Compressor weak magnetic area stable operation method and system based on angle compensation

Abstract

The invention discloses a compressor weak magnetic area stable operation method and system based on angle compensation, and relates to the technical field of compressor weak magnetic area control, and the method comprises the steps that a rotor angle is estimated through a phase-locked loop, and an estimated angle is generated; processing the operation parameters of the target compressor, the estimated angle of the rotor and the electromagnetic data generated by finite element analysis by using a machine learning model to generate a prediction angle; and adjusting operation parameters of the compressor based on the predicted angle. And the machine learning model compensates the influence of multiple frequency components on angle prediction by combining the space harmonic component, the high-frequency component of the back electromotive force signal and the noise signal. In addition, the electromagnetic data generated by finite element analysis further enhances the adaptability of the machine learning model to the complex electromagnetic environment. Operation parameters of the compressor can be dynamically adjusted under complex working conditions, and operation stability and energy efficiency are improved.

Description

Technical Field

[0001] The present application relates to the technical field of compressor weak magnetic region control, and in particular to a method and system for stable operation of a compressor weak magnetic region based on angle compensation. Background Art

[0002] In modern compressor motor control systems, accurate estimation of the rotor angle is critical to achieving stable operation of the compressor. Especially when operating in the weak magnetic field region, the torque output of the motor may become unstable due to the weakening of the rotor magnetic field, resulting in reduced operating efficiency of the compressor. This situation usually occurs when the motor speed is high and the voltage is close to the rated voltage upper limit. Traditional control systems rely on a phase-locked loop to estimate the rotor angle and adjust it based on the phase error to maintain stable operation of the motor. However, under complex operating conditions, especially in the presence of multiple frequency interference signals such as spatial harmonics, high-frequency components and noise, traditional phase error detection methods may be difficult to cope with, resulting in inaccurate rotor angle estimation, which in turn affects the overall control performance of the system.

[0003] For example, the Chinese patent with the authorization announcement number CN114738977B discloses a method, device and storage medium for controlling the weak magnetic angle of an air conditioner. The method includes: in response to the air conditioner entering the weak magnetic control mode, obtaining the operating frequency of the air conditioner compressor, and monitoring the outdoor ambient temperature during the operation of the air conditioner compressor; adjusting the weak magnetic angle of the air conditioner compressor according to the operating frequency and the outdoor ambient temperature. Through this disclosure, different weak magnetic angles can be distinguished and added through different outdoor ambient temperature intervals and air conditioner compressor operating frequency difference intervals, so as to achieve refined control of weak magnetic control.

[0004] This technical solution is limited to rough adjustment based on the working conditions of the compressor, and does not fully consider the influence of multiple frequency components in the rotor angle estimation, resulting in limited control accuracy. In addition, the control system of the prior art usually lacks intelligent learning ability and is difficult to adapt to different operating conditions and complex environmental conditions. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present application discloses a method and system for stable operation of a compressor in a weak magnetic field region based on angle compensation.

[0006] In a first aspect, the present application provides a method for stable operation of a compressor in a weak magnetic field region based on angle compensation, comprising:

[0007] estimating a rotor angle of a target compressor based on a phase-locked loop to generate an estimated angle of the rotor;

[0008] Generate a predicted angle based on the operating parameters of the target compressor and the estimated angle of the rotor using a preset machine learning model; wherein the machine learning model is used to compensate for the influence of multiple frequency components on the predicted angle; the multiple frequency components include: spatial harmonic components, high frequency components of back electromotive force signals, and noise signals;

[0009] Based on the predicted angle, operating parameters of the target compressor are adjusted.

[0010] As an optional implementation manner, generating the estimated angle of the rotor further includes:

[0011] injecting a high frequency voltage signal into a stator winding of the target compressor motor;

[0012] Collecting a current response signal of the stator winding, and extracting a spatial harmonic component in the current response signal;

[0013] The spatial harmonic component is weightedly combined with the back electromotive force signal, and the result is provided to a phase-locked loop for phase error detection to generate the estimated angle.

[0014] As an optional implementation, it further includes: using a finite element analysis method to model the electromagnetic characteristics of the target compressor to generate electromagnetic data of the target compressor under a target operating state;

[0015] The predicted angle is generated using a preset machine learning model based on the operating parameters of the target compressor, the estimated angle of the rotor, and electromagnetic data generated by finite element analysis.

[0016] As an optional implementation, the finite element analysis method adopts a pre-calculation method to pre-calculate the electromagnetic characteristics of the target compressor motor and generate electromagnetic data under different operating conditions;

[0017] The electromagnetic data include: magnetic flux density distribution, current density distribution, flux linkage coupling and electromagnetic loss characteristics;

[0018] The pre-calculated electromagnetic data is stored via a lookup table or a database and is dynamically called during real-time operation according to the current operating state of the compressor.

[0019] As an optional implementation, it also includes:

[0020] Determining the frequency and amplitude of the high-frequency voltage signal based on electromagnetic data pre-calculated and generated by finite element analysis;

[0021] When the magnetic flux density is lower than a first threshold, a high-frequency voltage signal with a frequency within a first preset range is selected for injection;

[0022] When the current density is higher than a second threshold, the amplitude of the high-frequency voltage signal is adjusted to be within a second preset range.

[0023] As an optional implementation, the input data of the machine learning model includes: the speed, load, stator current, back electromotive force signal, estimated angle of the rotor, and electromagnetic data generated by finite element analysis pre-calculation of the target compressor;

[0024] The machine learning model includes: a long short-term memory network model;

[0025] The long short-term memory network model is used to process the input data in real time and generate a prediction angle;

[0026] Among them, the long short-term memory network model generates a prediction angle by analyzing historical data and current operating parameters.

[0027] As an optional implementation, the machine learning model is trained by a truncated back propagation algorithm, including:

[0028] Dividing the historical operating parameters of the target compressor into segments based on a preset length;

[0029] Wherein, the preset length is determined based on the system calculated load and real-time target; the historical operating parameters include: speed, load, stator current and timing data of back electromotive force signal;

[0030] Perform local gradient calculation on the time series data of historical operating parameters of each compressor, and update the weight of the machine learning model based on the local gradient calculation results;

[0031] Wherein, the local gradient calculation limits the number of time steps of each back propagation by truncating the back propagation;

[0032] After the calculation of each segment of time series data is completed, the hidden state and memory state of the current segment of time series data are used as the initial state for the calculation of the next segment of time series data.

[0033] As an optional implementation manner, the preset length is determined based on the system computing load and the real-time target and includes:

[0034] Monitoring a system computing load of a target compressor; wherein the system computing load is determined based on a weighted average of the processor usage rate and the memory occupancy rate;

[0035] When the weighted average value of the load calculated by the system is greater than or equal to a preset first load threshold, shortening the preset length;

[0036] When the weighted average value of the load calculated by the system is less than the first load threshold, the preset length is extended.

