A hub motor heat dissipation monitoring method, system, device and medium thereof

Abstract

The application relates to a wheel hub motor heat dissipation monitoring method, system, device and medium thereof, comprising the following steps: acquiring temperature sensors arranged on three components of a stator winding, a bearing end cover and a shell surface of a wheel hub motor; continuously collecting temperature data according to a preset sampling period to obtain temperature time sequence data corresponding to each component; taking a temperature value at a historical sampling time point with a preset time length from a current sampling time point as a moving average value, performing weighted moving average calculation on temperature values corresponding to temperature time sequence data within the preset time length to obtain a weighted moving temperature value; determining a heat dissipation monitoring state grade of the wheel hub motor according to the weighted moving temperature value corresponding to each component at the current sampling time point and real-time operation parameters of the wheel hub motor; and controlling a corresponding heat dissipation fan of the wheel hub motor to perform air supply heat dissipation according to the heat dissipation monitoring state grade.

Description

Technical Field

[0001] This application relates to the field of motor control technology, and in particular to a method, system, device and medium for monitoring heat dissipation of a hub motor. Background Technology

[0002] As a key power component in electric vehicles and other equipment, the operational stability of in-wheel motors is of paramount importance. During operation, components such as the stator windings, bearing end caps, and housing surfaces generate heat due to current flow and mechanical friction. If this heat cannot be dissipated in time, it can lead to excessively high motor temperatures, affecting performance and lifespan, and even causing safety accidents.

[0003] Currently, although temperature data can be obtained by placing temperature sensors on motor components, real-time temperature values ​​are often used directly for judgment without fully considering historical temperature trends. This makes the data susceptible to interference from random factors, leading to inaccurate monitoring results. Furthermore, it is difficult to comprehensively assess the heat dissipation status based on the motor's real-time operating parameters, making it impossible to accurately determine the heat dissipation monitoring status level and consequently, to reasonably control the cooling fan for targeted airflow and heat dissipation. Summary of the Invention

[0004] The purpose of this invention is to provide a method, system, device, and medium for monitoring the heat dissipation of a hub motor. This invention aims to solve the technical problems of hub motor heat dissipation monitoring being susceptible to accidental interference, failing to accurately reflect temperature trends, and lacking the ability to comprehensively evaluate the heat dissipation status based on operating parameters, which leads to the inability to reasonably control the cooling fan. This invention can improve the accuracy and reliability of hub motor temperature monitoring, and at the same time, improve the accuracy and precision of cooling fan control.

[0005] To achieve the above objectives, a first aspect of this disclosure provides a method for monitoring heat dissipation of a hub motor, comprising: Temperature sensors arranged on three components of the hub motor—stator winding, bearing end cover, and outer shell surface—are used to continuously collect temperature data according to a preset sampling period to obtain temperature time series data corresponding to each component. For the temperature time series data of each component, the temperature value of the historical sampling time with a preset time distance from the current sampling time is used as the moving average. The temperature value corresponding to the temperature time series data within the preset time distance is calculated by weighted moving average to obtain the weighted moving temperature value of each component at the current sampling time. The further away from the current sampling time, the smaller the weight value of the weighted moving average. The heat dissipation monitoring status level of the hub motor is determined based on the weighted moving temperature value of each component at the current sampling time and the real-time operating parameters of the hub motor. Based on the heat dissipation monitoring status level, the cooling fan corresponding to the hub motor is controlled to deliver airflow for heat dissipation.

[0006] In one optional approach, the step of using the temperature values ​​of historical sampling times within a preset time interval from the current sampling time as moving averages, and performing a weighted moving average calculation on the temperature values ​​corresponding to the temperature time series data within the preset time interval to obtain the weighted moving temperature value of each component at the current sampling time includes: Using the temperature values ​​of historical sampling times with a preset time interval from the current sampling time as the moving average, and based on the temperature values ​​corresponding to each sampling time within the preset time interval and the corresponding weight values, calculate the weighted moving average value corresponding to each sampling time within the preset time interval for each component. The weighted moving average value of each sampling moment within the preset time period is used as the observation value to establish the state equation of the hub motor thermal model. The state equation is used to describe the linear relationship between the temperature state at any sampling moment and the temperature state at the corresponding previous sampling moment and the motor power loss. The prediction step in the Kalman filter algorithm is used to predict the prior temperature estimate at the current sampling time based on the state estimate and the corresponding state equation at the previous sampling time, and to calculate the covariance matrix of the prior estimation error. The update step in the Kalman filter algorithm is adopted. The weighted moving average value at the current sampling time is used as the measurement value to calculate the Kalman gain. The temperature value sampled at the current sampling time of each component is used to correct the corresponding prior temperature estimate to obtain the posterior temperature estimate at the current sampling time of each component. The posterior estimate of the temperature at the current sampling time of each component is used as the representative temperature value at the current sampling time, and the calculated weighted moving average value is replaced to obtain the weighted moving temperature value of each component at the current sampling time.

[0007] In one optional approach, the step of correcting the corresponding prior temperature estimate using the temperature value sampled at the current sampling time of each component to obtain the posterior temperature estimate at the current sampling time of each component includes: A multi-source temperature observation vector is formed by using the temperature value sampled at the current sampling time of each component and the corresponding prior temperature estimate. The joint probability distribution modeling method in Bayesian estimation is used to model the measurement error of each temperature sensor as a Gaussian distribution, and the joint likelihood function of the multi-source observation vector under the preset real temperature conditions is calculated. According to Bayes' theorem, the joint likelihood function is multiplied by the prior probability density function of the temperature state, and then normalized by a normalization constant to obtain the posterior probability density function corresponding to the temperature state. The temperature value corresponding to the maximum posterior probability is extracted from the posterior probability density function. The extracted temperature value is used as the substitute input of the measured value in the Kalman filter algorithm to correct the prior temperature estimate and obtain the posterior temperature estimate of each component at the current sampling time.

[0008] In one optional approach, the update step in the Kalman filter algorithm involves using the weighted moving average of the current sampling time as the measured value, calculating the Kalman gain, and correcting the corresponding prior temperature estimate using the temperature value sampled at the current sampling time for each component, thereby obtaining the posterior temperature estimate for each component at the current sampling time. This includes: The weighted moving average is set as the alternative input for the measurement at the current sampling time in the Kalman filter algorithm. This alternative input is associated with a pre-stored measurement noise covariance matrix and replaces the original sensor output measurement data. Read the error covariance matrix of the posterior temperature estimate obtained from the previous sampling time, and combine it with the state transition matrix of the system thermal model to calculate the prior temperature estimate and its corresponding prior estimate error covariance matrix at the current sampling time. Calculate the Kalman gain matrix for each component based on the prior estimation error covariance matrix and the measurement noise covariance matrix corresponding to each component at the current sampling time. Multiply the Kalman gain matrix by the residual between the substitute input and the prior temperature estimate, and then add the product to the corresponding prior temperature estimate to obtain the posterior temperature estimate for each component at the current sampling time.

[0009] In one optional approach, determining the heat dissipation monitoring status level of the hub motor based on the weighted moving temperature value corresponding to each component at the current sampling time and the real-time operating parameters of the hub motor includes: The input power and speed values ​​at the current sampling time in the real-time operating parameters of the hub motor are substituted into the multiple linear regression equation to calculate the predicted steady-state temperature. The multiple linear regression equation is established using the linear regression analysis method, with the input power and speed of the hub motor as two independent variables and the steady-state temperature of the outer shell surface of the hub motor as the dependent variable. The weighted moving temperature value corresponding to each component at the current sampling time is compared with the steady-state temperature prediction value to calculate the temperature deviation value of the corresponding component. If the temperature deviation value corresponding to any component exceeds the preset deviation tolerance limit, the location corresponding to that component is marked as an abnormal hot spot location; The weighted moving temperature value corresponding to each component at the current sampling time is compared with a plurality of preset temperature thresholds to determine the initial temperature level corresponding to each component. If the location corresponding to any component is marked as the abnormal hot spot location, a penalty additional weight greater than 1 is applied to the corresponding weighted moving temperature value. The highest initial temperature level among the three components at the current acquisition time is determined as the heat dissipation monitoring status level of the hub motor.

[0010] In one alternative approach, the multiple linear regression equation is constructed as follows: The sample operating parameters of the hub motor under multiple different operating conditions are obtained to construct a sample dataset. Each set of sample operating parameters includes the value of input electric power, the value of rotational speed, and the actual observed temperature value when the outer shell surface reaches a steady state. Using the input electric power and rotational speed in each sample's operating parameters as two independent variables and the steady-state temperature of the outer shell surface as the dependent variable, the basic form of a multiple linear regression equation is constructed. Using the least squares estimation method, regression coefficients and constant intercept terms that minimize the sum of squared residuals between the predicted value of the dependent variable and the actual observed temperature value are calculated based on all the sample operating parameters. Substituting the regression coefficients and the constant intercept term into the basic form of the multiple linear regression equation yields the multiple linear regression equation.

