A control method for a motor used in a variable frequency compressor

Abstract

The present invention belongs to the field of motor control technology and discloses a control method for a motor for a variable frequency compressor, comprising embedding a distributed thermocouple array in the stator slot of the motor to collect temperature data of the motor, synchronously collecting motor operating parameters, generating a dynamic thermal field characteristic map through a data fusion algorithm, marking the coordinates of hot spot areas with abnormally high temperatures, establishing a dynamic correlation model, and updating the heat conduction characteristic parameters in the dynamic correlation model through a sliding window algorithm; outputting a predicted temperature rise curve through the dynamic correlation model, adjusting the PWM control strategy of the inverter, and inputting it into the motor drive unit to perform preliminary regulation, while collecting feedback data to generate a multi-dimensional feedback feature vector, and then iteratively optimizing the combined weights of the switching interval compensation amount and the harmonic injection strategy through a gradient descent algorithm to form a global optimal strategy that takes into account both heat dissipation efficiency and operational stability, thereby significantly improving the operational stability and energy efficiency level of the variable frequency compressor under various working conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of motor control, and in particular to a control method for a motor for a variable frequency compressor. Background Art

[0002] As the core power unit of refrigeration equipment, variable-frequency compressors are widely used in household air conditioners, cold chain logistics, and industrial refrigeration systems. With the continued advancement of energy conservation and emission reduction policies, efficient motor speed regulation and thermal management capabilities have become key to improving system energy efficiency. Under harsh operating conditions such as high-temperature, enclosed environments and frequent starts and stops, localized overheating of motor windings and electromagnetic parameter mismatch are particularly prominent. Traditional heat dissipation designs struggle to balance the contradiction between suppressing rapid temperature rise and responding to dynamic loads, resulting in energy efficiency degradation and even the risk of device failure.

[0003] In the existing technology, the control methods for variable frequency compressor motors mostly adopt temperature threshold triggered heat dissipation strategies, such as adjusting the PWM switching frequency through a preset temperature threshold or forcibly starting an auxiliary heat dissipation device. Although such methods can temporarily alleviate overheating, they have the core defect of decoupling heat dissipation regulation from electromagnetic control: their temperature feedback is based only on limited measurement point data and cannot accurately identify the thermal field distribution characteristics inside the winding, resulting in a disconnect between the global heat dissipation strategy and local hot spot suppression; at the same time, the single-dimensional adjustment of PWM parameters often leads to secondary problems such as increased torque pulsation and worsening current harmonics, forming a vicious cycle of "heat dissipation optimization-stability degradation." The essence of this defect lies in the failure to establish a dynamic coupling mechanism between thermal field evolution and electromagnetic control parameters, resulting in the inability to coordinately optimize heat dissipation efficiency and operational stability.

[0004] Therefore, an intelligent control method capable of achieving coordinated regulation is required to solve the technical problem that the existing variable frequency compressor motor has a single heat dissipation regulation capability and cannot establish coordinated optimization between heat dissipation and stability. Summary of the Invention

[0005] The object of the present invention is to provide a control method for a motor for a variable frequency compressor to solve the above technical problems.

[0006] To achieve this object, the present invention adopts the following technical solutions:

[0007] A method for controlling a motor for a variable frequency compressor, comprising:

[0008] A distributed thermocouple array is embedded in the motor stator slots to collect the motor's temperature data and synchronously collect the motor's operating parameters. The temperature data and the motor operating parameters are combined through a data fusion algorithm to generate a dynamic thermal field characteristic map, and the coordinates of hot spots with abnormally high temperatures are marked. The motor operating parameters include the real-time temperature sequence of each winding segment and the three-phase current spectrum during motor operation.

[0009] Based on the dynamic thermal field characteristic map, a dynamic correlation model between winding temperature and motor operating parameters is established, and the heat conduction characteristic parameters in the dynamic correlation model are updated through a sliding window algorithm;

[0010] The predicted temperature rise curve is output by the dynamic correlation model to adjust the PWM control strategy of the inverter; the PWM control strategy includes:

[0011] For the marked hotspot area, increase the switching interval compensation of the corresponding phase winding according to the temperature rise gradient ratio;

[0012] According to the energy ratio of the fundamental wave and the third harmonic of the three-phase current spectrum, the harmonic injection strategy of the voltage vector is adjusted;

[0013] The adjusted PWM control strategy is input into the motor drive unit to perform preliminary control. At the same time, the winding temperature change rate and torque fluctuation data after control are collected to generate a multi-dimensional feedback feature vector.

[0014] Based on the multi-dimensional feedback feature vector, the combined weight of the switching interval compensation amount and the harmonic injection strategy is iteratively optimized through the gradient descent algorithm, and the motor control strategy matrix adapted to the current working condition is output and loaded into the control system in real time.

[0015] Optionally, the process of embedding the distributed thermocouple array in the stator slot of the motor is specifically as follows:

[0016] During the design phase of the motor stator slots, a laser micromachining process is used to form an array of micro grooves with equal spacing in the stator slots;

[0017] Flexible thin-film thermocouples are embedded in the grooves in a serpentine routing manner. The flexible thin-film thermocouples include a nickel-chromium-nickel-silicon composite layer and a polyimide insulation layer. The node spacing between adjacent flexible thin-film thermocouples is consistent with the groove spacing, forming a distributed temperature measurement network covering the circumference of the stator.

[0018] Optionally, the temperature data and the motor operating parameters are combined through a data fusion algorithm to generate a dynamic thermal field characteristic map, and the coordinates of hot spots where the temperature rises abnormally are marked, specifically including:

[0019] Extracting current harmonic features of the three-phase current spectrum by fast Fourier transform, wherein the current harmonic features include spectrum energy of fundamental wave, third harmonic and fifth harmonic components;

[0020] Performing joint spatiotemporal preprocessing on the temperature data, using a wavelet threshold denoising algorithm, selecting the sym4 wavelet basis for 5-layer decomposition, and setting an adaptive threshold function to eliminate electromagnetic interference noise of the temperature data;

[0021] A three-dimensional polar coordinate system is established based on the geometric parameters of the motor, and the discrete temperature points are mapped to the spatial topological structure of the stator winding to generate an initial temperature field distribution cloud map.