[0037] In a second aspect, the present application also provides a compressor weak magnetic region stable operation system based on angle compensation, including:

[0038] an estimating unit, estimating a rotor angle of a target compressor based on a phase-locked loop, and generating an estimated angle of the rotor;

[0039] A prediction unit, using a preset machine learning model, generates a predicted angle based on the operating parameters of the target compressor and the estimated angle of the rotor; wherein the machine learning model is used to compensate for the influence of multiple frequency components on the predicted angle; the multiple frequency components include: spatial harmonic components, high frequency components of back electromotive force signals, and noise signals;

[0040] The control unit adjusts the operating parameters of the target compressor based on the predicted angle.

[0041] Compared with the prior art, the beneficial effects of the present invention are: using a machine learning model to process multiple input parameters in real time, including speed, load, stator current, back electromotive force signal and electromagnetic data, effectively compensating for the influence of spatial harmonics, high-frequency signals and noise, and generating a more accurate rotor prediction angle. Through the comprehensive analysis of multi-source data, it can dynamically adapt to complex electromagnetic environments and operating conditions, especially improve the stable operation performance of the compressor in the weak magnetic area. The speed, load and current of the compressor are adjusted in real time based on the predicted angle to ensure that the system operates in the optimal state under different load conditions, further improving energy efficiency and operational reliability. The pre-calculation method of finite element analysis and the truncated back propagation algorithm in the machine learning training process reduce the computational burden in real-time operation, so that the system can maintain high-precision prediction while ensuring real-time performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A flow chart of a method for stable operation of a compressor in a weak magnetic field region based on angle compensation provided in an embodiment of the present application;

[0043] Figure 2 A flow chart for generating the estimated angle of the rotor provided in an embodiment of the present application;

[0044] Figure 3 A schematic diagram of a compressor weak magnetic region stable operation system based on angle compensation provided in an embodiment of the present application. DETAILED DESCRIPTION

[0045] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0046] The present invention relates to a stable operation control method and system for a compressor in a weak magnetic field region, and in particular to precise control through rotor angle compensation. When the compressor is operating in a weak magnetic field region, the error of the rotor angle may cause unstable operation, thereby affecting the efficiency and life of the system. In order to solve this problem, the present invention proposes a rotor angle prediction method based on a machine learning model and multiple frequency component compensation.

[0047] The applicable target compressor of the present invention can use a permanent magnet synchronous motor (PMSM) as a power source, and the rotor of the motor is rotated by electromagnetic induction. The compressor is usually in a high-speed and high-load state. Under this working condition, the actual angle of the rotor needs to be accurately estimated to achieve precise torque control.

[0048] The stator winding of the compressor motor is connected to the control system, and the dynamic operation status of the motor can be obtained by collecting the stator current and voltage signals. Specifically, during the operation of the motor, the back electromotive force (Back-EMF) signal on the stator winding of the motor is detected to provide a real-time estimation basis for the rotor position.

[0049] Angle compensation refers to the process of adjusting the relative angle (i.e. phase difference) between the rotor and stator magnetic fields during motor control to compensate for insufficient flux or torque caused by various factors (such as high-speed motor operation or load changes), thereby ensuring that the motor can maintain stable operation in specific areas (such as weak magnetic field areas).

[0050] The weak magnetic field area refers to the situation when the motor enters high-speed operation. Due to the increase in speed, the stator voltage approaches the upper limit of the motor's rated voltage, and the voltage cannot be increased to maintain the magnetic flux, resulting in a weakened magnetic field. At this time, the motor's magnetic flux is insufficient, which affects the output torque and enters the so-called weak magnetic field area.

[0051] See also Figure 1 As shown, it is a flow chart of a method for stable operation of a compressor in a weak magnetic field region based on angle compensation provided by an embodiment of the present application, the method comprising steps S101 to S103, wherein:

[0052] S101: Estimate the rotor angle of the target compressor based on a phase-locked loop to generate an estimated angle of the rotor.

[0053] S102: Generate a predicted angle based on the operating parameters of the target compressor and the estimated angle of the rotor using a preset machine learning model; wherein the machine learning model is used to compensate for the influence of multiple frequency components on the predicted angle; the multiple frequency components include: spatial harmonic components, high-frequency components of back electromotive force signals, and noise signals.

[0054] S103: Adjusting operating parameters of the target compressor based on the predicted angle.

[0055] A phase-locked loop is a feedback control system that can detect and adjust a signal phase error, and is used in the present invention to estimate the rotor angle of the target compressor.

[0056] In a specific implementation, the real-time operating signal can be collected through the stator sensor of the compressor motor. The phase-locked loop system compares the phase of the input signal with the reference signal and calculates the phase error in the signal. The system then makes real-time adjustments based on the phase error to generate an estimated angle corresponding to the actual operating angle of the rotor.

[0057] In order to further improve the operating stability of the compressor in the weak magnetic field area, the present invention uses a preset machine learning model to compensate and predict the rotor angle. The input parameters of the machine learning model include the operating parameters of the compressor (such as speed, stator current, back electromotive force signal) and the estimated angle of the rotor. By processing these input data in real time, the machine learning model generates an angle prediction value to compensate for the angle error caused by multiple frequency components (spatial harmonic components, high-frequency components of the back electromotive force signal, and noise signals).

[0058] Based on the predicted angle, the system adjusts the operating parameters of the compressor in real time to ensure stable operation of the compressor in the weak magnetic field area. For example, when the predicted angle deviation is large, the system can ensure the normal operation of the motor by adjusting the magnitude of the stator current or the frequency of the voltage.

[0059] It should be noted that the phase-locked loop in the present disclosure may also correspond to a phase-locked loop system or a phase-locked loop module.

[0060] As an optional implementation, the phase-locked loop in the present disclosure includes: a phase detector, a low-pass filter, a voltage-controlled oscillator, and a feedback loop.

[0061] Among them, the phase detector (PD) is used to compare the phase difference between the reference signal and the feedback signal and generate a phase error signal; the low-pass filter (LPF) is used to filter the phase error signal, remove high-frequency noise, and obtain a smooth error signal; the voltage-controlled oscillator (VCO) controls its output frequency through the error signal, so that the phase of the output signal is gradually locked to the phase of the back-electromotive force signal; the feedback loop feeds the output signal of the VCO back to the phase detector to form a closed-loop control.

[0062] In a specific implementation, when the compressor starts, the stator winding begins to be energized, and the motor rotor gradually enters a stable rotation state. At this time, the back electromotive force signal generated by the stator winding is collected and transmitted to the phase detector of the phase-locked loop as an input signal. The phase detector also receives the feedback signal from the VCO, calculates the phase difference between the feedback signal and the back electromotive force signal, and outputs the corresponding phase error signal. The error signal is processed by a low-pass filter to remove high-frequency interference, and outputs a smooth error signal to control the voltage-controlled oscillator so that its frequency and phase gradually approach the frequency and phase of the back electromotive force signal. When the phase-locked loop enters the locked state, the output signal of the VCO is synchronized with the back electromotive force signal, and the output phase of the VCO is the real-time angle of the motor rotor.