[0011] In one optional approach, controlling the cooling fan corresponding to the hub motor to deliver airflow for heat dissipation based on the heat dissipation monitoring status level includes: Based on the heat dissipation monitoring status level corresponding to the current collection time, the target speed level of the cooling fan is determined, wherein each heat dissipation monitoring status level is pre-associated with a fan speed level. The target speed setting is converted into a duty cycle parameter of a pulse width modulation signal. The duty cycle parameter is output to the control terminal of the cooling fan through the drive circuit, so that the cooling fan delivers air at the speed corresponding to the target speed setting.

[0012] A second aspect of this disclosure provides a hub motor heat dissipation monitoring system, the system comprising: The acquisition module is configured to acquire temperature data continuously collected by temperature sensors arranged on three components of the hub motor: the stator winding, the bearing end cover, and the outer surface of the housing, according to a preset sampling period, and obtain temperature time series data corresponding to each component. The first determining module is configured to perform a weighted moving average calculation on the temperature values ​​corresponding to the temperature time series data within the preset time period for each component, using the temperature values ​​of historical sampling times with a preset time period away from the current sampling time as the moving average, to obtain the weighted moving temperature value of each component at the current sampling time. The further away from the current sampling time is, the smaller the weight value of the weighted moving average. The second determining module is configured to determine the heat dissipation monitoring status level of the hub motor based on the weighted moving temperature value corresponding to each component at the current sampling time and the real-time operating parameters of the hub motor. The control module is configured to control the cooling fan corresponding to the hub motor to deliver airflow for heat dissipation based on the heat dissipation monitoring status level.

[0013] A third aspect of this disclosure provides an electronic device, comprising: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the steps of the method of any one of the first aspects.

[0014] A fourth aspect of this disclosure provides a computer-readable storage medium storing program code that is executed by a processor to implement the steps of the method described in any of the first aspects.

[0015] This invention provides a method, system, device, and medium for monitoring the heat dissipation of a hub motor. Compared with the prior art, it has the following advantages: By acquiring time-series temperature data from three components—the stator winding, bearing end cover, and housing surface—and performing a weighted moving average calculation on the temperature time-series data of each component, the weight of historical data further removed from the current sampling time is reduced. This effectively filters out instantaneous noise interference and highlights recent trends in temperature changes. Furthermore, by combining the weighted moving average temperature value of each component with the real-time operating parameters of the hub motor, a heat dissipation monitoring status level is determined, and the cooling fan is controlled accordingly to deliver airflow for heat dissipation. The introduction of a weighted moving average mechanism improves the stability and reliability of temperature data, avoiding misjudgments caused by accidental factors. Simultaneously, the comprehensive evaluation of real-time operating parameters and historical temperature data enhances the accuracy of heat dissipation status judgment, enabling refined fan control based on heat dissipation levels. This not only solves the problems of existing monitoring methods being susceptible to interference and unable to accurately classify data, but also improves the timeliness and specificity of heat dissipation response, ensuring the operational stability and safety of the hub motor under different operating conditions and extending the motor's service life. By constructing a complete closed-loop logic from data acquisition, trend analysis, status assessment to execution control, the systematization and intelligence of hub motor heat dissipation monitoring have been realized, improving the effectiveness and robustness of motor thermal management.

[0016] Other features and advantages of this disclosure will be described in detail in the following detailed description section. Attached Figure Description

[0017] The accompanying drawings are provided to further illustrate the present disclosure and form part of the specification. They are used together with the following detailed description to explain the present disclosure, but do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating a hub motor heat dissipation monitoring method according to an embodiment of the specification.

[0018] Figure 2 This is one implementation shown in the embodiment of the specification. Figure 1 A flowchart illustrating step S12.

[0019] Figure 3 This is one implementation shown in the embodiment of the specification. Figure 1 A flowchart of step S13.

[0020] Figure 4 A block diagram of a hub motor heat dissipation monitoring system is shown in the embodiment of the specification.

[0021] Figure 5 This is a block diagram of another hub motor heat dissipation monitoring system shown in the embodiment of the specification. Detailed Implementation

[0022] 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.

[0023] This disclosure provides a method for monitoring the heat dissipation of a hub motor. Figure 1 This is a flowchart illustrating a method for monitoring heat dissipation of a hub motor according to an embodiment. The method includes: In step S10, temperature sensors arranged on three components of the hub motor—stator winding, bearing end cover, and outer shell surface—are acquired, and temperature data are continuously collected according to a preset sampling period to obtain temperature time series data corresponding to each component. The hub motor can be a drive motor installed in an electric vehicle or robot. Temperature time-series data can be a dataset arranged chronologically, reflecting the temperature changes of various components. It can be acquired in real-time by temperature sensors located at three key heat-generating locations: the stator windings, bearing end caps, and the outer casing surface, and then converted from analog to digital. Specifically, a first temperature sensor is embedded inside the stator windings of the hub motor, a second temperature sensor is installed at the bearing end cap, and a third temperature sensor is placed near the heat dissipation fins on the outer casing surface. These three components cover the internal heat sources, mechanical friction heat sources, and external heat dissipation interfaces of the motor, comprehensively characterizing the motor's thermal state.

[0024] For example, if the preset sampling period is 1 second, then at the current time t, N+1 temperature data points are continuously collected from tN seconds to t seconds, corresponding to the temperature readings of the stator winding, bearing end cover, and outer casing surface, respectively, thus forming three parallel time series curves. Through multi-point layout and continuous acquisition, a complete spatiotemporal distribution basis of the motor thermal field can be constructed, avoiding the local deviations that may be caused by single-point measurement.

[0025] In step S20, for the temperature time series data of each component, the temperature value of the historical sampling time with a preset time distance from the current sampling time is used as the moving average. The temperature value corresponding to the temperature time series data within the preset time distance is calculated by weighted moving average to obtain the weighted moving temperature value of each component at the current sampling time. The further away from the current sampling time, the smaller the weight value of the weighted moving average. The weighted moving average temperature value is a smoothed representative temperature value used to eliminate random noise and highlight the true trend of temperature change. This value is obtained by performing a non-linear weighted calculation on historical data within a preset time window. Specifically, a sliding time window (e.g., the past 60 seconds) is set, containing temperature values ​​from multiple historical sampling times. During the calculation, a weight coefficient is assigned to each data point within the window. This weight coefficient is negatively correlated with the time interval between the data point and the current sampling time. That is, data closer to the current time has a larger weight value, and data farther away from the current time has a smaller weight value.

[0026] For example, assuming the preset time period includes the 5 most recent sampling points, the weights can be set to 0.3, 0.25, 0.2, 0.15, and 0.1 (after normalization) respectively, so that the most recently collected temperature data dominates the calculation results, while the influence of older data gradually decreases.

[0027] By introducing this time-decaying weighting mechanism, false alarms of temperature spikes caused by electromagnetic interference or sudden load changes can be suppressed, while the potential overheating trend of slow temperature rise can be accurately detected, thereby improving the stability and reliability of temperature data.

[0028] In step S30, the heat dissipation monitoring status level of the hub motor is determined based on the weighted moving temperature value of each component at the current sampling time and the real-time operating parameters of the hub motor. The heat dissipation monitoring status level is a quantitative indicator characterizing the degree of match between the motor's current thermal load and heat dissipation requirements. It can be divided into multiple levels, such as normal, attention, warning, and critical. The heat dissipation monitoring status level is determined by comprehensively considering the fused temperature data and the motor's real-time operating parameters. Real-time operating parameters may include the motor's input power, speed, torque, and ambient temperature. The weighted moving average temperature value of each component is compared with the theoretical steady-state temperature calculated from the current operating parameters or a preset safety threshold. If the weighted moving average temperature value is higher than the expected temperature under the same operating conditions, or the absolute temperature value exceeds the threshold of a specific level, the motor is determined to be in a higher heat dissipation risk level.

[0029] For example, when the weighted moving temperature of the stator winding is detected to be 140°C, and the motor is operating under low speed and low power no-load conditions, the temperature can be determined to be abnormally high, thus raising the heat dissipation monitoring status level to the warning level. Conversely, if the same temperature is measured under high power and full load conditions, it may only be determined to be at the attention level. By combining historical temperature trends with real-time operating conditions, a dynamic and multi-dimensional assessment of the heat dissipation status is achieved, avoiding the limitations of judging with a single temperature threshold under different operating conditions and ensuring the accuracy of status level determination.

[0030] In step S40, the cooling fan corresponding to the hub motor is controlled to deliver airflow for heat dissipation according to the heat dissipation monitoring status level.

[0031] In this system, airflow cooling can be achieved by adjusting the operating status of the cooling fan to change the airflow across the motor surface, thereby removing heat. Airflow cooling is a closed-loop control operation based on the previously determined heat dissipation monitoring status level. A mapping relationship between the heat dissipation monitoring status level and the fan control strategy can be pre-established. Different status levels correspond to different target fan speeds or start / stop commands. When the status level increases, the control system outputs a corresponding drive signal to increase the fan speed and thus increase airflow; when the status level decreases or returns to normal, the fan speed is reduced or stopped to save energy and reduce noise.