[0022] Optionally, the step of mapping the discrete temperature points to the spatial topology of the stator winding to generate an initial temperature field distribution cloud map further includes:

[0023] The initial temperature field distribution cloud map and the three-phase current spectrum are input into the DS evidence theory fusion model, where the temperature data is used as the first evidence body, the third harmonic energy ratio is used as the second evidence body, and the fifth harmonic energy ratio is used as the third evidence body;

[0024] The credibility weight of each evidence body is calculated through the correction algorithm, and the thermal field confidence distribution matrix containing the temperature-current coupling characteristics is output;

[0025] The interpolation algorithm is used to reconstruct the global temperature gradient surface of the winding using the thermal field confidence distribution matrix. When the local temperature confidence is ≥0.92 and the gradient change rate is >3℃ / mm 2 When an abnormal hotspot is detected, its polar coordinate position (r, θ, z) and confidence level are recorded to calibrate the hotspot area.

[0026] Optionally, the polar coordinate position (r, θ, z) and confidence level are recorded to mark the hotspot area, and then the following steps are further included:

[0027] The coordinates of the calibrated hotspot area are aligned with the current harmonic characteristics in time and space to establish a three-dimensional feature vector set including temperature amplitude, gradient vector, and harmonic energy weight;

[0028] Through the PCA dimensionality reduction processing of the three-dimensional feature vector set, the first three principal components are extracted to form the encoding matrix of the thermal field feature map.

[0029] Optionally, the step of establishing a dynamic correlation model between winding temperature and motor operating parameters based on the dynamic thermal field characteristic map and updating the heat conduction characteristic parameters in the dynamic correlation model through a sliding window algorithm specifically includes:

[0030] Analyze the encoding matrix of the dynamic thermal field characteristic map and extract the temperature amplitude principal component vector T_pca, harmonic energy weight vector H_wei and gradient vector field G_xyz as the model input data set;

[0031] A winding temperature rise rate prediction model is constructed as a dynamic correlation model. The model input variables are defined as: the current temperature amplitude T_pca(k), the harmonic energy vector of the three historical cycles [H_wei(k-1), H_wei(k-2), H_wei(k-3)], the modulus value ||G|| of the gradient vector field G_xyz(k), and the model output is the temperature rise rate ΔT_pred of the next cycle.

[0032] Initial heat conduction characteristic parameters, including heat conduction coefficient α_t, harmonic influence factor β_t and gradient weight δ_t;

[0033] Set the sliding window duration and the number of sampling groups within the window duration. Based on the residual between the measured temperature rise rate ΔT_meas and the predicted value ΔT_pred within the window, use the recursive least squares method to update the thermal conductivity characteristic parameters of the winding temperature rise rate prediction model.

[0034] The residual root mean square error (RMSE) within the window is calculated. When RMSE is less than 0.12°C / s, the updated thermal conductivity characteristic parameters are accepted. If RMSE is greater than or equal to 0.2°C / s after two consecutive updates, the previous valid parameter set is returned to.

[0035] Optionally, the adjusted PWM control strategy is input into the motor drive unit to perform preliminary regulation, and at the same time, the regulated winding temperature change rate and torque fluctuation data are collected to generate a multi-dimensional feedback feature vector, specifically including:

[0036] The adjusted PWM control parameter set is loaded into the motor driver's FPGA control unit for preliminary control. The actual dead time T_phase of each phase is written to the timer register in 16 bits via the SPI interface, and the composite voltage vector waveform data is stored in the dual-port RAM buffer.

[0037] The distributed thermocouple array sampling rate is used to capture the temperature sequence of each winding section after regulation, and the temperature change rate ΔT_rate = ΔT / Δt is calculated with a time window Δt = 100ms;

[0038] The motor rotor angular displacement signal is collected by a magnetic encoder, and the second-order derivative is obtained after Savitzky-Golay filtering to obtain the torque fluctuation component ΔTe.

[0039] Optionally, the second-order derivative is obtained after Savitzky-Golay filtering to obtain the torque fluctuation component ΔTe, and then the following is further included:

[0040] Extracting temperature feedback indicators, including: maximum temperature rise rate ΔT_max in the hotspot area, average temperature rise gradient T_avg, and temperature distribution uniformity index U_T = σ(T) / μ(T);

[0041] Extracting electrical parameter indicators, including: total harmonic distortion (THD) of current and third harmonic distortion (H3%);

[0042] The temperature feedback index and the electrical parameter index are normalized and encoded into a 5-dimensional feature vector in the order of [ΔT_max, T_avg, U_T, THD, H3%] to construct a multi-dimensional feedback feature vector.

[0043] Optionally, based on the multi-dimensional feedback feature vector, a gradient descent algorithm is used to iteratively optimize the combined weight of the switching interval compensation amount and the harmonic injection strategy, output a motor control strategy matrix adapted to the current working condition, and load it into the control system in real time, specifically including:

[0044] Analyzing the multidimensional feedback eigenvector, extracting the maximum temperature rise rate ΔT_max, the current total harmonic distortion rate THD, and the torque ripple peak ΔTe_pp as optimization targets, and constructing a normalized cost function;

[0045] Define the optimization variables and constraints: the switching interval compensation weight k_comp ranges from [0.1, 0.4], the harmonic injection ratio weight k_h3 ranges from [0.05, 0.25], set the learning rate α to 0.02, the maximum number of iterations to 50, and the convergence threshold ΔJ to < 0.5%.

