Intelligent control method and device for high-speed motor of cooling fan

Abstract

The invention relates to the technical field of high-speed motor intelligent control, and discloses a high-speed motor intelligent control method and device for a cooling fan, wherein the method comprises: collecting parameters of a high-speed fan motor to obtain a basic motor parameter group; calculating the motor overmodulation boundary to obtain a linear PWM region threshold and an overmodulation region threshold; performing multi-objective optimization operations on voltage distortion, current harmonics and system losses to obtain an initial control parameter group; performing current harmonic analysis to obtain a current harmonic compensation value and a system delay compensation value; performing online state evaluation operations to obtain a PWM control parameter correction value, a current controller parameter correction value and a voltage vector distribution correction value; performing fault feature analysis on the operating state of the high-speed fan motor, and outputting a fault phase isolation instruction and a fault-tolerant control instruction. The invention improves the overall operating effect of the high-speed fan motor.

Description

Technical Field

[0001] The present invention relates to the technical field of high-speed motor intelligent control, and in particular to a high-speed motor intelligent control method and device for a cooling fan. Background Art

[0002] Traditional cooling fan motors have problems such as low control accuracy and slow dynamic response when running at high speeds, and the control strategy in the overmodulation area is imperfect, resulting in reduced system performance. In high-speed fan motors with a single neutral point symmetrical winding structure, although they have strong fault tolerance, the single neutral point will limit the DC bus voltage utilization within the linear PWM range.

[0003] Existing cooling fan motor control methods mainly deal with this limitation by injecting non-torque current components, but lack a systematic comparison of control complexity, inverter efficiency, and copper loss. In addition, existing technologies generally lack effective multi-objective optimization methods and cannot simultaneously take into account multiple objectives such as system efficiency, control accuracy, and dynamic performance. Especially under fault conditions, existing control methods are difficult to perform fault diagnosis and fault-tolerant control in a timely manner, affecting the reliability and safety of the system. Summary of the invention

[0004] The present invention provides a method and device for intelligently controlling a high-speed motor of a heat dissipation fan, and the present invention improves the overall operation effect of the high-speed fan motor.

[0005] In a first aspect, the present invention provides a method for intelligently controlling a high-speed motor of a cooling fan, the method comprising:

[0006] Collect parameters of high-speed fan motor to obtain basic motor parameter group;

[0007] Calculate the motor overmodulation boundary based on the motor basic parameter group to obtain a linear PWM region threshold and an overmodulation region threshold;

[0008] Inputting the linear PWM region threshold and the overmodulation region threshold into the Kriging proxy model, performing multi-objective optimization calculations on voltage distortion, current harmonics and system loss, and obtaining an initial control parameter group;

[0009] Performing current harmonic analysis based on the initial control parameter group to obtain a current harmonic compensation value and a system delay compensation value;

[0010] Performing an online state evaluation operation on the current harmonic compensation value and the system delay compensation value to obtain a PWM control parameter correction value, a current controller parameter correction value, and a voltage vector distribution correction value;

[0011] The fault characteristics of the operating state of the high-speed fan motor are analyzed according to the PWM control parameter correction value, the current controller parameter correction value and the voltage vector distribution correction value, and a fault phase isolation instruction and a fault-tolerant control instruction are output.

[0012] In a second aspect, the present invention provides a high-speed motor intelligent control device for a cooling fan, the high-speed motor intelligent control device for a cooling fan comprising:

[0013] The acquisition module is used to acquire parameters of the high-speed fan motor and obtain a basic parameter group of the motor;

[0014] A calculation module, used for calculating the motor overmodulation boundary based on the motor basic parameter group to obtain a linear PWM region threshold and an overmodulation region threshold;

[0015] An optimization module, used for inputting the linear PWM region threshold and the overmodulation region threshold into a Kriging proxy model, performing multi-objective optimization operations on voltage distortion, current harmonics and system loss, and obtaining an initial control parameter group;

[0016] An analysis module, configured to perform current harmonic analysis based on the initial control parameter group to obtain a current harmonic compensation value and a system delay compensation value;

[0017] An evaluation module, used for performing an online state evaluation operation on the current harmonic compensation value and the system delay compensation value to obtain a PWM control parameter correction value, a current controller parameter correction value and a voltage vector distribution correction value;

[0018] The output module is used to perform fault characteristic analysis on the operating state of the high-speed fan motor according to the PWM control parameter correction value, the current controller parameter correction value and the voltage vector distribution correction value, and output fault phase isolation instructions and fault-tolerant control instructions.

[0019] The third aspect of the present invention provides a high-speed motor intelligent control device for a cooling fan, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the high-speed motor intelligent control device for the cooling fan executes the above-mentioned high-speed motor intelligent control method for the cooling fan.

[0020] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned intelligent control method for a high-speed motor of a cooling fan.

[0021] In the technical solution provided by the present invention, a multi-objective optimization framework based on a Kriging proxy model is established, combined with a comprehensive evaluation of voltage distortion, current harmonics and system losses, thereby significantly improving the control accuracy and energy utilization efficiency of the system; an improved PWM control strategy and harmonic compensation method are adopted to effectively solve the voltage distortion and current harmonic problems during high-speed operation, making the motor run more smoothly; an online state evaluation algorithm is designed, and the dynamic response capability and anti-interference ability of the system are enhanced by adjusting the control parameters in real time; based on a single neutral point symmetrical winding structure, an accurate fault diagnosis and fault-tolerant control strategy is developed, thereby improving the reliability and safety of the system; through a multi-level parameter optimization and correction mechanism, global optimization of system performance is achieved, and the overall operation effect of the high-speed fan motor is significantly improved.

[0022] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0023] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A schematic diagram of an embodiment of a method for intelligently controlling a high-speed motor of a heat dissipation fan according to an embodiment of the present invention;

[0025] Figure 2 A schematic diagram of an embodiment of a high-speed motor intelligent control device for a heat dissipation fan in an embodiment of the present invention;

[0026] Figure 3 Schematic diagram of an embodiment of a high-speed motor intelligent control device for a heat dissipation fan in an embodiment of the present invention. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0028] The terms "including" and "having" and any variations thereof mentioned in the embodiments of the present invention are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device end including a series of steps or units is not limited to the listed steps or units, but may optionally include other steps or units that are not listed, or may optionally include other steps or units that are inherent to these processes, methods, products or device ends.

[0029] To facilitate understanding of this embodiment, firstly, a high-speed motor intelligent control method for a heat dissipation fan disclosed in an embodiment of the present invention is described in detail. Figure 1 As shown, the method comprises the following steps:

[0030] 101. Collect parameters of the high-speed fan motor to obtain a basic parameter group of the motor;

[0031] It is understandable that the execution subject of the present invention may be a high-speed motor intelligent control device for a cooling fan, or a terminal or a server, which is not specifically limited here. The embodiment of the present invention is described by taking a server as the execution subject as an example.

[0032] Specifically, a DC excitation voltage is applied to the stator winding of the high-speed fan motor for measurement and calculation, and the resistance measurement value and stator leakage inductance measurement value of the stator winding are obtained. The stator resistance directly affects the thermal effect and energy efficiency of the motor, while the stator leakage inductance is related to the electromagnetic characteristics and response time of the motor. According to these two measurement values, the winding impedance is calculated to obtain the impedance characteristic parameters of the stator. These parameters reflect the resistance of the motor stator to current during actual operation. The magnetic field strength of the permanent magnet is measured by multi-point sampling. Through multi-point sampling, the magnetic field distribution data of the permanent magnet is obtained, which reflects the change of the magnetic field when the motor is running, thereby affecting the torque output and overall performance of the motor. Based on the magnetic field distribution data, the flux linkage is calculated to obtain the characteristic value of the permanent magnet flux linkage. The flux linkage characteristic value of the permanent magnet is closely related to the magnetic coupling and torque output of the motor. The accurate flux linkage characteristic value can improve the accuracy and response speed of the motor control. The three-phase current signal of the motor is sampled and processed. The three-phase current signal of the motor is input into the sampling processing unit, and the instantaneous value of the three-phase current is obtained through analog-to-digital conversion operation. According to these instantaneous values, the current vector synthesis calculation is performed to obtain the space vector value of the current. The current space vector value is an important parameter to describe the current state of the motor. It is effectively used to analyze the load condition of the motor and the dynamic response of the current, thereby optimizing the operating efficiency and stability of the motor. In the operation control of the motor, the output signal of the rotor position sensor of the high-speed fan motor is sampled and processed in multiple cycles. Through sampling, the rotor position angle value and the rotor speed value are obtained. Accurate measurement of the rotor position and speed is crucial for the real-time control of the motor. It can help the control system determine the motion state of the motor and make corresponding adjustments to ensure that the motor runs in the optimal state. The stator impedance characteristic parameters, permanent magnet flux characteristic values, current space vector values, rotor position angle values, and rotor speed values ​​obtained through the above steps are used as the basic parameter group of the motor.