[0032] For example, when the thermal monitoring status level is normal, the fan keeps rotating at a low speed or operates intermittently; when the level jumps to critical, the fan immediately switches to full-speed operation mode to maximize heat dissipation efficiency.

[0033] This graded response mechanism enables the on-demand allocation of heat dissipation resources, ensuring timely cooling of the motor under high-temperature risks and extending its lifespan, while avoiding energy waste caused by excessive heat dissipation under low-temperature conditions.

[0034] This approach allows for the acquisition of time-series temperature data from three components: the stator windings, bearing end caps, and the outer casing. A weighted moving average is then calculated for each component, reducing the weight of historical data further removed from the current sampling time. This effectively filters out transient noise interference and highlights recent temperature trends. Furthermore, by combining the weighted moving average temperature values ​​of each component with the real-time operating parameters of the hub motor, a heat dissipation monitoring status level is determined, which is then used to control the cooling fan for heat dissipation. The introduction of the weighted moving average mechanism improves the stability and reliability of temperature data, avoiding misjudgments caused by random factors. Simultaneously, the comprehensive evaluation of real-time operating parameters and historical temperature data enhances the accuracy of heat dissipation status judgment, enabling refined fan control based on heat dissipation levels. This not only solves the problems of existing monitoring methods being susceptible to interference and unable to accurately classify data but also improves the timeliness and specificity of heat dissipation response, ensuring the operational stability and safety of the hub motor under different operating conditions and extending the motor's service life. By constructing a complete closed-loop logic from data acquisition, trend analysis, status assessment to execution control, the systematization and intelligence of hub motor heat dissipation monitoring have been realized, improving the effectiveness and robustness of motor thermal management.

[0035] Preferably, referring to Figure 2, in step S20, the step of using the temperature values ​​of historical sampling times within a preset time interval from the current sampling time as moving averages, and performing a weighted moving average calculation on the temperature values ​​corresponding to the temperature time series data within the preset time interval to obtain the weighted moving temperature value of each component at the current sampling time, includes: In step S201, the temperature value of historical sampling times with a preset time interval from the current sampling time is used as the moving average value. Based on the temperature value and weight value corresponding to each sampling time within the preset time interval, the weighted moving average value corresponding to each sampling time within the preset time interval for each component is calculated. The preset duration can be the time span used for sliding window calculations, such as the past 10 seconds or 20 sampling periods. The weight value is dynamically set based on the time interval between the current sampling time and the previous sampling time; the greater the time interval, the smaller the weight value, reflecting the influence of recent temperature data on the current state. Specifically, for each of the three components—stator winding, bearing end cover, and housing surface—the system extracts its temperature time series data within the preset duration. Assume the preset duration includes N sampling points, and the temperature value of the i-th sampling point is T. i The corresponding weight is W. i Then, the weighted moving average is obtained by weighted calculation.

[0036] For example, if the preset duration is 5 sampling periods, the weights can be set sequentially to 0.1, 0.15, 0.2, 0.25, and 0.3, so that the temperature value of the most recently sampled temperature has the largest proportion. This weighted calculation method can effectively smooth out high-frequency random noise in the sensor acquisition process while preserving the overall trend of temperature change.

[0037] In step S202, the weighted moving average value of each sampling time within a preset time period is used as the observation value to establish the state equation of the hub motor thermal model. The state equation is used to describe the linear relationship between the temperature state at any sampling time and the temperature state at the corresponding previous sampling time as well as the motor power loss. The state equation is a mathematical model based on the thermal conductivity characteristics of the hub motor, used to characterize the evolution of the system's internal state. The core variables of this equation include the temperature state vector at the current sampling moment, the temperature state vector at the previous sampling moment, and the motor power loss. The motor power loss mainly comes from copper losses and iron losses, and its value can be calculated based on real-time current, voltage, and speed parameters. The specific form of the state equation can be X... k =A·X k-1 +B·U K +W k , where X kX represents the temperature status of each component at the current moment. k-1 U represents the temperature state at the previous moment. K Let A represent the current motor power loss input, A be the state transition matrix describing the hysteresis effect of heat transfer between the stator, bearings, and housing, and B be the control input matrix describing the efficiency of converting power loss into temperature rise. W k This is process noise.

[0038] For example, when the motor is operating under high load, the power loss U K With this addition, the state equation will predict the upward trend of the temperature state at the next moment through matrix B. By using a weighted moving average as the basis for observations and combining it with a physical model, a shift from simple data fitting to mechanism-driven modeling is achieved, enhancing the model's adaptability to complex operating conditions.

[0039] In step S203, the prediction step in the Kalman filter algorithm is used to predict the prior temperature estimate at the current sampling time based on the state estimate and the corresponding state equation at the previous sampling time, and to calculate the covariance matrix of the prior estimation error. The prediction step, the first half of the Kalman filter loop, aims to infer the theoretical state at the current moment using the system model. It reads the posterior temperature estimate and its error covariance matrix from the previous iteration, substitutes them into the established state equation, and calculates the prior temperature estimate at the current moment without any new measurements. Simultaneously, it updates the covariance matrix of the prior estimation error based on the statistical characteristics of process noise; this matrix quantifies the uncertainty of the prediction result.

[0040] For example, if drastic fluctuations in motor operating conditions lead to increased process noise, the diagonal elements of the covariance matrix will increase accordingly, indicating a decrease in the reliability of the predicted values. For instance, during vehicle acceleration, the motor's power loss changes abruptly. Although the state equation can predict the temperature rise, due to thermal inertia, there may be a deviation between the predicted and actual values. In this case, a larger covariance matrix calculated will provide an adjustment basis for subsequent correction steps, ensuring that the filter can flexibly respond to dynamic changes.

[0041] In step S204, the update step in the Kalman filter algorithm is adopted. The weighted moving average value at the current sampling time is used as the measured value to calculate the Kalman gain. The temperature value sampled at the current sampling time of each component is used to correct the corresponding prior temperature estimate to obtain the posterior temperature estimate at the current sampling time of each component. The update step, the latter half of the Kalman filter loop, is used to fuse model predictions with actual observation information. Instead of directly using the original temperature values ​​collected by the sensors, it uses the weighted moving average calculated above as the measurement input. This helps suppress the impact of measurement noise on the filtering results.

[0042] First, the Kalman gain is calculated based on the prior estimation error covariance matrix and the measurement noise covariance matrix. This gain determines the proportion of trust between the model prediction and the observed data. Then, the residual between the measured value (i.e., the weighted moving average) and the prior temperature estimate is calculated. The residual is then multiplied by the Kalman gain and added to the prior estimate to obtain the corrected posterior temperature estimate.

[0043] For example, when the sensor generates a spike pulse due to transient electromagnetic interference, the filter will automatically reduce its confidence in the abnormal data because it uses a weighted moving average as the measurement input and the Kalman gain will be dynamically adjusted according to the covariance. It will mainly rely on model prediction to maintain output stability. Conversely, when the temperature undergoes a continuous real change, the filter will quickly adjust the gain so that the posterior estimate can quickly track the real temperature.

[0044] In step S205, the posterior temperature estimate of each component at the current sampling time is used as the representative temperature value at the current sampling time, replacing the calculated weighted moving average value, to obtain the weighted moving temperature value of each component at the current sampling time.

[0045] The posterior temperature estimate is the optimal state estimate after model prediction and multi-source information fusion optimization, which has higher accuracy and robustness compared to the original weighted moving average. The calculated posterior temperature estimates of the stator winding, bearing end cover, and housing surface are directly defined as the weighted moving average temperature values ​​corresponding to that moment, and are used to replace the preliminary weighted average result.

[0046] For example, under frequent braking conditions on long downhill slopes, the hub motor continuously generates heat. The raw sensor data may be slow to respond due to thermal hysteresis caused by the installation location. However, the posterior temperature estimate after Kalman filtering correction can more accurately reflect the true temperature rise of the internal windings. The resulting weighted moving temperature value not only retains the trend characteristics of historical data but also eliminates random noise interference and compensates for system hysteresis.

[0047] Thus, by introducing the Kalman filter algorithm to deeply optimize the weighted moving average process, the organic synergy of technical features is achieved. By establishing a state equation describing the linear relationship between temperature state and motor power loss, the system can predict temperature change trends using physical mechanisms. Furthermore, the prediction step generates prior estimates, and the update step uses the weighted moving average as the observation input for correction, effectively solving the problems of lag and insufficient response to sudden changes inherent in simple weighted averaging.

[0048] Furthermore, using a weighted moving average instead of the raw instantaneous value as the measurement input for the Kalman filter leverages the smoothing effect of the weighted average on high-frequency noise and utilizes the dynamic adjustment mechanism of the Kalman gain to maintain stable estimation when the sensor experiences brief drift or noise interference, while also enabling rapid response to sudden changes in actual temperature. Finally, outputting the high-precision posterior temperature estimate as the weighted moving average temperature value improves the accuracy and anti-interference capability of hub motor heat dissipation monitoring, ensuring the timeliness and effectiveness of the heat dissipation control strategy.