[0046] Optionally, the definition of optimization variables and constraints further includes:

[0047] Calculate the cost function value J_current based on the current weight combination (k_comp, k_h3);

[0048] Calculate partial derivatives by perturbation analysis: apply ±0.02 perturbation to k_comp to obtain the J change ΔJ_comp; apply ±0.005 perturbation to k_h3 to obtain the J change ΔJ_h3;

[0049] The current weight combination (k_comp, k_h3) is updated by the calculated partial derivatives. If k_comp is out of range, the boundary value is taken. If k_h3 is out of range, the step size is 0.005 to perform gradient descent iterative optimization.

[0050] When the decrease of J is less than 0.5% for three consecutive iterations, the current weight combination is extracted;

[0051] Verify the constraints of ΔT_max≤5℃ / s, THD≤5%, and ΔTe_pp≤2N·m. If any indicator exceeds the limit, reiterate according to the weight priority of the cost function component;

[0052] Map the optimized weights to the actual control parameters to generate 12 sets of discretized parameter combinations, which are sorted in ascending order of J values and stored in the strategy matrix;

[0053] The strategy matrix is transferred to the DSP coprocessor of the motor controller via the PCIe bus. At the beginning of the next control cycle, the strategy group with the smallest J value is selected for activation and execution.

[0054] Compared with the existing technology, the present invention has the following beneficial effects: the method is constructed based on the principle of multi-physical field coupling control, realizes the synchronous perception of temperature and electrical parameters through a distributed thermocouple array, and uses a data fusion algorithm to establish a dynamic thermal field characteristic map to locate the hot spot area; combined with the dynamic correlation model of the sliding window algorithm, the thermal field characteristics and the PWM control parameters are formed into a closed-loop mapping; the dual-path control of the switch interval compensation and harmonic injection is driven by the predicted temperature rise curve to achieve the preliminary adaptation of the thermal-electric parameters; based on multi-dimensional feedback data, a coupling evaluation index of thermal control effect and mechanical performance is constructed, and the gradient descent algorithm is used to optimize the weights of multi-objective conflict parameters to form a global optimal strategy that takes into account both heat dissipation efficiency and operation stability; the method upgrades the traditional single electronic control logic to an intelligent control system that coordinates thermal-electric-mechanical multi-physical fields, while ensuring efficient heat dissipation of the motor, significantly improving the operating stability and energy efficiency of the variable frequency compressor under various working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0056] The structures, proportions, sizes, etc. depicted in the drawings of this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with this technology. They are not intended to limit the conditions under which the present invention can be implemented and therefore have no substantive technical significance. Any structural modifications, changes in proportional relationships, or adjustments in size should still fall within the scope of the technical contents disclosed in the present invention without affecting the effects and objectives that can be achieved by the present invention.

[0057] Figure 1 2 is a flow chart of a method for controlling a motor for a variable frequency compressor according to this embodiment. DETAILED DESCRIPTION

[0058] In order to make the purpose, features, and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described below are only 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 making creative work are within the scope of protection of the present invention.

[0059] In the description of the present invention, it should be understood that the terms "upper," "lower," "top," "bottom," "inner," "outer," etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate the description of the present invention and simplify the description. They are not intended to indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limitations on the present invention. It should be noted that when a component is considered to be "connected" to another component, it may be directly connected to the other component or there may be a centrally located component.

[0060] The technical solution of the present invention will be further described below with reference to the accompanying drawings and through specific implementation methods.

[0061] Combine Figure 1 As shown, an embodiment of the present invention provides a method for controlling a motor for a variable frequency compressor, comprising:

[0062] S1: A distributed thermocouple array is embedded in the stator slots of the motor to collect the motor's temperature data and synchronously collect the motor's operating parameters. The temperature data and motor operating parameters are combined through a data fusion algorithm to generate a dynamic thermal field characteristic map, and the coordinates of hot spots with abnormally high temperatures are marked. The motor operating parameters include the real-time temperature sequence of each winding segment and the three-phase current spectrum when the motor is running.

[0063] S2, based on the dynamic thermal field characteristic map, establishes a dynamic correlation model between winding temperature and motor operating parameters, and updates the heat conduction characteristic parameters in the dynamic correlation model through a sliding window algorithm; the dynamic correlation model includes the mapping relationship between current harmonic components, PWM switching frequency and local temperature rise rate.

[0064] S3, using the dynamic correlation model to output the predicted temperature rise curve, adjusts the inverter's PWM control strategy; the PWM control strategy includes:

[0065] For the marked hotspot area, increase the switching interval compensation of the corresponding phase winding according to the temperature rise gradient ratio;

[0066] The harmonic injection strategy of the voltage vector is adjusted according to the energy ratio of the fundamental wave and the third harmonic of the three-phase current spectrum.

[0067] S4, inputting the adjusted PWM control strategy into the motor drive unit to perform preliminary regulation, while collecting the winding temperature change rate and torque fluctuation data after regulation to generate a multi-dimensional feedback feature vector.

[0068] S5, based on the multi-dimensional feedback feature vector, iteratively optimizes the combined weight of the switching interval compensation and the harmonic injection strategy through the gradient descent algorithm, outputs the motor control strategy matrix adapted to the current working conditions and loads it into the control system in real time.

[0069] The working principle of the present invention is as follows: this control method is constructed based on the principle of multi-physical field coupling control, and realizes synchronous perception of temperature and electrical parameters through a distributed thermocouple array, and uses a data fusion algorithm to establish a dynamic thermal field characteristic map to locate the hot spot area; combined with the dynamic correlation model of the sliding window algorithm, the thermal field characteristics and PWM control parameters are formed into a closed-loop mapping; by predicting the temperature rise curve to drive the dual-path control of switch interval compensation and harmonic injection, the preliminary adaptation of thermal-electric parameters is achieved; based on multi-dimensional feedback data, a coupling evaluation index of thermal control effect and mechanical performance is constructed, and the gradient descent algorithm is used to optimize the weight of multi-objective conflict parameters to form a global optimal strategy that takes into account both heat dissipation efficiency and operation stability; this method upgrades the traditional single electronic control logic to an intelligent control system that coordinates thermal-electric-mechanical multi-physical fields, while ensuring efficient heat dissipation of the motor, significantly improving the operating stability and energy efficiency of the variable frequency compressor under various working conditions.

