Motor thermal state recognition method, device and controller

Abstract

This application discloses a method, apparatus, and controller for identifying the thermal state of a motor, relating to the field of vehicle technology. The method includes: acquiring multi-dimensional electrical data of the motor in a vehicle at the current moment; determining a feature vector reflecting the thermal state of the motor based on the multi-dimensional electrical data; inputting the feature vector into a motor thermal state classification model to obtain the target thermal state category corresponding to the motor at the current moment; the target thermal state category is one of a set of preset thermal state categories, each representing a different temperature state of the motor's stator and / or rotor; wherein the motor thermal state classification model is obtained by training an initial multi-classification model based on multiple sample data pairs; each sample data pair consists of a sample feature vector corresponding to the sample multi-dimensional electrical data and a thermal state category label; the initial multi-classification model consists of multiple binary sub-models containing kernel functions. This improves the accuracy of motor thermal state identification.

Description

Technical Field

[0001] This application relates to the field of vehicle technology, and in particular to a method, device and controller for identifying the thermal state of an electric motor. Background Technology

[0002] In related technologies, methods for identifying the temperature rise of motor stators and rotors mainly include direct measurement methods based on physical sensors and soft measurement methods based on mathematical models. However, direct measurement methods based on physical sensors are difficult to install sensors on high-speed rotating rotors and achieve reliable signal transmission, resulting in an inability to fully reflect the overall thermal state of the motor, thus leading to low accuracy in identifying the motor's thermal state. Soft measurement methods based on mathematical models heavily rely on the precise mathematical model of the motor, while motor parameters (such as flux linkage and resistance) themselves change with temperature and operating point, thus reducing the accuracy of identifying the motor's thermal state.

[0003] Therefore, improving the accuracy of motor thermal condition identification has become an urgent problem to be solved. Summary of the Invention

[0004] This application provides a method, apparatus, and controller for identifying the thermal condition of a motor, which can improve the accuracy of motor thermal condition identification.

[0005] In a first aspect, embodiments of this application provide a method for identifying the thermal state of a motor, the method comprising:

[0006] Acquire multi-dimensional electrical data of the motors in the vehicle at the current moment;

[0007] Based on multidimensional electrical data, a feature vector is determined to reflect the thermal state of the motor.

[0008] The feature vector is input into the motor thermal state classification model to obtain the target thermal state category of the motor at the current moment. The target thermal state category is one of a number of preset thermal state categories, which respectively represent different temperature states of the stator and / or rotor of the motor.

[0009] The motor thermal state classification model is obtained by training an initial multi-classification model based on multiple sample data pairs. Each sample data pair consists of a sample feature vector corresponding to the multidimensional electrical data of the sample and a thermal state category label. The initial multi-classification model consists of multiple binary sub-models containing kernel functions.

[0010] In this embodiment, the motor temperature monitoring problem, which relies on physical sensors or mathematical models in related technologies, is transformed into a machine learning pattern recognition problem that classifies the motor thermal state category based on multi-dimensional electrical data of the motor through a multi-classification model. This model (motor thermal state classification model) is trained on an initial multi-classification model composed of multiple binary sub-models containing kernel functions, based on sample data pairs consisting of sample feature vectors corresponding to sample multi-dimensional electrical data and thermal state category labels. Therefore, on the one hand, it is not necessary to accurately measure the rotor temperature, nor is it necessary to pay attention to the impact of changes in motor parameters with temperature and operating point on motor thermal state identification. On the other hand, by introducing kernel functions into the model, the similarity of different groups of multi-dimensional electrical data in causing motor heating can be measured, thereby improving the accuracy of motor thermal state identification.

[0011] In one optional implementation of the first aspect, the multidimensional electrical data includes the three-phase current of the motor, the DC bus voltage, and the motor speed; based on the multidimensional electrical data, a feature vector reflecting the thermal state of the motor is determined, including: based on the three-phase current, determining the direct-axis current for controlling the motor's magnetic field and the quadrature-axis current for generating electromagnetic torque; based on the DC bus voltage, determining the direct-axis voltage for generating the direct-axis current and the quadrature-axis voltage for generating the quadrature-axis current; based on the instantaneous input power and instantaneous output power of the motor, determining the instantaneous total power loss of the motor, and based on multiple instantaneous total power losses within a preset time window, determining the average power loss; wherein, the instantaneous input power is determined based on the DC bus voltage and the DC bus current, and the instantaneous output power is determined based on the quadrature-axis current and the motor speed; the feature vector reflecting the thermal state of the motor is determined based on the direct-axis current, quadrature-axis current, direct-axis voltage, quadrature-axis voltage, motor speed, the motor's mechanical angular velocity, and the average power loss.

[0012] In this implementation method, since power loss is a direct measure of the total heat generation rate inside the motor, and instantaneous power loss fluctuates greatly, determining the average power loss can better reflect the heat accumulation trend of the motor within a preset time window. Therefore, based on the direct-axis current, quadrature-axis current, direct-axis voltage, quadrature-axis voltage, motor speed, motor mechanical angular velocity, and average power loss, the feature vector reflecting the thermal state of the motor can be determined more accurately.

[0013] In one optional implementation of the first aspect, the motor thermal state classification model is trained as follows: multiple sets of operating data of the motor under different operating conditions are acquired, each set of operating data including sample multidimensional electrical data and corresponding temperature data; the sample feature vector corresponding to each set of sample multidimensional electrical data is determined; based on the temperature data corresponding to each set of sample multidimensional electrical data, the thermal state category label corresponding to each set of sample multidimensional electrical data is determined according to a preset thermal state category determination strategy; multiple sample data pairs are constructed based on the sample feature vector and thermal state category label corresponding to each set of sample multidimensional electrical data, and the initial multi-classification model is trained based on the multiple sample data pairs to obtain the motor thermal state classification model.

[0014] This implementation method can improve the classification accuracy of the motor thermal condition classification model used to determine the motor thermal condition category.

[0015] In one optional implementation of the first aspect, the temperature data corresponding to the multidimensional electrical data of the samples includes the stator temperature and rotor temperature of the motor; based on the temperature data corresponding to each group of multidimensional electrical data of the samples, a thermal state category label corresponding to each group of multidimensional electrical data of the samples is determined according to a preset thermal state category determination strategy, including: for each group of multidimensional electrical data of the samples, if the stator temperature included in the temperature data corresponding to the multidimensional electrical data of the samples is lower than a preset first warning temperature and the rotor temperature is lower than a preset second warning temperature, the thermal state category label corresponding to the multidimensional electrical data of the samples is determined to be a first thermal state category; if the stator temperature is within a preset first warning temperature range and the rotor temperature is not higher than the second warning temperature, the thermal state category label corresponding to the multidimensional electrical data of the samples is determined to be a second thermal state category; the temperature in the first warning temperature range is higher than the first warning temperature; if the stator temperature is not higher than the first warning temperature and the rotor temperature is within a preset second warning temperature range, the thermal state category label corresponding to the samples is determined to be a second thermal state category. The thermal state category label corresponding to the multidimensional electrical data of the sample is determined as the third thermal state category; the temperature in the second warning temperature range is higher than the second warning temperature; when the stator temperature is in the first warning temperature range and the rotor temperature is in the second warning temperature range, the thermal state category label corresponding to the multidimensional electrical data of the sample is determined as the fourth thermal state category; when the stator temperature is not lower than the preset first danger temperature, the thermal state category label corresponding to the multidimensional electrical data of the sample is determined as the fifth thermal state category; the first danger temperature is higher than the temperature in the first warning temperature range; when the rotor temperature is not lower than the preset second danger temperature, the thermal state category label corresponding to the multidimensional electrical data of the sample is determined as the sixth thermal state category; the second danger temperature is higher than the temperature in the second warning temperature range; wherein, the first warning temperature, the first warning temperature range, and the first danger temperature are determined based on the insulation level of the stator winding of the motor; the second warning temperature, the second warning temperature range, and the second danger temperature are determined based on the demagnetization characteristics of the motor.