[0049] Preferably, in step S204, the step of correcting the corresponding prior temperature estimate using the temperature value sampled at the current sampling time of each component to obtain the posterior temperature estimate at the current sampling time of each component includes: In step S2041, the temperature value sampled at the current sampling time of each component and the corresponding prior temperature estimate are used to form a multi-source temperature observation vector; The multi-source temperature observation vector can be a set of vectors formed by integrating the measured temperature data collected at the same sampling moment from three components of the hub motor—stator windings, bearing end caps, and outer casing surface—with the prior temperature estimates of the corresponding components predicted by the Kalman filter algorithm. This vector originates from the state estimation results output in the Kalman filter prediction stage of the previous step, as well as the raw sensor readings acquired in real time in this step, serving as multi-dimensional observation information characterizing the system state at the current moment.

[0050] For example, assuming three components are component A, component B, and component C, the multi-source temperature observation vector can be represented as a mathematical vector containing the difference or combination of the measured and predicted values ​​of these three components. For instance, if the measured temperature of the stator winding is 85℃, and its corresponding prior temperature estimate is 82℃; the measured temperature of the bearing end cover is 60℃, and its prior estimate is 58℃; and the measured temperature of the outer casing surface is 45℃, and its prior estimate is 44℃, then the resulting multi-source temperature observation vector not only contains these absolute values ​​but also implicitly includes the deviation information between the sensor readings and the model predictions. This vectorized composition method allows sensor data scattered in different physical locations to be uniformly incorporated into a single mathematical framework for processing.

[0051] In step S2042, the joint probability distribution modeling method in Bayesian estimation is used to model the measurement error of each temperature sensor as a Gaussian distribution, and the joint likelihood function of the multi-source observation vector under the preset real temperature conditions is calculated. The Gaussian distribution can be a normal distribution model used to describe the statistical characteristics of temperature sensor measurement errors. It includes two key parameters: mean and variance. The mean is usually set to zero to represent unbiased error, while the variance reflects the measurement accuracy or noise level of a specific sensor. This distribution is determined using historical calibration data or sensor specifications and is used to quantify the probability that each temperature sensor reading deviates from the true temperature. Different weights are assigned to sensors with different accuracies; sensors with higher accuracy (smaller variance) receive greater weight. The joint likelihood function is the product of the probability densities of all values ​​in the current multi-source temperature observation vector simultaneously observed by all sensors, given a preset true temperature value. This function is obtained by multiplying the probability density functions of the Gaussian distributions of each independent sensor and is used to evaluate the overall fit between the current observation data and a hypothetical true temperature value.

[0052] For example, if the measurement error variance of the stator winding sensor is small (high precision), while the variance of the housing surface sensor is large due to environmental interference (low precision), then when calculating the joint likelihood function, the temperature reading of the stator winding will have a greater impact on the final probability peak than the reading of the housing surface. This reflects the probability distribution characteristics during multi-source data fusion. In this way, the influence of noisy sensor data can be automatically identified and suppressed, thereby improving the reliability of the overall observation data.

[0053] In step S2043, the joint likelihood function is multiplied with the prior probability density function of the temperature state according to Bayes' theorem, and then normalized by a normalization constant to obtain the posterior probability density function corresponding to the temperature state. The prior probability density function of the temperature state can be a probability distribution of the current temperature state derived from the state estimate of the previous moment and the evolution law of the system's thermal model before the current observation is made. It originates from the prediction step of the Kalman filter. The posterior probability density function is the latest probability description of the current true temperature state after integrating the current multi-source observation information (i.e., the joint likelihood function) and the original prediction information (i.e., the prior probability density function). This function is calculated using Bayes' theorem formula, specifically by multiplying the joint likelihood function and the prior probability density function point by point, and then dividing by a normalization constant to ensure a total probability of 1. This is used to synthesize information from both model prediction and actual observation to obtain a temperature state estimate distribution that is more accurate and has less uncertainty than using either information source alone.

[0054] For example, if the Kalman filter predicts the current temperature to be 80℃ with high confidence, while the multi-source observation vector indicates a temperature of 85℃ but some sensors have significant noise, the peak value of the posterior probability density function after Bayesian updating will be between 80℃ and 85℃, and the distribution width (uncertainty) will be significantly narrower than the prior distribution or the simple observation distribution. This achieves a dynamic balance between the theoretical model and actual data, enabling the obtained posterior probability density function to more objectively reflect the true thermal state of each component of the hub motor.

[0055] In step S2044, the temperature value corresponding to the maximum posterior probability is extracted from the posterior probability density function. The extracted temperature value is used as the substitute input of the measured value in the Kalman filter algorithm to correct the prior temperature estimate and obtain the posterior temperature estimate of each component at the current sampling time. The temperature value corresponding to the maximum posterior probability can be the temperature value corresponding to the maximum probability density on the posterior probability density function curve. This value is obtained by searching for the peak point within the domain of the posterior probability density function using mathematical optimization methods, and is used to replace the measurement data that may contain noise or bias in the original sensor output. The posterior temperature estimate can be the final temperature estimate obtained by correcting the prior temperature estimate in the Kalman filter using the above-mentioned alternative input, representing the optimal estimate of the temperature of each component at the current sampling time.

[0056] For example, in a certain sampling, if a sensor experiences a sudden malfunction causing a reading jump, the peak value of its posterior probability density function does not shift significantly with this jump but remains within a reasonable range. In this case, the extracted maximum posterior probability temperature value will effectively filter out the abnormal jump, resulting in a smoother and more accurate posterior temperature estimate. This extraction and correction mechanism not only utilizes the time-series prediction capabilities of Kalman filtering but also combines the spatial multi-source fusion advantages of Bayesian estimation, improving the accuracy and robustness of temperature data in hub motor heat dissipation monitoring.

[0057] In this way, by employing this multi-source information fusion strategy, the reliability of different sensors can be effectively distinguished. When the measurement error of a sensor increases due to aging or environmental interference, its corresponding Gaussian distribution variance increases, and its contribution to the joint likelihood function automatically decreases, thereby reducing the negative impact on the final estimation result. Conversely, data from high-precision sensors are given higher weight. This not only solves the problem of single-sensor data being susceptible to interference from random factors, but also improves the scientific rigor and reliability of multi-source temperature information fusion, making the posterior temperature estimates of each component at the current sampling time more accurate.

[0058] Preferably, in step S204, the update step in the Kalman filter algorithm, which uses the weighted moving average of the current sampling time as the measured value, calculates the Kalman gain, and uses the temperature value sampled at the current sampling time of each component to correct the corresponding prior temperature estimate, to obtain the posterior temperature estimate of each component at the current sampling time, includes: In step S204a, the weighted moving average is set as the substitute input for the measurement value at the current sampling time in the Kalman filter algorithm. This substitute input is associated with the pre-stored measurement noise covariance matrix and replaces the measurement data output by the original sensor. The calculated weighted moving temperature value is directly assigned to the measurement vector Z in the Kalman filter update step. k By introducing a pre-stored measurement noise covariance matrix R, this alternative input is associated with the system's prior knowledge of the sensor's noise characteristics, thereby establishing a mapping relationship between the observed values ​​and the noise statistics at the mathematical model level.

[0059] For example, if the temperature sensor at the stator winding has significant high-frequency random noise, the diagonal element corresponding to that channel in the pre-stored measurement noise covariance matrix will be set to a larger value. In this case, using the weighted smoothed temperature value as the alternative input can effectively suppress abrupt interference in the original data. Replacing the original sensor output data with a weighted moving average makes the observed information entering the filter more stable, reducing the risk of filter divergence due to single-point sampling errors.

[0060] In step S204b, the error covariance matrix of the temperature posterior estimate obtained from the previous sampling time is read, and combined with the state transition matrix of the system thermal model, the temperature prior estimate and its corresponding prior estimate error covariance matrix at the current sampling time are calculated. The error covariance matrix of the posterior temperature estimate obtained from the previous sampling time can be P, a matrix generated at the end of the previous Kalman filter cycle, representing the uncertainty of the temperature estimate at the previous time. k-1 The state transition matrix F of the system thermal model is a linearized model parameter established based on the thermodynamic characteristics of the hub motor, used to describe the dynamic evolution of the temperature state over time. The prior temperature estimate at the current sampling time is calculated. The process involves using the state transition matrix F to estimate the posterior temperature value from the previous time step. The deduction is performed; and the prior estimation error covariance matrix at the current sampling time is calculated. The process is based on The matrix operation performed is given, where Q is the process noise covariance matrix.