[0070] In this embodiment, the process of embedding the distributed thermocouple array in the stator slot of the motor is specifically described as follows:

[0071] During the design phase of the motor stator slot, a laser micromachining process is used to form an array of equally spaced micro grooves in the stator slot; preferably, the groove spacing is set to 2.8±0.1 mm, and the depth is 1 / 3 of the stator slot depth.

[0072] Flexible thin-film thermocouples are embedded in the grooves in a serpentine routing manner. The flexible thin-film thermocouples include a nickel-chromium-nickel-silicon composite layer and a polyimide insulation layer. The node spacing between adjacent flexible thin-film thermocouples is consistent with the groove spacing, forming a distributed temperature measurement network covering the circumference of the stator.

[0073] In this embodiment, it is specifically explained that step S1 specifically includes:

[0074] S11, collecting temperature data of the motor through a distributed thermocouple array and synchronously collecting motor operating parameters;

[0075] A high-density layout strategy (four measurement points per slot) is employed to synchronously collect temperature data and three-phase current signals from each winding segment at a 1kHz sampling rate. The temperature data is filtered through an isolation amplifier to eliminate common-mode interference, while the current signal is converted via a 16-bit ADC and cached in the FPGA, ensuring strict temporal and spatial data alignment.

[0076] S12, extracting current harmonic features of the three-phase current spectrum by fast Fourier transform, where the current harmonic features include spectrum energy of the fundamental wave, the third harmonic, and the fifth harmonic components;

[0077] Fast Fourier transform analysis was performed on the three-phase current signals to calculate the energy contributions of the fundamental, third harmonic, and fifth harmonic. A bandpass filter (150 ± 5 Hz) was specifically set for the third harmonic to suppress noise interference near the switching frequency (15 kHz). The extracted harmonic eigenvectors [H3%, H5%] reflect the electromagnetic loss distribution and are strongly correlated with the temperature field evolution.

[0078] S13 performs joint spatiotemporal preprocessing on the temperature data. A wavelet threshold denoising algorithm is used, the sym4 wavelet basis is selected for 5-layer decomposition, and an adaptive threshold function is set to eliminate electromagnetic interference noise from the temperature data. Soft threshold processing is performed on coefficients exceeding the threshold. After reconstruction, the signal-to-noise ratio is improved by ≥18dB, effectively filtering out the pulse noise caused by PWM switching.

[0079] S14, establishing a three-dimensional polar coordinate system based on the geometric parameters of the motor, mapping the discrete temperature points to the spatial topological structure of the stator winding, and generating an initial temperature field distribution cloud map.

[0080] A non-uniform grid mapping model was established based on the three-dimensional polar coordinate parameters of the motor stator. Discrete temperature measurement points were interpolated into a grid system consisting of multiple cells (432). This generated an initial temperature field cloud map with a resolution of 0.5°C, visualizing the thermal distribution of the windings.

[0081] S15: Input the initial temperature field distribution cloud map and the three-phase current spectrum into the DS evidence theory fusion model, where the temperature data is used as the first evidence body, the third harmonic energy ratio is used as the second evidence body, and the fifth harmonic energy ratio is used as the third evidence body; the joint confidence is calculated through the evidence synthesis rule to break through the limitations of a single data source.

[0082] S16, calculates the credibility weight of each evidence body through the correction algorithm, and outputs the thermal field confidence distribution matrix containing the temperature-current coupling characteristics; ensures the dominance of temperature evidence in high-conflict scenarios (w1≥0.6), and outputs the thermal field confidence matrix.

[0083] S17, the interpolation algorithm is used to reconstruct the global temperature gradient surface of the winding using the thermal field confidence distribution matrix. When the local temperature confidence is ≥0.92 and the gradient change rate is >3℃ / mm2 When an abnormal hotspot is detected, its polar coordinate position (r, θ, z) and confidence level are recorded to calibrate the hotspot area.

[0084] S18, align the calibrated hotspot area coordinates with the current harmonic characteristics in time and space, and establish a three-dimensional feature vector set containing temperature amplitude, gradient vector, and harmonic energy weight; ensure data causality through the timestamp synchronization mechanism to form a time series feature library.

[0085] S19, through the PCA dimensionality reduction processing of the three-dimensional feature vector set, the first three principal components are extracted to form the encoding matrix of the thermal field feature map as the digital representation of the thermal field feature map.

[0086] In this embodiment, it is specifically described that step S2 specifically includes:

[0087] S21: Parse the encoding matrix of the dynamic thermal field characteristic map and extract the temperature amplitude principal component vector T_pca, the harmonic energy weight vector H_wei, and the gradient vector field G_xyz as the model input dataset. T_pca has a dimension of [1×64], and H_wei contains the energy contribution coefficients of the fundamental wave, the third harmonic, and the fifth harmonic. Data normalization eliminates dimensional differences and forms the model input dataset.

[0088] Temperature amplitude principal component vector T_pca: The dimension is [1×64], representing 64 standardized temperature principal components, reflecting the dimensionality reduction characteristics of the global temperature distribution of the winding;

[0089] Harmonic energy weight vector H_wei: Contains the normalized energy weight coefficients of the fundamental wave (accounting for 65%-85%), the third harmonic (8%-25%), and the fifth harmonic (2%-10%), representing the electromagnetic loss distribution;

[0090] Gradient vector field G_xyz: a three-dimensional matrix structure, each unit stores a gradient vector, and quantifies the spatial variation trend of the temperature field.

