Permanent Magnet Motor Current Control System Based on Prediction Compensation

Abstract

The present invention relates to the technical field of motor control, specifically a permanent magnet motor current control system based on predictive compensation. The system includes a thermal field identification module, a response offset identification module, a correction trajectory generation module, a limit range decision module, and a compensation instruction output module. In the present invention, a dynamic prediction and correction mechanism is established based on the thermal field characteristics and the trend of response errors to achieve real-time adjustment of the current trajectory of the permanent magnet motor. By constructing a three-dimensional space thermal load surface with the stator temperature rise curve, the temperature change rate of the permanent magnet, and the driving torque fluctuation, the transient influence of the thermal condition change on the magnetic performance of the motor is reflected. Overall, a multi-dimensional prediction and compensation path is constructed through the interaction between multiple thermal field parameters, electromagnetic response variables, and time sequence characteristic points, improving the response accuracy and control stability of the permanent magnet motor under dynamic conditions, reducing the influence of magnetic performance attenuation caused by thermal drift, and at the same time improving the tracking consistency under current trajectory disturbance.

Description

Technical Field

[0001] The present invention relates to the technical field of motor control, and particularly to a permanent magnet motor current control system based on predictive compensation. Background Art

[0002] The technical field of motor control mainly involves the regulation and control of the operating states of various motors, specifically including speed control, torque control, current control, voltage control, etc. In this field, input electrical signals are usually modulated by control devices to control the input current or voltage of the motor, thereby achieving predetermined motion behaviors or output characteristics. This technology is widely used in occasions with high requirements for motion performance, such as automation equipment, industrial robots, new energy vehicles, wind power generation systems, and numerical control machine tools. Motor control technology continues to develop towards high responsiveness, high precision, low energy consumption, and intelligence. The core methods include algorithms such as vector control, direct torque control, and space vector pulse width modulation, as well as the design of closed-loop feedback structures of current loops, speed loops, and position loops that cooperate with them.

[0003] Among them, the core of the permanent magnet motor current control system lies in the precise control of the stator winding current of permanent magnet motors such as permanent magnet synchronous motors or brushless DC motors. Its main purpose is to adjust the magnitude and direction of the drive current to achieve linear control of the motor output torque and optimize the dynamic response. This system is widely used in fields such as electric vehicle traction drive, industrial servo systems, aerospace equipment, high-performance fans, and compressors, and plays a key role in improving system efficiency, reducing electromagnetic noise and vibration, and extending the service life of equipment.

[0004] Traditional control systems lack an in-depth modeling and processing mechanism for the coupling relationship between the dynamic characteristics of the thermal field and the electromagnetic response offset. During continuous operation, the temperature rise of the stator winding and the temperature rise of the permanent magnet will cause non-linear attenuation of the magnetic properties. Traditional systems do not perform real-time identification and compensation for the thermally induced magnetic property reduction state, which is likely to cause current control offset and unstable torque output. The current control path is mostly based on fixed parameter settings and static feedback adjustment. In scenarios of rapid dynamic loads or high-speed switching, the response lag problem is serious, and it is difficult to effectively handle large fluctuations within a short period, resulting in oscillations, increased energy consumption, and decreased control accuracy during high-frequency operation of the motor, affecting the execution efficiency and service life of the overall system. Summary of the Invention

[0005] The purpose of the present invention is to solve the deficiencies existing in the prior art, and to propose a permanent magnet motor current control system based on predictive compensation.

[0006] To achieve the above purpose, the present invention adopts the following technical solution: A permanent magnet motor current control system based on predictive compensation, the system includes:

[0007] The thermal field recognition module obtains the operating state parameters of the permanent magnet motor, constructs a spatial coordinate set at each point for the stator temperature rise curve slope, temperature change rate, and torque fluctuation value, determines whether the trajectory undergoes a curvature inversion change, extracts the coordinate points as magnetic property reduction characteristic points, and generates a magnetic reduction characteristic trajectory;

[0008] Based on the magnetic reduction characteristic trajectory, the response offset recognition module determines whether there is a reverse difference in the change phase between the target current trend curve and the actual sampling curve before and after the magnetic reduction characteristic points in terms of cycles. If there is a reverse change, the relative drift amount is calculated, and a predicted response drift window is established;

[0009] The correction trajectory generation module calls the predicted response drift window to determine whether there are segments with continuous trend offsets. If so, the relative vector between the actual sampled current value and the target value is differentially calculated point by point, and the offset amount superposition trajectory of the corresponding cycle correction path is obtained to generate a predicted corrected current trajectory;

[0010] Based on the predicted corrected current trajectory, the amplitude limit interval decision module combines the amplitude upper limit point in the current correction trajectory and the inductance change rate within the corresponding time interval to constraint and clip the maximum amplitude of the correction trajectory, generating a dynamic amplitude-limited current trajectory.

[0011] The improvements of the present invention are that the magnetic reduction characteristic trajectory includes the time series of magnetic property change critical points, the peak path coordinates of the thermal load mapping surface, and the spatial distribution of curvature inversion points; the predicted response drift window includes the current amplitude deviation data set, the current trend phase difference section, and the response drift duration range; the predicted corrected current trajectory includes the current vector offset path, the amplitude superposition sequence, and the trend shift nodes; and the dynamic amplitude-limited current trajectory includes the amplitude limit upper boundary, the edge points of the clipped section, and the inductance amplitude change control calibration value.