[0033] 102. Calculate the motor overmodulation boundary based on the motor basic parameter group to obtain a linear PWM region threshold and an overmodulation region threshold;

[0034] Specifically, the voltage vector is calculated using the stator impedance characteristic parameter and the current space vector value to obtain the reference voltage vector value. The reference voltage vector value is an important reference quantity in motor control and provides a basis for subsequent PWM modulation. The reference voltage vector value is processed by six-phase space segmentation to obtain six basic voltage vector components, which reflect the change of the motor voltage under different modulation states. The six basic voltage vector components are combined with the rotor position angle value, and the voltage space vector synthesis value is obtained through vector synthesis calculation. The voltage space vector synthesis value is a key parameter to describe the motor voltage state and comprehensively reflects the voltage characteristics of the motor in actual operation. The voltage space vector synthesis value is normalized and converted into a standardized quantity to facilitate comparison and operation with the upper limit of the voltage modulation ratio. By comparison, the linear PWM area threshold is obtained. The linear PWM area threshold marks the normal working range of the motor voltage modulation. Once the voltage space vector synthesis value exceeds this threshold, the system will enter the overmodulation area, resulting in motor instability or reduced efficiency. The calculated linear PWM area threshold can provide a reasonable working range for motor control. The linear PWM area threshold and the rotor speed value are nonlinearly mapped to obtain the overmodulation starting point data. The rotor speed is an important dynamic parameter of the motor's operating state. Different speeds correspond to different voltage modulation strategies. Combining the linear PWM area threshold with the rotor speed value can accurately determine the starting point of the overmodulation area to avoid motor overload or performance degradation during the overmodulation process. According to the overmodulation starting point data, the voltage vector trajectory is segmented. According to the actual operation of the motor, the overmodulation interval boundary values ​​are accurately divided. These boundary values ​​determine the working range of the motor in the overmodulation area. Based on the overmodulation interval boundary values, the maximum voltage utilization is calculated to obtain the voltage limit value. The voltage limit value is the maximum voltage that the motor can withstand. Exceeding this voltage will cause the motor to be damaged or the efficiency to drop sharply. The voltage limit value is matched with the current space vector value to obtain the overmodulation area threshold. Through the matching operation, it is ensured that the motor will not exceed its voltage limit during the overmodulation process, thereby maintaining the stable operation and efficient operation of the motor.

[0035] 103. Inputting the linear PWM region threshold and the overmodulation region threshold into the Kriging agent model, performing multi-objective optimization calculations on the voltage distortion, current harmonics and system loss, and obtaining an initial control parameter group;

[0036] Specifically, the linear PWM region threshold and the overmodulation region threshold are input into the Kriging proxy model for multi-objective optimization operations to obtain an initial control parameter group. The two thresholds are input into the initial layer of the Kriging proxy model, and feature mapping is performed through the Gaussian kernel function in the layer to obtain high-dimensional feature mapping data. High-dimensional feature mapping is a key step in the Kriging proxy model. It converts low-dimensional input data into a higher-dimensional feature space so that the model can better fit complex nonlinear relationships and facilitate subsequent optimization calculations. The high-dimensional feature mapping data is input into the Kriging interpolation model to calculate the initial sampling points of the optimization parameters. The Kriging interpolation model consists of two parts: the average response term and the correlation term. The average response term is constructed by fitting a second-order polynomial and can reflect the global trend of the system, while the correlation term is constructed using an exponential variation function to describe the local correlation between input variables. Through the operation of these models, a set of initial sampling points of the optimization parameters is obtained. The initial sampling points of the optimization parameters are input into the three-layer response surface network. The first layer of the network contains 32 neurons and uses the ReLU activation function, the second layer contains 16 neurons and uses the Sigmoid activation function, and the third layer contains 8 neurons and uses the Tanh activation function. Through the calculation of the multi-layer neural network, the response values ​​of voltage distortion and current harmonics are obtained, which reflect the changes in voltage and current quality during the operation of the motor, help optimize the performance of the system, and reduce losses and unstable factors. The voltage distortion and harmonic response values ​​are adaptively weighted to obtain the voltage quality evaluation index. Through weighted calculation, the influence of voltage distortion and current harmonics is comprehensively considered to obtain the overall voltage quality evaluation index. Based on the initial sampling point of the optimization parameters, the copper loss and inverter loss are iteratively calculated to obtain the system loss evaluation index. Copper loss and inverter loss are the main sources of energy loss in the motor system. Optimizing these losses helps to improve system efficiency. Based on the voltage quality evaluation index and the system loss evaluation index, a multi-objective optimization function is constructed, and a Pareto optimal solution search is performed. The purpose of Pareto optimal solution search is to find the optimal solution set under multiple optimization objectives to ensure that the operating efficiency and stability of the motor are maximized while meeting the dual requirements of voltage quality and system loss. Through this search process, the optimal solution candidate set is obtained, which represents the best operating scheme of the system under different constraints. Based on the optimal solution candidate set, dominance sorting and congestion calculation are performed. By sorting the candidate solutions, the solution vector on the Pareto front is selected as the final optimal control parameter. The solution on the Pareto front represents the solution that cannot improve another objective by changing one objective under the existing conditions, and is the optimal solution. Based on the selected optimal control parameters, the space vector PWM parameter optimization is performed to obtain the initial control parameter group. The parameter group includes the d-axis current command value, the q-axis current command value, the PWM switching frequency value, and the voltage vector action time value.The d-axis and q-axis current command values ​​represent the current control commands for the direct and quadrature axes in the motor control system, the PWM switching frequency value controls the frequency of the PWM signal, and the voltage vector action time value determines the action time of each voltage vector. By optimizing these control parameters, the efficiency and stability of the motor during operation are ensured, thereby improving the overall performance of the cooling fan motor.

[0037] 104. Perform current harmonic analysis based on the initial control parameter group to obtain a current harmonic compensation value and a system delay compensation value;

[0038] Specifically, spectrum analysis is performed based on the d-axis current command value and the q-axis current command value in the initial control parameter group to obtain the current harmonic spectrum. Spectral analysis obtains the distribution of the current signal at different frequencies by performing a fast Fourier transform on the time domain data of the motor current signal, thereby identifying various harmonic components in the current. The current harmonic spectrum provides basic data for subsequent harmonic compensation and delay compensation, revealing the nonlinear characteristics and frequency characteristics of the motor operation. The current harmonic spectrum is classified and calculated by odd and even harmonics to obtain classified harmonic data. The harmonics in the current signal are divided into two categories: odd harmonics and even harmonics, where odd harmonics include harmonic components of frequencies such as the 3rd, 5th, and 7th, and even harmonics include harmonic components of frequencies such as the 2nd, 4th, and 6th. Odd harmonics and even harmonics have different characteristics and different effects on the motor system. Classifying them separately helps to make more accurate compensation. The classified harmonic data is input into the neural network harmonic estimator for feature extraction to obtain a harmonic feature vector. The neural network harmonic estimator contains 8 neurons in the input layer, 16 neurons in the hidden layer, and 4 neurons in the output layer. After this layer-by-layer processing, it can effectively extract the harmonic features in the current signal and convert these features into vector form for subsequent PWM compensation calculation. The multi-layer structure of the neural network can handle complex nonlinear problems, automatically learn the pattern of harmonic signals, and improve the analysis accuracy. The harmonic compensation calculation of the PWM switching sequence is performed using the harmonic feature vector to obtain the PWM compensation sequence. By compensating the harmonic components in the PWM signal, the harmonics in the current can be effectively reduced, and the operating efficiency and stability of the motor can be improved. The PWM compensation sequence is processed using a band-stop filter to remove the harmonic components of certain specific frequencies and obtain the final current harmonic compensation value. The band-stop filter can effectively filter specific frequencies and suppress unnecessary high-frequency components, thereby optimizing the operating characteristics of the motor. The system delay model is constructed using a discrete state observer. The state variables of the model include current and speed. The delay characteristic parameters are obtained through sampling delay analysis. There is a delay phenomenon in the motor system caused by factors such as sensor delay and controller response delay. By analyzing the sampling delay of the PWM switching frequency value, the delay characteristic parameters related to the control process are obtained. Based on the delay characteristic parameters, prediction compensation is performed. The second-order Runge Kutta method is used to predict the system state in advance, and the prediction step is one control cycle. The second-order Runge Kutta method is a high-precision numerical solution method that can effectively predict the system state and provide a more accurate compensation signal for the controller. Through this process, the system state at future moments is predicted and compensated in advance, thereby reducing the performance loss caused by system delay. The delay compensation value of the system is obtained, which helps to improve the real-time and accuracy of motor control and enhance the dynamic response capability of the system.