[0016] By adopting this implementation method, the continuous stator and rotor temperatures are mapped into six heat dissipation states (from the first thermal state category to the sixth thermal state category) based on the insulation level of the motor and the demagnetization characteristics of the permanent magnet. This facilitates the subsequent construction of multiple sample data pairs based on the sample feature vector and thermal state category label corresponding to the multidimensional electrical data of each sample. The trained motor thermal state classification model can simultaneously reflect the motor heat generation trend and the absolute thermal safety threshold.

[0017] In one optional implementation of the first aspect, an initial multi-classification model is trained based on multiple sample data pairs to obtain a motor thermal state classification model, including: determining training sample data pairs from multiple sample data pairs; determining a target hyperparameter group corresponding to the initial multi-classification model from multiple preset candidate hyperparameter groups based on the training sample data pairs; each candidate hyperparameter group includes a kernel parameter and a penalty parameter in a kernel function; the penalty parameter is used to represent the tolerance of the motor thermal state classification model to classification errors; the kernel function is used to measure the similarity of multidimensional electrical data from different groups in causing motor heating; the kernel parameter is used to determine the width of the curve corresponding to the kernel function; and the initial multi-classification model is trained based on the target hyperparameter group and multiple sample data pairs to obtain the motor thermal state classification model.

[0018] In this implementation method, since the hyperparameter set includes a kernel function and a penalty parameter, the kernel function can be used to measure the similarity of different sets of multidimensional electrical data in terms of causing motor heating, and the penalty parameter can be used to represent the tolerance of the motor thermal state classification model to classification error. Therefore, by training the initial multi-classification model based on the target hyperparameter set, a motor thermal state classification model with clear classification boundaries and strong generalization ability can be obtained.

[0019] In one optional implementation of the first aspect, based on training sample data pairs, determining the target hyperparameter group corresponding to the initial multi-classification model from a set of preset candidate hyperparameter groups includes: traversing each candidate hyperparameter group in the set of preset candidate hyperparameter groups, and for each traversed candidate hyperparameter group, performing multiple rounds of training on the initial multi-classification model based on the candidate hyperparameter group and the training sample data pairs to obtain multiple intermediate multi-classification models, and determining the average classification accuracy of the multiple intermediate multi-classification models; selecting the candidate hyperparameter group that results in the highest average classification accuracy from the multiple candidate hyperparameter groups as the target hyperparameter group corresponding to the initial multi-classification model.

[0020] This implementation method can help to determine a motor thermal state classification model with clear classification boundaries and strong generalization ability.

[0021] In an optional implementation of the first aspect, the method further includes: determining a target thermal management control strategy that matches the target thermal state category based on the correspondence between multiple thermal state categories and multiple thermal management control strategies; the target thermal management control strategy is one of multiple thermal management control strategies; wherein, different thermal management control strategies among the multiple thermal management control strategies include different torque derating control strategies and / or different cooling system power control strategies; and performing thermal management on the motor based on the target thermal management control strategy.

[0022] This implementation method uses a target thermal management control strategy that matches the target thermal state category to perform thermal management on the motor. In this way, compared with the related technology that determines an accurate temperature value and then protects the motor after the temperature value reaches a certain condition, the delayed protection of the motor can be transformed into advance thermal management, that is, thermal management of the motor can be performed in advance.

[0023] Secondly, embodiments of this application provide a motor thermal state identification device, the device comprising:

[0024] The acquisition module is used to acquire the multi-dimensional electrical data of the motor in the vehicle at the current moment;

[0025] The determination module is used to determine the feature vector that reflects the thermal state of the motor based on multidimensional electrical data;

[0026] The thermal state identification module is used to input the feature vector into the motor thermal state classification model to obtain the target thermal state category of the motor at the current moment. The target thermal state category is one of a number of preset thermal state categories, which respectively represent different temperature states of the stator and / or rotor of the motor.

[0027] The motor thermal state classification model is obtained by training an initial multi-classification model based on multiple sample data pairs. Each sample data pair consists of a sample feature vector corresponding to the multidimensional electrical data of the sample and a thermal state category label. The initial multi-classification model consists of multiple binary sub-models containing kernel functions.

[0028] Thirdly, embodiments of this application provide a controller, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method provided in the first aspect above.

[0029] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method provided in the first aspect above.

[0030] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method provided in the first aspect above.

[0031] Regarding the beneficial effects of any of the technical solutions in the second to fifth aspects mentioned above, refer to the beneficial effects of the corresponding technical solutions in the first aspect; repeated examples will not be listed here. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 This is an optional flowchart illustrating a method for identifying the thermal state of a motor provided in an embodiment of this application;

[0034] Figure 2 This is another optional flowchart illustrating a method for identifying the thermal state of a motor provided in an embodiment of this application;

[0035] Figure 3 This is an optional schematic diagram of the training process of a motor thermal state classification model provided in an embodiment of this application;

[0036] Figure 4 This is an optional structural schematic diagram of a motor thermal condition identification device provided in an embodiment of this application;

[0037] Figure 5 This is a schematic diagram of an optional structure of a controller provided in an embodiment of this application. Detailed Implementation

[0038] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0039] The method for identifying the thermal state of a motor provided in the embodiments of this application is described below.

[0040] Please see Figure 1 , Figure 1This is a schematic flowchart of an optional method for identifying the thermal state of a motor, provided in an embodiment of this application. The method can be executed by a controller (e.g., a motor controller in a vehicle, or a vehicle controller with integrated motor control functions). The following explanation uses the execution of this motor thermal state identification method by a motor controller as an example. Figure 1 As shown, the method for identifying the thermal state of a motor may include, but is not limited to, the following steps:

[0041] S101. Obtain the multi-dimensional electrical data of the motor in the vehicle at the current moment.

[0042] In this application, the motor may include a permanent magnet motor. A permanent magnet motor is a high-efficiency motor with its own magnetic field.

[0043] In some embodiments, multidimensional electrical data may include, but is not limited to, the three-phase current of the motor, the DC bus voltage, and the motor speed.

[0044] Three-phase current refers to the three-phase alternating current flowing into / out of the motor stator windings, including phase A, phase B, and phase C currents. The effective value, frequency, phase, and balance of the three-phase currents directly determine the electromagnetic torque and losses. Specifically, an increase in the effective value of any of the three-phase currents will directly lead to an increase in winding temperature, thus affecting the motor's thermal condition. An imbalance in the three-phase current will cause an excessively large current in one phase or a localized winding, creating localized hot spots, accelerating insulation aging, and thus affecting the motor's thermal condition.

[0045] The DC bus voltage refers to the DC link voltage after rectification and filtering within the frequency converter. When the DC bus voltage drops, the frequency converter will increase the current in order to maintain the same torque as the motor at the current moment. This will cause the winding temperature to rise, thus affecting the thermal state of the motor.

[0046] Motor speed refers to the actual mechanical speed of the rotor in the motor. When the motor speed is 0 or extremely low, if the current is very large, the motor's heat dissipation is almost zero, and the winding temperature rise will increase, which will affect the motor's thermal state.

[0047] S102. Based on multidimensional electrical data, determine the feature vector that reflects the thermal state of the motor.

[0048] Among them, the thermal state of the motor can be used to describe the dynamic situation of internal temperature distribution, heat generation and transfer, which directly affects the performance, life and reliability of the motor.

[0049] When a motor is running, losses are the only source of heat, and different types of losses will ultimately be converted into heat. These different types of losses include, but are not limited to, copper losses (or winding losses) and iron losses (or core losses). Copper losses refer to the heat generated by the resistance of the stator and rotor windings when current flows through them. Copper losses are proportional to the square of the current and are the primary heat source. Iron losses are caused by the alternating magnetic field in the stator and rotor cores, and include hysteresis losses and eddy current losses. Iron losses are related to the magnetic field frequency and magnetic flux density.

[0050] S103. Input the feature vector into the motor thermal state classification model to obtain the target thermal state category of the motor at the current moment; wherein, the target thermal state category is one of a number of preset thermal state categories, and the multiple thermal state categories respectively represent different temperature states of the stator and / or rotor of the motor.

[0051] The motor thermal state classification model is obtained by training an initial multi-classification model based on multiple sample data pairs. Each sample data pair consists of a sample feature vector corresponding to the multidimensional electrical data of the sample and a thermal state category label. The initial multi-classification model consists of multiple binary sub-models containing kernel functions.