[0012] The improvements of the present invention are that the thermal field recognition module includes:

[0013] The thermal parameter extraction sub-module obtains the operating state parameters of the permanent magnet motor, including the stator winding temperature rise curve, the surface temperature change rate of the permanent magnet, and the drive torque fluctuation range per unit time. The points of the three data items are synchronously paired according to the time period, and the time segments with the temperature rise slope inflection point, the sudden increase in the temperature change rate, and the sudden jump in the torque fluctuation amplitude are screened, and the fluctuation amplitude is jointly normalized to generate the periodic thermal response rate value;

[0014] Based on the periodic thermal response rate value, the thermal field construction sub-module constructs a spatial point set with the stator temperature rise curve slope as the X-axis, the permanent magnet temperature change rate as the Y-axis, and the torque fluctuation amplitude as the Z-axis, calculates the surface connection angle and gradient change rate between each point within each cycle, identifies the continuous change range and extreme value distribution range of the gradient curvature, and establishes the thermal load mapping surface trend value;

[0015] Based on the thermal load mapping surface trend value, the critical trajectory recognition sub-module extracts the coordinate sequence of the intersection points of the principal curvature poles and the time axis within a unit period, determines whether there is a trajectory segment with direction reversal and amplitude mutation within two consecutive periods in the sequence, screens out the spatial coordinates with prominent changes as the magnetic property change recognition points, and generates the magnetic reduction characteristic trajectory.

[0016] The improvement of the present invention is that the response offset recognition module includes:

[0017] Based on the magnetic reduction characteristic trajectory, the current trend determination sub-module obtains the target current vector amplitude, the actual sampling current change trend and the time position of the characteristic points in the magnetic reduction characteristic trajectory, calculates the change phase difference between the target current and the actual current in the two periods before and after the characteristic points respectively, determines whether the change directions of the two curves before and after the characteristic point period show a reverse trend. If there is a reverse relationship, the position of the phase turning point is recorded to generate a reverse trend recognition position sequence;

[0018] The deviation calculation sub-module calls the reverse trend recognition position sequence, collects the target current amplitude curve and the actual sampling current amplitude curve in the corresponding period, calculates the amplitude difference and the occurrence time difference between each corresponding point along the time axis, constructs a set of offset distance vectors within the same period, combines the rotor angular velocity change curve, uses the fluctuation amplitude of the angular velocity at the corresponding deviation point as a scaling factor to adjust the amplitude difference, and uses the formula:

[0019] ;

[0020] Calculate the response relative drift amount through operation;

[0021] Wherein, represents the response relative drift amount, represents the target current amplitude at the th point, represents the actual sampling current amplitude at the th point, represents the time difference at the th point, represents the rotor angular velocity corresponding to the th point, is the total number of paired points participating in the calculation;

[0022] Based on the response relative drift amount, the time window construction sub-module combines the speed volatility of the corresponding section in the rotor angular velocity change curve, sets the angular velocity fluctuation threshold to twice the standard deviation of the speed median, screens out the sections where the response relative drift amount is higher than the threshold, extracts the corresponding time segments, and arranges and combines them in chronological order to establish a predicted response drift window.

[0023] The improvement of the present invention is that the correction trajectory generation module includes:

[0024] The slope extraction sub-module obtains the predicted response drift window, extracts the component slope of each data point on the time axis, the amplitude difference between consecutive points, and the corresponding drift time value based on the target current vector trajectory within the corresponding time period, calculates the amplitude change range of the component slope within a unit period based on the amplitude change rate and time offset ratio between adjacent points, and generates the current slope change amplitude.

[0025] The trend recognition sub-module calls the current slope change amplitude, detects whether there are three consecutive monotonic change sequences within the slope change section, if so, determines whether the fluctuation direction has a double reversal, extracts the corresponding start and end drift time intervals according to the position index of the reversal section, establishes the time range of the offset section, and obtains the trend offset time interval.

[0026] The trajectory generation sub-module calls the target current vector trajectory and the actual sampled current value within the interval according to the trend offset time interval, performs point-by-point vector difference on the two sets of current data at the corresponding time points, and sequentially superimposes the difference results, establishes a correction path vector sequence and supplements the head and tail smoothing sections to generate a predicted correction current trajectory.

[0027] The improvement of the present invention is that the amplitude limiting interval decision module includes:

[0028] The inductance fluctuation recognition sub-module obtains the stator inductance trend change curve within the current period based on the predicted correction current trajectory, calculates the inductance change slope value and the range difference of the change interval within three consecutive periods, determines whether the maximum slope value exceeds the inductance stability bandwidth threshold, and if the judgment is established, records the section as a high-variation section and generates an inductance abnormal variation interval.

[0029] The current peak positioning sub-module calls the inductance abnormal variation interval, obtains the amplitude distribution sequence of the predicted correction current trajectory, extracts the point with the maximum amplitude within the variation interval, and calculates the amplitude amplification ratio by combining the inductance slope value at the time index corresponding to the point with the maximum amplitude and the average inductance value of the previous period to generate the current amplitude slope ratio.

[0030] The clipping boundary generation sub-module sets the control boundary point and calibrates the section response time according to the current amplitude slope ratio, in combination with the point with the maximum amplitude and the corresponding inductance change rate in the predicted correction current trajectory, using the formula:

[0031] ;

[0032] Performs operations to obtain the clipping boundary value, compresses the amplitude of the peak section in the correction trajectory, and establishes a dynamic amplitude-limited current trajectory.

[0033] Wherein, Represents the cropping boundary value, Represents the maximum amplitude point in the predicted corrected current trajectory, Represents the target limit setting value, Represents the rate of change of inductance at the current moment, Represents the response duration at which the amplitude peak is located, Represents the difference between the current correction amplitudes before and after.

[0034] The improvement of the present invention is that the system further includes:

[0035] The compensation instruction output module calls the dynamic limit current trajectory, collects the sampling error bias value in the current cycle and the current response delay value generated by the PWM dead time, determines whether the error bias value shows an amplification trend within the time corresponding to the peak of the limit trajectory wave, and if so, accumulatively offsets and corrects the bias value and the delay response value, adjusts the command vector, and generates a predicted compensation current execution instruction;

[0036] The predicted compensation current execution instruction includes an end command vector sequence, a response delay correction amount, and an execution output offset structure.