[0039] 105. Perform online state evaluation calculation on the current harmonic compensation value and the system delay compensation value to obtain a PWM control parameter correction value, a current controller parameter correction value, and a voltage vector distribution correction value;

[0040] Specifically, the current harmonic compensation value and the system delay compensation value are input into the fuzzy evaluation network for online state evaluation to obtain compensation effect scoring data. The fuzzy evaluation network contains 7 input nodes, 12 hidden nodes and 3 output nodes. The triangular membership function is used for membership mapping to fuzzify the input variables and obtain quantitative compensation effect scoring data, which reflects the effect of current harmonic compensation and delay compensation in the current state. The compensation effect scoring data is segmented by multi-level thresholds to screen out the optimal segmentation threshold. The scoring data is divided into different intervals, and the weight coefficient is calibrated for each interval to obtain the segmented evaluation interval. These segmented intervals represent different levels of compensation effects. Through segmented evaluation, the quality of the compensation effect is defined, thereby guiding the optimization of subsequent control parameters. A PWM control parameter optimization model is constructed based on the segmented evaluation interval. The optimization model calculates the optimal PWM switch timing according to the different compensation effects and obtains the PWM control parameter correction value. The optimization of the PWM control parameter correction value can make the motor control signal more accurate, avoid the influence of current harmonics or system delay, and improve the stability and response speed of motor control. The PWM control parameter correction value is input into the active disturbance rejection controller for parameter correction to obtain the current controller parameter correction value. The active disturbance rejection controller includes a tracking differentiator, a nonlinear state error feedback law and an extended state observer. Through this controller, efficient compensation of the current control system is achieved, external interference and system instability factors are eliminated, and the current controller can maintain good performance under complex working conditions. The active disturbance rejection controller accurately corrects the current controller parameters through the combined effects of tracking error, nonlinear feedback and state observer, and can dynamically adjust during system operation to adapt to different loads and working conditions. Based on the current controller parameter correction value, a state space model is constructed, and closed-loop stability verification is performed through characteristic root analysis and Lyapunov stability criterion to obtain the system stability boundary. Through this verification process, it is evaluated whether the system has sufficient stability to ensure that the system can operate stably within the expected range under the control strategy. Characteristic root analysis can reveal the dynamic characteristics of the system, and Lyapunov stability criterion verifies whether the system is stable under all possible working conditions through mathematical means. The voltage vector sequence within the stability boundary is optimized to obtain the voltage vector distribution correction value. According to the stability requirements of the system and the actual operation of the motor, the voltage vector distribution method is adjusted to ensure the maximum efficiency and minimum loss of the motor during operation. The optimization of the voltage vector distribution correction value can effectively improve the energy efficiency of the system and reduce the risk of motor operation failure, thereby ensuring the long-term stable operation of the cooling fan motor.

[0041] 106. Perform fault characteristic analysis on the operating state of the high-speed fan motor according to the PWM control parameter correction value, the current controller parameter correction value and the voltage vector distribution correction value, and output a fault phase isolation instruction and a fault-tolerant control instruction.

[0042] Specifically, feature fusion operation is performed based on the PWM control parameter correction value, the current controller parameter correction value and the voltage vector distribution correction value to obtain the correction parameter feature vector. A single neutral point symmetrical winding fault analysis operation is performed on the correction parameter feature vector. By setting 6 symmetrical component detection units, the imbalance of the three-phase current is effectively calculated to obtain the deviation vector of the phase current. This step identifies the possible fault points in the motor winding, especially the fault of the three-phase unbalanced or neutral point symmetrical winding. The deviation vector, as a manifestation of the imbalance of the current, can provide an early warning of the motor fault, which is of great significance for the accurate detection of the winding fault. Based on the phase current deviation vector, a fault phase location model is constructed. The root mean square deviation value of each phase current is calculated by the recursive least squares method to obtain the identification data of the fault phase. As an iterative optimization algorithm, the recursive least squares method can effectively deal with the noise and interference generated during the operation of the motor and improve the accuracy of fault location. The model can quickly identify the phase with a large current deviation. The fault phase identification data is input into the multi-layer decision tree network for fault level determination. The multi-layer decision tree network contains three decision layers, each with four decision nodes. The decision tree network can automatically determine the severity and type of fault based on historical data and real-time monitoring results combined with identification data. Through layer-by-layer judgment, the decision tree can accurately distinguish faults of different levels and generate corresponding fault isolation instructions, thereby ensuring that the system can respond quickly when a fault occurs and reduce the impact of the fault on the entire system. According to the fault phase isolation instruction, the open phase mathematical model is established, and the current distribution of the remaining phases is simulated and calculated. Through simulation, the current distribution characteristic quantity is obtained, which reflects the change of the remaining phase current after the motor fails in some phases. On this basis, a single neutral point reconstruction is performed based on the current distribution characteristic quantity. The reconstruction process simulates the current redistribution scheme when the motor is missing or fails, ensuring that the motor can still maintain a certain operating capacity after a fault occurs. The robustness analysis of the current reconstruction scheme is performed. By constructing the sliding mode surface of the exponential reaching law, the controller parameters are designed, and the fault-tolerant control instructions are obtained. Sliding mode control is a nonlinear control method that can effectively deal with uncertainties and external interference in the system and maintain stability in the case of motor failure. The exponential approaching sliding surface design ensures that the controller has strong adaptability and fault tolerance to the current reconstruction scheme, thereby achieving smooth operation of the motor under various fault scenarios. The output of the fault-tolerant control command can guide the motor system to adopt appropriate control strategies to deal with faults and maintain the overall stability of the system.

[0043] In the embodiment of the present invention, a multi-objective optimization framework based on the Kriging proxy model is established, combined with a comprehensive evaluation of voltage distortion, current harmonics and system losses, thereby significantly improving the control accuracy and energy utilization efficiency of the system; an improved PWM control strategy and harmonic compensation method are adopted to effectively solve the voltage distortion and current harmonic problems during high-speed operation, making the motor run more smoothly; an online state evaluation algorithm is designed, and the dynamic response capability and anti-interference ability of the system are enhanced by adjusting the control parameters in real time; based on a single neutral point symmetrical winding structure, an accurate fault diagnosis and fault-tolerant control strategy is developed, thereby improving the reliability and safety of the system; through a multi-level parameter optimization and correction mechanism, global optimization of system performance is achieved, and the overall operation effect of the high-speed fan motor is significantly improved.

[0044] In a specific embodiment, the process of executing step 101 may specifically include the following steps:

[0045] Applying a DC excitation voltage to the stator winding of the high-speed fan motor for measurement operation, obtaining a stator winding resistance measurement value and a stator leakage inductance measurement value, and performing winding impedance calculation based on the stator winding resistance measurement value and the stator leakage inductance measurement value to obtain a stator impedance characteristic parameter;

[0046] The permanent magnet magnetic field strength of the high-speed fan motor is measured by multi-point sampling to obtain the permanent magnet magnetic field distribution data, and the magnetic flux calculation is performed based on the permanent magnet magnetic field distribution data to obtain the permanent magnet magnetic flux characteristic value;

[0047] The three-phase current signal of the high-speed fan motor is input into the sampling processing unit for analog-to-digital conversion operation to obtain the instantaneous value of the three-phase current, and the current vector synthesis calculation is performed according to the instantaneous value of the three-phase current to obtain the current space vector value;

[0048] Perform multi-cycle sampling processing on the output signal of the rotor position sensor of the high-speed fan motor to obtain the rotor position angle value and the rotor speed value;

[0049] The stator impedance characteristic parameters, permanent magnet flux characteristic values, current space vector values, rotor position angle values ​​and rotor speed values ​​are taken as the basic motor parameter group.

[0050] Specifically, a DC excitation voltage is applied to the stator winding of the high-speed fan motor for measurement and calculation to obtain the stator winding resistance measurement value and the stator leakage inductance measurement value. By applying a known DC voltage to the stator winding and measuring the current, the resistance value of the stator winding is calculated. With the known voltage and current values, Ohm's law is used to calculate the stator winding resistance value. The measured value of the stator winding resistance is obtained, where is the stator winding resistance, is the applied DC voltage, is the measured current value. The stator winding leakage inductance is measured by applying an AC signal and monitoring the phase difference between the voltage and current. The leakage inductance measurement reflects the resistance of the motor winding to the changing current, and its calculation formula is:

[0051] ;

[0052] in, is the stator leakage inductance, is the voltage amplitude, is the rate of change of current. Using these two values, calculate the total impedance of the stator winding , which is composed of stator winding resistance and leakage inductance and is calculated by the following formula:

[0053] ;

[0054] in, is an imaginary unit, is the angular frequency of the signal. The stator impedance characteristic parameter describes the electrical characteristics of the winding through this impedance, which helps the optimization and fault diagnosis of the subsequent control system. In order to obtain the magnetic field characteristics of the permanent magnet, the magnetic field strength of the permanent magnet of the motor is measured by multi-point sampling. This process collects the magnetic field data of the permanent magnet in real time by arranging multiple sensors at different positions of the motor. Assume that the magnetic field strength measured by the sensor is The distribution data at each point is , these data are used to calculate the flux linkage by numerical integration method The formula for calculating magnetic flux is:

[0055] ;