[0063] After the phase-locked loop is locked, the phase signal output by the VCO can be sent to the control system or central processing unit deployed in the compressor as an estimated value of the rotor angle. The control system can perform further torque control or speed control based on the estimated angle to ensure that the compressor maintains stable torque output when running at high speed.

[0064] In order to further improve the accuracy of rotor angle estimation, the present invention adopts the method of injecting high-frequency voltage signals into the stator windings of the target compressor motor. By collecting the current response signals of the stator windings and extracting the spatial harmonic components therefrom, the accuracy of rotor angle estimation in the weak magnetic field region can be effectively improved.

[0065] See also Figure 2 As shown, it is a flow chart of generating the estimated angle of the rotor provided in an embodiment of the present application. As an optional implementation, the generating the estimated angle of the rotor further includes steps S201 to S203, wherein:

[0066] S201: injecting a high-frequency voltage signal into the stator winding of the target compressor motor;

[0067] S202: collecting a current response signal of the stator winding, and extracting a spatial harmonic component in the current response signal;

[0068] S203: Perform weighted combination of the spatial harmonic component and the back electromotive force signal, and provide the result to a phase-locked loop for phase error detection to generate the estimated angle.

[0069] The purpose of injecting the high-frequency voltage signal is to obtain a richer current response signal, especially including harmonic component information related to the rotor angle.

[0070] In the present invention, the high-frequency voltage signal refers to a voltage signal with a frequency higher than the operating frequency of the motor. In a preferred embodiment, the frequency of the high-frequency voltage is in the range of 1kHz to 10kHz. The signal is injected through the stator winding of the compressor motor to stimulate the motor to generate a current response signal in the high-frequency range, thereby obtaining harmonic components related to the rotor angle. The frequency and amplitude of the high-frequency voltage signal are dynamically adjusted according to the motor operating state and electromagnetic characteristics.

[0071] The current response signal refers to the current feedback signal collected by the current sensor after the high-frequency voltage signal is injected into the motor stator winding. This signal reflects the transient electromagnetic state of the motor and contains harmonic components related to the rotor position. The acquisition frequency of the current response signal should be at least twice the frequency of the high-frequency voltage signal to ensure effective capture of the high-frequency components. By further processing the signal, the spatial harmonic components can be extracted to generate an estimate of the rotor angle.

[0072] For example, the control system collects the current response signal of the stator winding through the current sensor at a sampling rate of at least 20kHz. The signal is processed by a digital bandpass filter, and the passband of the filter is set to correspond to the frequency of the injected high-frequency voltage signal. The processed signal is spectrally analyzed by fast Fourier transform (FFT) to extract the main spatial harmonic components for subsequent angle estimation.

[0073] The spatial harmonic component refers to the high-frequency component related to the rotor angle in the motor current response signal, which is usually caused by the structural characteristics of the motor and the electromagnetic interaction between the stator and the rotor. In the present invention, the spatial harmonic component is obtained by performing spectrum analysis and filtering on the current response signal. The spatial harmonic component can reflect the change of the rotor angle and further improve the accuracy of the rotor angle estimation by weighted combination with the back electromotive force signal.

[0074] In a specific implementation, a high-frequency voltage signal of preset frequency and amplitude is injected into the stator winding through the control system of the compressor motor. After the signal is injected, the stator winding of the motor generates a current response signal, which contains high-frequency harmonic components related to the rotor angle. After the high-frequency voltage signal is injected, the system collects the current response signal of the stator winding through a current sensor. In order to ensure the accuracy of signal acquisition, the sampling rate of the current sensor is at least 20kHz to capture high-frequency components. The collected current response signal is filtered to extract the spatial harmonic components.

[0075] In a specific implementation, the control system of the target compressor can have an embedded or external signal generator to generate a high-frequency voltage signal of a preset frequency and amplitude. The output signal of the signal generator is applied to the stator winding of the motor through the power inverter of the control system. The control system can modulate the output voltage signal by controlling the power inverter to generate a high-frequency voltage signal. Usually the switching frequency of the inverter can be adjusted to meet the required frequency range, such as 1kHz to 10kHz. In addition, the frequency and amplitude of the high-frequency voltage signal can be dynamically adjusted through a closed-loop control algorithm. The control system can control the frequency and amplitude based on the real-time monitored motor operating status (such as voltage, current, speed, load). For example, the frequency suitable for the current operating state can be dynamically selected through a lookup table, and the amplitude of the signal can be automatically adjusted according to the flux density or current density.

[0076] For example, the compressor control system has a built-in digital signal processor (DSP), which generates a high-frequency voltage signal of a preset frequency. The high-frequency voltage signal in the range of 1kHz to 10kHz is injected into the stator winding by pulse width modulation (PWM) control of the switching state of the inverter. The amplitude of the signal is dynamically adjusted based on the real-time load and magnetic flux density of the motor.

[0077] In addition, the extracted spatial harmonic components are weightedly combined with the back-EMF signal to further improve the accuracy of rotor angle estimation. The back-EMF signal comes from the self-induced voltage of the motor stator winding during operation, which contains information related to the rotor motion. The system performs weighted processing on the spatial harmonic components and the back-EMF signal, where the weighting coefficient is dynamically adjusted according to the operating status. The combined signal is then provided to the phase-locked loop system for phase error detection.

[0078] Among them, the back-EMF signal can be obtained through the same acquisition circuit as the spatial harmonic component, and the high-frequency component can be extracted after filtering. The weighting coefficient can be dynamically adjusted according to the current operating state. For example, if the load of the motor is large, the weight of the back-EMF signal can be increased to reflect the impact of the change in rotor position. The weighting coefficient can be obtained through a preset lookup table or real-time calculation.

[0079] In this way, the extracted spatial harmonic components and back-EMF signals are linearly weighted combined using weighting coefficients. The combined signal will contain angle information from both the harmonic components and the back-EMF, reducing the error caused by a single signal.

[0080] The combined signal is input to a phase-locked loop (PLL) system for real-time phase error detection. The phase-locked loop generates an estimated angle of the rotor by adjusting the signal phase to minimize the phase error between the input signal and the reference signal. This estimated angle reflects the actual rotor position of the motor in the field weakening region and can be used for further angle compensation and control.

[0081] Exemplarily, after the combined signal is input into the phase-locked loop, the phase detector calculates the phase difference between the signal and the expected rotor angle reference signal. The phase error signal is smoothed by the loop filter and passed to the voltage-controlled oscillator. The VCO adjusts the output angle in real time according to the phase error signal to ensure that the estimated angle matches the actual rotor position. Finally, an accurate rotor estimated angle is generated for subsequent control steps.