[0052] Optionally, the kernel function may include, but is not limited to, conventional kernel functions such as radial basis function (RBF), linear kernel, polynomial kernel, and sigmoid kernel; no limitation is made here.

[0053] In this embodiment, the motor temperature monitoring problem, which relies on physical sensors or mathematical models in related technologies, is transformed into a machine learning pattern recognition problem that classifies the motor thermal state category based on multi-dimensional electrical data of the motor through a multi-classification model. This model (motor thermal state classification model) is trained on an initial multi-classification model composed of multiple binary sub-models containing kernel functions, based on sample data pairs consisting of sample feature vectors corresponding to sample multi-dimensional electrical data and thermal state category labels. Therefore, on the one hand, it is not necessary to accurately measure the rotor temperature, nor is it necessary to pay attention to the impact of changes in motor parameters with temperature and operating point on motor thermal state identification. On the other hand, by introducing kernel functions into the model, the similarity of different groups of multi-dimensional electrical data in causing motor heating can be measured, thereby improving the accuracy of motor thermal state identification.

[0054] In one alternative implementation, Figure 1The multi-dimensional electrical data in the motor thermal state identification method shown may include the motor's three-phase current, DC bus voltage, and motor speed. Step S102, where the motor controller determines the feature vector reflecting the motor's thermal state based on the multi-dimensional electrical data, may involve: determining the direct-axis current for controlling the motor's magnetic field and the quadrature-axis current for generating electromagnetic torque based on the three-phase current; determining the direct-axis voltage for generating the direct-axis current and the quadrature-axis voltage for generating the quadrature-axis current based on the DC bus voltage; determining the motor's instantaneous total loss based on the motor's instantaneous input power and instantaneous output power, and determining the average power loss based on multiple instantaneous total losses within a preset time window; wherein, the instantaneous input power is determined based on the DC bus voltage and DC bus current, and the instantaneous output power is determined based on the quadrature-axis current and motor speed; and determining the feature vector reflecting the motor's thermal state based on the direct-axis current, quadrature-axis current, direct-axis voltage, quadrature-axis voltage, motor speed, motor mechanical angular velocity, and average power loss.

[0055] In this context, the direct-axis current is the current component along the rotor's magnetic pole axis (direct axis), while the quadrature-axis current is the current component along the direction perpendicular to the magnetic pole axis (quadrature axis). In the case of a permanent magnet motor, the quadrature-axis current can also be called the torque current. In other words, the direct axis is the direction of the rotor's magnetic field, and the quadrature axis is the direction of the rotor's rotation.

[0056] In some embodiments, the motor controller determines the direct-axis current for controlling the motor magnetic field and the quadrature-axis current for generating electromagnetic torque based on the three-phase current. This can be achieved by converting the three-phase current in the three-phase coordinate system to a rotating coordinate system that rotates with the rotor, thereby obtaining the direct-axis current for controlling the motor magnetic field and the quadrature-axis current for generating electromagnetic torque.

[0057] In some embodiments, the motor controller determines the direct-axis voltage for generating the direct-axis current and the quadrature-axis voltage for generating the quadrature-axis current based on the DC bus voltage. This can be achieved by: determining the maximum allowable voltage amplitude based on the DC bus voltage; determining the original direct-axis voltage and the original quadrature-axis voltage based on the current loop proportional-integral model; determining the combined voltage amplitude of the original direct-axis voltage and the original quadrature-axis voltage; and constraining the combined voltage amplitude based on the maximum allowable voltage amplitude to obtain the direct-axis voltage for generating the direct-axis current and the quadrature-axis voltage for generating the quadrature-axis current.

[0058] In some embodiments, the instantaneous input power of the motor can be determined by the motor controller by taking the product of the DC bus voltage and the DC bus current as the instantaneous input power of the motor.

[0059] In some embodiments, the instantaneous output power of the motor can be determined by the motor controller in the following manner: determining the electromagnetic torque based on the quadrature axis current; and taking the product of the electromagnetic torque and the motor speed as the instantaneous output power.

[0060] In this implementation method, since power loss is a direct measure of the total heat generation rate inside the motor, and instantaneous power loss fluctuates greatly, determining the average power loss can better reflect the heat accumulation trend of the motor within a preset time window. Therefore, based on the direct-axis current, quadrature-axis current, direct-axis voltage, quadrature-axis voltage, motor speed, motor mechanical angular velocity, and average power loss, the feature vector reflecting the thermal state of the motor can be determined more accurately.

[0061] In one alternative implementation, Figure 1 The motor thermal state identification method shown can be trained by the motor controller in the following way: acquiring multiple sets of operating data of the motor under different operating conditions, each set of operating data including sample multidimensional electrical data and corresponding temperature data; determining the sample feature vector corresponding to each set of sample multidimensional electrical data; determining the thermal state category label corresponding to each set of sample multidimensional electrical data according to a preset thermal state category determination strategy based on the temperature data corresponding to each set of sample multidimensional electrical data; constructing multiple sample data pairs based on the sample feature vector and thermal state category label corresponding to each set of sample multidimensional electrical data, and training the initial multi-classification model based on the multiple sample data pairs to obtain the motor thermal state classification model.

[0062] Different operating conditions can include steady-state operating conditions, transient operating conditions, and extreme operating conditions. These different operating conditions correspond to various combinations of motor torque and speed. For example, steady-state operating conditions may include those corresponding to low speed and low torque, high speed and low torque, etc.

[0063] The multidimensional electrical data of the samples may include, but is not limited to, the sample three-phase current, sample DC bus voltage, and sample motor speed of the motor; the temperature data corresponding to each set of multidimensional electrical data may include the stator temperature and rotor temperature of the motor. The stator temperature refers to the temperature of the stator winding; the rotor temperature refers to the temperature of the conductive parts of the rotor (such as the excitation winding or permanent magnet of a synchronous motor).

[0064] In some embodiments, each set of sample multidimensional electrical data includes sample three-phase current, sample DC bus voltage, and sample motor speed. The motor controller determines the sample feature vector corresponding to each set of sample multidimensional electrical data by: determining the sample direct-axis current and sample quadrature-axis current corresponding to each set of sample multidimensional electrical data based on the sample three-phase current in each set of sample multidimensional electrical data; determining the sample direct-axis voltage and sample quadrature-axis voltage corresponding to each set of sample multidimensional electrical data based on the sample DC bus voltage in each set of sample multidimensional electrical data; and determining the sample instantaneous input power and sample instantaneous output power corresponding to each set of sample multidimensional electrical data. The sample instantaneous total loss corresponding to the multidimensional electrical data of the samples is determined, and the average power loss corresponding to each group of multidimensional electrical data of the samples is determined based on the sample DC bus voltage and sample DC bus current within a preset time window; wherein, the sample instantaneous input power is determined based on the sample DC bus voltage and sample DC bus current, and the sample instantaneous output power is determined based on the sample quadrature-axis current and sample motor speed; based on the sample direct-axis current, sample quadrature-axis current, sample direct-axis voltage, sample quadrature-axis voltage, sample motor speed, sample motor mechanical angular velocity and sample power loss average corresponding to each group of multidimensional electrical data of the samples, the sample feature vector corresponding to each group of multidimensional electrical data of the samples is determined.