[0091] S22, construct a winding temperature rise rate prediction model as a dynamic correlation model, and define the model input variables as: the current temperature amplitude T_pca(k), the harmonic energy vector of the three historical cycles [H_wei(k-1), H_wei(k-2), H_wei(k-3)], the modulus value ||G|| of the gradient vector field G_xyz(k), and the model output is the temperature rise rate ΔT_pred of the next cycle; the winding temperature rise rate prediction model reveals the dynamic correlation between temperature evolution and multi-physical fields through linear combination.

[0092] S23, initial heat conduction characteristic parameters, which include heat conduction coefficient α_t, harmonic influence factor β_t and gradient weight δ_t;

[0093] When setting the initial heat conduction characteristic parameters, preferably, the initial heat conduction coefficient α_0 is set to 0.75 W / (m·K), the harmonic influence factor β_0 is set to 0.02, the gradient weight δ_0 is set to 0.15, and the historical harmonic attenuation coefficients γ_1 are defined to be 0.5, γ_2 is defined to be 0.3, and γ_3 is defined to be 0.2.

[0094] S24, setting the sliding window duration and the number of sampling groups within the window duration, and updating the thermal conductivity characteristic parameters of the winding temperature rise rate prediction model by recursive least squares method based on the residual between the measured temperature rise rate ΔT_meas and the predicted value ΔT_pred within the window;

[0095] Preferably, the sliding window duration is set to 300ms, including 20 consecutive groups of sampled data, with an interval of 15ms between each group.

[0096] S25, calculate the residual root mean square error (RMSE) in the window, and accept the updated heat conduction characteristic parameters when RMSE is less than 0.12°C / s; if RMSE is greater than or equal to 0.2°C / s after two consecutive updates, return to the previous valid parameter set.

[0097] Calculate the prediction error RMSE within the window = √(Σ(ΔT_meas-ΔT_pred) 2 / N), the threshold RMSE < 0.12℃ / s corresponds to a temperature rise prediction accuracy of ±1.2℃ / min.

[0098] Abnormal handling: When RMSE is ≥ 0.2°C / s for two consecutive times, it is determined to be a model mismatch (such as a cooling system failure), and the system automatically falls back to the mean parameters (α_avg, β_avg, δ_avg) of the previous 10 windows, and a warning signal is triggered.

[0099] In this embodiment, it is specifically explained that step S4 specifically includes:

[0100] S41, loading the adjusted PWM control parameter set into the FPGA control unit of the motor driver to perform preliminary control, wherein the actual dead time T_phase of each phase is written into the timer register via the SPI interface in 16 bits, and the composite voltage vector waveform data is stored in the dual-port RAM buffer;

[0101] The adjusted PWM control parameters are written to the motor driver's FPGA control unit via the SPI interface with 16-bit accuracy, ensuring that the timer register receives the accurate dead time parameter (T_phase). The actual dead time of each phase is configured differently based on the spatial angle of the hotspot area; for example, when the U-phase hotspot is at θ = 120°:

[0102] T_phase_U=T_dead+ΔT_comp·cos (120°);

[0103] The composite voltage vector waveform data (including the fundamental and third harmonic components) is transferred to a dual-port RAM buffer via a DMA channel. The dual-port parallel access feature enables zero-latency switching of waveform data, ensuring seamless update of the modulation strategy at the start of the next PWM cycle (typically 50μs) and avoiding control interruption.

[0104] S42, using a distributed thermocouple array sampling rate to capture the temperature sequence of each winding segment after regulation, and calculating the temperature change rate ΔT_rate = ΔT / Δt, with a time window Δt = 100ms;

[0105] Real-time monitoring of overheating risks. When the ΔT_rate in a certain area exceeds the limit three times in a row, a warning signal is triggered, providing key input for subsequent optimization.

[0106] S43, the motor rotor angular displacement signal is collected through a magnetic encoder, and the second-order derivative is obtained after Savitzky-Golay filtering to obtain the torque fluctuation component ΔTe. Its spectrum analysis uses STFT transformation with a resolution of 0.5 Hz;

[0107] The rotor angular displacement signal is collected at a frequency of 20kHz by a magnetic encoder, and the 5th-order Savitzky-Golay filter (window length 201 points) is used to smooth the angle data to eliminate the slight oscillation caused by the cogging torque. The second-order derivative (d 2 θ / dt 2 ), combined with the motor's moment of inertia J = 0.02kg·m 2 Calculate the torque fluctuation component ΔTe=J·d 2 θ / dt 2 The short-time Fourier transform (STFT, window length 4096 points, overlap rate 75%) was used to analyze the ΔTe spectrum, extract the main frequency energy distribution (such as the 150Hz switching frequency sideband) with a precision of 0.5Hz, and quantify the mechanical vibration characteristics.

[0108] S44, extracting temperature feedback indicators, which include: maximum temperature rise rate ΔT_max in the hotspot area, average temperature rise gradient T_avg, and temperature distribution uniformity index U_T = σ(T) / μ(T);

[0109] Three core indicators are extracted from the temperature data: the maximum temperature rise rate ΔT_max in the hotspot (the 95th percentile of the ΔT_rate across all measurement points), the average temperature rise gradient T_avg (the arithmetic mean of the global gradient modulus ||G||), and the temperature distribution uniformity index U_T (the ratio of the standard deviation to the mean). For example, when U_T > 0.15, the temperature distribution is considered uneven, and the cooling strategy needs to be adjusted first. ΔT_max reflects the severity of local overheating, T_avg indicates overall cooling efficiency, and U_T assesses thermal field uniformity. Together, these three indicators constitute the cooling performance evaluation system.