[0037] The improvement of the present invention is that the compensation instruction output module includes:

[0038] The bias trend recognition sub-module calls the dynamic limit current trajectory, obtains the time period corresponding to the peak in the dynamic limit current trajectory, collects the sampling error bias value sequence in the current cycle, performs a moving average process on the sequence, recognizes the mean difference change trend of three consecutive points, determines whether there is a continuously increasing segment, and if the determination is established, extracts the maximum increment value of the corresponding segment to generate a bias amplification trend value;

[0039] The delay correction amount generation sub-module calls the bias amplification trend value, collects the PWM dead time in the current cycle, the current response delay change value within the time period, and the sampling current change rate at the end of the limit trajectory, calculates the relative amplification amount of the current response delay and the PWM dead time compensation ratio respectively, and uses the formula:

[0040] ;

[0041] Obtain the correction offset through calculation;

[0042] Wherein, is the correction offset, represents the amplification increment of the sampling bias, represents the PWM dead time duration, represents the current response slope, represents the current change amplitude, represents the response peak duration;

[0043] The vector instruction adjustment submodule performs amplitude supplementation correction on each point in the last data sequence of the limited current trajectory according to the correction offset, uses the supplemented current trajectory as the corrected execution current input, combines the vector angle distribution interval with the instruction angle of the previous cycle, calculates the corrected vector angle sequence, and establishes the predicted compensation current execution instruction.

[0044] Compared with the prior art, the advantages and positive effects of the present invention are:

[0045] In the present invention, a dynamic prediction and correction mechanism is established based on the thermal field characteristics and the response error trend to achieve real-time adjustment of the permanent magnet motor current trajectory. A three-dimensional thermal load surface is constructed with the help of the stator temperature rise curve, the permanent magnet temperature change rate and the driving torque fluctuation, and the magnetic performance reduction point is extracted to reflect the transient impact of thermal conditions on the motor magnetic performance. Combined with the response phase change trend of the target current and the actual current, the relative drift is derived and the synchronous offset window is constructed in conjunction with the rotor angular velocity to improve the dynamic adaptability of the current prediction capability. By performing point-by-point differential processing on the target current vector within the error window, the drift offset superposition trajectory is obtained, and then a correction path is formed to ensure the continuity of current control. In order to improve the response stability, the stator inductance change trend and the maximum slope of the correction path are collected, the dynamic amplitude clipping boundary is set, the risk of control instability caused by inductance abnormality is effectively constrained, the current response delay and error offset caused by PWM dead time are considered, and correction compensation is superimposed at the end of the limiting trajectory to enhance the responsiveness of the compensation command to rapid disturbances. The overall multi-dimensional predictive compensation path is constructed through the interaction between multiple thermal field parameters, electromagnetic response variables and timing feature points to improve the response accuracy and control stability of the permanent magnet motor under dynamic conditions, reduce the influence of magnetic performance attenuation caused by thermal drift, and at the same time improve the tracking consistency under current trajectory disturbance, and enhance the linear controllability and high-frequency response characteristics of the output torque. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 is a system flow chart of the present invention;

[0047] Figure 2 It is a flow chart of the thermal field identification module of the present invention;

[0048] Figure 3 A flow chart of a response deviation identification module of the present invention;

[0049] Figure 4 A flowchart of a modified trajectory generation module of the present invention;

[0050] Figure 5 It is a flow chart of the clipping interval decision module of the present invention;

[0051] Figure 6 This is a flow chart of the compensation instruction output module of the present invention. Detailed implementation manner

[0052] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0053] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the accompanying drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention. In addition, in the description of the present invention, the meaning of "a plurality of" is two or more unless otherwise specifically defined.

[0054] Please refer to Figure 1 , the present invention provides a technical solution: a permanent magnet motor current control system based on prediction compensation. The system includes:

[0055] The thermal field identification module obtains the operating state parameters of the permanent magnet motor, including the stator winding temperature rise curve, the surface temperature change rate of the permanent magnet and the driving torque fluctuation range per unit time. Synchronously pair the points of the three data items according to the time period, and construct a spatial coordinate set for the slope of the stator temperature rise curve, the temperature change rate and the torque fluctuation value at each point, establish a thermal load mapping surface, and judge whether the trajectory formed by the intersection of the principal curvature poles and the time axis within a unit period in the surface undergoes a curvature inversion change. If a change occurs, extract the current coordinate point as the magnetic property reduction feature point and generate a magnetic reduction feature trajectory;

[0056] The response offset identification module, based on the magnetic reduction feature trajectory, obtains the target current vector amplitude, the actual sampled current change trend and the rotor angular velocity curve within the same period, and judges whether the change phase of the target current trend curve and the actual sampled curve shows a reverse difference before and after the magnetic reduction feature point period. If there is a reverse change, calculate the relative drift amount and establish a prediction response drift window;

[0057] The correction trajectory generation module calls the prediction response drift window, extracts the component slope, amplitude change range and time drift value of the target current vector trajectory corresponding to the error window, and judges whether there is a segment with a continuous trend offset. If so, perform a point-by-point difference on the relative vector between the actual sampled current value and the target value, obtain the offset amount superposition trajectory of the corresponding period correction path, and generate a predicted correction current trajectory;

[0058] The clipping range decision module, based on the predicted and corrected current trajectory, collects the stator inductance trend change curve, the winding conductor temperature rise rate, and the maximum slope of the corrected trajectory within the current period, and determines whether the inductance change breaks through the upper limit of the inductance stability bandwidth. If so, it combines the amplitude upper limit point in the current corrected trajectory with the inductance change rate within the corresponding time interval to set the upper limit value of the constraint, performs constraint clipping on the maximum amplitude of the corrected trajectory, and generates a dynamically clipped current trajectory.