[0056] in, is the magnetic field strength, It is the area through which the magnetic field passes. The flux characteristic value reflects the total energy of the permanent magnet magnetic field and can provide an important basis for the evaluation and optimization of the motor's operating status. The magnitude of the flux directly affects the electromagnetic torque generated by the motor. After obtaining the resistance and leakage inductance characteristics of the stator winding and the flux characteristics of the permanent magnet, the three-phase current signal of the motor is sampled and processed through the current sensor installed on the motor. After the current signal is converted by the sampling and processing unit, the instantaneous value of the three-phase current is obtained, which is recorded as , these values ​​represent the three-phase current at time The current vector synthesis calculation is performed on these three instantaneous current values ​​to obtain the current space vector value The calculation of the current vector is based on the complex representation of the three-phase current, and its calculation formula is:

[0057] ;

[0058] in, is an imaginary unit, They are the three-phase current, is a complex number representing the rotating vector. The current space vector value It reflects the comprehensive effect of the three-phase current and can provide important data support for the motor's load status, power control, etc. The rotor position sensor outputs the rotor position angle through multi-cycle sampling. and speed . Rotor angle refers to the position of the rotor relative to the stator, while the rotor speed It represents the rotation speed of the rotor. The speed is calculated by sampling the change in the rotor position in each cycle, and the formula is:

[0059] ;

[0060] These data are the basis for important operations such as calculating electromagnetic torque, estimating load conditions, and adjusting drive strategies in the motor control system. Real-time monitoring of rotor angle and speed helps optimize motor control so that the motor can maintain stable operation under various working conditions. The stator impedance characteristic parameters, permanent magnet flux characteristic values, current space vector values, rotor position angle values, and rotor speed values ​​are combined to form a basic parameter group for the motor.

[0061] In a specific embodiment, the process of executing step 102 may specifically include the following steps:

[0062] The voltage vector is calculated based on the stator impedance characteristic parameter and the current space vector value to obtain a reference voltage vector value;

[0063] The reference voltage vector value is processed into six-phase space segments to obtain six basic voltage vector components;

[0064] Perform vector synthesis calculation based on six basic voltage vector components and rotor position angle value to obtain voltage space vector synthesis value;

[0065] After normalizing the voltage space vector synthesis value, the value is compared with the upper limit of the voltage modulation ratio to obtain the linear PWM region threshold;

[0066] Perform nonlinear mapping operation on the linear PWM region threshold and the rotor speed value to obtain the overmodulation starting point data;

[0067] The voltage vector trajectory is segmented according to the overmodulation starting point data to obtain the overmodulation interval boundary value;

[0068] The maximum voltage utilization rate is calculated for the boundary value of the overmodulation interval to obtain the voltage limit value, and the voltage limit value is matched with the current space vector value to obtain the overmodulation area threshold.

[0069] Specifically, the voltage vector is calculated based on the stator impedance characteristic parameters and the current space vector value to obtain the reference voltage vector value. The stator impedance characteristic parameters reflect the electrical characteristics of the motor, and the current space vector provides a comprehensive representation of the motor's three-phase current in space. The voltage vector is calculated using the voltage space vector modulation technology. On this basis, the current space vector is combined with the motor impedance to obtain the ideal voltage vector. This process is expressed by the following formula:

[0070] ;

[0071] in, is the reference voltage vector, is the stator impedance of the motor, is the current space vector. In this formula, the product of the motor's resistance and the current space vector gives the reference value of the voltage vector, reflecting the electrical characteristics of the motor under different operating conditions. The reference voltage vector is processed by six-phase space segmentation, and the voltage vector is decomposed into multiple basic components to provide accurate parameters for the subsequent modulation process. In the process of six-phase space segmentation, the reference voltage vector is decomposed into six basic voltage vector components. These basic voltage vector components are derived based on the geometric layout of the motor and the spatial distribution of the phase current, and are achieved by performing the following transformations on the reference voltage vector:

[0072] ;

[0073] in, It is based on six basic vectors in voltage space. By decomposing the reference voltage vector into these six components, the output voltage of the motor can be controlled more accurately and the voltage modulation efficiency during motor operation can be ensured. Vector synthesis calculation is performed based on the six basic voltage vector components and the rotor position angle value. The rotor position angle value It is very important for the control of the motor, and determines the rotation direction and amplitude of the voltage vector in space. The voltage space vector is synthesized by the following formula to obtain a new voltage space vector:

[0074] ;

[0075] In this formula, are the six basic voltage vector components, is the rotor angle, is the phase angle of each basic voltage vector. By synthesizing these components, a voltage space vector that reflects the actual operating state of the motor is obtained. After the voltage space vector is synthesized, the synthesized value is normalized and compared with the upper limit of the voltage modulation ratio to obtain the linear PWM area threshold. The normalization of the voltage space vector is to control the voltage within the allowable range to prevent exceeding the maximum voltage carrying capacity of the motor. The normalized voltage space vector value is expressed by the following formula:

[0076] ;

[0077] in, is the normalized voltage space vector, is the maximum voltage value of the motor. The normalized voltage space vector is compared with the upper limit of the voltage modulation ratio to obtain the threshold value of the linear PWM region. , and its calculation formula is:

[0078] ;

[0079] In this formula, It is the upper limit of the voltage modulation ratio and determines the maximum output range of PWM modulation. According to this threshold, the necessary voltage control parameters are provided for the subsequent PWM modulation process. The linear PWM region threshold and the rotor speed value are nonlinearly mapped to obtain the overmodulation starting point data. Rotor speed value It is an important parameter of motor speed, which affects the dynamic response and modulation strategy of the motor. In this step, the nonlinear mapping relationship between the rotor speed and the PWM region threshold is described by the following formula:

[0080] ;

[0081] in, It is a nonlinear mapping function. It adjusts the overmodulation starting point data according to the relationship between the motor speed and the PWM threshold to ensure that the modulation effect of the motor can maintain the best state at different speeds, avoid overmodulation during low-speed operation, and thus affect the efficiency and stability of the motor. According to the overmodulation starting point data, the voltage vector trajectory is segmented to obtain the boundary value of the overmodulation interval. The overmodulation interval refers to the area where the voltage space vector can continue to modulate after exceeding the linear PWM area. By segmenting the voltage vector trajectory, the maximum voltage value and current response of each stage are calculated to determine the boundary of the overmodulation interval. The maximum voltage utilization rate is calculated for the boundary value of the overmodulation interval to obtain the voltage limit value. The voltage limit value refers to the maximum voltage value that the motor can withstand under given working conditions. By calculating the maximum voltage utilization rate, the operation of the motor is ensured to be within a safe range while improving efficiency. The calculation formula for the voltage limit value is:

[0082] ;

[0083] The voltage limit value is matched with the current space vector value to obtain the overmodulation region threshold, ensuring that the motor voltage output can match the load demand and maintain the best operating state under different current space vectors. The calculation formula for the overmodulation region threshold is:

[0084] ;

[0085] The overmodulation region threshold is obtained through calculation.

[0086] In a specific embodiment, the process of executing step 103 may specifically include the following steps:

[0087] Inputting the linear PWM region threshold and the overmodulation region threshold into the initial layer of the Kriging proxy model, and performing feature mapping through the Gaussian kernel function in the initial layer to obtain high-dimensional feature mapping data;

[0088] The high-dimensional feature mapping data is input into the Kriging interpolation model to calculate the initial sampling points of the optimization parameters, and the initial sampling points of the optimization parameters are obtained. The Kriging interpolation model includes an average response term and a correlation term. The average response term is constructed by a second-order polynomial fitting, and the correlation term is constructed by an exponential variation function.

[0089] The initial sampling points of the optimization parameters are input into a three-layer response surface network to obtain the voltage distortion and harmonic response values. The first layer of the three-layer response surface network contains 32 neurons and uses the ReLU activation function, the second layer contains 16 neurons and uses the Sigmoid activation function, and the third layer contains 8 neurons and uses the Tanh activation function.

[0090] Adaptively weight the voltage distortion and harmonic response values ​​to obtain the voltage quality evaluation index, and iteratively calculate the copper loss and inverter loss based on the initial sampling point of the optimized parameters to obtain the system loss evaluation index;

[0091] A multi-objective optimization function is constructed based on voltage quality evaluation indicators and system loss evaluation indicators, and a Pareto optimal solution search is performed to obtain the optimal solution candidate set;

[0092] Based on the optimal solution candidate set, dominance sorting and crowding calculation are performed, and the solution vector on the Pareto front is selected as the optimal control parameter to obtain the initial value of the d-axis current command and the initial value of the q-axis current command;

[0093] The space vector PWM parameters are optimized based on the d-axis current command initial value and the q-axis current command initial value to obtain an initial control parameter group, which includes a d-axis current command value, a q-axis current command value, a PWM switching frequency value, and a voltage vector action time value.

