataset, train a PMP-ANN hybrid prediction model, generate multiple sets of preliminary control schemes, and select multi-level adaptive control schemes through expert semantic differentiation evaluation. S4. Based on the multi-level adaptive control scheme output in step S3, and combined with the data reliability weights in step S1, construct a multi-objective optimization model of energy consumption, control accuracy, component lifespan, and robustness. Optimize the model using the NSGA-II algorithm and verify it using the AD axiom design to determine the optimal control parameter scheme.

[0005] Therefore, the present invention employs the above-mentioned intelligent control method for variable frequency motors, which has the following beneficial effects: 1. Optimal energy consumption: Compared with traditional PID control, the normalized energy consumption is minimized through a multi-objective optimization model, which significantly reduces the energy consumption of motor operation; 2. High control precision: Adapts to the nonlinear characteristics of motor-load, and combines rotor slot harmonic feature extraction and hybrid prediction model to keep the control error within the maximum allowable error range, resulting in accurate dynamic load response; 3. Extended component lifespan: By real-time monitoring of losses in core components such as IGBTs and capacitors and optimization of control parameters, losses caused by overload and high-frequency fluctuations are reduced, extending the overall lifespan of components and motors. 4. Strong robustness: It integrates fault constraints, hardware constraints and multi-source data credibility weights, and has outstanding resistance to sensor noise, load fluctuations and operating condition changes. 5. Excellent dynamic adaptability: The MLP-XGBoost prediction model and multi-level adaptation scheme screening can quickly respond to dynamic requirements such as load torque fluctuations and pressure changes, and adapt to complex working conditions. 6. Practical application: Generates PLC executable code and standardized parameter files, combined with expert semantic evaluation and AD axiomatic design verification, taking into account both algorithm optimization and practical industrial application needs; 7. Closed-loop self-optimization: Based on real-time data from IoT sensors, an adaptive adjustment model is trained, the control parameter library is dynamically updated, and the system continuously adapts to component aging and changes in operating conditions to maintain optimal long-term performance.

[0006] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0007] Figure 1 This is a flowchart of an intelligent control method for a variable frequency motor according to the present invention. Detailed Implementation

[0008] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely illustrative of the embodiments of the present invention and are not intended to limit the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of this application. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout.

[0009] It should be noted that the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, such as a process, method, system, product, or server that includes a series of steps or units, not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such process, method, product, or device.

[0010] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0011] like Figure 1 As shown, a method for intelligent control of a variable frequency motor includes the following steps: S1. Combining the built-in monitoring data of the frequency converter and the historical industrial operation data of the motor operating parameters, an original multi-source state dataset is constructed. The credibility weight of the original multi-source state data is calculated based on the AHP-entropy weight method. The rotor slot harmonic features are extracted by sub-band decomposition. After calibration, noise reduction, normalization and fusion, normalized state parameters and data credibility weights are obtained. Step S1 specifically includes the following steps: S11. Multi-source data synchronous acquisition: Based on the Modbus TCP / IP protocol, synchronously acquire the inverter's built-in parameters, acquire motor operating parameters through the industrial bus, and simultaneously extract operating data from the historical database to obtain the original multi-source status dataset. Among them, the built-in parameters of the frequency converter include the stator voltage acquired by the internal circuitry of the frequency converter. Stator current Motor speed Motor core temperature DC bus voltage and cumulative running time The motor operating parameters include the stator current collected by external sensors. Motor core temperature and the operating pressure of the load equipment driven by the motor. Historical industrial operating data includes historical load torque. Historical energy consumption and historical fault records ; S12. Calculate the credibility weight of the original data using the AHP-entropy weight method. : ; in, ; ; ; ; In the formula, This represents the weighting coefficient between subjective and objective factors; and These represent the subjective weights of AHP and the objective weights of the entropy weight method, respectively. Indicates the first Information entropy of class data; Indicates the first Class Data The normalized values ​​of each sample, and The built-in monitoring data of the strain gauge, motor operating parameters, and historical industrial operating data were respectively analyzed. Indicates the number of historical data samples; Indicates the first Class Data The original observations of each sample; Indicates the first Class data relative to the first Each evaluation criterion is used to compare and judge the elements of the matrix; S13. Based on the inverter's built-in parameters and motor operating data collected from external sensors, zero-point drift correction and gain calibration are performed on the motor operating parameters. Simultaneously, a first-order low-pass filter is used to suppress random noise. Weighted fusion of multi-source, similar data is then performed using confidence weights to obtain the filtered stator current. Motor speed Core temperature DC bus voltage and load operating pressure ; S14. Using 9-level binary sub-band decomposition, the filtered stator current... Decomposed into 512 narrowband sub-signals, retaining only the frequency band containing rotor slot harmonics. This achieves signal bandwidth compression and noise isolation, with the following sub-band selection criteria: ; in, ; ; In the formula, This indicates the rotor slot harmonic sub-carrier current; This represents the frequency variable of the current signal; Indicates the bandwidth of the rotor slot harmonic subband; Indicates the base frequency of the power supply; Indicates the sampling frequency; Indicates the slip ratio, and ; Represents a rectangular window function; Meanwhile, based on the frequency band where the rotor slot harmonics are located And reverse the real-time speed of the motor : ; S15, Motor speed Stator current Core temperature and load operating pressure Normalization is performed to obtain normalized state parameters. and data credibility weight ;in, These represent the normalized real-time motor speed, filtered current, motor core temperature, and power, respectively.