[0061] In step S204c, the Kalman gain matrix corresponding to each component is calculated based on the prior estimation error covariance matrix and the measurement noise covariance matrix corresponding to each component at the current sampling time. Wherein, the Kalman gain matrix K k It is a weighting matrix used to determine, at the current sampling time, the proportion of the input to be trusted by the model's predictions or by trusting the observations. Kalman gain matrix K k It is based on the prior estimation error covariance matrix at the current sampling time. The associated pre-stored measurement noise covariance matrix R is determined. For example: , where H is the observation matrix.

[0062] For example, when the prior estimate error covariance matrix When the measured noise covariance matrix R is relatively small, the calculated Kalman gain matrix has smaller element values, indicating that the system is more inclined to believe the predictions of the thermal model. Conversely, if the observed data is very reliable (small R), the gain value increases, and the system corrects the predicted values ​​more to better approximate the observed values. For example, in situations where the high-speed operation of the hub motor leads to increased nonlinearity in the thermal model and potentially increased prediction bias, if the weighted moving average shows extremely high stability (corresponding to a small equivalent R), the algorithm will automatically adjust the gain, increasing the weight of the observed data. Through this adaptive gain calculation mechanism, optimal weighted fusion of model predictions and measured information under different noise environments is achieved.

[0063] In step S204d, the Kalman gain matrix is ​​multiplied by the residual between the alternative input and the prior temperature estimate, and the product is superimposed on the corresponding prior temperature estimate to obtain the posterior temperature estimate for each component at the current sampling time.

[0064] The residual can be the difference between the aforementioned alternative input (weighted moving average) and the calculated prior temperature estimate, reflecting the degree of deviation between the actual observed trend and the model's predicted trend. The product result represents the weighted correction of the residual by the Kalman gain matrix, indicating the adjustment required to the predicted value based on the current confidence level analysis.

[0065] Specifically, first calculate the residual vector. Then calculate the correction term K. k y k Finally, the correction term is linearly added to the prior estimate, i.e. .

[0066] Correcting the prior temperature estimate not only eliminates the accumulated error of the model but also smooths out observation noise. The final output posterior temperature estimate retains the physical continuity of the thermal model and responds promptly to the actual temperature change trend.

[0067] In this way, by setting the weighted moving average as the substitute input for the measured value and associating it with the measurement noise covariance matrix, the high-frequency noise interference from the original sensor is effectively isolated. The current prior estimate and error covariance are derived using the posterior error covariance matrix from the previous time step combined with the state transition matrix, ensuring that the state evolution over time conforms to the thermodynamic laws of the motor. Furthermore, the Kalman gain matrix is ​​dynamically calculated based on the prior error covariance and the measurement noise covariance, enabling real-time quantitative evaluation of the model prediction reliability and the reliability of the observed data. Finally, the residual between the substitute input and the prior estimate is weighted and compensated using the Kalman gain and then superimposed onto the prior estimate to obtain a high-precision posterior temperature estimate. This allows the system to maintain the stability and response speed of temperature estimation even when the motor operating conditions change drastically or the sensor is subjected to instantaneous interference, avoiding misjudgments caused by fluctuations in a single data source.

[0068] Preferably, referring to Figure 3, in step S30, determining the heat dissipation monitoring status level of the hub motor based on the weighted moving temperature value corresponding to each component at the current sampling time and the real-time operating parameters of the hub motor includes: In step S301, the input power and speed values ​​at the current sampling time in the real-time operating parameters of the hub motor are substituted into the multiple linear regression equation to calculate the predicted steady-state temperature. The multiple linear regression equation is established using the linear regression analysis method, with the input power and speed of the hub motor as two independent variables and the steady-state temperature of the hub motor's outer shell surface as the dependent variable. The predicted steady-state temperature is the theoretically expected surface temperature of the hub motor's casing under specific input power and speed conditions, in a state of thermal equilibrium. The multiple linear regression equation reflects the linear mapping relationship between the motor's operating parameters and the steady-state temperature. Specifically, instantaneous data from current and speed sensors are read in real-time and input as independent variables into the regression model stored in the controller.

[0069] For example, when the current input power is detected to be 2.5kW and the speed is 1200rpm, the system calls equation T. pred= a×P+b×N+c (where a and b are regression coefficients, and c is the intercept term), the theoretical steady-state temperature is calculated to be 65℃. As shown in Figure 5, this process begins by acquiring real-time operating parameters, then enters the regression calculation module, outputting the corresponding predicted steady-state temperature value. This value provides a dynamic benchmark for subsequent temperature deviation analysis. By introducing operating parameters for dynamic prediction, misjudgments caused by differences in the absolute temperature value under different load conditions can be eliminated, allowing the monitoring standard to adaptively adjust with the motor's operating state.

[0070] In step S302, the weighted moving temperature value corresponding to each component at the current sampling time is compared with the steady-state temperature prediction value to calculate the temperature deviation value of the corresponding component. The temperature deviation value characterizes the difference between the actual heat dissipation performance of the motor and the theoretical expectation. This value is obtained by subtracting, or taking the absolute difference of, the measured temperature value after weighted moving average processing of each component (stator winding, bearing end cover, and housing surface) from the steady-state temperature prediction value calculated above as a reference. Specifically, if the actual weighted moving average temperature of a component is higher than the steady-state prediction value under the same operating conditions, it indicates that there is heat dissipation obstruction or abnormal heating at that location.

[0071] For example, if the predicted steady-state temperature of the outer casing surface is 65°C, while the actual temperature calculated using a weighted moving average is 78°C, the calculated temperature deviation is 13°C. This quantifies the degree to which the actual temperature deviates from the normal thermal model, providing data support for identifying potential local overheating risks.

[0072] In step S303, if the temperature deviation value corresponding to any component exceeds the preset deviation tolerance limit, the location corresponding to that component is marked as an abnormal hot spot location. Abnormal hotspots can be specific component areas where temperature deviations exceed the allowable fluctuation range, indicating potential faults such as poor contact, blocked air ducts, or aging insulation. The preset deviation tolerance limit is a threshold set based on the motor material characteristics and safety margins, used to distinguish between normal temperature fluctuations and abnormal heat dissipation failures.

[0073] For example, a preset deviation tolerance limit is set to 10°C. When the temperature deviation of the bearing end cover is detected to be 13°C (exceeding 10°C), the control logic immediately marks the bearing end cover as an abnormal hot spot location; conversely, if the deviation of the stator winding is 5°C, the normal marking is maintained. In this way, early warning of local overheating risk is achieved, and even if the absolute temperature has not reached the danger threshold, abnormal heat dissipation can be identified through relative deviation.

[0074] In step S304, the weighted moving temperature value corresponding to each component at the current sampling time is compared with multiple preset temperature thresholds to determine the initial temperature level corresponding to each component. If the location corresponding to any component is marked as an abnormal hot spot location, a penalty additional weight greater than 1 is applied to the corresponding weighted moving temperature value. The initial temperature level is a risk level classification based on temperature, such as normal, caution, warning, and danger. The punitive additional weight is a correction coefficient used to amplify the impact of abnormal hotspots. First, the marking results are checked: if a component is not marked as an abnormal hotspot, its original weighted moving temperature value is used directly in the level determination; if it is marked as an abnormal hotspot, its weighted moving temperature value is multiplied by a coefficient greater than 1 (such as 1.2 or 1.5) to obtain the corrected temperature value, which is then compared with the threshold.

[0075] For example, the preset temperature threshold ranges are: less than 70℃ is Level 1, 70℃-90℃ is Level 2, and greater than 90℃ is Level 3. If the measured weighted moving average temperature of the casing surface is 68℃ (which would normally be Level 1), but because it is marked as an abnormal hotspot, a penalty weight of 1.2 is applied, correcting the temperature to 81.6℃, thus classifying it as Level 2 (Caution / Warning). This ensures that localized heat dissipation anomalies receive higher priority in the overall condition assessment, preventing serious heat dissipation efficiency degradation from being masked by the absolute temperature not exceeding the limit.

[0076] In step S305, the temperature level with the highest initial temperature level among the three components at the current acquisition time is determined as the heat dissipation monitoring status level of the hub motor.

[0077] The heat dissipation monitoring status level represents the overall heat dissipation health of the hub motor. Following the "weakest link" principle, the overall status is determined by the hottest or most dangerous component. The stator windings, bearing end caps, and housing surface are compared horizontally at the aforementioned initial temperature levels, and the level with the highest value or highest risk is selected as the final output.

[0078] For example, if the stator winding is classified as Level 1, the bearing end cover as Level 2, and the outer casing surface as Level 1, then the final heat dissipation monitoring status level of the hub motor is determined to be Level 2. This system integrates the monitoring results from multiple sensors and the penalty mechanism for abnormal hot spots, outputting a comprehensive index that fully reflects the motor's heat dissipation status. This provides a direct basis for subsequent precise control of the cooling fan's speed, thereby achieving closed-loop control of the hub motor's thermal management.