[0094] Specifically, the linear PWM region threshold and the overmodulation region threshold are input into the initial layer of the Kriging proxy model. In this process, the initial layer uses the Gaussian kernel function for feature mapping. The Kriging proxy model is a proxy model for optimization. By mapping the features of the input data, it can predict and approximate the output response of a complex system. The role of the Gaussian kernel function in feature mapping is to convert the low-dimensional data of the input space into a high-dimensional feature space, providing a more accurate model for subsequent interpolation calculations. The mathematical expression of the Gaussian kernel function is:

[0095] ;

[0096] in, and are two data points in the input space, is the bandwidth parameter of the Gaussian kernel function, which controls the influence range between data points. Through the action of the Gaussian kernel function, the input data is mapped to a high-dimensional space, so that complex nonlinear relationships can be better approximated and processed. The high-dimensional feature mapping data is input into the Kriging interpolation model to calculate the initial sampling points of the optimization parameters. The Kriging interpolation model is a process of finding the optimal solution by modeling the input data and calculating its response. It includes an average response term and a correlation term, in which the average response term uses a second-order polynomial fitting to approximate the global trend of the model, while the correlation term captures local changes through an exponential variation function. The formula for the average response term of the second-order polynomial is:

[0097] ;

[0098] in, are the coefficients of the second-order polynomial, is the input variable. The exponential variogram is expressed by the following formula:

[0099] ;

[0100] in, It is a parameter that controls the range of correlation. By adjusting this parameter, the sensitivity of the model to the input data can be adjusted. The Kriging interpolation model provides a response prediction of the input space through the combination of these two items, thereby calculating the initial optimized sampling points. The initial sampling points are input into the three-layer response surface network to calculate the response values ​​of voltage distortion and harmonic quantity. The function of the response surface network is to process the input data through a series of neural network layers to obtain system performance indicators. The network consists of three levels, the first layer contains 32 neurons and uses the ReLU activation function, the second layer contains 16 neurons and uses the Sigmoid activation function, and the third layer contains 8 neurons and uses the Tanh activation function. The formula of the ReLU activation function is:

[0101] ;

[0102] Its function is to ensure that the output value is non-negative by setting negative values ​​to zero. The Sigmoid activation function is:

[0103] ;

[0104] Its output value is limited between 0 and 1, which is suitable for processing binary classification problems. The formula of Tanh activation function is:

[0105] ;

[0106] It compresses the output value to between -1 and 1 and is used to process polarized data. After being processed by the three-layer response surface network, the response values ​​of the voltage distortion and harmonic quantity are adaptively weighted to obtain the voltage quality evaluation index. The voltage quality evaluation index is the weighted sum of multiple performance indicators, reflecting the voltage quality of the motor operation. In this process, based on the initial sampling point of the optimization parameters, the copper loss and inverter loss are iteratively calculated to obtain the system loss evaluation index. The copper loss and inverter loss are calculated using the following formula:

[0107] ;

[0108] ;

[0109] in, is the current, is the motor winding resistance, is the efficiency of the inverter, is the maximum power of the inverter. Based on these evaluation indicators, a multi-objective optimization function is constructed and a Pareto optimal solution search is performed. The multi-objective optimization function aims to optimize multiple objectives at the same time, such as voltage quality, system loss, etc. Its goal is to find a solution that performs best in multiple aspects. Pareto optimal solution search is a multi-objective optimization method, the purpose of which is to find a solution that cannot be optimized on a certain objective and cannot be further improved on other objectives. Through the calculated optimal solution candidate set, dominance sorting and congestion calculation are performed, and the solution vector on the Pareto front is selected as the optimal control parameter to obtain the initial value of the d-axis current command and the initial value of the q-axis current command. Based on the initial values ​​of the d-axis and q-axis current commands, the space vector PWM parameters are optimized to obtain the initial control parameter group. The space vector PWM optimization adjusts the voltage vector action time of each switching cycle according to the actual needs of the motor, thereby achieving efficient motor control. The initial control parameter group obtained includes the d-axis current command value, the q-axis current command value, the PWM switching frequency value, and the voltage vector action time value.

[0110] Among them, the high-dimensional feature mapping data is input into the Kriging interpolation model to calculate the initial sampling points of the optimization parameters to obtain the initial sampling points of the optimization parameters. The Kriging interpolation model includes an average response term and a correlation term. The average response term is constructed by a second-order polynomial fitting, and the correlation term is constructed using an exponential variation function, including: performing a polynomial trend decomposition operation on the high-dimensional feature mapping data, calculating the second-order polynomial coefficients by the least squares method, and obtaining an average response polynomial model; constructing an exponential variation function based on the average response polynomial model, and using the maximum likelihood estimation method to determine the variation function parameters to obtain a spatial correlation model; The spatial correlation model is input into the Kriging weight calculation unit, and the Lagrange multiplier method is used to solve the Kriging equations to obtain the interpolation weight coefficients; the interpolation weight coefficients are normalized, and the cross-validation method is used to evaluate the interpolation accuracy to obtain the weight optimization parameters; the weight optimization parameters are input into the goodness of fit evaluation module, and the goodness of fit evaluation results are obtained by calculating the root mean square error and the coefficient of determination; based on the goodness of fit evaluation results, the sampling point selection criteria are constructed, and the Latin hypercube sampling method is used for spatial point distribution to obtain the candidate sampling point set; the candidate sampling point set is subjected to spatial correlation analysis, and the spatial distribution characteristics are calculated using the Ripley K function to obtain the spatial distribution characteristic values; based on the spatial distribution characteristic values, the candidate sampling point set is screened and optimized, and the final sampling points are determined using the maximum and minimum distance criteria to obtain the initial sampling points of the optimization parameters.

[0111] In a specific embodiment, the process of executing step 104 may specifically include the following steps:

[0112] Perform spectrum analysis based on the d-axis current command value and the q-axis current command value in the initial control parameter group to obtain a current harmonic spectrum;

[0113] Perform odd and even harmonic classification calculations on the current harmonic spectrum to obtain classified harmonic data, wherein odd harmonics include 3rd, 5th, and 7th harmonic components, and even harmonics include 2nd, 4th, and 6th harmonic components;

[0114] The classified harmonic data is input into a neural network harmonic estimator for feature extraction to obtain a harmonic feature vector. The neural network harmonic estimator includes 8 neurons in the input layer, 16 neurons in the hidden layer, and 4 neurons in the output layer.

[0115] The harmonic compensation of the PWM switching sequence is calculated according to the harmonic characteristic vector to obtain the PWM compensation sequence, and the PWM compensation sequence is processed by a band-stop filter to obtain the current harmonic compensation value;

[0116] A discrete state observer is used to construct a system delay model. The state variables include current and speed. The sampling delay analysis is performed on the PWM switching frequency value in the initial control parameter group to obtain the delay characteristic parameters.

[0117] Prediction compensation is performed based on delay characteristic parameters, and the second-order Runge-Kutta method is used to predict the system state in advance. The prediction step is one control cycle, and the system delay compensation value is obtained.

[0118] Specifically, a spectrum analysis operation is performed based on the d-axis current command value and the q-axis current command value in the initial control parameter group to obtain the current harmonic spectrum. The current harmonic spectrum is calculated by fast Fourier transform. Assume that the time domain expressions of the d-axis and q-axis currents are and , whose spectrum is expressed by the following formula:

[0119] ;

[0120] ;

[0121] in, and are the spectrum of the d-axis and q-axis currents, respectively. is the frequency, is an imaginary unit. The current signal in the time domain is converted into a frequency domain signal through fast Fourier transform, revealing the amplitude and phase information of each frequency component, forming the current harmonic spectrum. The current harmonic spectrum is classified and calculated by odd and even orders. Odd harmonics include 3rd, 5th, 7th and other components, while even harmonics include 2nd, 4th, 6th and other components. The classification process is implemented through the following logic: for each frequency component , if the harmonic order is an odd number (i.e., satisfies ,in is an integer), it is classified as an odd harmonic; if its harmonic order is an even number (i.e., it satisfies ), it is divided into even harmonics. The classified data is expressed as:

[0122] ;

[0123] ;

[0124] These classified harmonic data can reflect the characteristics of different types of harmonics in the current. The classified harmonic data are input into the neural network harmonic estimator for feature extraction to obtain the harmonic feature vector. The harmonic estimator is a multi-layer neural network, which includes an input layer, a hidden layer and an output layer. The input layer contains 8 neurons, the hidden layer contains 16 neurons, and the output layer contains 4 neurons. Assume that the input classified harmonic data is , the neural network performs nonlinear mapping on it, and calculates the harmonic feature vector through the weight matrix and activation function The calculation formula for each layer is:

[0125] ;

[0126] in, It is The output of the layer, is the weight matrix, is the bias vector, is the activation function. The first layer uses the ReLU activation function, defined as:

[0127] ;

[0128] The second layer uses the Sigmoid activation function, which is defined as:

[0129] ;

[0130] The third layer uses the Tanh activation function, defined as:

[0131] ;

[0132] Through the feature extraction of neural network, the harmonic feature vector obtained can effectively describe the distribution characteristics of current harmonics. According to the harmonic feature vector, the harmonic compensation calculation is performed on the PWM switching sequence to generate a PWM compensation sequence. The compensation calculation reduces the harmonic component by adjusting the switching time of the original PWM signal, thereby optimizing the current quality. The generated PWM compensation sequence needs to be processed by a band-stop filter to suppress the harmonic components of a specific frequency. The transfer function of the band-stop filter is expressed as

[0133] ;

[0134] in, is the complex frequency variable in the Laplace transform, and are the low frequency and center frequency parameters of the filter respectively. The filtered PWM signal can significantly reduce the impact of harmonics and finally obtain the current harmonic compensation value. At the same time, a discrete state observer is used to build a system delay model. The state variables of the state observer include current and speed, which are used to determine the delay characteristic parameters of the system through sampling delay analysis. For the PWM switching frequency value , the sampling delay is calculated by the following formula:

[0135] ;

[0136] in, It's the delay. yes Switching frequency. The delay characteristic parameter reflects the transmission delay of the control signal in the motor system. Based on the delay characteristic parameter, prediction compensation is performed, and the second-order Runge Kutta method is used to predict the system state in advance. The calculation formula of the second-order Runge Kutta method is:

[0137] ;

[0138] in, is the current state, is the step size (one control cycle), It is a state change function. Through prediction, the control signal can be adjusted in advance before the system response lags, and the delay compensation value of the system can be obtained.