[0012] S2. Based on the normalized state parameters and data confidence weights output in step S1, and combined with the nonlinear dynamic characteristics of the motor-load, construct a modular control parameter library of control law-prediction model-fault constraint. Step S2 specifically includes the following steps: S21, to minimize energy consumption To optimize the objective, a nonlinear state-space model of the motor-load is constructed, and an analytical control law is derived based on the Pontryagin maximum principle to determine the control constraints. in, The optimization objective expression is as follows: ; In the formula, This represents the normalized motor stator voltage; Indicates the power factor angle; The integral time interval representing the optimization objective; The expression for the nonlinear state-space model of the motor-load is as follows: ; In the formula, Represents the state vector The derivative of, and ; This represents the state transition function of the state-space model. Indicates the control input quantity; This represents the moment of inertia of the motor. This represents the motor torque coefficient; Represents the load torque function; Indicates the damping coefficient; Indicates the electrical time constant of the motor; Indicates the voltage gain coefficient; Indicates the back electromotive force coefficient; Indicates the temperature-time constant; This represents the normalized ambient temperature. Control Law The expression is as follows: ; in, ; In the formula, This represents the Hamiltonian function in Pontryagin's maximum principle; , and Both represent conjugate variables; The time derivative of the normalized motor speed; This represents the time derivative of the normalized filter current. The time derivative of the normalized motor temperature; The control constraint expressions are as follows: ; In the formula, and These represent control input quantities. Frequency constraints upper and lower limits; S22. Construct an MLP-XGBoost prediction model, where the MLP model is used to predict the adaptive switching frequency. Its expression is as follows: ; In the formula, Indicates the activation function; , This represents the weight matrices of the input-hidden-output layer and the hidden-output layer of an MLP model. , This represents the bias of the hidden and output layers of an MLP model; This represents the input feature vector of the MLP model; The XGBoost model is used to predict velocity profile parameters, and its expression is as follows: ; get: ; In the formula, Indicates the velocity profile type; Indicates the normalized velocity; This represents the prediction mapping function of the XGBoost model; Let represent the input feature vector of the XGBoost model, and ; This represents the optimal velocity profile function predicted by XGBoost; Indicates the number of modulation cycles; Indicates normalized time; S23. Based on historical fault data, using the 3σ criterion and statistical distribution fitting, the fault thresholds for each state parameter are calibrated, and a fault constraint set is constructed. , These represent the fault thresholds for stator current, core temperature, and rotor slot frequency, respectively. , , , and These represent the mean and standard deviation of the normalized effective current during normal operation, respectively. and These represent the mean and standard deviation of the normalized core temperature during normal operation, respectively. and These represent the mean and standard deviation of the rotor slot frequency during normal operation, respectively. S24. Based on the inverter hardware parameters and motor rated parameters, construct a hardware constraint set. , These represent the upper limits of the normalized stator current, core temperature, and switching frequency change rate, respectively. , , , and These represent the maximum allowable current and maximum surface temperature of the motor, respectively. S25. Integrate control law parameters, MLP-XGBoost prediction model parameters, fault constraint thresholds, and hardware constraint quantification indicators according to control law, prediction model, and constraint conditions to construct a structured parameter library. , These represent the sub-libraries in the parameter library that store PMP control law parameters, MLP-XGBoost prediction model parameters, fault constraint threshold parameters, and hardware constraint quantification indicators, respectively.