[0065] In some embodiments, the temperature data corresponding to the multidimensional electrical data of the samples includes the stator temperature and rotor temperature of the motor. Based on the temperature data corresponding to each set of multidimensional electrical data of the samples, the motor controller determines the thermal state category label corresponding to each set of multidimensional electrical data of the samples according to a preset thermal state category determination strategy. This can be as follows: For each set of multidimensional electrical data of the samples, if the stator temperature included in the temperature data corresponding to the multidimensional electrical data of the samples is lower than a preset first warning temperature and the rotor temperature is lower than a preset second warning temperature, the thermal state category label corresponding to the multidimensional electrical data of the samples is determined to be the first thermal state category; if the stator temperature is within the preset first warning temperature range and the rotor temperature is not higher than the second warning temperature, the thermal state category label corresponding to the multidimensional electrical data of the samples is determined to be the second thermal state category; if the temperature in the first warning temperature range is higher than the first warning temperature; if the stator temperature is not higher than the first warning temperature and the rotor temperature is within the preset second warning temperature range, the motor controller determines... The thermal state category label corresponding to the sample multidimensional electrical data is the third thermal state category; the temperature in the second warning temperature range is higher than the second warning temperature; when the stator temperature is in the first warning temperature range and the rotor temperature is in the second warning temperature range, the thermal state category label corresponding to the sample multidimensional electrical data is determined to be the fourth thermal state category; when the stator temperature is not lower than the preset first danger temperature, the thermal state category label corresponding to the sample multidimensional electrical data is determined to be the fifth thermal state category; the first danger temperature is higher than the temperature in the first warning temperature range; when the rotor temperature is not lower than the preset second danger temperature, the thermal state category label corresponding to the sample multidimensional electrical data is determined to be the sixth thermal state category; the second danger temperature is higher than the temperature in the second warning temperature range; wherein, the first warning temperature, the first warning temperature range, and the first danger temperature are determined based on the insulation level of the motor stator winding; the second warning temperature, the second warning temperature range, and the second danger temperature are determined based on the demagnetization characteristics of the motor. In this way, by using the insulation class of the motor and the demagnetization characteristics of the permanent magnet, the continuous stator and rotor temperatures are mapped into six heat dissipation states (from the first thermal state category to the sixth thermal state category). This facilitates the subsequent construction of multiple sample data pairs based on the sample feature vectors and thermal state category labels corresponding to the multidimensional electrical data of each sample. The trained motor thermal state classification model can simultaneously reflect the motor's heat generation trend and absolute thermal safety threshold.

[0066] The first thermal state category can also be called the safe state category. In other words, in the first thermal state category, both the stator and rotor temperatures of the motor are lower than the corresponding warning temperatures.

[0067] The second thermal state category can also be called the stator warning category. In other words, in the second thermal state category, the stator temperature of the motor is within the stator warning temperature range, and the rotor temperature is within the rotor safe temperature range. The stator warning category usually corresponds to motor operating conditions of continuous high torque (i.e., high quadrature-axis current of the motor) and medium to low speed (poor heat dissipation).

[0068] The third thermal state category can also be called the rotor warning category. In other words, in the third thermal state category, the stator temperature of the motor is within the stator safe temperature range, and the rotor temperature is within the rotor warning temperature range. Among them, the rotor warning category usually corresponds to motor operating conditions with high speed (possibly with large direct-axis current) or severe rotor eddy current losses.

[0069] The fourth thermal state category can also be called the comprehensive warning category. In other words, in the fourth thermal state category, the stator temperature of the motor is within the stator warning temperature range, and the rotor temperature is within the rotor warning temperature range. If the motor's thermal state category is the fourth thermal state category, it indicates that both the stator and rotor of the motor are overheating simultaneously, and the motor is facing comprehensive thermal stress.

[0070] The fifth thermal condition category can also be called the stator danger category. In other words, in the fifth thermal condition category, the stator temperature of the motor is not lower than the stator danger temperature. If the motor's thermal condition category is the fifth thermal condition category, it indicates that the stator temperature of the motor has reached its physical limit, and permanent damage to the motor is imminent or has already occurred.

[0071] The sixth thermal condition category can also be called the rotor danger category. In other words, in the sixth thermal condition category, the rotor temperature of the motor is not lower than the rotor danger temperature. If the motor's thermal condition category is the sixth thermal condition category, it indicates that the rotor temperature of the motor has reached its physical limit, and the permanent magnets are about to demagnetize or have already completely demagnetized.

[0072] In summary, the thermal state category labels mentioned in this application may include the six types listed in Table 1 below.

[0073] Table 1

[0074]

[0075] Among them, the first warning temperature can also be called the stator warning temperature, the first warning temperature range can also be called the stator warning temperature range, and the first danger temperature can also be called the stator danger temperature; the second warning temperature can also be called the rotor warning temperature, the second warning temperature range can also be called the rotor warning temperature range, and the second danger temperature can also be called the rotor danger temperature.

[0076] For example, assuming the insulation class of the motor stator winding is H, and the temperature corresponding to H is 180℃ (referring to the highest allowable temperature at which the insulation material can work stably for a long time without accelerating aging), in this case, for the H-class stator winding, the stator warning temperature can be 150℃ (meaning that once the stator temperature exceeds 150℃, although it is safe in the short term, attention should be paid and preventive measures should be taken to allow time and space for the control system to respond); the stator danger temperature can be 170℃ (meaning that once the stator temperature exceeds 170℃, it indicates that the stator temperature is very close to or has reached the highest allowable temperature corresponding to H); the stator warning temperature range can be [150℃, 170℃].

[0077] For example, assuming the motor is a permanent magnet, the permanent magnet will undergo irreversible demagnetization at high temperatures, which will lead to a decrease in the motor's performance. Therefore, the rotor warning temperature can be much lower than the Curie temperature. Thus, the rotor warning temperature can be the temperature below the highest temperature at which the permanent magnet can remain stable under the maximum demagnetizing magnetic field (meaning that at this temperature, the risk of demagnetization of the permanent magnet begins to increase), for example, 145°C; the rotor danger temperature can be the critical temperature at which the permanent magnet undergoes irreversible demagnetization (meaning that once the rotor temperature reaches or exceeds this critical temperature, the permanent magnet will be permanently damaged), for example, 165°C; the rotor warning temperature range can be [145°C, 165°C].

[0078] In some embodiments, the motor controller trains an initial multi-classification model based on multiple sample data pairs to obtain a motor thermal state classification model. This can be achieved by: determining training sample data pairs from multiple sample data pairs; determining a target hyperparameter group corresponding to the initial multi-classification model from multiple preset candidate hyperparameter groups based on the training sample data pairs; each candidate hyperparameter group includes a kernel parameter and a penalty parameter in the kernel function; the penalty parameter represents the tolerance of the motor thermal state classification model to classification errors; the kernel function measures the similarity of different groups of multidimensional electrical data in causing motor heating; the kernel parameter determines the width of the curve corresponding to the kernel function; and the initial multi-classification model is trained based on the target hyperparameter group and multiple sample data pairs to obtain the motor thermal state classification model. Thus, since the hyperparameter group includes a kernel function and a penalty parameter, and the kernel function can be used to measure the similarity of different groups of multidimensional electrical data in causing motor heating, and the penalty parameter can be used to represent the tolerance of the motor thermal state classification model to classification errors, training the initial multi-classification model based on the target hyperparameter group yields a motor thermal state classification model with clear classification boundaries and strong generalization ability.

[0079] Optionally, the initial multi-class classification model can be an initial multi-class support vector machine model.

[0080] Optionally, the motor controller can determine the number K of thermal state categories corresponding to the preset thermal state category determination strategy; for each of the K thermal state categories, a binary classification SVM model is trained for every two thermal state categories. For example, assuming K is 6, corresponding to the aforementioned first thermal state category (Class 0), second thermal state category (Class 1), third thermal state category (Class 2), fourth thermal state category (Class 3), fifth thermal state category (Class 4), and sixth thermal state category (Class 5), the motor controller can train... Examples of binary classification SVM models, such as Class0 vs Class1, Class0 vs Class2, Class0 vs Class3, ..., Class4 vs Class5, are provided.

[0081] Optionally, the kernel function can be a radial basis function (RBF). A radial basis function is a real-valued function whose value depends only on the radial distance (usually Euclidean distance) from the input point x to a center point c. This function is a key tool in SVM models for handling nonlinear problems. It measures the similarity between samples by the square of the distance between samples; the closer the samples are, the higher the similarity (close to 1); the farther the samples are, the lower the similarity (close to 0).

[0082] The RBF function can be expressed as the following formula (1).

[0083] (1)

[0084] In formula (1), x i This represents the feature vector of the i-th sample; x j This represents the feature vector of the j-th sample; It represents the square of the Euclidean norm (i.e., geometric distance) between the i-th sample feature vector and the j-th sample feature vector, or the square of the straight-line distance between the i-th sample feature vector and the j-th sample feature vector in the original feature space. This represents the kernel parameter of RBF, which acts as a scaling factor to control the impact of the geometric distance between two sample feature vectors on the classification performance of the multi-class model; exp() represents the exponential function, which maps inputs from negative infinity to 0 to outputs from 0 to 1. It represents x i and x j The similarity between them.