[0059] The compensation instruction output module calls the dynamically clipped current trajectory, collects the sampling error offset value in the current period and the current response delay value generated by the PWM dead time, and determines whether the error offset value shows an amplification trend within the time corresponding to the peak of the clipped trajectory. If so, it accumulates and offsets the offset value and the delay response value, adjusts the instruction vector, and generates a predicted compensation current execution instruction.

[0060] The magnetic reduction characteristic trajectory includes the time series of magnetic performance change critical points, the peak path coordinates of the thermal load mapping surface, and the spatial distribution of curvature inversion points. The predicted response drift window includes the current amplitude deviation data set, the current trend phase difference section, and the response drift duration range. The predicted corrected current trajectory includes the current vector offset path, the amplitude superposition sequence, and the trend shift nodes. The dynamically clipped current trajectory includes the upper boundary of the amplitude limit, the edge points of the clipped section, and the inductance amplitude change control calibration value. The predicted compensation current execution instruction includes the end instruction vector sequence, the response delay correction amount, and the execution output offset structure.

[0061] Please refer to Figure 2 , the thermal field identification module includes:

[0062] The thermal parameter extraction sub-module obtains the operating state parameters of the permanent magnet motor, including the stator winding temperature rise curve, the surface temperature change rate of the permanent magnet, and the drive torque fluctuation range per unit time. It synchronizes and pairs the points of the three data items according to the time period, screens the time segments with the inflection point of the temperature rise slope, the sudden increase in the temperature change rate, and the sudden jump in the torque fluctuation amplitude, and performs joint normalization processing on the fluctuation amplitude to generate the periodic thermal response rate value.

[0063] The thermal parameter extraction sub-module obtains the operating state parameters of the permanent magnet motor. First, the stator winding temperature rise curve is collected based on the thermocouple sensors embedded in the stator slots. The voltage signal is read at an interval of every 0.5 seconds and converted into a temperature value through the thermocouple linear calibration formula, and a complete temperature rise curve is constructed in combination with the time stamp. Subsequently, the surface temperature change rate of the permanent magnet is obtained through the patch-type thermistor attached to the surface of the permanent magnet. The thermistor records the temperature change once per second, and the temperature change rate per unit time is calculated based on the difference method. , such as when the temperature is , For , the change rate is ; the third item of driving torque fluctuation range is detected by a torque sensor with high-frequency sampling. During the sampling period, the maximum and minimum torques of each time period are collected and the difference is calculated as the fluctuation amplitude. For example, within to , the maximum torque is , the minimum is , then the fluctuation amplitude is . Subsequently, within a window with as the period, the above three items of data are time-stamped aligned, and the asynchronous recorded points are corrected by linear interpolation to ensure that three groups of data points with consistent time stamps are obtained within each period. Then, the point where the derivative change slope in the temperature rise curve appears to reverse is used as the temperature rise slope inflection point. For example, before the slope gradually increases, but after that, it significantly decreases, then is taken as the inflection point. At the same time, the increase rate of the change rate in the previous and subsequent two unit time periods is calculated for the temperature change rate. For example and are respectively and , then the mutation is . The reference value for determining the mutation of the temperature change rate is set to . This value is obtained from the average stable change rate recorded during the operation of the permanent magnet motor under medium load conditions for 10 minutes, and is set to times the maximum normal change rate under stable conditions, and increases linearly with the increase of the load frequency. When the change rate exceeds this reference value, it is determined as a rate mutation. Then, the jump section with an increase amplitude of more than in the torque fluctuation amplitude and lasting for more than two periods is screened. For example, period 1 is , period 2 is , the jump is , meeting the conditions. The torque fluctuation amplitude threshold based on which this jump determination is made is set to . Its setting reference is times the maximum steady-state torque fluctuation in the low-speed area (<500 rpm). By measuring the maximum torque fluctuations of the same type of motor under no-load, 50% load, and 100% load conditions as 1.9, 2.1, and 2.3, the maximum value is taken and multiplied by to obtain the threshold as , and rounding to an integer gives . After screening, the three items of data are uniformly normalized. Among them, the normalization range of the temperature rise slope is set to , the temperature change rate is set to , and the torque fluctuation amplitude is set to , the unified normalization method adopts the min-max scaling method. For example, if the temperature rise slope of a certain period is , then the normalized value is , and the three normalized values are respectively 、 、 Multiply and sum them to form the periodic heat response rate value . The setting of these three weights refers to their influence degrees in the overheat response of the motor. Through the statistics of the abnormal points in the operation process of 20 permanent magnet motors, it is found that the proportion of the temperature rise slope triggering the abnormal warning is , the temperature change rate is , and the torque fluctuation is . Accordingly, the set weights are respectively 、 、 , and the overall normalization is 1. For example, if the normalized values are successively 、 、 , then . Finally, the characteristic quantity sequence of each period is formed with this periodic heat response rate value.

[0064] Based on the periodic heat response rate value, the thermal field construction sub-module constructs a spatial point set with the slope of the stator temperature rise curve as the X-axis, the temperature change rate of the permanent magnet as the Y-axis, and the torque fluctuation amplitude as the Z-axis, calculates the surface connection angle and gradient change rate between each point within each period, identifies the continuously changing range of the gradient curvature and the extreme value distribution range, and establishes the trend value of the thermal load mapping surface;