[0013] S3. Based on the modular control parameter library built in step S2, and combined with dynamic load demand parameters, construct a state-control-effect labeled dataset, train a PMP-ANN hybrid prediction model, generate multiple sets of preliminary control schemes, and select multi-level adaptive control schemes through expert semantic differentiation evaluation. Step S3 specifically includes the following steps: S31. Construct a state-control-effect labeled dataset: Label the state parameters, control parameters, and operational effects from historical running data to construct a training dataset. , These represent the first and second elements in the labeled dataset. The state parameter vector, control parameters, and running effect of each sample; This represents the total number of samples in the labeled dataset; and , These represent the labeled datasets at the 1st, 2nd, and 3rd positions, respectively. Energy consumption indicators, control response time, and robustness indicators for each sample; S32. Construct a PMP-ANN hybrid prediction model and input the state-control-effect labeled dataset into the PMP-ANN hybrid prediction model for training. The loss function for model training is expressed as follows: ; In the formula, Indicates the loss value; , and These represent the loss weights for energy consumption, response time, and robustness, respectively. , and These represent the target performance indicators for energy consumption, response time, and robustness, respectively. This represents the energy consumption value predicted by the PMP-ANN hybrid prediction model; This represents the control response time predicted by the PMP-ANN hybrid prediction model; This represents the robustness index of the PMP-ANN hybrid prediction model. PMP-ANN hybrid prediction model output : ; In the formula, and These represent the weights of the PMP model and the ANN model, respectively; (represented by...) This represents the control input quantity output by the ANN; S33. Input the current dynamic load demand into the trained PMP-ANN hybrid prediction model to generate multiple preliminary control schemes; S34. Invite experts to score the preliminary control schemes based on semantic evaluation indicators, and sort the scoring results in descending order. Take the top three preliminary control schemes as multi-level adaptation schemes.

[0014] S4. Based on the multi-level adaptive control scheme output in step S3, and combined with the data reliability weights in step S1, construct a multi-objective optimization model of energy consumption, control accuracy, component lifespan, and robustness. Optimize the model using the NSGA-II algorithm and verify it using the AD axiom design to determine the optimal control parameter scheme.

[0015] Step S4 specifically includes the following steps: S41. Construct a normalized multi-objective function with the objectives of minimizing energy consumption, maximizing control precision, extending component lifespan, and enhancing robustness. ; in, Minimum energy consumption target : ; Highest control precision target : ; Maximum component life target : ; The most robust target : ; In the formula, This represents normalized energy consumption; This indicates the energy consumption under PID control. This represents the normalized control error; and These represent the actual control error and the maximum allowable control error, respectively. Indicates the normalized component lifespan; and These represent the actual lifespan and rated lifespan of the component, respectively. Indicates normalization robustness; Indicates the first The robustness contribution of the data type; S42. Integrate hardware constraints, fault constraints, and operating condition constraints to clarify the value boundaries of optimization variables; Hardware constraint C1: ; Fault constraint C2: ; Control precision constraint C3: ; Switching frequency constraint C4: ; In the formula, This indicates the optimized switching control frequency; Indicates the temperature fault threshold; Indicates the current fault threshold; and They represent Time and The switching frequency at any given time; S43. A non-dominated sorting genetic algorithm is used to optimize a multi-objective function and obtain the Pareto optimal solution set. ; S44. Verify the Pareto optimal solution set based on axiomatic design theory. By considering functional independence and information minimization, the optimal solution is selected. S45, Combining Pareto optimal solutions Based on the verification results of axiomatic design theory, the globally optimal control scheme is determined. , These represent the optimal control frequency, optimal adaptive switching frequency, optimal control parameter set, and optimal velocity profile function in the global optimal control scheme, respectively.

[0016] In step S44, the functional independence axiom is set as follows: the functional requirement matrix is ​​a diagonal matrix; The axiom of information minimization is set as follows: Select the combination of design parameters that minimizes information content.

[0017] Step S4 is followed by S5, which generates PLC executable code and standardized parameter files based on the optimal control parameter scheme output in step S4, deploys IoT sensors to collect motor operating status and component wear data in real time, trains parameter adaptive adjustment model, and dynamically updates the modular control parameter library and control law parameters described in step S2 to form a closed-loop control system.