[0085] By squaring the geometric distance between two sample feature vectors, the difference between the two sample feature vectors can be amplified, making the classification model pay more attention to the distance; on the other hand, open computation can be avoided, thus facilitating calculation.

[0086] Among them, two sample feature vectors x i and x j The similarity between them decreases exponentially with the square of their Euclidean distance. In x i and x j Under completely identical conditions, x i and x j The square of the distance (i.e., ||x) i -x j If ||²) is 0, then exp(-γ*0)=exp(0)=1; in x i and x j The more different the cases, the better. i and x j The larger the square of the distance, the more -γ*||x i -x j ||² will approach negative infinity, at which point exp() will approach 0.

[0087] For example, suppose the sample feature vector x i Characterized by high torque and low speed, its corresponding thermal state category label is Class1, and the sample feature vector x j Characterizing samples with the same torque and high speed, their corresponding thermal state category is Class 0. In this case, using a linear model might lead to misclassification due to similar currents. However, after calculation using the RBF function, it can be found that the speed / loss difference between the two is significant, resulting in low similarity. This is beneficial for classifying the sample feature vector x. i and sample feature vector x j They are classified into two different thermal state categories.

[0088] Due to nuclear parameters If the kernel parameter is too large, the RBF function curve will be too narrow. This means a single sample can only influence the decision boundary within a very small neighborhood, leading to an extremely complex decision boundary that heavily relies on individual sample points, potentially causing overfitting. If the kernel parameter is too small, the RBF function curve will be too wide. This means that the influence range of a single sample is very large, resulting in an overly smooth decision boundary, even approximating a linear model, which cannot effectively distinguish complex categories and leads to underfitting. Therefore, it is necessary to determine the optimal kernel parameter.

[0089] A larger penalty parameter (denoted as C) results in the multi-class classification model severely penalizing the feature vector of each misclassified sample in an effort to achieve 100% accuracy during training, potentially leading to overfitting. Conversely, a smaller penalty parameter allows the model to tolerate misclassification more readily, potentially resulting in underfitting. Therefore, it is necessary to determine the optimal penalty parameter to strike a balance between classifying the model as accurately as possible and maintaining a simple and smooth decision boundary, i.e., controlling the complexity of the multi-class classification model to improve its generalization ability.

[0090] Optionally, the preset multiple candidate hyperparameter sets can be each pair of candidate hyperparameter sets in a parameter grid formed based on preset candidate penalty parameters and preset candidate kernel parameters. The preset candidate penalty parameters may, for example, include 0.1, 1, 10, and 100; the preset kernel parameters may, for example, include 0.01, 0.1, 1, and 10.

[0091] In this embodiment, the motor controller determines the target hyperparameter group corresponding to the initial multi-classification model from a set of preset candidate hyperparameter groups based on training sample data pairs. This can be achieved by: traversing each of the preset candidate hyperparameter groups, and for each traversed candidate hyperparameter group, training the initial multi-classification model multiple times based on the candidate hyperparameter group and the training sample data pairs to obtain multiple intermediate multi-classification models, and determining the average classification accuracy of the multiple intermediate multi-classification models; selecting the candidate hyperparameter group that results in the highest average classification accuracy from the multiple candidate hyperparameter groups as the target hyperparameter group corresponding to the initial multi-classification model. The target hyperparameter group can be denoted as (C... best , best ).

[0092] Optionally, for each iteration of the candidate hyperparameter set, the motor controller performs multiple rounds of training on the initial multi-classification model based on the candidate hyperparameter set and the training sample data pair to obtain multiple intermediate multi-classification models, and determines the average classification accuracy of the multiple intermediate multi-classification models. This can be achieved by dividing the training sample data pair into K parts, using K-1 parts as the training subset, and using the remaining part as the validation subset; during each round of training of the initial multi-classification model, training the initial multi-classification model based on the K-1 training subset to obtain intermediate multi-classification models, and determining the classification accuracy of the intermediate multi-classification models based on the validation subset; and determining the average classification accuracy of the multiple intermediate multi-classification models based on the classification accuracy corresponding to the multiple intermediate multi-classification models obtained from multiple rounds of training.

[0093] For example, assuming K is 5, the motor controller can train the initial multi-class model for 5 rounds. In each round of training, one sample data pair is selected as the validation subset, and the remaining 4 sample data pairs are selected as the training subset. The initial multi-class model is trained based on the candidate hyperparameter group traversed and the training subset of the current training round to obtain intermediate multi-class models. The accuracy of the intermediate multi-class models obtained in the current training round is determined based on the validation subset of the current training round. Then, based on the accuracy of the 5 intermediate multi-class models obtained in 5 rounds of training, the average classification accuracy of the 5 intermediate multi-class models is determined.

[0094] Optionally, the motor controller determines the classification accuracy of the intermediate multi-classification model based on the validation subset. This can be achieved by: inputting the sample feature vectors of each sample data pair included in the validation subset into the intermediate multi-classification model to obtain the predicted thermal state category corresponding to each sample feature vector; and determining the classification accuracy of the intermediate multi-classification model based on the predicted thermal state category and thermal state category label corresponding to each sample feature vector.

[0095] In this embodiment, the motor controller trains an initial multi-classification model based on a target hyperparameter set and multiple sample data pairs to obtain a motor thermal state classification model. This can be achieved by: inputting the sample feature vector from each of the multiple sample data pairs into the initial multi-classification model based on the target hyperparameter set to obtain the predicted thermal state category corresponding to each sample feature vector; and training the initial multi-classification model based on the target hyperparameter set in a direction that reduces the difference between the predicted thermal state category corresponding to each sample feature vector and the thermal state category label to obtain the motor thermal state classification model.

[0096] Optionally, the initial multi-class classification model can be a multi-class SVM model, and the kernel function can be an RBF function. The motor controller trains the initial multi-class classification model based on the target hyperparameter set in the direction of reducing the difference between the predicted thermal state category and the thermal state category label corresponding to each sample feature vector to obtain a motor thermal state classification model. Alternatively, the initial multi-class classification model based on the target hyperparameter set can be trained in the direction of reducing the difference between the predicted thermal state category and the thermal state category label corresponding to each sample feature vector to obtain an intermediate motor thermal state classification model. The hyperparameter set corresponding to the intermediate motor thermal state classification model (including the optimal penalty parameter C and the optimal kernel parameter) is then used to obtain the intermediate motor thermal state classification model. The motor thermal state classification model is determined by using support vectors and kernel functions. In this way, the subsequent inference process mainly involves vector multiplication and exponential operations, with controllable computational load. This allows the motor controller to quickly identify the motor's thermal state in a short time, meeting the real-time requirements of the motor controller.

[0097] In some embodiments, the kernel function can also be a conventional kernel function such as a linear kernel, polynomial kernel, or sigmoid kernel; this is not limited here. It should be noted that the model training process remains unchanged after changing the kernel function; only the kernel function expression and grid search optimization parameters need to be adjusted.

[0098] This implementation method can improve the classification accuracy of the motor thermal condition classification model used to determine the motor thermal condition category.

[0099] In one alternative implementation, Figure 1 In the motor thermal state identification method shown, the motor controller can also determine a target thermal management control strategy that matches the target thermal state category based on the correspondence between multiple thermal state categories and multiple thermal management control strategies; the target thermal management control strategy is one of multiple thermal management control strategies; wherein, the different thermal management control strategies among the multiple thermal management control strategies include different torque derating control strategies and / or different cooling system power control strategies; based on the target thermal management control strategy, the motor is thermally managed.

[0100] In some embodiments, the correspondence between multiple thermal state categories and multiple thermal management control strategies can be a table stored locally on the motor controller (denoted as the thermal state category-control strategy correspondence table), or a table stored in a database and accessible to the motor controller (denoted as the thermal state category-control strategy correspondence table), etc., without limitation here. The thermal state category-control strategy correspondence table includes the correspondence between multiple thermal state categories and multiple thermal management control strategies, and the different thermal management control strategies among the multiple thermal management control strategies, including different torque derating control strategies and / or different cooling system power control strategies. In this way, the target thermal management control strategy matching the target thermal state category can be determined simply and quickly.