[0079] In this way, by constructing a multiple linear regression equation using input electrical power and rotational speed, a steady-state temperature prediction value that varies with operating conditions can be dynamically generated, overcoming the shortcomings of the traditional fixed threshold method in terms of poor adaptability under different loads. By comparing the actual weighted moving temperature with this dynamic prediction value and setting a deviation tolerance limit, early anomalies deviating from the normal thermal model can be keenly detected, and abnormal hot spots can be marked in a timely manner. Furthermore, a punitive additional weight mechanism is introduced, giving components marked as abnormal hot spots a higher weight in temperature level determination. Even if their absolute temperature does not reach the high-risk threshold, a higher level of heat dissipation response can be triggered, effectively preventing low-temperature data in non-hot spot areas from masking serious local heat dissipation problems. Finally, the highest level is selected as the overall state level, ensuring that the heat dissipation strategy is always executed on the weakest link, which can improve the foresight, sensitivity, and safety of hub motor heat dissipation monitoring.

[0080] Preferably, in step S301, the multiple linear regression equation is constructed in the following manner: In step S3011, sample operating parameters of the hub motor under multiple different operating conditions are obtained, and a sample dataset is constructed. Each set of sample operating parameters includes the value of input electric power, the value of rotational speed, and the actual observed temperature value when the corresponding outer shell surface reaches steady state. The sample operating parameters can serve as the foundational dataset for training and validating the regression model. These parameters are collected by controlling the hub motor under various preset operating conditions during bench tests or real-vehicle tests. Specifically, the multiple operating conditions cover the motor's operating range from low to high load and from low to high speed, ensuring data representativeness. Each set of sample operating parameters comprises three core elements: first, the input electrical power value, reflecting the motor's current energy input level, typically calculated as the product of voltage and current; second, the rotational speed value, characterizing the motor's mechanical motion and the strength of the air-cooling effect; and third, the actual observed temperature value corresponding to the steady-state condition of the outer casing surface. This value is the stable temperature reading obtained by a temperature sensor placed on the casing surface after the motor has run for a sufficient period and all components have reached thermal equilibrium.

[0081] For example, a set of test conditions is set as follows: input power 5kW, speed 300rpm. After the motor runs continuously for 30 minutes until the casing temperature fluctuates by no more than 0.5℃ within 5 minutes, the casing temperature at this time is recorded as 65℃, thus forming a set of sample data {5, 300, 65}. By constructing a sample dataset, the true temperature rise characteristics of the motor under different thermo-coupling conditions can be comprehensively reflected.

[0082] In step S3012, the input electric power and rotation speed in the operating parameters of each sample are taken as two independent variables, and the steady-state temperature of the outer shell surface is taken as the dependent variable to construct the basic form of the multiple linear regression equation. The basic form of the multiple linear regression equation is a mathematical expression describing the linear mapping relationship between independent and dependent variables. Specifically, in this equation, the input electrical power and rotational speed are defined as two independent variables, denoted as p and n respectively, which together affect the heating and cooling process of the motor; the steady-state temperature of the outer casing surface is defined as the dependent variable, denoted as T. steady The multiple linear regression equation can be expressed as T steady =β0 + β1p + β2n + ∈, where β0 is the constant intercept term, representing the reference temperature rise under the basic ambient temperature or zero load, β1 and β2 are the regression coefficients corresponding to the input electric power and speed, respectively, characterizing the degree of influence of the unit increase in electric power or the unit change in speed on the steady-state temperature of the shell, and ∈ is the random error term.

[0083] In step S3013, the least squares estimation method is used to calculate the regression coefficients and constant intercept terms that minimize the sum of squared residuals between the predicted value of the dependent variable and the actual observed temperature value based on all sample operating parameters. The least squares estimation method is used to determine the best estimates of the unknown parameters in the regression equation, and to find a set of regression coefficients (β) through mathematical derivation. 1, The constant intercept term (β2) and constant intercept term (β0) make the sum of squares of the difference (i.e., residuals) between the predicted value of the dependent variable calculated by the model and the actual observed temperature value collected above reach the global minimum.

[0084] Specifically, first, construct the residual sum of squares function S = ∑(T observed -T predicted ) 2 Then, we take the partial derivatives of this function with respect to β0, β1, and β2 respectively, and set the partial derivatives to zero, thus obtaining the normal equation system. Solving this equation system yields a unique parametric solution.

[0085] For example, if the sample set contains 100 sets of data, the algorithm will traverse all data points and adjust the coefficients until the sum of the squares of the vertical distances between the predicted curve and all measured points is minimized, thereby eliminating the interference of individual outlier measurements and ensuring that the model is the best fit in a statistical sense.

[0086] In step S3014, the regression coefficients and constant intercept terms are substituted into the basic form of the multiple linear regression equation to obtain the multiple linear regression equation.

[0087] In this embodiment, the calculated regression coefficients and constant intercept terms in their specific numerical form are filled into the corresponding positions in the basic form of the constructed equation, thereby generating a specific, executable mathematical equation. This equation no longer contains unknown parameters but becomes a definite functional relationship that can directly receive real-time input electrical power and rotational speed values ​​and output the corresponding steady-state temperature prediction value.

[0088] For example, if we calculate β0 = 25, β1 = 8.5, and β2 = -0.02, then the final multiple linear regression equation is T. steady =25 + 8.5p - 0.02n. The multiple linear regression equation can be deployed in the controller of the hub motor to quickly estimate the theoretical steady-state temperature under current operating conditions during real-time operation.

[0089] In this way, by acquiring sample operating parameters covering multiple operating conditions, the sufficiency and diversity of model training data are ensured, avoiding the one-sidedness of modeling under a single operating condition. Modeling with input electrical power and speed as independent variables and the steady-state casing temperature as the dependent variable accurately captures the combined influence of the motor's main heat sources and heat dissipation factors on temperature. Furthermore, the least squares estimation method is used to solve for the regression coefficients and constant intercept term, mathematically ensuring that the sum of squared residuals between the predicted and actual observed values ​​is minimized, eliminating the influence of random measurement noise, and making the obtained regression coefficients statistically optimal. Finally, the determined parameters are substituted into the equation form to generate a high-confidence multiple linear regression equation. This construction method not only improves the accuracy of steady-state temperature prediction but also enhances the model's generalization ability under different load and speed conditions, making subsequent temperature deviation calculations based on the predicted values ​​more reliable, thereby enabling more accurate identification of abnormal hot spots and determination of the heat dissipation monitoring status level.

[0090] Preferably, in step S40, controlling the cooling fan corresponding to the hub motor to deliver airflow for heat dissipation according to the heat dissipation monitoring status level includes: In step S401, the target speed of the cooling fan is determined according to the heat dissipation monitoring status level corresponding to the current acquisition time. Each heat dissipation monitoring status level is pre-associated with a fan speed level. The target speed setting can be the specific rotational speed level that the cooling fan needs to reach under a specific control strategy. This is used to dynamically adjust the airflow based on the motor's thermal state to achieve precise heat dissipation. The correlation between each heat dissipation monitoring status level and the fan speed setting is determined through a preset mapping table or lookup algorithm, which is stored in the control system's memory.

[0091] For example, if the thermal monitoring status is divided into four levels—normal, caution, warning, and danger—then the fan speeds of four speed settings—stop, low speed (e.g., 1000 RPM), medium speed (e.g., 2000 RPM), and high speed (e.g., 3000 RPM)—can be pre-associated. When the system determines that it is currently at the warning level, it directly locks the medium speed as the target speed setting by looking up a table. This hierarchical association ensures that the thermal response matches the degree of thermal risk, avoiding energy waste caused by excessive heat dissipation or equipment damage caused by insufficient heat dissipation.

[0092] Abstract thermal state assessment results can be transformed into specific execution command parameters. Based on the aforementioned determined heat dissipation monitoring state level, the control system quickly locks the matching fan operating speed by retrieving pre-stored configuration data. This achieves logical connection between thermal monitoring logic and mechanical control actions.

[0093] In step S402, the target speed gear is converted into the duty cycle parameter of the pulse width modulation signal. The duty cycle parameter is output to the control terminal of the cooling fan through the drive circuit, so that the cooling fan delivers air at the speed corresponding to the target speed gear.

[0094] The pulse width modulation (PWM) signal is a digital signal that controls the average voltage output by adjusting the pulse width. The duty cycle parameter is the ratio of the high-level duration to the total cycle time within one signal period, typically ranging from 0% to 100%. The duty cycle parameter is calculated based on the ratio between the target speed setting and the fan's maximum rated speed, and is used to precisely control the average power applied to the fan motor. The drive circuit is the power amplification stage connecting the controller and the cooling fan, used to convert weak logic level signals into a high-current signal capable of driving the fan motor coils.

[0095] For example, the control system internally stores conversion formulas or lookup tables between speed gears and duty cycles. For instance, if the target speed gear corresponds to 60% of the maximum speed, the PWM waveform parameters with a duty cycle of 60% are calculated. This duty cycle parameter is output to the control terminal of the cooling fan via the drive circuit. The electronic commutation circuit inside the fan adjusts the energizing time of the coil according to the received duty cycle, thereby causing the fan blades to rotate at the speed corresponding to the target speed gear.