[0139] In a specific embodiment, the process of executing step 105 may specifically include the following steps:

[0140] The current harmonic compensation value and system delay compensation value are input into the fuzzy evaluation network for online status evaluation to obtain the compensation effect scoring data. The fuzzy evaluation network contains 7 input nodes, 12 hidden nodes and 3 output nodes, and adopts a triangular membership function.

[0141] Perform multi-level threshold segmentation on the compensation effect scoring data to obtain the optimal segmentation threshold, and calibrate the weight coefficient of each segment interval to obtain the segment evaluation interval;

[0142] Based on segmented evaluation interval construction Control parameter optimization model and perform optimal Switching timing calculation, get Control parameter correction value;

[0143] Inputting the PWM control parameter correction value into the active disturbance rejection controller for parameter correction to obtain the current controller parameter correction value, the active disturbance rejection controller includes a tracking differentiator, a nonlinear state error feedback law and an extended state observer;

[0144] The state space model is constructed according to the correction values ​​of the current controller parameters, and the closed-loop stability is verified by characteristic root analysis and Lyapunov stability criterion to obtain the system stability boundary.

[0145] The voltage vector sequence within the system stability boundary is optimized to obtain the voltage vector distribution correction value.

[0146] Specifically, the current harmonic compensation value and the system delay compensation value are input into the fuzzy evaluation network for online status evaluation to obtain the compensation effect score data. The fuzzy evaluation network has 7 input nodes, 12 hidden nodes and 3 output nodes, and adopts a triangular membership function. The input nodes of the fuzzy evaluation network represent the current harmonic compensation value, the system delay compensation value and possible system status characteristics, while the output nodes represent different scores of the compensation effect. The fuzzy membership function is used to convert the actual value of the input into a fuzzy value. The triangular membership function is defined as:

[0147] ;

[0148] in, is the input variable, is the center of the membership function, is the width of the membership function. This function is used to map the input data to the fuzzy set and obtain the value of each output node through weighted averaging and inference process. After the fuzzy evaluation is completed, the compensation effect score data obtained enters the multi-level threshold segmentation process. By dividing the compensation effect score data into intervals, the data is divided into multiple levels, each level represents a different control strategy. Each segmented interval will be calibrated with a weight coefficient, and the selection of the weight coefficient is based on the response characteristics of the system and the needs of the control strategy. Assume that the compensation effect score data is , segment these data according to the preset segmentation threshold. For example, define the following three intervals:

[0149] ;

[0150] ;

[0151] ;

[0152] in, and is the threshold used to divide the compensation effect. In this way, the compensation effect is graded and the corresponding weight coefficient is defined in each level, for example:

[0153] ;

[0154] The data of each segment interval is weighted and summed by these weight coefficients to obtain the optimization effect of each segment interval. Based on the segmented evaluation interval, a PWM control parameter optimization model is constructed. This model adjusts the PWM control parameters according to the evaluation results, optimizes the PWM switching timing, and ensures the performance and stability of the motor. The optimization of PWM switching timing needs to consider multiple factors, including control response time, power loss, thermal management, etc. Assuming that the PWM switching timing optimization goal is to minimize the mismatch between voltage and current, the control model is expressed as:

[0155] ;

[0156] in, It is The voltage value at the moment, is the corresponding current value, is the total number of PWM cycles. Through this optimization process, the optimal PWM switching timing is obtained, and the PWM control parameter correction value is obtained. The PWM control parameter correction value is input into the ADRC for parameter correction. The ADRC is used to eliminate disturbances and nonlinear effects in the system to ensure the stability of the system under various working conditions. The ADRC includes a tracking differentiator, a nonlinear state error feedback law, and an extended state observer. The tracking differentiator is used to estimate the control error, and the control error formula is:

[0157] ;

[0158] in, is the desired control target, is the actual output of the system. By tracking the error at high frequency, the system response is ensured to be accurate. The nonlinear state error feedback law controls the error by adjusting the gain, and the formula is expressed as:

[0159] ;

[0160] in, is the feedback gain, is the error, and are the first and second derivatives of the error, respectively. In this way, the controller can effectively suppress disturbances and improve the robustness of the system. The output of the ADRC is the corrected current controller parameter value. Assume that the corrected current controller parameter is Based on the corrected current controller parameters, the state space model of the system is constructed. The state space model describes the dynamic behavior of the motor and provides a method to analyze the stability of the system. Assume that the state vector of the system is , the state equation is:

[0161] ;

[0162] ;

[0163] in, is the system matrix, is the input, is the output. By analyzing these matrices, the stability of the system is determined. Eigenvalue analysis and Lyapunov stability criterion are commonly used tools to verify whether the system is stable. The eigenvalue analysis method evaluates the stability of the system by solving the eigenvalues ​​of the system matrix. If the eigenvalues ​​of the system matrix are all in the left half complex plane, the system is asymptotically stable:

[0164] ;

[0165] in, is the eigenvalue of the system matrix. The Lyapunov stability criterion verifies the stability of the system by constructing the Lyapunov function, selecting the quadratic function as the Lyapunov function, and verifying whether it satisfies the inequality:

[0166] ;

[0167] If this condition is met, the system is stable. Within the system stability boundary, the voltage vector sequence is optimized to obtain the voltage vector distribution correction value. The optimization goal of the voltage vector is to minimize the energy loss of the system and improve efficiency. The optimization method is carried out through the gradient descent method or genetic algorithm, and the objective function is expressed as:

[0168] ;

[0169] in, is the optimized voltage vector value, is the current voltage vector value. Through the optimization method, the final voltage vector distribution correction value is obtained.

[0170] In a specific embodiment, the process of executing step 106 may specifically include the following steps:

[0171] Based on the PWM control parameter correction value, the current controller parameter correction value and the voltage vector distribution correction value, a feature fusion operation is performed to obtain a correction parameter feature vector;

[0172] Perform single neutral point symmetrical winding fault analysis on the correction parameter characteristic vector, calculate the three-phase current imbalance by setting 6 symmetrical component detection units, and obtain the phase current deviation vector;

[0173] A fault phase location model is constructed based on the phase current deviation vector, and the root mean square deviation value of each phase current is calculated by recursive least square method to obtain the fault phase identification data;

[0174] The fault phase identification data is input into the multi-layer decision tree network to determine the fault level and obtain the fault phase isolation instruction. The multi-layer decision tree network includes 3 decision layers, and each layer has 4 decision nodes.

[0175] According to the fault phase isolation instruction, an open-circuit phase mathematical model is established, and the remaining phase current distribution is simulated and calculated to obtain the current distribution characteristic quantity, and a single neutral point reconstruction is performed based on the current distribution characteristic quantity to obtain a current reconstruction scheme;

[0176] The robustness analysis of the current reconstruction scheme is carried out, and the controller parameters are designed by constructing a sliding mode surface with an exponential reaching law to obtain fault-tolerant control instructions.