[0065] The thermal field construction sub-module processes based on the periodic heat response rate value. First, a three-dimensional point set is established in the data space, where the three coordinates of each point are respectively the slope of the stator temperature rise curve, the temperature change rate of the permanent magnet, and the driving torque fluctuation amplitude. For example, the three values of period A are respectively 、 、 , then it represents point A . The points formed by all periods are constructed into a discrete point set in the three-dimensional coordinate space. Subsequently, a smooth surface segment is established between any two adjacent points through three-dimensional linear fitting, and the angle between this segment of surfaces is calculated. The angle is obtained from the dot product of the normal vectors by the three-point method. For example, points A, B, and C form plane ABC, and its normal vector is obtained from the cross product of vectors AB and AC. Then, calculate the angle between plane ABC and the adjacent plane BCD, and obtain the angle through . Further calculate the gradient of the change in the heat response rate value between each period point to the three coordinate dimensions, that is, calculate for 、 、 the axis directions respectively , , ,in is the periodic thermal response rate value. The above gradient curvature is used to determine whether the overall thermal field structure has a continuous unidirectional change segment in multiple cycles. If the continuous gradient change rate of five cycles in a certain direction is greater than When , the change rate threshold is The setting basis is that the average slope change under normal conditions in the normalized values of the three variables is Considering that the significant point of thermal field gradient evolution needs to be higher than the mean times, set to , rounded up to , if the gradient in a certain direction changes from positive to negative within a certain period and the rate of change is greater than When , it is defined as the extreme point, for example, the change rate from period 6 to 7 is If the judgment condition is met, the continuous areas and extreme value areas in all surface structures are screened, and the coordinate point set is extracted to construct the trend value of the heat load mapping surface. The trend value is calculated as the weighted average of the double factor weights of the surface connection angle and the gradient, that is:

[0066] ;

[0067] in is the angle (unit: °), is the average gradient, , The weight setting here is based on the fact that the change of surface angle plays a greater role in the switching of thermal load direction. By tracing back the statistics of the event at the surface mutation, we can find There is a significant angle fluctuation rather than a gradient change before the warning point, so the face angle weight is set to , for example, the average angle is , the mean gradient is ,but

[0068] ;

[0069] Finally, the spatial heat load trend surface is formed.

[0070] The critical trajectory identification submodule extracts the coordinate sequence of the intersection of the main curvature pole and the time axis within a unit period according to the trend value of the thermal load mapping surface, determines whether there is a trajectory segment with reversed direction and sudden amplitude change within two consecutive periods, selects the spatial coordinates with prominent changes as the identification points of magnetic property changes, and generates magnetic reduction characteristic trajectories;

[0071] The critical trajectory recognition sub-module extracts the principal curvature poles in the three-dimensional structure of the thermal field corresponding to each period at intervals of a unit period in the thermal load mapping surface, that is, the points with the maximum curvature in the thermal field trend surface. The acquisition method is to sort the curvature values of each point in the surface within each period, take the maximum value point, and record its position coordinates in the surface space. and match the corresponding cycle time of this point. to form a four-dimensional point. For example, the point with the maximum curvature in cycle 9 is , and the time is , then the point is formed. Arrange the poles of each period in chronological order to form a pole sequence. Judge the direction of the pole vector between every two consecutive periods in the sequence. The judgment of the direction reversal is through calculating the included angle of the displacement vector of the principal curvature poles between cycle and . If the included angle is greater than and the increase in the spatial distance between the two points exceeds , it is defined as direction reversal and amplitude mutation. Among them, the setting of the included angle threshold is based on the geometric analysis of the curvature reversal characteristic segment, and the judgment greater than or equal to the reverse direction angle standard between three-dimensional coordinates is used as the criterion. The judgment of the distance increase is based on the standard value of the average Euclidean distance between points , and the increase standard is set to . For example, the included angle between point and is , and the spatial distance increases from to , and the increase is , meeting the judgment conditions. Identify this pair of points, and further continuously judge whether the above changes occur continuously in two cycles. If so, it is determined as a critical trajectory segment. Finally, mark the coordinate points in the above trajectory segment, select several points with the largest change amplitude as magnetic property change recognition points, construct a magnetic reduction characteristic trajectory and record its three-dimensional space trajectory data set for subsequent map construction and evolution comparison.

[0072] Please refer to Figure 3 , the response offset recognition module includes:

[0073] Based on the magnetic reduction characteristic trajectory, the current trend determination sub-module obtains the target current vector amplitude, the actual sampling current change trend and the time position of the characteristic points in the magnetic reduction characteristic trajectory, calculates the change phase difference of the target current and the actual current in the two periods before and after the characteristic point respectively, and judges whether the change directions of the two curves before and after the characteristic point period show a reverse trend. If there is a reverse relationship, record the position of the phase turning point and generate a reverse trend recognition position sequence.

[0074] Based on the data obtained from the magnetic reduction characteristic trajectory, the current trend determination sub-module first extracts the time coordinates corresponding to each magnetic reduction characteristic point , and then combines the target current vector amplitude sequence with the actual sampled current sequence . Respectively extract the current amplitude change sequences within two complete cycles before and after each characteristic point time . The cycle length is set to 20 ms according to the system. If a certain characteristic point exists, then extract the current data from to . Among them, the target current is recorded from the current command curve of the control system, and the actual sampled current is obtained by linearly converting the voltage after conversion by the sampling resistor through an operational amplifier to obtain the amplitude. Subsequently, trend fitting is performed on the current data within two cycles, and the change slope is calculated using differential values. For example, the target current is during , with a slope of , while the actual current changes during the same period as , with a slope of . Since the two slopes are in opposite directions, it is recorded as reverse. When judging the reverse trend, set the direction reversal determination threshold to that the slope signs are opposite and the absolute value product is not less than 25. According to the actual current slope fluctuation range under typical operating loads, use the slope pair with opposite changes and larger amplitudes as the key inflection point criterion, and set the threshold to approximately the median product of the high slope pair . If the judgment is established, record the current cycle number as the phase turning point, and generate a reverse trend recognition sequence indexed by time points. For example, in the above example, record the cycle number as 63, corresponding to the time point . This sequence will be used as the basis for subsequent deviation calculations.