[0018] In step S5, the execution code is first generated, and then the motor operating speed is collected in real time using IoT sensors. Current ,temperature ,pressure IGBT junction temperature and capacitor ripple current And calculate component losses: IGBT loss : ; Capacitor loss : ; In the formula, Indicates the IGBT loss coefficient; Indicates the capacitor loss coefficient; Then, an adaptive adjustment model is trained based on the motor operating data to dynamically adjust the control parameters: ; In the formula, This indicates the updated control input value; This represents the original globally optimal control input. This indicates an adaptive adjustment of the step size coefficient; Indicates the target indicators of operational effectiveness; Indicates real-time performance metrics; The inverse of the Jacobian matrix is ​​represented by . Finally, based on the output of the adaptive adjustment model, the modular control parameter library is updated regularly to optimize the PMP control law parameters, ANN model weights, and constraint thresholds.

[0019] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for intelligent control of a variable frequency motor, characterized in that: Includes the following steps: S1. Combining the built-in monitoring data of the frequency converter and the historical industrial operation data of the motor operating parameters, an original multi-source state dataset is constructed. The credibility weight of the original multi-source state data is calculated based on the AHP-entropy weight method. The rotor slot harmonic features are extracted by sub-band decomposition. After calibration, noise reduction, normalization and fusion, normalized state parameters and data credibility weights are obtained. S2. Based on the normalized state parameters and data confidence weights output in step S1, and combined with the nonlinear dynamic characteristics of the motor-load, construct a modular control parameter library of control law-prediction model-fault constraint. S3. Based on the modular control parameter library built in step S2, and combined with dynamic load demand parameters, construct a state-control-effect labeled dataset, train a PMP-ANN hybrid prediction model, generate multiple sets of preliminary control schemes, and select multi-level adaptive control schemes through expert semantic differentiation evaluation. S4. Based on the multi-level adaptive control scheme output in step S3, and combined with the data reliability weights in step S1, construct a multi-objective optimization model of energy consumption, control accuracy, component lifespan, and robustness. Optimize the model using the NSGA-II algorithm and verify it using the AD axiom design to determine the optimal control parameter scheme.

2. The intelligent control method for a variable frequency motor according to claim 1, characterized in that: Step S1 specifically includes the following steps: S11. Multi-source data synchronous acquisition: Based on the Modbus TCP / IP protocol, synchronously acquire the inverter's built-in parameters, acquire motor operating parameters through the industrial bus, and simultaneously extract operating data from the historical database to obtain the original multi-source status dataset. Among them, the built-in parameters of the frequency converter include the stator voltage acquired by the internal circuitry of the frequency converter. Stator current Motor speed Motor core temperature DC bus voltage and cumulative running time The motor operating parameters include the stator current collected by external sensors. Motor core temperature and the operating pressure of the load equipment driven by the motor. Historical industrial operating data includes historical load torque. Historical energy consumption and historical fault records ; S12. Calculate the credibility weight of the original data using the AHP-entropy weight method. : ； in, ； ； ； ； In the formula, This represents the weighting coefficient between subjective and objective factors; and These represent the subjective weights of AHP and the objective weights of the entropy weight method, respectively. Indicates the first Information entropy of class data; Indicates the first Class Data The normalized values ​​of each sample, and The built-in monitoring data of the strain gauge, motor operating parameters, and historical industrial operating data were respectively analyzed. Indicates the number of historical data samples; Indicates the first Class Data The original observations of each sample; Indicates the first Class data relative to the first Each evaluation criterion is used to compare and judge the elements of the matrix; S13. Based on the inverter's built-in parameters and motor operating data collected from external sensors, zero-point drift correction and gain calibration are performed on the motor operating parameters. Simultaneously, a first-order low-pass filter is used to suppress random noise. Weighted fusion of multi-source, similar data is then performed using confidence weights to obtain the filtered stator current. Motor speed Core temperature DC bus voltage and load operating pressure ; S14. Using 9-level binary sub-band decomposition, the filtered stator current... Decomposed into 512 narrowband sub-signals, retaining only the frequency band containing rotor slot harmonics. This achieves signal bandwidth compression and noise isolation, with the following sub-band selection criteria: ； in, ； ； In the formula, This indicates the rotor slot harmonic sub-carrier current; This represents the frequency variable of the current signal; Indicates the bandwidth of the rotor slot harmonic subband; Indicates the base frequency of the power supply; Indicates the sampling frequency; Indicates the slip ratio, and ; Represents a rectangular window function; Meanwhile, based on the frequency band where the rotor slot harmonics are located And reverse the real-time speed of the motor : ； S15, Motor speed Stator current Core temperature and load operating pressure Normalization is performed to obtain normalized state parameters. and data credibility weight ;in, These represent the normalized real-time motor speed, filtered current, motor core temperature, and power, respectively.