[0096] Based on the current heat dissipation monitoring status level, the target speed setting is first determined, establishing a logical mapping between thermal risk and mechanical response. This target speed setting is then converted into the duty cycle parameter of a pulse width modulation signal, realizing the quantitative conversion of analog speed requirements into digital control commands. Subsequently, the duty cycle parameter is output to the control terminal of the cooling fan via the drive circuit, completing the final execution of the energy transfer from electricity to wind power. This not only ensures the continuous adjustability of the fan speed under different heat loads but also improves the system's response speed and adjustment accuracy. For example, when a slight temperature rise is detected, the system automatically switches to a low-duty-cycle, low-speed setting to save energy; while when severe overheating is detected, it immediately switches to a high-duty-cycle, high-speed setting for intensive cooling, thereby optimizing the vehicle's energy management efficiency while ensuring the safe operation of the hub motor.

[0097] See Figure 4 As shown in the figure, this disclosure also provides a hub motor heat dissipation monitoring system 400, wherein the hub motor heat dissipation monitoring system 400 includes: The acquisition module 410 is configured to acquire temperature data continuously collected by temperature sensors arranged on three components of the hub motor, namely the stator winding, bearing end cover and housing surface, according to a preset sampling period, and obtain temperature time series data corresponding to each component. The first determining module 420 is configured to perform a weighted moving average calculation on the temperature values ​​corresponding to the temperature time series data within the preset time period for each component, using the temperature values ​​of historical sampling times with a preset time period away from the current sampling time as the moving average, to obtain the weighted moving temperature value of each component at the current sampling time. The further away from the current sampling time is, the smaller the weight value of the weighted moving average. The second determining module 430 is configured to determine the heat dissipation monitoring status level of the hub motor based on the weighted moving temperature value corresponding to each component at the current sampling time and the real-time operating parameters of the hub motor. The control module 440 is configured to control the cooling fan corresponding to the hub motor to deliver airflow for heat dissipation based on the heat dissipation monitoring status level.

[0098] In one alternative embodiment, the first determining module 420 is configured as follows: Using the temperature values ​​of historical sampling times with a preset time interval from the current sampling time as the moving average, and based on the temperature values ​​corresponding to each sampling time within the preset time interval and the corresponding weight values, calculate the weighted moving average value corresponding to each sampling time within the preset time interval for each component. The weighted moving average value of each sampling moment within the preset time period is used as the observation value to establish the state equation of the hub motor thermal model. The state equation is used to describe the linear relationship between the temperature state at any sampling moment and the temperature state at the corresponding previous sampling moment and the motor power loss. The prediction step in the Kalman filter algorithm is used to predict the prior temperature estimate at the current sampling time based on the state estimate and the corresponding state equation at the previous sampling time, and to calculate the covariance matrix of the prior estimation error. The update step in the Kalman filter algorithm is adopted. The weighted moving average value at the current sampling time is used as the measurement value to calculate the Kalman gain. The temperature value sampled at the current sampling time of each component is used to correct the corresponding prior temperature estimate to obtain the posterior temperature estimate at the current sampling time of each component. The posterior estimate of the temperature at the current sampling time of each component is used as the representative temperature value at the current sampling time, and the calculated weighted moving average value is replaced to obtain the weighted moving temperature value of each component at the current sampling time.

[0099] In one alternative embodiment, the first determining module 420 is configured as follows: A multi-source temperature observation vector is formed by using the temperature value sampled at the current sampling time of each component and the corresponding prior temperature estimate. The joint probability distribution modeling method in Bayesian estimation is used to model the measurement error of each temperature sensor as a Gaussian distribution, and the joint likelihood function of the multi-source observation vector under the preset real temperature conditions is calculated. According to Bayes' theorem, the joint likelihood function is multiplied by the prior probability density function of the temperature state, and then normalized by a normalization constant to obtain the posterior probability density function corresponding to the temperature state. The temperature value corresponding to the maximum posterior probability is extracted from the posterior probability density function. The extracted temperature value is used as the substitute input of the measured value in the Kalman filter algorithm to correct the prior temperature estimate and obtain the posterior temperature estimate of each component at the current sampling time.

[0100] In one alternative embodiment, the first determining module 420 is configured as follows: The weighted moving average is set as the alternative input for the measurement at the current sampling time in the Kalman filter algorithm. This alternative input is associated with a pre-stored measurement noise covariance matrix and replaces the original sensor output measurement data. Read the error covariance matrix of the posterior temperature estimate obtained from the previous sampling time, and combine it with the state transition matrix of the system thermal model to calculate the prior temperature estimate and its corresponding prior estimate error covariance matrix at the current sampling time. Calculate the Kalman gain matrix for each component based on the prior estimation error covariance matrix and the measurement noise covariance matrix corresponding to each component at the current sampling time. Multiply the Kalman gain matrix by the residual between the substitute input and the prior temperature estimate, and then add the product to the corresponding prior temperature estimate to obtain the posterior temperature estimate for each component at the current sampling time.

[0101] In one alternative embodiment, the second determining module 430 is configured to: The input power and speed values ​​at the current sampling time in the real-time operating parameters of the hub motor are substituted into the multiple linear regression equation to calculate the predicted steady-state temperature. The multiple linear regression equation is established using the linear regression analysis method, with the input power and speed of the hub motor as two independent variables and the steady-state temperature of the outer shell surface of the hub motor as the dependent variable. The weighted moving temperature value corresponding to each component at the current sampling time is compared with the steady-state temperature prediction value to calculate the temperature deviation value of the corresponding component. If the temperature deviation value corresponding to any component exceeds the preset deviation tolerance limit, the location corresponding to that component is marked as an abnormal hot spot location; The weighted moving temperature value corresponding to each component at the current sampling time is compared with a plurality of preset temperature thresholds to determine the initial temperature level corresponding to each component. If the location corresponding to any component is marked as the abnormal hot spot location, a penalty additional weight greater than 1 is applied to the corresponding weighted moving temperature value. The highest initial temperature level among the three components at the current acquisition time is determined as the heat dissipation monitoring status level of the hub motor.

[0102] In one alternative approach, the multiple linear regression equation is constructed as follows: The sample operating parameters of the hub motor under multiple different operating conditions are obtained to construct a sample dataset. Each set of sample operating parameters includes the value of input electric power, the value of rotational speed, and the actual observed temperature value when the outer shell surface reaches a steady state. Using the input electric power and rotational speed in each sample's operating parameters as two independent variables and the steady-state temperature of the outer shell surface as the dependent variable, the basic form of a multiple linear regression equation is constructed. Using the least squares estimation method, regression coefficients and constant intercept terms that minimize the sum of squared residuals between the predicted value of the dependent variable and the actual observed temperature value are calculated based on all the sample operating parameters. Substituting the regression coefficients and the constant intercept term into the basic form of the multiple linear regression equation yields the multiple linear regression equation.

[0103] In an alternative embodiment, the control module 440 is configured as follows: Based on the heat dissipation monitoring status level corresponding to the current collection time, the target speed level of the cooling fan is determined, wherein each heat dissipation monitoring status level is pre-associated with a fan speed level. The target speed setting is converted into a duty cycle parameter of a pulse width modulation signal. The duty cycle parameter is output to the control terminal of the cooling fan through the drive circuit, so that the cooling fan delivers air at the speed corresponding to the target speed setting.

[0104] Specific limitations regarding the hub motor heat dissipation monitoring system can be found in the limitations of the hub motor heat dissipation monitoring method described above, and will not be repeated here. Each module in the aforementioned hub motor heat dissipation monitoring system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0105] This disclosure also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the methods described in the foregoing embodiments.

[0106] This disclosure also provides an electronic device, including: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the steps of any of the methods described in the foregoing embodiments.

[0107] Figure 5 The hub motor heat dissipation monitoring system 100 shown includes a processor 1001 and a memory 1003. The processor 1001 and the memory 1003 are connected, for example, via a bus 1002. Optionally, the hub motor heat dissipation monitoring system 100 may further include a communication component, which can be used for data interaction between the device 100 and other devices, such as sending or receiving data. It should be noted that in actual scheduling, the communication component is not limited to one, and the structure of this hub motor heat dissipation monitoring system 100 does not constitute a limitation on the embodiments of this application.

[0108] Processor 1001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 1001 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0109] Bus 1002 may include a pathway for transmitting information between the aforementioned components. Bus 1002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 1002 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 5 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0110] The memory 1003 may be ROM (Read Only Memory) or other types of static storage devices capable of storing static information and instructions, RAM (Random Access Memory) or other types of dynamic storage devices capable of storing information and instructions, or EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media, other magnetic storage devices, or any other medium capable of carrying or storing program code and capable of being read by a computer, without limitation herein.

[0111] The memory 1003 is used to store program code for executing the embodiments of this disclosure, and its execution is controlled by the processor 1001. The processor 1001 is used to execute the program code stored in the memory 1003 to implement the steps shown in the aforementioned embodiments of the hub motor heat dissipation monitoring method.