[0075] The deviation calculation sub-module calls the reverse trend recognition position sequence, collects the target current amplitude curve and the actual sampled current amplitude curve within the corresponding cycle, calculates the amplitude difference and the time difference between each corresponding point along the time axis, constructs a set of offset distance vectors within the same cycle, and combines the rotor angular velocity change curve. Use the fluctuation amplitude of the angular velocity at the corresponding deviation point as a scaling factor to adjust the amplitude difference, using the formula:

[0076] ;

[0077] Calculate the response relative drift amount through the operation;

[0078] Among them, represents the response relative drift amount, represents the target current amplitude at the th point, represents the actual sampled current amplitude at the th point, Indicates the time difference at the th point, and indicates the rotor angular velocity corresponding to the th point;

[0079] The deviation calculation sub-module calls each cycle index recorded in the reverse trend recognition position sequence , and extracts the current amplitudes corresponding to each time point within the cycle from the target current curve and the actual sampled current curve respectively. The two curves are mapped point by point, and the current difference between each pair of data points is recorded along with its offset on the time axis . The time difference calculation method is as follows: If the time difference between the points with the same amplitude of the target current and the actual current is and , then . At the same time, the rotor angular velocity at this moment is obtained, with the unit of . Its value is obtained by sampling the rotor position with an encoder and differentiating the position increment. For example, if the rotor angle change per sampling period is and the sampling interval is , then the angular velocity is . Subsequently, the response relative drift is calculated according to the following formula:

[0080] ;

[0081] The advantage of this formula is that the influence of the time difference is normalized to the time scale of the speed change through the term, enhancing the influence of the response time delay of the offset. Now, an example calculation is performed with actual data. Suppose the sampling pairing points for a certain cycle are as follows: , then:

[0082] ;

[0083] ;

[0084] Taking the average of the groups of data. For example, if each is successively , then:

[0085] ;

[0086] This result indicates that there is a response relative drift of approximately magnitude between the target current and the actual current within this cycle. This value is used as the drift intensity index for subsequent time window construction.

[0087] The time window construction sub-module sets the angular velocity fluctuation threshold to twice the standard deviation of the velocity median according to the response relative drift amount and the velocity volatility of the corresponding section in the rotor angular velocity change curve, screens the sections where the response relative drift amount is higher than the threshold, extracts the corresponding time segments, arranges and combines them in chronological order, and establishes a predicted response drift window;

[0088] The time window construction sub-module synchronizes and aligns the above sequence with the rotor angular velocity change curve, and extracts the time index corresponding to each response drift value , and at the same time calculates the statistical volatility of each section of the angular velocity curve, that is, calculates the standard deviation with a sliding window , the median is , if the current section of the velocity sequence is , then , the standard deviation , set the angular velocity fluctuation threshold to , the multiple of the standard deviation is selected according to the robust fluctuation judgment standard commonly used in industry. Taking 3σ as strong disturbance and 2σ as medium disturbance, it is set to 2σ according to the sensitivity tuning of this system to ensure full screening of high-fluctuation sections. Screen the drift amount and the corresponding speed fluctuation is greater than of the periodic section as a candidate, extract its time period. For example, if the period number is 64 - 66 and the corresponding time is 1.28 - 1.34s, it is recorded as one of the predicted response drift windows. Finally, all the time periods that meet the conditions are arranged in chronological order to generate a set of non-overlapping predicted response drift windows.

[0089] Please refer to Figure 4 , the corrected trajectory generation module includes:

[0090] The slope extraction sub-module obtains the predicted response drift window, extracts the component slope, the amplitude difference between consecutive points and the corresponding drift time value of each data point on the time axis according to the target current vector trajectory within the corresponding time period, and calculates the change amplitude value of the component slope per unit period based on the amplitude change rate and time offset ratio between adjacent points, generating the current slope change amplitude;

[0091] After the slope extraction sub-module obtains the predicted response drift window, it first extracts the target current vector trajectory sequence within each drift time period interval. Each data point is composed of a timestamp and the corresponding current amplitude component . This trajectory can be represented as a sequence of time series vectors with equally spaced sampling in the three-phase symmetric coordinate system. Then, it calculates the component slope of each data point in the time axis direction, that is, takes two adjacent points and , and uses the difference quotient form to obtain the slope , if at the start time of the drift window , with data points , , and the time interval between the two points is 0.002s, then , and then continue to extract the amplitude difference between each pair of adjacent data points, that is, , and record the drift time interval corresponding to the difference , and then calculate the amplitude change rate and time offset ratio between adjacent points, specifically taking the unit period All The difference As the value of the component slope change amplitude, and aggregated into an array sequence for subsequent trend analysis, for example, if , , then the variation between the two points is , all the changes constitute the slope change sequence To avoid the influence of numerical jitter on judgment, the minimum difference threshold is introduced and set to The threshold is derived from the maximum deviation statistic of the slope change in each cycle under steady-state operation, 17.4 A / s, and its rounded-up value of 20 A / s is taken as the critical judgment to filter out the interference of small disturbances on trend identification. This sequence will be called by subsequent sub-modules to identify the change trend.

[0092] The trend identification submodule calls the current slope change amplitude to detect whether there are three continuous monotonic change sequences in the slope change section. If so, it determines whether the fluctuation direction produces a double reversal. According to the position index of the reversal section, the corresponding start and end drift time intervals are extracted, the offset section time range is established, and the trend offset time interval is obtained.