[0110] S45, extracting electrical parameter indicators, which include: total harmonic distortion rate THD of current and third harmonic distortion rate H3%;

[0111] Perform harmonic analysis on the current signal and calculate the total harmonic distortion THD = √(ΣH_n 2 ) / H_1×100% (n=2-50), focusing on the third harmonic distortion rate H3%=H_3 / H_1×100%. A Blackman-Harris windowed FFT (65536 points) is used to improve spectral resolution and accurately separate dense harmonic components. The THD safety threshold is set at 5%. When THD exceeds the limit, the system automatically records the number, amplitude, and phase of the exceeding harmonics, providing data support for optimizing harmonic suppression strategies.

[0112] S46 , normalizing the temperature feedback index and the electrical parameter index and encoding them into a 5-dimensional feature vector in the order of [ΔT_max, T_avg, U_T, THD, H3%] to construct a multi-dimensional feedback feature vector.

[0113] The numerical range of each dimension is compressed to the interval [0, 1]. For example, under certain operating conditions, the eigenvector is [0.8, 0.35, 0.12, 0.28, 0.18], indicating that the current heat dissipation pressure is high (ΔT_max_norm = 0.8) but the electromagnetic interference is controllable (THD_norm = 0.28). This vector serves as a multidimensional feedback signal and is input into the subsequent optimization algorithm to drive dynamic tuning of the control strategy.

[0114] In this embodiment, it is specifically described that step S5 specifically includes:

[0115] S51, analyzing the multi-dimensional feedback feature vector, extracting the maximum temperature rise rate ΔT_max, the current total harmonic distortion rate THD and the torque fluctuation peak ΔTe_pp as optimization targets, and constructing a normalized cost function;

[0116] By analyzing the multidimensional feedback eigenvector, the maximum temperature rise rate ΔT_max (reflecting the risk of local overheating), the current total harmonic distortion (THD) (indicating the level of electromagnetic interference), and the torque ripple peak ΔTe_pp (indicating mechanical stability) were extracted as optimization targets. When constructing the normalized cost function, each metric was scaled to a safety threshold: ΔT_max was divided by 5°C / s (maximum allowable temperature rise rate), THD by 5% (limit), and ΔTe_pp by 2 N·m (mechanical tolerance threshold). These metrics were then linearly combined with a weighting ratio of 0.6:0.3:0.1 to prioritize heat dissipation. For example, when ΔT_max = 4°C / s, THD = 3%, and ΔTe_pp = 1.5 N·m, the calculation J = 0.6*(4 / 5)+0.3*(3 / 5)+0.1*(1.5 / 2)=0.717, providing a quantitative benchmark for subsequent optimization.

[0117] S52, define optimization variables and constraints: the switching interval compensation weight k_comp ranges from [0.1, 0.4], the harmonic injection ratio weight k_h3 ranges from [0.05, 0.25], set the learning rate α to 0.02, the maximum number of iterations to 50, and the convergence threshold ΔJ to < 0.5%.

[0118] Optimization variables and their physical constraints were defined: the switching interval compensation weight k_comp (which adjusts the dead-time compensation strength) was limited to the range of [0.1, 0.4] to prevent overcompensation and a sharp increase in switching losses. The harmonic injection ratio weight k_h3 (which controls the third harmonic voltage amplitude) was limited to [0.05, 0.25] to avoid electromagnetic compatibility issues caused by excessive harmonics. The learning rate α was set to 0.02 (calibrated through previous simulations) to ensure that the weight adjustment step size balanced convergence speed and stability. The maximum number of iterations was 50 (taking approximately 10ms) and the convergence threshold ΔJ was less than 0.5% (when the objective function is saturated) to balance real-time requirements and optimization accuracy.

[0119] S53, calculating the cost function value J_current based on the current weight combination (k_comp, k_h3);

[0120] For example, when k_comp = 0.25 and k_h3 = 0.15, the system obtains ΔT_max = 3.8°C / s, THD = 4.2%, and ΔTe_pp = 1.2 N·m based on the actual control effect. Substituting these into the cost function formula yields J_current = 0.6*(3.8 / 5)+0.3*(4.2 / 5)+0.1*(1.2 / 2)=0.672. This value reflects the overall performance of the current strategy and serves as the starting point for optimization.

[0121] S54, calculate partial derivatives by perturbation analysis method: apply ±0.02 perturbation to k_comp to obtain J change ΔJ_comp; apply ±0.005 perturbation to k_h3 to obtain J change ΔJ_h3;

[0122] For example, if k_comp = 0.25 and J = 0.672, after the perturbation, J_+ = 0.665 and J_- = 0.680, then ΔJ_comp = (0.665 - 0.680) / 0.04 = -0.375. Similarly, applying a ±0.005 perturbation to k_h3 calculates ΔJ_h3.

[0123] S55, update the current weight combination (k_comp, k_h3) using the calculated partial derivatives. If k_comp is out of range, the boundary value is taken. If k_h3 is out of range, the step size is 0.005 to perform gradient descent iterative optimization.

[0124] The weights are adjusted based on the gradient direction. For example, when ΔJ_comp = -0.375, k_comp_new = 0.25 - 0.02*(-0.375) = 0.2575, rounded to 0.258. If k_comp exceeds the range [0.1, 0.4], it is forcibly truncated (e.g., 0.405 → 0.4). If k_h3 exceeds the limit, it is retraced in steps of 0.005 (e.g., 0.26 → 0.255) to prevent oscillation caused by sudden parameter changes. This process ensures that the optimization path remains within the feasible region.

[0125] In step S56, if the J value decreases by less than 0.5% for three consecutive iterations, extract the current weight combination. If the J value decreases by less than 0.5% for three consecutive iterations (e.g., 0.672 → 0.670 → 0.669 → 0.668), convergence is determined. Extract the current weight combination (e.g., k_comp = 0.28, k_h3 = 0.12) and record its J value (e.g., 0.668) as a candidate solution.