[0082] In this way, by injecting high-frequency voltage signals into the stator windings, obtaining the spatial harmonic components related to the rotor angle, and weightedly combining them with the back-EMF signal, a more accurate rotor angle estimation is achieved. The frequency and amplitude of the high-frequency voltage signal are dynamically adjusted to adapt to different motor operating states, improving the adaptability and robustness of the system. Phase error detection through a phase-locked loop ensures real-time and accurate angle estimation, especially improving the operating stability and control accuracy of the compressor in the weak magnetic field area.

[0083] As an optional implementation, the method further includes: modeling the electromagnetic characteristics of the target compressor using a finite element analysis method to generate electromagnetic data of the target compressor under a target operating state;

[0084] The predicted angle is generated using a preset machine learning model based on the operating parameters of the target compressor, the estimated angle of the rotor, and electromagnetic data generated by finite element analysis.

[0085] Among them, finite element analysis (FEA) is a computational tool based on numerical methods to solve complex physical problems. It is often used to analyze electromagnetic field distribution and electromagnetic performance in the field of motors. By inputting the structure, material, geometry and working conditions of the compressor motor into the finite element analysis model, electromagnetic data of the motor under different operating conditions can be generated.

[0086] The electromagnetic data mentioned in this disclosure refers to the electromagnetic performance parameters obtained through finite element analysis to describe the motor under specific operating conditions, including: magnetic flux density distribution, flux linkage coupling, induced electromotive force, current density distribution, electromagnetic loss, etc. These data can accurately reflect the operation of the motor under different load, speed and voltage conditions.

[0087] In the specific implementation, the geometric structure of the compressor motor is first modeled in three dimensions, and then specific boundary conditions are applied to the model through finite element analysis software, such as the way of applying voltage or current. Next, the electromagnetic field simulation is performed using the material properties of the motor, such as the conductivity and magnetic permeability of the stator and rotor. Finally, the electromagnetic data under the target operating state is obtained through calculation.

[0088] In addition, a machine learning model is used to generate a predicted angle for the rotor. The inputs to the model include the operating parameters of the compressor (such as speed, load, stator current), the estimated rotor angle generated by the phase-locked loop, and electromagnetic data generated by finite element analysis. By combining these input information, the machine learning model is able to output a more accurate predicted angle.

[0089] The input variables include the compressor's real-time operating parameters, the estimated rotor angle generated by the phase-locked loop (PLL), and electromagnetic data obtained from finite element analysis. These data are combined to provide a more comprehensive angle prediction input that can make up for the shortcomings of a single signal source.

[0090] Exemplarily, the geometric structure of the compressor motor can be obtained in a variety of ways, for example, directly from the CAD model of the motor, or extracted from technical drawings and design documents provided by the motor manufacturer.

[0091] Exemplary finite element analysis software commonly used for electromagnetic simulation includes: ANSYS Maxwell, COMSOL Multiphysics, JMAG, and Altair Flux, etc. In specific implementation, the above software can be selected according to the usage scenario and the different characteristics of the above software.

[0092] It should be noted that in finite element simulation, it is very important to accurately set boundary conditions, which usually include: voltage or current excitation, periodic boundary conditions, air gap settings, and material properties;

[0093] Among them, AC voltage or current is usually applied to the stator winding of the compressor motor. Usually, the voltage / current can be applied by defining an AC source in the simulation software. Common boundary condition setting methods include: setting the voltage waveform of the stator winding, such as a sine wave voltage (the amplitude is set according to the rated voltage of the motor).

[0094] In addition, for motors with symmetrical structures, periodic boundary conditions can be used to reduce the amount of simulation calculations. For example, for a 12-slot stator, only 1 / 12 of the part needs to be simulated, and the rest is automatically mapped out by periodic boundary conditions.

[0095] Among them, material properties are important input data for simulation, including the conductivity and magnetic permeability of materials such as stators, rotors, and coils. Specific parameters can be obtained from the following methods:

[0096] Material database: Most simulation software comes with a material database. The conductivity, magnetic permeability and other data of common motor materials (such as silicon steel sheets, copper wires, permanent magnet materials, etc.) can be directly extracted from the software library.

[0097] Before simulation, the geometric structure of the motor needs to be discretized, that is, divided into finite element meshes. The fineness of the mesh determines the simulation accuracy and calculation time. For example, key areas such as air gaps require smaller mesh units to improve accuracy, while flatter areas can use larger mesh units to reduce the amount of calculation.

[0098] During the simulation, the software calculates the electromagnetic field of the motor, solving parameters such as the magnetic flux between the stator and rotor, the current density distribution in the stator coil, and the induced electromotive force of the motor. During the calculation process, the dynamic characteristics of the electromagnetic field over time are also simulated, especially when the rotor rotates or the load changes.

[0099] In this way, by combining finite element analysis with a machine learning model, the prediction accuracy of the rotor angle can be greatly improved. The electromagnetic data generated by the finite element analysis method provides the machine learning model with richer physical information, and the machine learning model can generate accurate predicted angles in real time by combining electromagnetic data, operating parameters, and the estimated angle of the rotor. Compared with traditional methods that rely only on empirical formulas or a single signal source, the present invention can provide more robust prediction results under different operating conditions, especially in the weak magnetic field area, the control of the compressor is more stable, and the overall energy efficiency and dynamic response performance of the system are improved.

[0100] It is understandable that the finite element analysis method used for real-time data simulation analysis will bring a certain computational burden, which may have a certain impact on the built-in control system of the compressor with weaker computing power.

[0101] As an optional implementation, the finite element analysis method adopts a pre-calculation method to pre-calculate the electromagnetic characteristics of the target compressor motor and generate electromagnetic data under different operating conditions; wherein the electromagnetic data includes: magnetic flux density distribution, current density distribution, flux linkage coupling and electromagnetic loss characteristics; the pre-calculated electromagnetic data is stored in a lookup table or a database, and is dynamically called according to the current operating state of the compressor during real-time operation;

[0102] Pre-calculation refers to calculating the electromagnetic characteristics of the motor under various operating conditions in advance during the design phase through finite element simulation, and storing the calculation results. The benefit of pre-calculation is that it reduces the computational burden in real-time operation, especially in real-time control systems, reducing the time delay of complex electromagnetic calculations.

[0103] In a specific implementation, the electromagnetic data generated by pre-calculation is stored in a lookup table or database. When the compressor is running in real time, the system can dynamically call the pre-stored electromagnetic data according to the current operating status (such as speed, load and other parameters) to quickly respond to system requirements.

[0104] Among them, the lookup table is a common and efficient storage method. It can pre-calculate and store the results of complex calculations, and directly obtain the results through the lookup table at runtime, thereby greatly improving the efficiency of system operation. Databases can also be used for larger-scale storage, especially when the amount of data is large, the database can be managed and retrieved more efficiently.

[0105] For example, during the real-time control process, the system automatically retrieves pre-calculated electromagnetic data based on the current operating state of the compressor measured by the sensor (such as motor speed, load, voltage, etc.). These data have been generated and stored in the finite element simulation stage, so the system no longer needs to perform complex finite element calculations in real time, thereby improving the response speed and real-time performance of the system.