3. The intelligent control method for a variable frequency motor according to claim 2, characterized in that: Step S2 specifically includes the following steps: S21, to minimize energy consumption To optimize the objective, a nonlinear state-space model of the motor-load is constructed, and an analytical control law is derived based on the Pontryagin maximum principle to determine the control constraints. in, The optimization objective expression is as follows: ； In the formula, This represents the normalized motor stator voltage; Indicates the power factor angle; The integral time interval representing the optimization objective; The expression for the nonlinear state-space model of the motor-load is as follows: ； In the formula, State vector The derivative of, and ; This represents the state transition function of the state-space model. Indicates the control input quantity; This represents the moment of inertia of the motor. Indicates the motor torque coefficient; Represents the load torque function; Indicates the damping coefficient; Indicates the electrical time constant of the motor; Indicates the voltage gain coefficient; Indicates the back electromotive force coefficient; Indicates the temperature-time constant; This represents the normalized ambient temperature. Control Law The expression is as follows: ； in, ； In the formula, This represents the Hamiltonian function in Pontryagin's maximum principle; , and Both represent conjugate variables; The time derivative of the normalized motor speed; This represents the time derivative of the normalized filter current. The time derivative of the normalized motor temperature; The control constraint expressions are as follows: ； In the formula, and These represent control input quantities. Frequency constraints upper and lower limits; S22. Construct an MLP-XGBoost prediction model, where the MLP model is used to predict the adaptive switching frequency. Its expression is as follows: ； In the formula, Indicates the activation function; , This represents the weight matrices of the input-hidden-output layer and the hidden-output layer of an MLP model. , This represents the bias of the hidden and output layers of an MLP model; This represents the input feature vector of the MLP model; The XGBoost model is used to predict velocity profile parameters, and its expression is as follows: ； get: ; In the formula, Indicates the velocity profile type; Indicates the normalized velocity; This represents the prediction mapping function of the XGBoost model; Let represent the input feature vector of the XGBoost model, and ; This represents the optimal velocity profile function predicted by XGBoost; Indicates the number of modulation cycles; Indicates normalized time; S23. Based on historical fault data, using the 3σ criterion and statistical distribution fitting, the fault thresholds for each state parameter are calibrated, and a fault constraint set is constructed. , These represent the fault thresholds for stator current, core temperature, and rotor slot frequency, respectively. , , , and These represent the mean and standard deviation of the normalized effective current during normal operation, respectively. and These represent the mean and standard deviation of the normalized core temperature during normal operation, respectively. and These represent the mean and standard deviation of the rotor slot frequency during normal operation, respectively. S24. Based on the inverter hardware parameters and motor rated parameters, construct a hardware constraint set. , These represent the upper limits of the normalized stator current, core temperature, and switching frequency change rate, respectively. , , , and These represent the maximum allowable current and maximum surface temperature of the motor, respectively. S25. Integrate control law parameters, MLP-XGBoost prediction model parameters, fault constraint thresholds, and hardware constraint quantification indicators according to control law, prediction model, and constraint conditions to construct a structured parameter library. , These represent the sub-libraries in the parameter library that store PMP control law parameters, MLP-XGBoost prediction model parameters, fault constraint threshold parameters, and hardware constraint quantification indicators, respectively.