[0177] Specifically, feature fusion operation is performed based on the PWM control parameter correction value, the current controller parameter correction value and the voltage vector allocation correction value to obtain the correction parameter feature vector. The PWM control parameter correction value, the current current controller parameter correction value and the voltage vector allocation correction value jointly determine the response performance of the system. Through feature fusion operation, these three are integrated into a correction parameter feature vector, which contains the comprehensive characteristics of all control inputs and can reflect the current state of the control system. Assume that the PWM control parameter correction value is , the current controller parameter correction value is , the voltage vector distribution correction value is , through weighted averaging or other fusion algorithms, the modified parameter feature vector is obtained:

[0178] ;

[0179] in, are weighting coefficients, the selection of which optimizes system performance by adjusting the control effect. In this way, the modified parameter eigenvector Contains information about all major control parameters. Perform single neutral point symmetrical winding fault analysis on the eigenvector to detect and diagnose the imbalance of the three-phase current of the motor. Set up 6 symmetrical component detection units to calculate the imbalance of each phase current. The current imbalance is calculated using the following formula:

[0180] ;

[0181] in, It is The instantaneous value of the phase current, is the average value of the current. By calculating the imbalance, the phase current deviation vector is obtained , which represents the deviation between each phase current and the expected value. Based on the phase current deviation vector , build a fault phase location model. The key to fault location is to accurately locate the specific phase where the fault occurs by analyzing the current deviation. The recursive least squares method is used to calculate the root mean square deviation value of each phase current. The recursive least squares method continuously corrects the estimated value to accurately track the change of current. The formula is:

[0182] ;

[0183] ;

[0184] in, is the error, is the weighting coefficient, is the covariance matrix, It's the forgetting factor. is the input vector. Through the recursive least squares method, the root mean square deviation value of each phase is obtained in the process of continuous updating, and by calculating these values, the fault phase is identified. After calculating the root mean square deviation value of each phase, the fault phase identification data is obtained. Assume that the fault phase identification data is , which indicates the phase in the motor where the fault occurred. For example, if Indicates that the first phase is faulty, and further processing is performed based on this information. The fault phase identification data is input into the multi-layer decision tree network for fault level determination. The multi-layer decision tree network includes 3 determination layers, each layer contains 4 determination nodes, and each node corresponds to a decision rule. The decision tree network makes decisions on the input data and outputs the isolation instruction of the faulty phase. The specific determination process is achieved by traversing the path of the decision tree, and finally outputs which phase should be isolated. Each node of the decision tree is selected based on a condition. For example, when the phase current deviation is greater than a certain threshold, the phase is determined to be a faulty phase. In this way, the decision tree can accurately determine the faulty phase and output an isolation instruction. A mathematical model of the open phase is established based on the fault phase isolation instruction. The model of the open phase assumes that the current of the faulty phase is zero, so the current of the remaining phases needs to be redistributed. Assume that the current distribution of the remaining phases is calculated by the following model:

[0185] ;

[0186] in, is the total current, is the fault phase current. On this basis, the current is simulated and calculated to obtain the current distribution characteristic quantity. Assume that the current distribution characteristic quantity is , which represents the distribution of the remaining phase current among the phases. After obtaining the current distribution characteristic, single neutral point reconstruction is performed. Based on the existing current distribution information, the normal operation state of the system after the fault occurs is restored. The current reconstruction is achieved by constructing a suitable mathematical model and control strategy. Assume that the current reconstruction scheme is , which is obtained by calculation. For example, the distribution of the remaining phase current is adjusted by the optimization algorithm so that the current is as balanced as possible between the phases. The robustness analysis of the current reconstruction scheme is carried out to verify the stability and reliability of the current reconstruction scheme under different working conditions. The controller parameters are designed by constructing the sliding surface of the exponential reaching law. The sliding surface of the exponential reaching law is constructed by the following formula:

[0187] ;

[0188] in, is the sliding surface, is the approach rate, is the power index, is the system error. By constructing the sliding surface, appropriate controller parameters are designed so that the system can still operate stably when a fault occurs. According to the robustness analysis results, fault-tolerant control instructions are obtained to ensure that the motor can still maintain its function when a fault occurs.

[0189] The above describes the intelligent control method of the high-speed motor of the heat dissipation fan in the embodiment of the present invention. The following describes the intelligent control device of the high-speed motor of the heat dissipation fan in the embodiment of the present invention. Figure 2 , an embodiment of the high-speed motor intelligent control device of the heat dissipation fan in the embodiment of the present invention includes:

[0190] The acquisition module 201 is used to acquire parameters of the high-speed fan motor to obtain a basic parameter group of the motor;

[0191] A calculation module 202 is used to calculate the motor overmodulation boundary based on the motor basic parameter group to obtain a linear PWM region threshold and an overmodulation region threshold;

[0192] The optimization module 203 is used to input the linear PWM region threshold and the overmodulation region threshold into the Kriging agent model, perform multi-objective optimization operations on the voltage distortion, current harmonics and system loss, and obtain an initial control parameter group;

[0193] An analysis module 204 is used to perform current harmonic analysis based on the initial control parameter group to obtain a current harmonic compensation value and a system delay compensation value;

[0194] An evaluation module 205 is used to perform an online state evaluation operation on the current harmonic compensation value and the system delay compensation value to obtain a PWM control parameter correction value, a current controller parameter correction value and a voltage vector distribution correction value;

[0195] The output module 206 is used to analyze the fault characteristics of the high-speed fan motor in the running state according to the PWM control parameter correction value, the current controller parameter correction value and the voltage vector distribution correction value, and output the fault phase isolation instruction and the fault-tolerant control instruction.

[0196] Through the coordinated cooperation of the above-mentioned components, by establishing a multi-objective optimization framework based on the Kriging proxy model, combined with the comprehensive evaluation of voltage distortion, current harmonics and system losses, the control accuracy and energy utilization efficiency of the system are significantly improved; the improved PWM control strategy and harmonic compensation method are adopted to effectively solve the voltage distortion and current harmonic problems during high-speed operation, making the motor run more smoothly; an online state evaluation algorithm is designed to enhance the dynamic response capability and anti-interference ability of the system by adjusting the control parameters in real time; based on the single neutral point symmetrical winding structure, an accurate fault diagnosis and fault-tolerant control strategy is developed to improve the reliability and safety of the system; through the multi-level parameter optimization and correction mechanism, the global optimization of system performance is achieved, which significantly improves the overall operation effect of the high-speed fan motor.

[0197] above Figure 2 The high-speed motor intelligent control device of the cooling fan in the embodiment of the present invention is described in detail from the perspective of modular functional entities. The high-speed motor intelligent control device of the cooling fan in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0198] Figure 3 1 is a schematic diagram of the structure of a high-speed motor intelligent control device for a heat dissipation fan provided by an embodiment of the present invention. The high-speed motor intelligent control device 300 for the heat dissipation fan may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 310 (for example, one or more processors) and a memory 320, and one or more storage media 330 (for example, one or more mass storage device terminals) storing application programs 333 or data 332. Among them, the memory 320 and the storage medium 330 may be short-term storage or permanent storage. The program stored in the storage medium 330 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the high-speed motor intelligent control device 300 for the heat dissipation fan. Furthermore, the processor 310 may be configured to communicate with the storage medium 330, and execute a series of instruction operations in the storage medium 330 on the high-speed motor intelligent control device 300 for the heat dissipation fan to implement the steps of the above-mentioned high-speed motor intelligent control method for the heat dissipation fan.

[0199] The high-speed motor intelligent control device 300 of the cooling fan may also include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input and output interfaces 360, and / or one or more operating systems 331, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. It can be understood by those skilled in the art that Figure 3 The structure of the high-speed motor intelligent control device for the cooling fan shown does not constitute a limitation on the high-speed motor intelligent control device for the cooling fan provided by the present invention, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0200] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the intelligent control method for the high-speed motor of the cooling fan.

[0201] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0202] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0203] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for intelligently controlling a high-speed motor of a cooling fan, characterized in that: The method comprises: Parameters of the high-speed fan motor are collected to obtain a basic parameter group of the motor; specifically, the method includes: applying a DC excitation voltage to the stator winding of the high-speed fan motor for measurement operation to obtain a stator winding resistance measurement value and a stator leakage inductance measurement value, and performing winding impedance calculation based on the stator winding resistance measurement value and the stator leakage inductance measurement value to obtain a stator impedance characteristic parameter; performing multi-point sampling measurement on the permanent magnet magnetic field strength of the high-speed fan motor to obtain permanent magnet magnetic field distribution data, and performing flux calculation based on the permanent magnet magnetic field distribution data to obtain a permanent magnet flux characteristic value; inputting the three-phase current signal of the high-speed fan motor into a sampling processing unit for analog-to-digital conversion operation to obtain a three-phase current instantaneous value, and performing current vector synthesis calculation based on the three-phase current instantaneous value to obtain a current space vector value; performing multi-cycle sampling processing on the output signal of the rotor position sensor of the high-speed fan motor to obtain a rotor position angle value and a rotor speed value; taking the stator impedance characteristic parameter, the permanent magnet flux characteristic value, the current space vector value, the rotor position angle value and the rotor speed value as the motor basic parameter group; Based on the motor basic parameter group, the motor overmodulation boundary is calculated to obtain the linear PWM area threshold and the overmodulation area threshold; specifically including: based on the stator impedance characteristic parameter and the current space vector value, the voltage vector is calculated to obtain the reference voltage vector value; the reference voltage vector value is segmented into six phases to obtain six basic voltage vector components; vector synthesis calculation is performed according to the six basic voltage vector components and the rotor position angle value to obtain the voltage space vector synthesis value; the voltage space vector synthesis value is normalized and compared with the upper limit of the voltage modulation ratio to obtain the linear PWM area threshold; the linear PWM area threshold is nonlinearly mapped with the rotor speed value to obtain the overmodulation starting point data; the voltage vector trajectory is segmented according to the overmodulation starting point data to obtain the overmodulation interval boundary value; the maximum voltage utilization rate is calculated for the overmodulation interval boundary value to obtain the voltage limit value, and the voltage limit value is matched with the current space vector value to obtain the overmodulation area threshold; Inputting the linear PWM region threshold and the overmodulation region threshold into the Kriging proxy model, performing multi-objective optimization calculations on voltage distortion, current harmonics and system loss, and obtaining an initial control parameter group; Performing current harmonic analysis based on the initial control parameter group to obtain a current harmonic compensation value and a system delay compensation value; Performing an online state evaluation operation on the current harmonic compensation value and the system delay compensation value to obtain a PWM control parameter correction value, a current controller parameter correction value, and a voltage vector distribution correction value; The fault characteristics of the operating state of the high-speed fan motor are analyzed according to the PWM control parameter correction value, the current controller parameter correction value and the voltage vector distribution correction value, and a fault phase isolation instruction and a fault-tolerant control instruction are output.