[0126] S57 , checking the constraints of ΔT_max≤5°C / s, THD≤5%, and ΔTe_pp≤2N·m. If any of the indicators exceeds the limit, reiterate according to the weight priority of the cost function components.

[0127] S58, mapping the optimized weights to actual control parameters, generating 12 sets of discretized parameter combinations, and storing them in the strategy matrix in ascending order of J value;

[0128] The dead time of each phase is adjusted based on k_comp (for example, T_phase_U = 3μs*(1+0.28cos(120°)) = 3.42μs), and harmonic voltages are generated based on k_h3 (for example, V_h3 = 220V0.12*e^{-j60°} = 26.4∠-60°V). Twelve discrete parameter combinations are generated (for example, k_comp = 0.25, 0.26, ..., 0.36; k_h3 = 0.10, 0.11, ..., 0.15), and stored in a matrix in ascending order of J value, forming an "optimal-suboptimal" strategy sequence that supports real-time fast switching.

[0129] S59, the strategy matrix is transmitted to the DSP coprocessor of the motor controller via the PCIe bus, and at the start of the next control cycle, the strategy group with the smallest J value is selected for activation and execution.

[0130] At the start of the next control cycle (every 200ms), the strategy group with the smallest J value (e.g., index 1, J=0.652) is selected and activated. The ambient temperature change rate (dT_env / dt) and load change rate (dLoad / dt) are simultaneously monitored. When dT_env / dt > 2°C / s or dLoad / dt > 15%, strategy reoptimization is immediately triggered to ensure continuous adaptability under dynamic operating conditions.

[0131] 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 above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features thereof can 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 controlling a motor for a variable frequency compressor, characterized in that: include: A distributed thermocouple array is embedded in the motor stator slots to collect the motor's temperature data and synchronously collect the motor's operating parameters. The temperature data and the motor operating parameters are combined through a data fusion algorithm to generate a dynamic thermal field characteristic map, and the coordinates of hot spots with abnormally high temperatures are marked. The motor operating parameters include the real-time temperature sequence of each winding segment and the three-phase current spectrum during motor operation. Based on the dynamic thermal field characteristic map, a dynamic correlation model between winding temperature and motor operating parameters is established, and the heat conduction characteristic parameters in the dynamic correlation model are updated through a sliding window algorithm; The dynamic correlation model outputs a predicted temperature rise curve to adjust the PWM control strategy of the inverter; The PWM control strategy includes: For the marked hotspot area, increase the switching interval compensation of the corresponding phase winding according to the temperature rise gradient ratio; According to the energy ratio of the fundamental wave and the third harmonic of the three-phase current spectrum, the harmonic injection strategy of the voltage vector is adjusted; The adjusted PWM control strategy is input into the motor drive unit to perform preliminary control. At the same time, the winding temperature change rate and torque fluctuation data after control are collected to generate a multi-dimensional feedback feature vector. Based on the multi-dimensional feedback feature vector, the combined weight of the switching interval compensation amount and the harmonic injection strategy is iteratively optimized through the gradient descent algorithm, and the motor control strategy matrix adapted to the current working condition is output and loaded into the control system in real time.

2. The control method for a variable frequency compressor motor according to claim 1, characterized in that: The process of embedding the distributed thermocouple array in the stator slot of the motor is specifically as follows: During the design phase of the motor stator slots, a laser micromachining process is used to form an array of micro grooves with equal spacing in the stator slots; Flexible thin-film thermocouples are embedded in the grooves in a serpentine routing manner. The flexible thin-film thermocouples include a nickel-chromium-nickel-silicon composite layer and a polyimide insulation layer. The node spacing between adjacent flexible thin-film thermocouples is consistent with the groove spacing, forming a distributed temperature measurement network covering the circumference of the stator.

3. The control method for a variable frequency compressor motor according to claim 1, characterized in that: The temperature data and the motor operating parameters are combined through a data fusion algorithm to generate a dynamic thermal field characteristic map, and the coordinates of the hot spot areas where the temperature rises abnormally are marked, specifically including: Extracting current harmonic features of the three-phase current spectrum by fast Fourier transform, wherein the current harmonic features include spectrum energy of fundamental wave, third harmonic and fifth harmonic components; Performing joint spatiotemporal preprocessing on the temperature data, using a wavelet threshold denoising algorithm, selecting the sym4 wavelet basis for 5-layer decomposition, and setting an adaptive threshold function to eliminate electromagnetic interference noise of the temperature data; A three-dimensional polar coordinate system is established based on the geometric parameters of the motor, and the discrete temperature points are mapped to the spatial topological structure of the stator winding to generate an initial temperature field distribution cloud map.

4. The control method for a variable frequency compressor motor according to claim 3, characterized in that: The discrete temperature points are mapped to the spatial topology of the stator winding to generate an initial temperature field distribution cloud map, and then the following steps are further included: The initial temperature field distribution cloud map and the three-phase current spectrum are input into the DS evidence theory fusion model, where the temperature data is used as the first evidence body, the third harmonic energy ratio is used as the second evidence body, and the fifth harmonic energy ratio is used as the third evidence body; The credibility weight of each evidence body is calculated through the correction algorithm, and the thermal field confidence distribution matrix containing the temperature-current coupling characteristics is output; The interpolation algorithm is used to reconstruct the global temperature gradient surface of the winding using the thermal field confidence distribution matrix. When the local temperature confidence is ≥0.92 and the gradient change rate is >3℃ / mm 2 When an abnormal hotspot is detected, its polar coordinate position (r, θ, z) and confidence level are recorded to calibrate the hotspot area.