[0106] As an optional implementation, the frequency and amplitude of the high-frequency voltage signal can also be determined based on the electromagnetic data pre-calculated and generated by finite element analysis; when the magnetic flux density is lower than a first threshold value, a high-frequency voltage signal with a frequency within a first preset range is selected for injection; when the current density is higher than a second threshold value, the amplitude of the high-frequency voltage signal is adjusted to within a second preset range.

[0107] In the specific implementation, the electromagnetic characteristics of the motor under different load, speed and other operating conditions are pre-calculated during the finite element analysis simulation process. Based on this data, the system can quickly find the optimal high-frequency voltage signal frequency and amplitude through a lookup table. For example, when the magnetic flux density is low, a higher frequency signal is required to excite the spatial harmonics; when the current density is high, the signal amplitude needs to be appropriately reduced to prevent the motor from overloading.

[0108] For example, the frequency and amplitude comparison data are stored in a lookup table, and when the system is running, the real-time operating status information obtained by the sensor automatically matches the corresponding frequency and amplitude. This method has a small amount of calculation, fast response speed, and is suitable for real-time control. In addition, dynamic algorithms (such as fuzzy control, PID control, etc.) can be used to calculate the appropriate high-frequency voltage signal frequency and amplitude in real time, and combined with electromagnetic data to provide a more flexible control solution.

[0109] For example, when the current magnetic flux density is detected to be low, the lookup table shows that the frequency should be adjusted to 5kHz and the amplitude should be 80% of the rated voltage. If the current density is high, the system automatically reduces the amplitude of the high-frequency voltage signal to the second preset range.

[0110] In a specific implementation, by monitoring the magnetic flux density of the motor in real time, when the magnetic flux density is lower than a set first threshold, a high-frequency voltage signal within a first preset range is automatically selected for injection.

[0111] The first preset range can be determined based on experimental or simulation data of the target compressor, and is usually a high frequency range that satisfies the stable operation of the motor. By selecting this frequency, it is possible to ensure that the magnetic field can effectively excite the spatial harmonics of the stator and rotor under low magnetic flux density, thereby improving the estimation accuracy of the rotor angle.

[0112] For example, in actual operation, the system monitors the magnetic flux density of the motor in real time through a magnetic flux sensor and determines whether it is lower than a first threshold value. If it is lower, the corresponding frequency is immediately selected through a lookup table.

[0113] The frequency of the high-frequency voltage signal in the stator winding is adjusted by a modulation controller (such as a frequency converter or inverter) to ensure that the injected signal meets the requirements of the current flux state of the motor.

[0114] In a specific implementation, the current density of the stator winding is also monitored in real time. When the current density is higher than a set second threshold, the system automatically adjusts the amplitude of the high-frequency voltage signal to within a second preset range to prevent the motor from overloading or overheating.

[0115] The second preset range is a safe amplitude interval, which can be set by experimental or empirical data for the target compressor, such as 50% to 80% of the rated voltage, to ensure stable operation of the motor under high load. The purpose of adjusting the amplitude is to control the amplitude of the signal to avoid excessive current on the motor winding without reducing the signal frequency.

[0116] For example, the current density of the motor winding is monitored in real time by a current sensor. If the current density exceeds a safety threshold (e.g., more than 10 A / mm 2), the system immediately reduces the amplitude of the high-frequency voltage signal.

[0117] The system automatically adjusts the amplitude of the high-frequency voltage signal through the inverter or power regulation module to ensure that the adjustment amplitude is within the preset safety range.

[0118] In this way, by dynamically adjusting the frequency and amplitude of the high-frequency voltage signal using electromagnetic data generated by finite element analysis, the present invention can automatically select and adjust the optimal parameters of the high-frequency voltage signal according to the current magnetic flux density and current density of the compressor. This dynamic adjustment mechanism effectively improves the adaptability of the system to different working conditions, especially when the magnetic flux density is low or the current density is high. By adjusting the frequency and amplitude, the validity of the signal can be ensured, and the motor can be avoided from being overloaded or excessively lost, thereby improving the stability and energy efficiency of the compressor operation.

[0119] As an optional implementation, the input data of the machine learning model includes: the speed, load, stator current, back electromotive force signal, estimated angle of the rotor, and electromagnetic data pre-calculated and generated by finite element analysis of the target compressor; the machine learning model includes: a long short-term memory network model; the long short-term memory network model is used to process the input data in real time and generate a predicted angle; wherein the long short-term memory network model generates a predicted angle by analyzing historical data and current operating parameters.

[0120] The input of the machine learning model includes multiple operating parameters of the target compressor and electromagnetic data generated by finite element analysis. The input data comes from multiple sources and is used to describe the current operating status and electromagnetic environment of the compressor:

[0121] The rotational speed is the real-time rotational speed of the target compressor, which can be monitored by sensors and used to capture the dynamic changes of the rotor through the input of the machine learning model.

[0122] The load is the mechanical load on the compressor, which can be obtained by monitoring the power or torque of the compressor. Load changes affect the electromagnetic characteristics and are therefore used as input parameters.

[0123] The stator current reflects the current condition of the stator winding and is monitored in real time by the current sensor. The change of stator current is directly related to the torque and electromagnetic loss of the motor.

[0124] The back-EMF signal represents the voltage induced in the motor stator coil and is an important feedback signal for the motor's operating status. The back-EMF signal is directly related to the rotor motion and can provide information about the rotor angle.

[0125] The estimated rotor angle is the preliminary rotor angle calculated by the phase-locked loop (PLL) system, which is used to assist the machine learning model in generating a more accurate predicted angle.

[0126] The electromagnetic data generated by finite element analysis is the pre-calculated electromagnetic data, which includes information such as magnetic flux density and current density under different operating conditions of the motor, and is used to enhance the machine learning model's understanding of changes in the electromagnetic environment.

[0127] These input data provide a detailed description of the compressor's real-time status and historical operation, helping the machine learning model capture the dynamic changes of the system.

[0128] In the specific implementation, the machine learning model adopts the long short-term memory network (LSTM), which has the advantage of being able to process input data with time correlation and memorize and analyze long-term historical data to generate prediction angles.

[0129] Among them, LSTM is a recurrent neural network (RNN) that can process time series data. By memorizing past historical information, it can handle the complex relationship between the current operating state and the historical state. In the operation of the compressor, not only the current parameters are crucial to the prediction of the rotor angle, but the past operating state also affects the future angle change. LSTM can handle this information well.

[0130] The LSTM model receives time series data including speed, load, stator current, etc., and combines the pre-calculated electromagnetic data from finite element analysis and the estimated angle of the rotor to generate a more accurate predicted angle. The model processes this data in real time and updates the prediction results at each time step. By processing the input data online, the LSTM model can generate a predicted angle during the real-time operation of the compressor, ensuring the real-time control performance of the system.