[0101] In some embodiments, the various thermal state categories can be as shown in Table 1 above. The control objective of the thermal management control strategy corresponding to the first thermal state category (safe state) can be to maximize power output, without thermal control intervention, and fully utilize motor performance. The control objective of the thermal management strategy corresponding to the second thermal state category (stator warning) can be to focus on stator protection, directionally suppress stator temperature rise, protect stator winding insulation, and avoid thermal state deterioration. The control objective of the thermal management control strategy corresponding to the third thermal state category (rotor warning) can be to specifically reduce rotor thermal load and avoid the risk of permanent magnet demagnetization. The control objective of the thermal management control strategy corresponding to the fourth thermal state category (comprehensive warning) can be to rapidly reduce the overall machine thermal load, actively suppress synchronous temperature rise of stator and rotor, and prevent entering a dangerous state. The control objective of the thermal management control strategy corresponding to the fifth thermal state category (stator danger) can be stator limit protection, minimize hardware damage, and forcibly and rapidly reduce stator thermal load. The control objective of the thermal management control strategy corresponding to the sixth thermal state category (rotor danger) can be rotor protection, prevent permanent demagnetization of permanent magnets, and avoid complete damage to motor hardware.

[0102] This implementation method uses a target thermal management control strategy that matches the target thermal state category to perform thermal management on the motor. In this way, compared with the related technology that determines an accurate temperature value and then protects the motor after the temperature value reaches a certain condition, the delayed protection of the motor can be transformed into advance thermal management, that is, thermal management of the motor can be performed in advance.

[0103] The following is combined Figure 2 This paper provides an overall description of the motor thermal state identification method provided in the embodiments of this application. Please refer to [link to relevant documentation]. Figure 2 , Figure 2 This is another optional flowchart illustrating a method for identifying the thermal state of a motor provided in an embodiment of this application. For example... Figure 2 As shown, the method for identifying the thermal state of a motor may include, but is not limited to, the following steps:

[0104] S201. Acquire multiple sets of operating data of the motor under different operating conditions. Each set of operating data includes sample multidimensional electrical data and corresponding temperature data.

[0105] S202. Determine the sample feature vector corresponding to the multidimensional electrical data of each group of samples.

[0106] In some embodiments, each set of sample multidimensional electrical data includes sample three-phase current, sample DC bus voltage, and sample motor speed. The motor controller determines the sample feature vector corresponding to each set of sample multidimensional electrical data by: determining the sample direct-axis current and sample quadrature-axis current corresponding to each set of sample multidimensional electrical data based on the sample three-phase current in each set of sample multidimensional electrical data; determining the sample direct-axis voltage and sample quadrature-axis voltage corresponding to each set of sample multidimensional electrical data based on the sample DC bus voltage in each set of sample multidimensional electrical data; and determining the sample instantaneous input power and sample instantaneous output power corresponding to each set of sample multidimensional electrical data. The sample instantaneous total loss corresponding to the multidimensional electrical data of the samples is determined, and the average power loss corresponding to each group of multidimensional electrical data of the samples is determined based on the sample DC bus voltage and sample DC bus current within a preset time window; wherein, the sample instantaneous input power is determined based on the sample DC bus voltage and sample DC bus current, and the sample instantaneous output power is determined based on the sample quadrature-axis current and sample motor speed; based on the sample direct-axis current, sample quadrature-axis current, sample direct-axis voltage, sample quadrature-axis voltage, sample motor speed, sample motor mechanical angular velocity and sample power loss average corresponding to each group of multidimensional electrical data of the samples, the sample feature vector corresponding to each group of multidimensional electrical data of the samples is determined.

[0107] S203. Based on the temperature data corresponding to the multidimensional electrical data of each group of samples, determine the thermal state category label corresponding to the multidimensional electrical data of each group of samples according to the preset thermal state category determination strategy.

[0108] In some embodiments, the motor controller determines the thermal state category label corresponding to each set of multidimensional electrical data based on the temperature data corresponding to each set of sample multidimensional electrical data according to a preset thermal state category determination strategy. The specific process can be found in the relevant parameters mentioned above, and will not be repeated here.

[0109] S204. Based on the sample feature vector and thermal state category label corresponding to the multidimensional electrical data of each group of samples, construct multiple sample data pairs.

[0110] S205. Determine the training sample data pair from multiple sample data pairs.

[0111] S206. Based on the training sample data pairs, determine the target hyperparameter set corresponding to the initial multi-classification model from multiple preset candidate hyperparameter sets.

[0112] Each candidate hyperparameter group includes a kernel parameter and a penalty parameter in the kernel function; the penalty parameter is used to represent the tolerance of the motor thermal state classification model to classification errors; the kernel function is used to measure the similarity of the multidimensional electrical data of different groups in terms of causing motor heating; and the kernel parameter is used to determine the width of the curve corresponding to the kernel function.

[0113] S207. Based on the target hyperparameter set and multiple sample data pairs, the initial multi-classification model is trained to obtain the motor thermal state classification model.

[0114] The following is combined Figure 3 The training process of the motor thermal state classification model provided in the embodiments of this application will be introduced. Figure 3 This is an optional schematic diagram illustrating the training process of a motor thermal state classification model provided in an embodiment of this application. For example... Figure 3 As shown, the motor controller can first determine the training sample data pairs from multiple sample data pairs, and define the hyperparameter grid (corresponding to multiple candidate hyperparameter groups, such as the penalty parameter C and the kernel function). The hyperparameter set is composed of K sets of training samples. The training sample data pairs are divided into K sets of samples, and the initial multi-class SVM model is trained for K rounds. In the i-th round of training, one set of samples is selected as the validation set, and the remaining K-1 sets of samples are used as the training set. i and K are positive integers, i is greater than or equal to 1 and less than or equal to K. Based on the candidate hyperparameter set and the training set, the initial SVM model is trained, and the classification accuracy of the model is evaluated based on the validation set. The classification accuracy is used as the score of the i-th round of training. The average classification accuracy of the model after K rounds of training is determined for the candidate hyperparameter set. Each candidate hyperparameter set in the hyperparameter grid is traversed, and the candidate hyperparameter set with the highest average classification accuracy is selected as the target hyperparameter set (C). best , best ); based on (C best , best The initial multi-class SVM model was retrained using multiple sample data pairs to obtain the final deployable motor thermal state classification model.

[0115] S208. Obtain the multi-dimensional electrical data of the motor in the vehicle at the current moment.

[0116] S209. Based on multidimensional electrical data, determine the feature vector that reflects the thermal state of the motor.

[0117] S210. Input the feature vector into the motor thermal state classification model to obtain the target thermal state category of the motor at the current moment.

[0118] In an optional implementation, the relevant descriptions of steps S208 to S210 can be found in the descriptions of steps S101 to S103 above, and will not be repeated here.

[0119] S211. Based on the correspondence between multiple thermal state categories and multiple thermal management control strategies, determine the target thermal management control strategy that matches the target thermal state category.

[0120] The target thermal management control strategy is one of a variety of thermal management control strategies; the different thermal management control strategies among the multiple thermal management control strategies include different torque derating control strategies and / or different cooling system power control strategies.

[0121] In some embodiments, the relevant description of step S211 can be found in the relevant description above, and will not be repeated here.

[0122] S212. Based on the target thermal management control strategy, perform thermal management on the motor.

[0123] In this embodiment, on the one hand, the motor temperature monitoring problem, which relies on physical sensors or mathematical models in related technologies, is transformed into a machine learning pattern recognition problem that classifies the motor's thermal state category based on multi-dimensional electrical data of the motor through a multi-classification model. This model (motor thermal state classification model) is trained on an initial multi-classification model composed of multiple binary sub-models containing kernel functions, based on sample data pairs consisting of sample feature vectors corresponding to sample multi-dimensional electrical data and thermal state category labels. Therefore, it eliminates the need for precise rotor temperature measurement and the impact of changes in motor parameters with temperature and operating point on motor thermal state identification. Furthermore, by introducing kernel functions into the model, the similarity of different groups of multi-dimensional electrical data in causing motor heating can be measured, thereby improving the accuracy of motor thermal state identification. On the other hand, based on a target thermal management control strategy matching the target thermal state category, thermal management is performed on the motor. Compared to the method in related technologies that determines an accurate temperature value and then protects the motor only after the temperature reaches a certain condition, this transforms delayed protection of the motor into proactive thermal management, allowing for early thermal management of the motor.