4. The intelligent control method for a variable frequency motor according to claim 3, characterized in that: Step S3 specifically includes the following steps: S31. Construct a state-control-effect labeled dataset: Label the state parameters, control parameters, and operational effects from historical running data to construct a training dataset. , These represent the first and second elements in the labeled dataset. The state parameter vector, control parameters, and running effect of each sample; This represents the total number of samples in the labeled dataset; and , These represent the labeled datasets at the 1st, 2nd, and 3rd positions, respectively. Energy consumption indicators, control response time, and robustness indicators for each sample; S32. Construct a PMP-ANN hybrid prediction model and input the state-control-effect labeled dataset into the PMP-ANN hybrid prediction model for training. The loss function for model training is expressed as follows: ； In the formula, Indicates the loss value; , and These represent the loss weights for energy consumption, response time, and robustness, respectively. , and These represent the target performance indicators for energy consumption, response time, and robustness, respectively. This represents the energy consumption value predicted by the PMP-ANN hybrid prediction model; This represents the control response time predicted by the PMP-ANN hybrid prediction model; This represents the robustness index of the PMP-ANN hybrid prediction model. PMP-ANN hybrid prediction model output : ； In the formula, and These represent the weights of the PMP model and the ANN model, respectively; (represented by...) This represents the control input quantity output by the ANN; S33. Input the current dynamic load demand into the trained PMP-ANN hybrid prediction model to generate multiple preliminary control schemes; S34. Invite experts to score the preliminary control schemes based on semantic evaluation indicators, and sort the scoring results in descending order. Take the top three preliminary control schemes as multi-level adaptation schemes.

5. The intelligent control method for a variable frequency motor according to claim 4, characterized in that: Step S4 Specifically, the following steps are included: S41. Construct a normalized multi-objective function with the objectives of minimizing energy consumption, maximizing control precision, extending component lifespan, and enhancing robustness. ; in, Minimum energy consumption target : ; Highest control precision target : ; Maximum component life target : ; The most robust target : ; In the formula, This represents normalized energy consumption; This indicates the energy consumption under PID control. This represents the normalized control error; and These represent the actual control error and the maximum allowable control error, respectively. Indicates the normalized component lifespan; and These represent the actual lifespan and rated lifespan of the component, respectively. Indicates normalization robustness; Indicates the first The robustness contribution of the data type; S42. Integrate hardware constraints, fault constraints, and operating condition constraints to clarify the value boundaries of optimization variables; Hardware constraint C1: ; Fault constraint C2: ; Control precision constraint C3: ; Switching frequency constraint C4: ; In the formula, This indicates the optimized switching control frequency; Indicates the temperature fault threshold; Indicates the current fault threshold; and They represent Time and The switching frequency at any given time; S43. A non-dominated sorting genetic algorithm is used to optimize a multi-objective function and obtain the Pareto optimal solution set. ; S44. Verify the Pareto optimal solution set based on axiomatic design theory. By considering functional independence and information minimization, the optimal solution is selected. S45, Combining Pareto optimal solutions Based on the verification results of axiomatic design theory, the globally optimal control scheme is determined. , These represent the optimal control frequency, optimal adaptive switching frequency, optimal control parameter set, and optimal velocity profile function in the global optimal control scheme, respectively.

6. The intelligent control method for a variable frequency motor according to claim 5, characterized in that: In step S44, the functional independence axiom is set as follows: the functional requirement matrix is ​​a diagonal matrix; The axiom of information minimization is set as follows: Select the combination of design parameters that minimizes information content.

7. The intelligent control method for a variable frequency motor according to claim 6, characterized in that: Step S4 is followed by S5, which generates PLC executable code and standardized parameter files based on the optimal control parameter scheme output in step S4, deploys IoT sensors to collect motor operating status and component wear data in real time, trains parameter adaptive adjustment model, and dynamically updates the modular control parameter library and control law parameters described in step S2 to form a closed-loop control system.

8. The intelligent control method for a variable frequency motor according to claim 7, characterized in that: In step S5, the execution code is first generated, and then the motor operating speed is collected in real time using IoT sensors. Current ,temperature ,pressure IGBT junction temperature and capacitor ripple current And calculate component losses: IGBT loss : ; Capacitor loss : ; In the formula, Indicates the IGBT loss factor; Indicates the capacitor loss coefficient; Then, an adaptive adjustment model is trained based on the motor operating data to dynamically adjust the control parameters: ； In the formula, This indicates the updated control input value; This represents the original globally optimal control input. This indicates an adaptive adjustment of the step size coefficient; Indicates the target indicators of operational effectiveness; Indicates real-time performance metrics; The inverse of the Jacobian matrix is ​​represented by . Finally, based on the output of the adaptive adjustment model, the modular control parameter library is updated regularly to optimize the PMP control law parameters, ANN model weights, and constraint thresholds.

Images