[0131] In addition, LSTM can retain important historical data for a long time through its internal memory unit and make full use of this data in current predictions. For example, the changing trends of speed and load have an important impact on the future changes of the rotor angle. LSTM can capture these trends and use them for predictions. At the same time, the LSTM model also processes current real-time data. By analyzing the real-time status of speed, load, stator current, etc., the model can be dynamically adjusted in combination with historical data to make the prediction results more accurate.

[0132] Exemplarily, the input layer in the present disclosure is used to accept multiple input parameters, including speed, load, stator current, back electromotive force signal, estimated angle of the rotor, and electromagnetic data generated by finite element analysis. These data constitute a time series input, and each time step provides a set of current state data. The LSTM hidden layer is composed of multiple units, each of which can capture the time dependency of the input data. The hidden layer is the core of the entire LSTM architecture, responsible for processing and memorizing time series data.

[0133] After the LSTM layer, a fully connected layer is used to map the output of the LSTM to the desired output space. In the present invention, the output of this layer is the predicted rotor angle. The output layer is used to provide the final prediction result, i.e. the predicted angle of the rotor. This layer will output the actual angle value through the calculation of the previous layer.

[0134] For example, the input dimension is equal to the number of input parameters, such as speed, load, stator current, back EMF signal, estimated rotor angle, and electromagnetic data, that is, the data dimension of each time step is 6 to correspond to 6 features.

[0135] The time step refers to how many time steps the LSTM model processes in a sequence. For the dynamic changes of the motor system, choosing the right time step is very critical.

[0136] In addition, the loss function is used to measure the error between the predicted angle and the true angle. For the prediction problem, the loss function can be the mean square error (MSE), which measures the squared error between the predicted value and the actual value.

[0137] For example, in the training of the LSTM model, the preparation of the data set may include: training data: historical operating parameters, including speed, load, stator current, etc., and the actual rotor angle. Through the time series input of these data, the model can learn the dynamic relationship between the data; input window: divide the time series data into multiple input windows, for example, input a preset time step of data each time as the input of the model, and correspond to an actual rotor angle as the output.

[0138] As an optional implementation, the input signal can also be preprocessed to extract frequency domain features including spatial harmonic components and high-frequency components of back-electromotive force signals; the long short-term memory network predicts the rotor angle by combining the frequency domain features with the time domain features; the input dimension of the machine learning model is expanded based on the time domain features and the frequency domain features, and the frequency features of the spatial harmonic components and the high-frequency components of the back-electromotive force are added to compensate for the influence of multiple frequency components on the predicted angle.

[0139] In specific implementations, the advantage of LSTM lies in processing time series data and its long-term and short-term dependencies, but in order to compensate for the impact of different frequency components, special feature processing and model architecture need to be designed.

[0140] The spatial harmonic components are closely related to the change in rotor angle. In order for LSTM to process this information, the harmonic components can be used as part of the time series input. By performing frequency domain analysis (such as Fourier transform) on the current response signal, the spatial harmonic components in the signal are extracted and used as the input of LSTM. In this way, the LSTM model can compensate for the impact of the harmonic components on the angle prediction by learning these time series harmonic features.

[0141] The high-frequency components in the back-EMF signal reflect the transient dynamic changes of the rotor. Therefore, by performing high-frequency filtering on the back-EMF signal, the high-frequency components are extracted and used as LSTM input. LSTM can learn the correlation between the changes in the high-frequency components and the rotor angle, thereby making the prediction of the rotor angle more accurate.

[0142] Noise is usually a random interference that affects the clarity of the signal. In order to make LSTM more resistant to noise, the input signal can be preprocessed, including denoising filtering (such as low-pass filtering or adaptive filtering) to reduce the noise component. Or the noise can be directly used as a time series input feature to let LSTM learn how to identify the impact of noise on angle prediction and compensate for it.

[0143] In the specific implementation, in order to allow the LSTM network to better handle the impact of different frequency components on the prediction angle, its architecture can be appropriately adjusted.

[0144] For example, the LSTM model can be designed with multiple input channels, each of which is responsible for processing different frequency features. For example, spatial harmonic components, high-frequency components of back EMF, and time domain signals (such as speed, load, etc.) can be passed to the LSTM through different input paths. With this structure, the LSTM can focus on the changes in the frequencies of different signals and learn how to combine this information.

[0145] In addition, a hierarchical LSTM architecture can also be used to allow different LSTM layers to process different frequency features. For example, the first layer of LSTM can specifically process time domain signals (such as speed, load, etc.), and the second layer of LSTM can process frequency domain signals (such as spatial harmonics, high-frequency components, etc.). This hierarchical structure helps the network better learn the contribution of different frequency components to the predicted angle.

[0146] In order to capture the local features of different frequency components, a convolutional network (CNN) can be added before LSTM to extract the spatial harmonic components and high-frequency components in the input signal. The convolutional layer can effectively capture the local frequency features in the signal, while LSTM is good at processing long-term dependencies. The combination of CNN-LSTM can effectively improve the prediction ability of the model.

[0147] Although LSTM is good at time series modeling, the compensation ability of the model can be significantly improved by converting the input signal into the frequency domain, extracting spatial harmonics and high-frequency components, and combining these frequency domain features with time domain data. For example, the time domain signal is converted into frequency domain features through Fourier transform or wavelet transform, and these features are used as the input of LSTM.

[0148] By optimizing the loss function, the influence of different frequency components can be weighted. For example, high-frequency components (such as spatial harmonics and back-EMF high-frequency components) are given higher weights to ensure that LSTM can pay more attention to these key frequency features when predicting angles.

[0149] It should be noted that when high-frequency components of spatial harmonics or back-EMF are extracted as independent features, the input dimension increases accordingly to ensure that the model can better compensate for the impact of these frequency components on angle prediction.

[0150] In this way, by adopting the long short-term memory network (LSTM) as a machine learning model, it is possible to effectively process time-series data, combine historical operating parameters and current electromagnetic characteristics, and generate accurate rotor prediction angles. Compared with traditional methods, LSTM can utilize past operating information and process the current state in real time, thereby providing more stable and accurate prediction results. In addition, by combining the pre-calculated electromagnetic data generated by finite element analysis and the rotor estimated angle of the phase-locked loop, the prediction accuracy and the real-time response of the system are further improved. The present invention greatly improves the operating stability and energy efficiency control of the compressor in the weak magnetic field area.

[0151] In the long short-term memory network or other recursive neural network (RNN) models, the traditional back propagation algorithm back propagates the error along the entire time series and updates the model weights. However, a long time series will lead to high computational costs and easily cause gradient vanishing or gradient exploding problems. To this end, the present disclosure adopts a truncated back propagation algorithm (TBPTT) to reduce the computational complexity by processing the time series data in segments, and limit the number of time steps of back propagation by local gradient calculation.