[0112] This disclosure also provides a computer-readable storage medium storing program code. When the program code is executed by a processor, it can implement the steps and corresponding content of the aforementioned hub motor heat dissipation monitoring method embodiment.

[0113] The preferred embodiments of the present disclosure have been described in detail above with reference to the accompanying drawings. However, the present disclosure is not limited to the specific details of the above embodiments. Within the scope of the technical concept of the present disclosure, various changes, modifications, substitutions and variations can be made to these embodiments, and all such changes, modifications, substitutions and variations fall within the protection scope of the present disclosure.

[0114] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction, and such combinations should also be considered as part of this disclosure. To avoid unnecessary repetition, this disclosure will not further describe the various possible combinations. The technical scope of this application is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A method for monitoring heat dissipation of a hub motor, characterized in that, The method includes: Temperature sensors arranged on three components of the hub motor—stator winding, bearing end cover, and outer shell surface—are used to continuously collect temperature data according to a preset sampling period to obtain temperature time series data corresponding to each component. For the temperature time series data of each component, the temperature value of the historical sampling time with a preset time distance from the current sampling time is used as the moving average. The temperature value corresponding to the temperature time series data within the preset time distance is calculated by weighted moving average to obtain the weighted moving temperature value of each component at the current sampling time. The further away from the current sampling time, the smaller the weight value of the weighted moving average. The heat dissipation monitoring status level of the hub motor is determined based on the weighted moving temperature value of each component at the current sampling time and the real-time operating parameters of the hub motor. Based on the heat dissipation monitoring status level, the cooling fan corresponding to the hub motor is controlled to deliver airflow for heat dissipation.

2. The method according to claim 1, characterized in that, The step involves using the temperature values ​​of historical sampling times within a preset time interval from the current sampling time as moving averages. A weighted moving average is then calculated for the temperature values ​​corresponding to the temperature time series data within the preset time interval to obtain the weighted moving temperature value of each component at the current sampling time. This includes: Using the temperature values ​​of historical sampling times with a preset time interval from the current sampling time as the moving average, and based on the temperature values ​​corresponding to each sampling time within the preset time interval and the corresponding weight values, calculate the weighted moving average value corresponding to each sampling time within the preset time interval for each component. The weighted moving average value of each sampling moment within the preset time period is used as the observation value to establish the state equation of the hub motor thermal model. The state equation is used to describe the linear relationship between the temperature state at any sampling moment and the temperature state at the corresponding previous sampling moment and the motor power loss. The prediction step in the Kalman filter algorithm is used to predict the prior temperature estimate at the current sampling time based on the state estimate and the corresponding state equation at the previous sampling time, and to calculate the covariance matrix of the prior estimation error. The update step in the Kalman filter algorithm is adopted. The weighted moving average value at the current sampling time is used as the measurement value to calculate the Kalman gain. The temperature value sampled at the current sampling time of each component is used to correct the corresponding prior temperature estimate to obtain the posterior temperature estimate at the current sampling time of each component. The posterior estimate of the temperature at the current sampling time of each component is used as the representative temperature value at the current sampling time, and the calculated weighted moving average value is replaced to obtain the weighted moving temperature value of each component at the current sampling time.

3. The method according to claim 2, characterized in that, The step of correcting the corresponding prior temperature estimate by using the temperature value sampled at the current sampling time of each component to obtain the posterior temperature estimate at the current sampling time of each component includes: A multi-source temperature observation vector is formed by using the temperature value sampled at the current sampling time of each component and the corresponding prior temperature estimate. The joint probability distribution modeling method in Bayesian estimation is used to model the measurement error of each temperature sensor as a Gaussian distribution, and the joint likelihood function of the multi-source observation vector under the preset real temperature conditions is calculated. According to Bayes' theorem, the joint likelihood function is multiplied by the prior probability density function of the temperature state, and then normalized by a normalization constant to obtain the posterior probability density function corresponding to the temperature state. The temperature value corresponding to the maximum posterior probability is extracted from the posterior probability density function. The extracted temperature value is used as the substitute input of the measured value in the Kalman filter algorithm to correct the prior temperature estimate and obtain the posterior temperature estimate of each component at the current sampling time.

4. The method according to claim 2, characterized in that, The update step in the Kalman filter algorithm uses the weighted moving average of the current sampling time as the measured value, calculates the Kalman gain, and corrects the corresponding prior temperature estimate using the temperature value sampled at the current sampling time for each component, to obtain the posterior temperature estimate for each component at the current sampling time, including: The weighted moving average is set as the alternative input for the measurement at the current sampling time in the Kalman filter algorithm. This alternative input is associated with a pre-stored measurement noise covariance matrix and replaces the original sensor output measurement data. Read the error covariance matrix of the posterior temperature estimate obtained from the previous sampling time, and combine it with the state transition matrix of the system thermal model to calculate the prior temperature estimate and its corresponding prior estimate error covariance matrix at the current sampling time. Calculate the Kalman gain matrix for each component based on the prior estimation error covariance matrix and the measurement noise covariance matrix corresponding to each component at the current sampling time. Multiply the Kalman gain matrix by the residual between the substitute input and the prior temperature estimate, and then add the product to the corresponding prior temperature estimate to obtain the posterior temperature estimate for each component at the current sampling time.

5. The method according to claim 1, characterized in that, The step of determining the heat dissipation monitoring status level of the hub motor based on the weighted moving temperature value of each component at the current sampling time and the real-time operating parameters of the hub motor includes: The input power and speed values ​​at the current sampling time in the real-time operating parameters of the hub motor are substituted into the multiple linear regression equation to calculate the predicted steady-state temperature. The multiple linear regression equation is established using the linear regression analysis method, with the input power and speed of the hub motor as two independent variables and the steady-state temperature of the outer shell surface of the hub motor as the dependent variable. The weighted moving temperature value corresponding to each component at the current sampling time is compared with the steady-state temperature prediction value to calculate the temperature deviation value of the corresponding component. If the temperature deviation value corresponding to any component exceeds the preset deviation tolerance limit, the location corresponding to that component is marked as an abnormal hot spot location; The weighted moving temperature value corresponding to each component at the current sampling time is compared with a plurality of preset temperature thresholds to determine the initial temperature level corresponding to each component. If the location corresponding to any component is marked as the abnormal hot spot location, a penalty additional weight greater than 1 is applied to the corresponding weighted moving temperature value. The highest initial temperature level among the three components at the current acquisition time is determined as the heat dissipation monitoring status level of the hub motor.

6. The method according to claim 5, characterized in that, The multiple linear regression equation is constructed as follows: The sample operating parameters of the hub motor under multiple different operating conditions are obtained to construct a sample dataset. Each set of sample operating parameters includes the value of input electric power, the value of rotational speed, and the actual observed temperature value when the outer shell surface reaches a steady state. Using the input electric power and rotational speed in each sample's operating parameters as two independent variables and the steady-state temperature of the outer shell surface as the dependent variable, the basic form of a multiple linear regression equation is constructed. Using the least squares estimation method, regression coefficients and constant intercept terms that minimize the sum of squared residuals between the predicted value of the dependent variable and the actual observed temperature value are calculated based on all the sample operating parameters. Substituting the regression coefficients and the constant intercept term into the basic form of the multiple linear regression equation yields the multiple linear regression equation.

7. The method according to any one of claims 1-6, characterized in that, The step of controlling the cooling fan corresponding to the hub motor to deliver airflow for heat dissipation based on the heat dissipation monitoring status level includes: Based on the heat dissipation monitoring status level corresponding to the current collection time, the target speed level of the cooling fan is determined, wherein each heat dissipation monitoring status level is pre-associated with a fan speed level. The target speed setting is converted into a duty cycle parameter of a pulse width modulation signal. The duty cycle parameter is output to the control terminal of the cooling fan through the drive circuit, so that the cooling fan delivers air at the speed corresponding to the target speed setting.

8. A hub motor heat dissipation monitoring system, characterized in that, The system includes: The acquisition module is configured to acquire temperature data continuously collected by temperature sensors arranged on three components of the hub motor: the stator winding, the bearing end cover, and the outer surface of the housing, according to a preset sampling period, and obtain temperature time series data corresponding to each component. The first determining module is configured to perform a weighted moving average calculation on the temperature values ​​corresponding to the temperature time series data within the preset time period for each component, using the temperature values ​​of historical sampling times with a preset time period away from the current sampling time as the moving average, to obtain the weighted moving temperature value of each component at the current sampling time. The further away from the current sampling time is, the smaller the weight value of the weighted moving average. The second determining module is configured to determine the heat dissipation monitoring status level of the hub motor based on the weighted moving temperature value corresponding to each component at the current sampling time and the real-time operating parameters of the hub motor. The control module is configured to control the cooling fan corresponding to the hub motor to deliver airflow for heat dissipation based on the heat dissipation monitoring status level.

9. An electronic device, characterized in that, include: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the steps of the method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores program code that is executed by a processor to implement the steps of the method according to any one of claims 1-7.

Images