[0124] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0125] Based on the same inventive concept, this application also provides a motor thermal condition identification device for implementing the motor thermal condition identification method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more embodiments of the motor thermal condition identification device provided below can be found in the limitations of the motor thermal condition identification method described above, and will not be repeated here.

[0126] Please see Figure 4 , Figure 4 This is a schematic diagram of an optional structure of a motor thermal condition identification device provided in an embodiment of this application. For example... Figure 4 As shown, the motor thermal condition identification device may include, but is not limited to:

[0127] The acquisition module 401 is used to acquire the multi-dimensional electrical data of the motor in the vehicle at the current moment;

[0128] The determination module 402 is used to determine a feature vector reflecting the thermal state of the motor based on multidimensional electrical data;

[0129] The thermal state recognition module 403 is used to input the feature vector into the motor thermal state classification model to obtain the target thermal state category of the motor at the current moment; the target thermal state category is one of a number of preset thermal state categories, and the multiple thermal state categories respectively represent different temperature states of the stator and / or rotor of the motor.

[0130] The motor thermal state classification model is obtained by training an initial multi-classification model based on multiple sample data pairs. Each sample data pair consists of a sample feature vector corresponding to the multidimensional electrical data of the sample and a thermal state category label. The initial multi-classification model consists of multiple binary sub-models containing kernel functions.

[0131] In some embodiments, the multidimensional electrical data includes the three-phase current of the motor, the DC bus voltage, and the motor speed. When determining the feature vector reflecting the thermal state of the motor based on the multidimensional electrical data, the determining module 402 specifically performs the following: based on the three-phase current, determining the direct-axis current used to control the motor's magnetic field and the quadrature-axis current used to generate electromagnetic torque; based on the DC bus voltage, determining the direct-axis voltage used to generate the direct-axis current and the quadrature-axis voltage used to generate the quadrature-axis current; based on the instantaneous input power and instantaneous output power of the motor, determining the instantaneous total power loss of the motor, and determining the average power loss based on multiple instantaneous total power losses within a preset time window; wherein the instantaneous input power is determined based on the DC bus voltage and DC bus current, and the instantaneous output power is determined based on the quadrature-axis current and the motor speed; and based on the direct-axis current, quadrature-axis current, direct-axis voltage, quadrature-axis voltage, motor speed, motor mechanical angular velocity, and the average power loss, determining the feature vector reflecting the thermal state of the motor.

[0132] In some embodiments, the device may further include a training module. The training module is configured to acquire multiple sets of operating data of the motor under different operating conditions, each set of operating data including sample multidimensional electrical data and corresponding temperature data; determine the sample feature vector corresponding to each set of sample multidimensional electrical data; based on the temperature data corresponding to each set of sample multidimensional electrical data, determine the thermal state category label corresponding to each set of sample multidimensional electrical data according to a preset thermal state category determination strategy; construct multiple sample data pairs based on the sample feature vector and thermal state category label corresponding to each set of sample multidimensional electrical data, and train the initial multi-classification model based on the multiple sample data pairs to obtain a motor thermal state classification model.

[0133] In some embodiments, the temperature data corresponding to the multidimensional electrical data of the samples includes the stator temperature and rotor temperature of the motor. When the training module determines the thermal state category label corresponding to each group of multidimensional electrical data based on the temperature data corresponding to each group of samples according to a preset thermal state category determination strategy, it specifically performs the following: For each group of multidimensional electrical data, if the stator temperature included in the temperature data corresponding to the multidimensional electrical data of the samples is lower than a preset first warning temperature and the rotor temperature is lower than a preset second warning temperature, the thermal state category label corresponding to the multidimensional electrical data of the samples is determined to be a first thermal state category; if the stator temperature is within a preset first warning temperature range and the rotor temperature is not higher than the second warning temperature, the thermal state category label corresponding to the multidimensional electrical data of the samples is determined to be a second thermal state category; if the temperature in the first warning temperature range is higher than the first warning temperature; if the stator temperature is not higher than the first warning temperature and the rotor temperature is within a preset second warning temperature range... The thermal state category label corresponding to the multidimensional electrical data of the sample is determined as the third thermal state category; the temperature in the second warning temperature range is higher than the second warning temperature; when the stator temperature is in the first warning temperature range and the rotor temperature is in the second warning temperature range, the thermal state category label corresponding to the multidimensional electrical data of the sample is determined as the fourth thermal state category; when the stator temperature is not lower than the preset first danger temperature, the thermal state category label corresponding to the multidimensional electrical data of the sample is determined as the fifth thermal state category; the first danger temperature is higher than the temperature in the first warning temperature range; when the rotor temperature is not lower than the preset second danger temperature, the thermal state category label corresponding to the multidimensional electrical data of the sample is determined as the sixth thermal state category; the second danger temperature is higher than the temperature in the second warning temperature range; wherein, the first warning temperature, the first warning temperature range, and the first danger temperature are determined based on the insulation level of the stator winding of the motor; the second warning temperature, the second warning temperature range, and the second danger temperature are determined based on the demagnetization characteristics of the motor.

[0134] In some embodiments, when the training module trains an initial multi-classification model based on multiple sample data pairs to obtain a motor thermal state classification model, it specifically performs the following steps: determining training sample data pairs from the multiple sample data pairs; determining the target hyperparameter group corresponding to the initial multi-classification model from a set of multiple candidate hyperparameter groups based on the training sample data pairs; each candidate hyperparameter group includes a kernel parameter and a penalty parameter in the kernel function; the penalty parameter represents the tolerance of the motor thermal state classification model to classification errors; the kernel function measures the similarity of different groups of multidimensional electrical data in terms of causing motor heating; the kernel parameter determines the width of the curve corresponding to the kernel function; and training the initial multi-classification model based on the target hyperparameter group and multiple sample data pairs to obtain the motor thermal state classification model.

[0135] In some embodiments, when the training module determines the target hyperparameter group corresponding to the initial multi-class classification model from a set of preset candidate hyperparameter groups based on training sample data pairs, it specifically performs the following: traversing each candidate hyperparameter group in the set of preset candidate hyperparameter groups, and for each traversed candidate hyperparameter group, training the initial multi-class classification model multiple times based on the candidate hyperparameter group and the training sample data pairs to obtain multiple intermediate multi-class classification models, and determining the average classification accuracy of the multiple intermediate multi-class classification models; selecting the candidate hyperparameter group that results in the highest average classification accuracy from the multiple candidate hyperparameter groups as the target hyperparameter group corresponding to the initial multi-class classification model.

[0136] In some embodiments, the device may further include a thermal management module; the determining module is further configured to determine a target thermal management control strategy that matches a target thermal state category based on the correspondence between multiple thermal state categories and multiple thermal management control strategies; the target thermal management control strategy is one of multiple thermal management control strategies; wherein, different thermal management control strategies among the multiple thermal management control strategies include different torque derating control strategies and / or different cooling system power control strategies; the thermal management module is configured to perform thermal management on the motor based on the target thermal management control strategy.

[0137] It is understood that the specific implementation of each module in the motor thermal condition identification device provided in this application embodiment and the beneficial effects that can be achieved can be referred to the description of the aforementioned motor thermal condition identification method embodiment, and will not be repeated here.

[0138] Each module in the aforementioned motor thermal condition identification device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of the vehicle-mounted terminal device in hardware form or stored in the memory of the motor thermal condition identification device in software form, so that the processor can call and execute the corresponding operations of each module.

[0139] In one exemplary embodiment, a controller is provided, the internal structure of which can be shown in the following diagram. Figure 5 As shown, the controller includes a processor, memory, input / output interfaces, a communication interface, and input devices. The processor, memory, and input / output interfaces are connected via a system bus, and the communication interface and input devices are also connected to the system bus via the input / output interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interfaces are used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When executed by the processor, the computer program implements a method for identifying the thermal state of a motor.