[0152] As an optional implementation, the machine learning model is trained using a truncated back propagation algorithm, including: segmenting the historical operating parameters of the target compressor based on a preset length; wherein the preset length is determined based on the system's calculated load and real-time targets; the historical operating parameters include: speed, load, stator current, and time series data of back electromotive force signals; performing local gradient calculations on the time series data of each segment of the compressor's historical operating parameters, and updating the weights of the machine learning model based on the local gradient calculation results; wherein the local gradient calculation limits the number of time steps for each back propagation by truncating back propagation; after the calculation of each time series data is completed, the hidden state and memory state of the current time series data are used as the initial state for the calculation of the next time series data.

[0153] This method divides long time series data into multiple shorter segments and performs back propagation and gradient updates in each segment. The number of time steps for each back propagation is limited, so there is no need to back propagate along the entire sequence, which can significantly reduce computational overhead. In addition, after each time series is calculated, the hidden state and memory state will be retained and passed to the next time series to ensure that information is not lost.

[0154] Each time series performs a separate local gradient calculation, and the weights are updated according to the local error of the segment. Eventually, the weights of the entire model gradually tend to the optimal value after multiple local updates.

[0155] In the specific implementation, the historical operating parameters of the target compressor (such as speed, load, stator current, back electromotive force, etc.) are segmented according to the time step length of 50. Each segment of data represents the operation of the compressor within 50 time steps. The segment length can be dynamically adjusted according to the calculation load and real-time requirements of the system.

[0156] For each 50 time steps of data, the model performs forward propagation, then calculates the error through local gradient, and performs back propagation and weight update within these 50 time steps. The time step of back propagation is limited to 50, and no longer historical data is traced back.

[0157] Whenever the calculation of 50 time steps of data is completed, the hidden state and memory state of LSTM will be retained and used as the input initial state for the next 50 time steps. This can maintain the context continuity of the time series and ensure that the model captures long-term dependencies.

[0158] In this way, by processing long time series in segments, the computational complexity in the back propagation process is significantly reduced. This segmented processing method avoids the need to calculate the entire time series, reduces memory usage and computing time, and improves the real-time performance of the system.

[0159] As an optional implementation, the preset length is determined based on the system calculated load and the real-time target and includes: monitoring the system calculated load of the target compressor; wherein the system calculated load is determined based on the weighted average of the processor usage rate and the memory occupancy rate; when the weighted average of the system calculated load is greater than or equal to a preset first load threshold, shortening the preset length; when the weighted average of the system calculated load is less than the first load threshold, extending the preset length.

[0160] In the present disclosure, the computational load of the system is evaluated by monitoring the weighted average of the processor usage and memory occupancy in real time. The processor (such as CPU or GPU) of the compressor system consumes a lot of computing resources when running the LSTM model. By monitoring the resource usage of the system, the operating parameters of the model can be dynamically adjusted to ensure that the system meets the real-time target.

[0161] In the specific implementation, the processor usage (CPU / GPU occupancy) can be collected in real time through the operating system or dedicated monitoring software to indicate the load of the current computing task on the processor. Memory usage is evaluated by monitoring the occupancy of the system memory. The training and reasoning process of the LSTM model consumes a lot of memory, especially when processing large-scale time series data, memory usage may become a bottleneck. The computing load is measured by the weighted average of the processor usage and memory occupancy. Specifically, the weights can be set according to the design requirements of the system, such as 70% and 30% for the processor usage and memory occupancy respectively.

[0162] According to the monitoring results of the system computing load, the time series segment length (i.e. preset length) of the LSTM model is dynamically adjusted. When the load is high, in order to avoid exceeding the computing resource limit, the length of each time series is shortened, thereby reducing the computing pressure of the model; when the load is low, the time step can be appropriately extended.

[0163] According to the historical operation data of the target compressor, the first load threshold is set as the critical point for calculating the load. If the calculated load is greater than or equal to this threshold, it means that the system resources are relatively tight and the length of the time series should be shortened; if the load is lower than the threshold, it means that the system resources have not been fully utilized, and the length of the series can be extended to improve the prediction accuracy.

[0164] For example, key performance indicators such as the response time, task processing time, and model inference time of the compressor system can be recorded under different load scenarios. In particular, attention is paid to whether the computing power of the system is fully utilized under high load, and whether there are delays, processor overheating, or memory overflow. Once this data is collected, the next step is to set the critical point of the computing load (the first load threshold) according to the real-time requirements of the system. This threshold should reflect that the system can still maintain the required response speed under high load without exceeding the capabilities of the computing resources.

[0165] Exemplarily, when the computational load is higher than the first load threshold (e.g., more than 70%), the system automatically shortens the time step. For example, a sequence of 50 time steps can be shortened to 30 time steps to reduce the amount of model calculations and improve real-time response. When the computational load is lower than the first load threshold, the system can appropriately extend the time step to increase the model's ability to process historical data. For example, extending the original 30 time steps to 50 time steps can improve the model's prediction accuracy.

[0166] In this way, by dynamically adjusting the preset length of the time series, the present invention can automatically adapt to system resources under different computing load conditions, ensuring that the system can still maintain real-time performance under high load and avoiding model calculation overload.

[0167] Regarding S103 above:

[0168] In specific implementation, the following key operating parameters can be adjusted according to the prediction angle:

[0169] Speed: The predicted rotor angle reflects the deviation of the motor speed. By providing feedback on this angle, the compressor system can adjust the speed to ensure that the compressor maintains stable operating efficiency under different load conditions.

[0170] Torque / load: If the predicted angle indicates that the rotor angle deviates from the desired value, the system can adjust the torque by adjusting the current or voltage to control the magnetic field in the stator windings, ensuring that the compressor load is adapted to the operating conditions.

[0171] Stator current: The predicted angle helps determine if the motor is currently under high load. Under high load conditions, the system can adjust the stator current based on the angle feedback to prevent overload or motor loss.

[0172] Operating frequency: If the system detects irregular fluctuations in the predicted rotor angle, it may indicate a large change in the system load. The system can automatically adjust the operating frequency to match the current operating requirements to ensure that the compressor operates at the optimal frequency.

[0173] In this way, by adjusting the compressor parameters such as speed, load and current based on the predicted rotor angle, the compressor can maintain efficient operation under different loads and operating conditions, avoiding system instability or overload problems caused by mismatched operating parameters.

[0174] It is important to emphasize that in the phase-locked loop system, the spatial harmonic components and the back-EMF signal have been weighted and combined to generate a preliminary rotor angle estimate through phase error detection. This process has effectively utilized these signals and corrected the phase error through the feedback control mechanism to generate a more accurate estimated angle. For the combination of spatial harmonics and back-EMF signals, the PLL can already process the noise and high-frequency components in the signal at this stage, thereby obtaining a reliable angle estimate.