5. The control method for a variable frequency compressor motor according to claim 4, characterized in that: The polar coordinate position (r, θ, z) and confidence level are recorded to mark the hotspot area, and then the following is included: The coordinates of the calibrated hotspot area are aligned with the current harmonic characteristics in time and space to establish a three-dimensional feature vector set including temperature amplitude, gradient vector, and harmonic energy weight; Through the PCA dimensionality reduction processing of the three-dimensional feature vector set, the first three principal components are extracted to form the encoding matrix of the thermal field feature map.

6. The control method for a motor for a variable frequency compressor according to claim 5, characterized in that: The method of establishing a dynamic correlation model between winding temperature and motor operating parameters based on the dynamic thermal field characteristic map and updating the heat conduction characteristic parameters in the dynamic correlation model through a sliding window algorithm specifically includes: Analyze the encoding matrix of the dynamic thermal field characteristic map and extract the temperature amplitude principal component vector T_pca, harmonic energy weight vector H_wei and gradient vector field G_xyz as the model input data set; A winding temperature rise rate prediction model is constructed as a dynamic correlation model. The model input variables are defined as: the principal component vector of the temperature amplitude at the current moment T_pca(k), the harmonic energy weight vector of the three historical cycles [H_wei(k-1), H_wei(k-2), H_wei(k-3)], and the modulus value ||G|| of the gradient vector field G_xyz(k). The model output is the temperature rise rate ΔT_pred of the next cycle. Initial heat conduction characteristic parameters, including heat conduction coefficient α_t, harmonic influence factor β_t and gradient weight δ_t; Set the sliding window duration and the number of sampling groups within the window duration. Based on the residual between the measured temperature rise rate ΔT_meas within the window and the temperature rise rate ΔT_pred of the next cycle, use the recursive least squares method to update the thermal conductivity characteristic parameters of the winding temperature rise rate prediction model. The residual root mean square error (RMSE) within the window is calculated. When RMSE is less than 0.12°C / s, the updated thermal conductivity characteristic parameters are accepted. If RMSE is greater than or equal to 0.2°C / s after two consecutive updates, the previous valid parameter set is returned to.

7. The control method for a variable frequency compressor motor according to claim 1, characterized in that: The adjusted PWM control strategy is input into the motor drive unit to perform preliminary regulation, and the regulated winding temperature change rate and torque fluctuation data are collected to generate a multi-dimensional feedback feature vector, specifically including: The adjusted PWM control parameter set is loaded into the motor driver's FPGA control unit for preliminary control. The actual dead time T_phase of each phase is written to the timer register in 16 bits via the SPI interface, and the composite voltage vector waveform data is stored in the dual-port RAM buffer. A distributed thermocouple array is used to capture the temperature sequence of each winding section after regulation, and the temperature change rate ΔT_rate = ΔT / Δt is calculated with a time window Δt = 100ms. The motor rotor angular displacement signal is collected by a magnetic encoder, and the second-order derivative is obtained after Savitzky-Golay filtering to obtain the torque fluctuation component ΔTe.

8. The control method for a motor for a variable frequency compressor according to claim 7, characterized in that: The second-order derivative is obtained after Savitzky-Golay filtering to obtain the torque fluctuation component ΔTe, which is then further included: Extract temperature feedback indicators, which include: the maximum temperature rise rate ΔT_max in the hotspot area, the average temperature rise gradient T_avg, and the temperature distribution uniformity index U_T = σ(T) / μ(T), where σ(T) represents the temperature standard deviation and μ(T) represents the temperature mean; Extracting electrical parameter indicators, including: total harmonic distortion (THD) of current and third harmonic distortion (H3%); The temperature feedback index and the electrical parameter index are normalized and encoded into a 5-dimensional feature vector in the order of [ΔT_max, T_avg, U_T, THD, H3%] to construct a multi-dimensional feedback feature vector.

9. The control method for a motor for a variable frequency compressor according to claim 1, characterized in that: Based on the multi-dimensional feedback feature vector, the combined weights of the switching interval compensation and the harmonic injection strategy are iteratively optimized through a gradient descent algorithm. The motor control strategy matrix adapted to the current working conditions is output and loaded into the control system in real time. Specifically, the following steps are performed: Analyze the multidimensional feedback feature vector, extract the maximum temperature rise rate ΔT_max, the current total harmonic distortion rate THD and the torque ripple peak ΔTe_pp as optimization targets, and construct a normalized cost function J; Define the optimization variables and constraints: the switching interval compensation weight k_comp ranges from [0.1, 0.4], the harmonic injection ratio weight k_h3 ranges from [0.05, 0.25], set the learning rate α to 0.02, the maximum number of iterations to 50, and the convergence threshold ΔJ to < 0.5%.

10. The control method for a motor for a variable frequency compressor according to claim 9, characterized in that: The definition of optimization variables and constraints also includes: Calculate the cost function value J_current based on the current weight combination (k_comp, k_h3); Calculate partial derivatives by perturbation analysis: apply ±0.02 perturbation to k_comp to obtain the J change ΔJ_comp; apply ±0.005 perturbation to k_h3 to obtain the J change ΔJ_h3; The current weight combination (k_comp, k_h3) is updated by the calculated partial derivatives. If k_comp is out of range, the boundary value is taken. If k_h3 is out of range, the step size is 0.005 to perform gradient descent iterative optimization. When the decrease of J is less than 0.5% for three consecutive iterations, the current weight combination is extracted; Verify the constraints of ΔT_max≤5℃ / s, THD≤5%, and ΔTe_pp≤2N·m. If any indicator exceeds the limit, reiterate according to the weight priority of the cost function component; Map the optimized weights to the actual control parameters to generate 12 sets of discretized parameter combinations, which are sorted in ascending order of J values and stored in the strategy matrix; The strategy matrix is transferred to the DSP coprocessor of the motor controller via the PCIe bus. At the beginning of the next control cycle, the strategy group with the smallest J value is selected for activation and execution.

Images