[0140] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the controller to which the present application is applied. A specific controller may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0141] In one exemplary embodiment, this application provides a controller, including a memory and a processor, wherein the memory stores a computer program; when the processor executes the computer program, it implements the steps in the above-described motor thermal state identification methods.

[0142] In one exemplary embodiment, this application provides a computer-readable storage medium having a computer program stored thereon. When executed by a processor, the computer program implements the steps in the above-described methods for identifying the thermal state of motors.

[0143] In one exemplary embodiment, this application provides a computer program product, including a computer program. When executed by a processor, the computer program implements the steps in the above-described methods for identifying the thermal state of motors.

[0144] It should be noted that the data involved in this application (including but not limited to data used for analysis, data stored, data displayed, etc.) are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0145] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0146] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0147] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for identifying the thermal state of a motor, characterized in that, The method includes: Acquire multi-dimensional electrical data of the motors in the vehicle at the current moment; Based on the multidimensional electrical data, a feature vector reflecting the thermal state of the motor is determined; The feature vector is input into the motor thermal state classification model to obtain the target thermal state category of the motor at the current time; the target thermal state category is one of a number of preset thermal state categories, which respectively characterize the different temperature states of the stator and / or rotor of the motor. The motor thermal state classification model is obtained by training an initial multi-classification model based on multiple sample data pairs; each sample data pair consists of a sample feature vector corresponding to the multidimensional electrical data of the sample and a thermal state category label; the initial multi-classification model consists of multiple binary sub-models containing kernel functions.

2. The method according to claim 1, characterized in that, The multidimensional electrical data includes the motor's three-phase current, DC bus voltage, and motor speed; the determination of a feature vector reflecting the motor's thermal state based on the multidimensional electrical data includes: Based on the three-phase current, the direct-axis current used to control the motor magnetic field and the quadrature-axis current used to generate electromagnetic torque are determined. Based on the DC bus voltage, determine the direct-axis voltage for generating the direct-axis current and the quadrature-axis voltage for generating the quadrature-axis current; Based on the instantaneous input power and instantaneous output power of the motor, the instantaneous total loss of the motor is determined, and based on multiple instantaneous total losses within a preset time window, the average power loss is determined; wherein, the instantaneous input power is determined based on the DC bus voltage and DC bus current, and the instantaneous output power is determined based on the quadrature axis current and the motor speed; Based on the direct-axis current, the quadrature-axis current, the direct-axis voltage, the quadrature-axis voltage, the motor speed, the motor's mechanical angular velocity, and the average power loss, a feature vector reflecting the thermal state of the motor is determined.

3. The method according to claim 1, characterized in that, The motor thermal state classification model was trained in the following way: Multiple sets of operating data are obtained when the motor is running under different operating conditions. Each set of operating data includes sample multidimensional electrical data and corresponding temperature data. Determine the sample feature vector corresponding to the multidimensional electrical data of each group of samples; Based on the temperature data corresponding to the multidimensional electrical data of each group of samples, the thermal state category label corresponding to the multidimensional electrical data of each group of samples is determined according to the preset thermal state category determination strategy. Based on the sample feature vector and thermal state category label corresponding to the multidimensional electrical data of each group of samples, multiple sample data pairs are constructed, and the initial multi-classification model is trained based on the multiple sample data pairs to obtain the motor thermal state classification model.

4. The method according to claim 3, characterized in that, The temperature data corresponding to the sample multidimensional electrical data includes the stator temperature and rotor temperature of the motor; based on the temperature data corresponding to each group of sample multidimensional electrical data, and according to a preset thermal state category determination strategy, the thermal state category label corresponding to each group of sample multidimensional electrical data is determined, including: For each set of sample multidimensional electrical data, if the stator temperature in the temperature data corresponding to the sample multidimensional electrical data is lower than a preset first warning temperature and the rotor temperature is lower than a preset second warning temperature, the thermal state category label corresponding to the sample multidimensional electrical data is determined to be the first thermal state category. When the stator temperature is within a preset first warning temperature range and the rotor temperature is not higher than the second warning temperature, the thermal state category label corresponding to the sample multidimensional electrical data is determined to be the second thermal state category; the temperature in the first warning temperature range is higher than the first warning temperature. If the stator temperature is not higher than the first warning temperature and the rotor temperature is within the preset second warning temperature range, the thermal state category label corresponding to the sample multidimensional electrical data is determined to be the third thermal state category; the temperature in the second warning temperature range is higher than the second warning temperature. When the stator temperature is in the first warning temperature range and the rotor temperature is in the second warning temperature range, the thermal state category label corresponding to the sample multidimensional electrical data is determined to be the fourth thermal state category. If the stator temperature is not lower than a preset first danger temperature, the thermal state category label corresponding to the sample multidimensional electrical data is determined to be the fifth thermal state category; the first danger temperature is higher than the first warning temperature range. If the rotor temperature is not lower than the preset second danger temperature, the thermal state category label corresponding to the sample multidimensional electrical data is determined to be the sixth thermal state category; the second danger temperature is higher than the second warning temperature range. The first warning temperature, the first warning temperature range, and the first danger temperature are determined based on the insulation class of the stator winding of the motor; the second warning temperature, the second warning temperature range, and the second danger temperature are determined based on the demagnetization characteristics of the motor.

5. The method according to claim 3, characterized in that, The process of training an initial multi-classification model based on multiple sample data pairs to obtain the motor thermal state classification model includes: Determine training sample data pairs from among the multiple sample data pairs; Based on the training sample data pairs, the target hyperparameter group corresponding to the initial multi-classification model is determined from a set of multiple candidate hyperparameter groups. Each candidate hyperparameter group includes a kernel parameter and a penalty parameter in the kernel function. The penalty parameter represents the tolerance of the motor thermal state classification model to classification errors. The kernel function measures the similarity of different groups of multidimensional electrical data in terms of causing motor heating. The kernel parameter determines the width of the curve corresponding to the kernel function. Based on the target hyperparameter set and multiple sample data pairs, the initial multi-classification model is trained to obtain the motor thermal state classification model.

6. The method according to claim 5, characterized in that, The step of determining the target hyperparameter set corresponding to the initial multi-classification model from a preset set of candidate hyperparameter sets based on the training sample data pairs includes: The initial multi-class classification model is trained in multiple rounds based on the candidate hyperparameter group and the training sample data pair, and multiple intermediate multi-class classification models are obtained by traversing each candidate hyperparameter group and for each traversed candidate hyperparameter group. The average classification accuracy of the multiple intermediate multi-class classification models is then determined. From the multiple candidate hyperparameter sets, the candidate hyperparameter set that results in the highest average classification accuracy is selected as the target hyperparameter set corresponding to the initial multi-classification model.

7. The method according to any one of claims 1 to 6, characterized in that, The method further includes: Based on the correspondence between various thermal state categories and various thermal management control strategies, a target thermal management control strategy matching the target thermal state category is determined; the target thermal management control strategy is one of the various thermal management control strategies; wherein, different thermal management control strategies among the various thermal management control strategies include different torque derating control strategies and / or different cooling system power control strategies; Based on the target thermal management control strategy, thermal management is performed on the motor.

8. A motor thermal condition identification device, characterized in that, The device includes: The acquisition module is used to acquire the multi-dimensional electrical data of the motor in the vehicle at the current moment; The determination module is used to determine a feature vector reflecting the thermal state of the motor based on the multidimensional electrical data; A thermal state identification module is used to input the feature vector into a motor thermal state classification model to obtain the target thermal state category of the motor at the current moment; the target thermal state category is one of a preset multiple thermal state categories, and the multiple thermal state categories respectively characterize different temperature states of the stator and / or rotor of the motor; The motor thermal state classification model is obtained by training an initial multi-classification model based on multiple sample data pairs; each sample data pair consists of a sample feature vector corresponding to the multidimensional electrical data of the sample and a thermal state category label; the initial multi-classification model consists of multiple binary sub-models containing kernel functions.

9. A controller comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images