[0175] The introduction of machine learning models is mainly to further optimize or compensate the estimated angle, especially to improve the prediction accuracy under complex working conditions. LSTM, a deep learning model, is good at processing time series data and can capture more complex time dependencies and historical information, making up for the shortcomings of traditional feedback control systems (such as PLL) in some nonlinear or highly dynamic change scenarios. Specifically, LSTM can: use a wider range of historical data and real-time inputs to compensate for errors in PLL estimated angles caused by complex nonlinear working conditions. Handle the complex effects of multi-source data from the system (such as electromagnetic characteristics, speed, load, etc.) on angle estimation, which may be features that the phase-locked loop system cannot effectively handle. Provide a more flexible processing method than a single feedback mechanism based on phase error, especially when the working conditions are complex or there is higher uncertainty.

[0176] Those skilled in the art will appreciate that, in the above method of specific implementation, the order in which the steps are written does not imply a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of the steps should be determined by their functions and possible internal logic.

[0177] Based on the same inventive concept, an embodiment of the present application also provides a compressor weak magnetic region stable operation system based on angle compensation. Since the principle of the system in the embodiment of the present application to solve the problem is similar to the above-mentioned compressor weak magnetic region stable operation method based on angle compensation in the embodiment of the present application, the implementation of the system can refer to the implementation of the method, and the repeated parts will not be repeated.

[0178] Reference Figure 3 , which is a schematic diagram of a compressor weak magnetic region stable operation system based on angle compensation provided by an embodiment of the present application, the system includes:

[0179] An estimation unit 10 estimates a rotor angle of a target compressor based on a phase-locked loop to generate an estimated angle of the rotor;

[0180] The prediction unit 20 generates a predicted angle based on the operating parameters of the target compressor and the estimated angle of the rotor using a preset machine learning model; wherein the machine learning model is used to compensate for the influence of multiple frequency components on the predicted angle; the multiple frequency components include: a spatial harmonic component, a high frequency component of a back electromotive force signal, and a noise signal;

[0181] The control unit 30 adjusts the operating parameters of the target compressor based on the predicted angle.

[0182] For descriptions of the processing flow of each module in the system and the interaction flow between each module, reference can be made to the relevant descriptions in the above method embodiments, which will not be described in detail here.

[0183] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, the specific working process of the system described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here. In the several embodiments provided in the present disclosure, it should be understood that the disclosed system and method can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of the device or unit can be electrical, mechanical or other forms.

[0184] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0185] In addition, each functional unit in each embodiment of the present disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0186] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present disclosure, which are used to illustrate the technical solutions of the present disclosure, rather than to limit them. The protection scope of the present disclosure is not limited thereto. Although the present disclosure is described in detail with reference to the above-mentioned embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-mentioned embodiments within the technical scope disclosed in the present disclosure, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be based on the protection scope of the claims.

Claims

1. A method for stable operation of a compressor in a weak magnetic field region based on angle compensation, characterized in that: include: estimating a rotor angle of a target compressor based on a phase-locked loop to generate an estimated angle of the rotor; Generate a predicted angle based on the operating parameters of the target compressor and the estimated angle of the rotor using a preset machine learning model; wherein the machine learning model is used to compensate for the influence of multiple frequency components on the predicted angle; the multiple frequency components include: spatial harmonic components, high frequency components of back electromotive force signals, and noise signals; Based on the predicted angle, operating parameters of the target compressor are adjusted.

2. The method according to claim 1, characterized in that Generating the estimated angle of the rotor further comprises: injecting a high frequency voltage signal into a stator winding of the target compressor motor; Collecting a current response signal of the stator winding, and extracting a spatial harmonic component in the current response signal; The spatial harmonic component is weightedly combined with the back electromotive force signal, and the result is provided to a phase-locked loop for phase error detection to generate the estimated angle.

3. The method according to claim 2, characterized in that The method further includes: using a finite element analysis method to model the electromagnetic characteristics of the target compressor to generate electromagnetic data of the target compressor under a target operating state; The predicted angle is generated using a preset machine learning model based on the operating parameters of the target compressor, the estimated angle of the rotor, and electromagnetic data generated by finite element analysis.

4. The method according to claim 3, characterized in that The finite element analysis method adopts a pre-calculation method to pre-calculate the electromagnetic characteristics of the target compressor motor and generate electromagnetic data under different operating conditions; The electromagnetic data include: magnetic flux density distribution, current density distribution, flux linkage coupling and electromagnetic loss characteristics; The pre-calculated electromagnetic data is stored via a lookup table or a database and is dynamically called during real-time operation according to the current operating state of the compressor.

5. The method according to claim 4, characterized in that Also includes: Determining the frequency and amplitude of the high-frequency voltage signal based on electromagnetic data pre-calculated and generated by finite element analysis; When the magnetic flux density is lower than a first threshold, a high-frequency voltage signal with a frequency within a first preset range is selected for injection; When the current density is higher than a second threshold, the amplitude of the high-frequency voltage signal is adjusted to be within a second preset range.

6. The method according to claim 5, characterized in that The input data of the machine learning model include: the speed, load, stator current, back electromotive force signal, estimated angle of the rotor, and electromagnetic data generated by finite element analysis pre-calculation of the target compressor; The machine learning model includes: a long short-term memory network model; The long short-term memory network model is used to process the input data in real time and generate a prediction angle; Among them, the long short-term memory network model generates a prediction angle by analyzing historical data and current operating parameters.

7. The method according to claim 6, characterized in that The machine learning model is trained using a truncated back-propagation algorithm, including: Dividing the historical operating parameters of the target compressor into segments based on a preset length; Wherein, the preset length is determined based on the system calculated load and real-time target; the historical operating parameters include: speed, load, stator current and timing data of back electromotive force signal; Perform local gradient calculation on the time series data of historical operating parameters of each compressor, and update the weight of the machine learning model based on the local gradient calculation results; Wherein, the local gradient calculation limits the number of time steps of each back propagation by truncating the back propagation; After the calculation of each segment of time series data is completed, the hidden state and memory state of the current segment of time series data are used as the initial state for the calculation of the next segment of time series data.

8. The method according to claim 7, characterized in that The preset length is determined based on the system computing load and real-time target and includes: Monitoring a system computing load of a target compressor; wherein the system computing load is determined based on a weighted average of the processor usage rate and the memory occupancy rate; When the weighted average value of the load calculated by the system is greater than or equal to a preset first load threshold, shortening the preset length; When the weighted average value of the load calculated by the system is less than the first load threshold, the preset length is extended.

9. The compressor weak magnetic area stable operation system based on angle compensation is characterized by: include: an estimating unit, estimating a rotor angle of a target compressor based on a phase-locked loop, and generating an estimated angle of the rotor; A prediction unit, using a preset machine learning model, generates a predicted angle based on the operating parameters of the target compressor and the estimated angle of the rotor; wherein the machine learning model is used to compensate for the influence of multiple frequency components on the predicted angle; the multiple frequency components include: spatial harmonic components, high frequency components of back electromotive force signals, and noise signals; The control unit adjusts the operating parameters of the target compressor based on the predicted angle.

Images