2. The intelligent control method for a high-speed motor of a cooling fan according to claim 1, characterized in that: The linear PWM region threshold and the overmodulation region threshold are input into the Kriging proxy model, and a multi-objective optimization operation is performed on the voltage distortion, the current harmonic amount and the system loss to obtain an initial control parameter group, including: Inputting the linear PWM region threshold and the overmodulation region threshold into the initial layer of the Kriging proxy model, and performing feature mapping through the Gaussian kernel function in the initial layer to obtain high-dimensional feature mapping data; Inputting the high-dimensional feature mapping data into a Kriging interpolation model to calculate the initial sampling points of the optimization parameters to obtain the initial sampling points of the optimization parameters, wherein the Kriging interpolation model includes an average response term and a correlation term, wherein the average response term is constructed by a second-order polynomial fitting, and the correlation term is constructed by an exponential variation function; Inputting the initial sampling points of the optimization parameters into a three-layer response surface network to obtain voltage distortion and harmonic response values, wherein the first layer of the three-layer response surface network includes 32 neurons and uses a ReLU activation function, the second layer includes 16 neurons and uses a Sigmoid activation function, and the third layer includes 8 neurons and uses a Tanh activation function; Performing adaptive weighted calculations on the voltage distortion and harmonic response values ​​to obtain a voltage quality evaluation index, and iteratively calculating the copper loss and inverter loss based on the initial sampling point of the optimization parameters to obtain a system loss evaluation index; Constructing a multi-objective optimization function based on the voltage quality evaluation index and the system loss evaluation index, and performing a Pareto optimal solution search to obtain an optimal solution candidate set; Based on the optimal solution candidate set, dominance sorting and congestion calculation are performed, and a solution vector on the Pareto front is selected as an optimal control parameter to obtain an initial value of a d-axis current command and an initial value of a q-axis current command; Based on the d-axis current command initial value and the q-axis current command initial value, space vector PWM parameter optimization is performed to obtain an initial control parameter group, wherein the initial control parameter group includes a d-axis current command value, a q-axis current command value, a PWM switching frequency value, and a voltage vector action time value.

3. The intelligent control method for a high-speed motor of a cooling fan according to claim 2, characterized in that: The performing current harmonic analysis based on the initial control parameter group to obtain a current harmonic compensation value and a system delay compensation value includes: Performing spectrum analysis calculation based on the d-axis current command value and the q-axis current command value in the initial control parameter group to obtain a current harmonic spectrum; Performing odd-order and even-order harmonic classification calculations on the current harmonic spectrum to obtain classified harmonic data, wherein odd-order harmonics include 3rd, 5th, and 7th harmonic components, and even-order harmonics include 2nd, 4th, and 6th harmonic components; Inputting the classified harmonic data into a neural network harmonic evaluator for feature extraction to obtain a harmonic feature vector, wherein the neural network harmonic evaluator comprises 8 neurons in an input layer, 16 neurons in a hidden layer, and 4 neurons in an output layer; Performing harmonic compensation calculation on the PWM switching sequence according to the harmonic characteristic vector to obtain a PWM compensation sequence, and processing the PWM compensation sequence through a band-stop filter to obtain a current harmonic compensation value; A system delay model is constructed by using a discrete state observer, the state variables include current and speed, and a sampling delay analysis is performed on the PWM switching frequency value in the initial control parameter group to obtain a delay characteristic parameter; Prediction compensation is performed based on the delay characteristic parameters, and the system state is predicted in advance using the second-order Runge-Kutta method, with a prediction step length of one control cycle, to obtain a system delay compensation value.

4. The intelligent control method for a high-speed motor of a cooling fan according to claim 3, characterized in that: The performing of online state evaluation operation on the current harmonic compensation value and the system delay compensation value to obtain a PWM control parameter correction value, a current controller parameter correction value and a voltage vector distribution correction value includes: Inputting the current harmonic compensation value and the system delay compensation value into a fuzzy evaluation network for online status evaluation to obtain compensation effect scoring data, wherein the fuzzy evaluation network comprises 7 input nodes, 12 hidden layer nodes and 3 output nodes, and adopts a triangular membership function; Perform multi-level threshold segmentation on the compensation effect score data to obtain the optimal segmentation threshold, and calibrate the weight coefficient of each segment interval to obtain the segment evaluation interval; Based on the segmented evaluation interval, a PWM control parameter optimization model is constructed, and an optimal PWM switch timing is calculated to obtain a PWM control parameter correction value; Inputting the PWM control parameter correction value into an active disturbance rejection controller for parameter correction to obtain a current controller parameter correction value, wherein the active disturbance rejection controller includes a tracking differentiator, a nonlinear state error feedback law and an extended state observer; A state space model is constructed according to the current controller parameter correction value, and closed-loop stability is verified by characteristic root analysis and Lyapunov stability criterion to obtain a system stability boundary; The voltage vector sequence within the system stability boundary is optimized to obtain a voltage vector distribution correction value.

5. The intelligent control method for a high-speed motor of a cooling fan according to claim 4, characterized in that: The method of performing a fault characteristic analysis on the operating state of the high-speed fan motor according to the PWM control parameter correction value, the current controller parameter correction value and the voltage vector distribution correction value, and outputting a fault phase isolation instruction and a fault-tolerant control instruction, comprises: Performing feature fusion operation based on the PWM control parameter correction value, the current controller parameter correction value and the voltage vector allocation correction value to obtain a correction parameter feature vector; Performing a single neutral point symmetrical winding fault analysis operation on the correction parameter characteristic vector, calculating the three-phase current imbalance by setting 6 symmetrical component detection units, and obtaining a phase current deviation vector; A fault phase location model is constructed based on the phase current deviation vector, and a root mean square deviation value of each phase current is calculated by a recursive least square method to obtain fault phase identification data; Inputting the fault phase identification data into a multi-layer decision tree network to determine the fault level and obtain a fault phase isolation instruction, wherein the multi-layer decision tree network comprises three determination layers, each layer having four determination nodes; Establishing an open-circuit phase mathematical model according to the fault phase isolation instruction, performing simulation calculation on the remaining phase current distribution to obtain a current distribution characteristic quantity, and performing single neutral point reconstruction based on the current distribution characteristic quantity to obtain a current reconstruction scheme; The robustness analysis of the current reconstruction scheme is carried out, and the controller parameters are designed by constructing a sliding mode surface of an exponential reaching law to obtain fault-tolerant control instructions.

6. A high-speed motor intelligent control device for a cooling fan, characterized in that: A method for intelligently controlling a high-speed motor of a cooling fan according to any one of claims 1 to 5, the device comprising: The acquisition module is used to acquire parameters of the high-speed fan motor and obtain a basic parameter group of the motor; A calculation module, used for calculating the motor overmodulation boundary based on the motor basic parameter group to obtain a linear PWM region threshold and an overmodulation region threshold; An optimization module, used for inputting the linear PWM region threshold and the overmodulation region threshold into a Kriging proxy model, performing multi-objective optimization operations on voltage distortion, current harmonics and system loss, and obtaining an initial control parameter group; An analysis module, configured to perform current harmonic analysis based on the initial control parameter group to obtain a current harmonic compensation value and a system delay compensation value; An evaluation module, used for performing an online state evaluation operation on the current harmonic compensation value and the system delay compensation value to obtain a PWM control parameter correction value, a current controller parameter correction value and a voltage vector distribution correction value; The output module is used to perform fault characteristic analysis on the operating state of the high-speed fan motor according to the PWM control parameter correction value, the current controller parameter correction value and the voltage vector distribution correction value, and output fault phase isolation instructions and fault-tolerant control instructions.

7. A high-speed motor intelligent control device for a cooling fan, characterized in that: The high-speed motor intelligent control device of the cooling fan comprises: a memory and at least one processor, wherein instructions are stored in the memory; The at least one processor calls the instruction in the memory so that the high-speed motor intelligent control device of the cooling fan executes the high-speed motor intelligent control method of the cooling fan according to any one of claims 1-5.

8. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the high-speed motor intelligent control method of the cooling fan according to any one of claims 1 to 5 is implemented.

Images