[0093] After receiving the current slope change amplitude sequence, the trend recognition submodule detects whether there are three continuous monotonically changing subsequences in the sampling order. The specific operation is to scan any subsegments with a length of not less than 3 in the full sequence, and determine whether the slope change value is in a strictly increasing or decreasing relationship. If it is satisfied, the start and end indexes are recorded, and then the overall direction of the three continuous sequences is determined. For example, if the three slope difference sequences are , , , it can be determined that the first segment rises, the second segment falls, and the third segment rises again, indicating that there is a double reversal. Further, according to the reversal start and end position index, the corresponding time value on the drift time axis is extracted. For example, the sequence index position 17 to 22 corresponds to the time to , this section is defined as the trend deviation time period. In trend judgment, if the difference between the slope direction switching points among three sections exceeds ±30 A / s, it is recognized as an effective reversal. This value is based on the empirical boundary of 35.7 A / s for typical current slope mutations in the motor's rapid change working conditions and is rounded down to 30 A / s to ensure the direction effectiveness of trend recognition. All eligible time periods will be sequentially combined to form a complete set of ranges for the offset section time period.

[0094] According to the trend deviation time period interval, the trajectory generation sub-module calls the target current vector trajectory and the actual sampled current value within the interval, performs point-by-point vector difference on the two sets of current data at the corresponding time sequence points, and sequentially superimposes the difference results to establish a corrected path vector sequence and supplement the head and tail smoothing sections to generate a predicted corrected current trajectory;

[0095] After receiving the trend deviation time period interval, the trajectory generation sub-module extracts the target current vector trajectory within each interval and the actual sampled current trajectory . After aligning the two trajectory points in time sequence, perform point-by-point vector difference. The calculation method is . The result forms a set of difference vector sequences . Subsequently, establish a corrected path vector sequence in an accumulative manner . For example, if the difference vectors of three sampling points within a certain period are (1.2, 0.8), (1.1, 0.7), (0.9, 0.6), then the cumulative vectors of its corrected path are (1.2, 0.8), (2.3, 1.5), (3.2, 2.1) in sequence. This sequence constitutes the main section of the predicted corrected path. To avoid sudden changes in the transition of the trajectory before and after, smooth sections need to be added to the head and tail of the corrected path. The length is set to 10% of the length of the main section. This proportional coefficient refers to the width of the convergence interval of the typical current trajectory amplitude change curve. If the main section contains 40 points, then 4 points are taken for each of the front and back smooth sections, and interpolation is performed according to the head and tail trends to construct the supplementary section to build the final trajectory. The corrected trajectory is the final predicted corrected current trajectory and is constructed and output according to the sequence .

[0096] Please refer to Figure 5 . The amplitude limit interval decision module includes:

[0097] Based on the predicted corrected current trajectory, the inductance fluctuation identification sub-module obtains the stator inductance trend change curve within the current period, calculates the inductance change slope value and the range difference of the change interval within three consecutive periods, and judges whether the maximum slope value exceeds the inductance stability bandwidth threshold. If the judgment is established, the section is recorded as a high-variation section, and an inductance abnormal variation interval is generated;

[0098] Based on the sampling time series given by the inductance fluctuation identification sub-module for the predicted and corrected current trajectory, first obtain the change trend curve of the stator inductance value within the current period. This curve is obtained by dividing the high-speed sampled excitation voltage by the current, that is, on the basis of the known voltage , current , use the formula to calculate the inductance value for each point. Subsequently, construct a group of inductance sequences for multiple consecutive periods according to the sampling period. Take the first-order difference of the inductance curve composed of several sampling points within each period to obtain the inductance change slope , and record the maximum value, minimum value, and range value within each period. The range value calculation formula is . For example, in period 1, the inductance value sequence is [2.3, 2.6, 2.1, 2.8, 2.5] mH, corresponding to a maximum value of 2.8 mH and a minimum value of 2.1 mH, then the range is 0.7 mH. If the maximum slope is / single-point time interval 0.002 s, then the slope is 150 mH / s. When judging whether the inductance change exceeds the limit, it is necessary to compare the maximum slope value with the inductance stability bandwidth threshold. This bandwidth threshold is set to . The setting basis is: based on the analysis of the average operation data of 30 types of permanent magnet synchronous motors, the maximum slope of the inductance fluctuation range in the stable state is about 95 mH / s. Considering the allowable increase amount during the working condition switching, the bandwidth is set to 1.25 times the maximum stable slope, that is , rounded up to 120 mH / s as the threshold. If the actual slope exceeds this value, this period is judged as a high inductance change period. Finally, output all the period time segments that meet the exceeding conditions, and summarize them to form an inductance abnormal change interval.

[0099] The current peak positioning sub-module calls the inductance abnormal change interval, obtains the amplitude distribution sequence of the predicted and corrected current trajectory, extracts the point with the maximum amplitude within the change interval, and calculates the amplitude amplification ratio by combining the inductance slope value at the time index corresponding to the point with the maximum amplitude and the average inductance of the previous period to generate a current amplitude slope ratio;

[0100] The current peak positioning sub-module calls all the time segments within the inductance abnormal change interval, obtains the amplitude distribution sequence of the predicted and corrected current trajectory within this segment, records the current vector modulus value at each time point, and obtains the instantaneous amplitude through the modulus length calculation formula . For example, at time , there are , , then . Then extract the maximum value from all the amplitude points of this segment sequence. If the maximum amplitude point of a certain segment is , it is recorded as the peak point; then obtain the inductance slope value at the time index position of the peak point. If it is calculated as in the previous sampling , and obtain the average value of the inductor from the previous complete cycle , and then combine it with the target limit value , the limit value is set to the maximum allowable current configured in the system, which is 15.0 A. The limit value setting reference is the critical current of 14.6 A calculated from the thermal capacity of the inverter bridge arm under the rated load condition plus a tolerance of 0.4 A, resulting in 15.0 A as the system peak tolerance limit value. Based on this, the amplitude amplification ratio is calculated as , and finally output this ratio as the current amplitude slope ratio

[0101] The clipping boundary generation sub-module sets the control boundary points and calibrates the section response time according to the current amplitude slope ratio, combined with the maximum amplitude point and the corresponding inductor change rate in the predicted corrected current trajectory, using the formula:

[0102] ;

[0103] Obtain the clipping boundary value through calculation, compress the amplitude of the peak section in the corrected trajectory, and establish a dynamic limited current trajectory

[0104] Among them, represents the clipping boundary value represents the maximum amplitude point in the predicted corrected current trajectory represents the target limit setting value represents the inductor change rate at the current moment represents the response duration at which the amplitude peak is located represents the difference between the current before and after the amplitude correction

[0105] The clipping boundary generation sub-module, according to the current amplitude slope ratio obtained in the previous step, combines the amplitude of the peak point in the predicted corrected current trajectory with the corresponding inductor change rate , and reads the response duration where the amplitude peak point is located and the difference between the current correction values before and after, and substitutes the above parameters into the formula:

[0106] ;

[0107] Now, a numerical calculation example is given. Let , , , , , the calculation steps are as follows:

[0108] The first item:

[0109] ;

[0110] Item 2:

[0111] ;

[0112] Combined to get:

[0113] ;

[0114] Finally, the cropping boundary value is calculated , which is used to perform a dynamic clipping operation on the amplitude compression of the peak section in the corrected trajectory later, ensuring that the current corrected trajectory does not exceed this limit value within the response interval, forming a dynamic clipped current trajectory. This result shows that the cropping boundary effectively identifies the basis for limit adjustment, quantifies the compression range, and can be used as a dynamic reference threshold for subsequent clipping logic.

[0115] Please refer to Figure 6 , the compensation instruction output module includes:

[0116] The bias trend identification sub-module calls the dynamic clipped current trajectory, obtains the time period corresponding to the wave peak in the dynamic clipped current trajectory, collects the sequence of sampling error bias values within the current period, performs a moving average process on the sequence, identifies the change trend of the mean difference of three consecutive points, and determines whether there is a continuously increasing segment. If the judgment is established, the maximum increment value of the corresponding segment is extracted to generate a bias amplification trend value;

[0117] After the bias trend identification sub-module calls the dynamic clipped current trajectory, it obtains the time period corresponding to the known amplitude wave peak in the dynamic clipped current trajectory, continuously samples the sequence of sampling error bias values within this period. Each bias value is the difference between the actual sampled current and the predicted corrected current, and constructs the sequence , for example, at time points , , , the bias values are respectively . Using a moving window size of 3, a moving average process is performed on the bias sequence to obtain the mean sequence , specifically , and then it advances backward in this way; subsequently, a difference operation is performed between every three consecutive moving averages, that is, the sequence is calculated, and it is judged whether the condition that the three consecutive differences are positive is satisfied. For example, if , , , it is determined that there is an increasing trend segment. The numerical constraint for "increasing" in the judgment condition is set such that the differences in the three segments are all greater than 0.01 A. The basis for setting this threshold is that the standard deviation of the normal error random perturbation distribution is 0.007 A, and 1.5 times of it is taken to obtain the upper bound of the tolerance band, which is approximately 0.0105 A. Rounding up to 0.01 A is used as the judgment basis. If the judgment is established, the difference between the maximum value and the minimum value of the mean in this segment is extracted as the maximum increment value. For example, if the maximum value is 0.27 A and the minimum value is 0.14 A, the offset amplification trend value is 0.13 A, and this value will be used as the input of the subsequent delay offset generation sub-module.

[0118] The delay correction amount generation sub-module calls the offset amplification trend value, collects the PWM dead time of the current cycle, the current response delay change value within the time period, and the sampling current change rate at the end of the amplitude limiting trajectory, and calculates the relative amplification amount of the current response delay and the PWM dead time compensation ratio respectively, using the formula:

[0119] ;

[0120] Calculate to obtain the correction offset;

[0121] Among them, is the correction offset, represents the amplified increment of the sampling offset, represents the PWM dead time, represents the current response slope, represents the current change amplitude, represents the response peak duration;

[0122] The delay correction amount generation sub-module calls the offset amplification trend value , and at the same time collects the PWM dead time in the current cycle, the sampling current response change rate at the end of the amplitude limiting trajectory, the current change amplitude, and the peak response duration, , , , , and substitute them into the following calculation formula:

[0123] ;

[0124] Execute the calculation:

[0125] The first item:

[0126] ;

[0127] The second item:

[0128] ;

[0129] Finally, we get:

[0130] ;

[0131] Although the first term of this offset correction amount is relatively small, the second term is greatly affected by the response of the limited amplitude tail section and constitutes the main adjustment amount. This result provides the basis for the current offset supplement required for the correction of the dynamic limited amplitude trajectory response.

[0132] The vector instruction adjustment sub-module corrects the amplitude of each point in the data sequence at the end of the limited amplitude current trajectory according to the correction offset, takes the current trajectory after supplementation as the corrected execution current input, combines the vector angle distribution interval with the instruction angle of the previous cycle to calculate the corrected vector angle sequence, and establishes a prediction compensation current execution instruction;

[0133] After receiving the correction offset, the vector instruction adjustment sub-module performs addition compensation on the amplitude sequence of all sampling points at the end of the limited amplitude current trajectory. The amplitude of each point follows the rule: , for example, the original sampling point amplitude sequence is , and after correction it is . Take the current trajectory after supplementation as the corrected execution current input, and then combine the target vector angle of the previous cycle with the current vector direction corresponding to the corrected trajectory to calculate the current vector angle , calculate the angle change . If the vector angle in the previous cycle is , and the current corrected point corresponds to , then . Build a set of angle difference sequences in this way, and synchronously generate a set of prediction compensation current execution instructions. These instructions are finally input into the execution unit to form a new round of feedback control reference value. The update of the prediction compensation angle sequence can dynamically respond to the vector direction disturbance changes caused by the correction offset, ensuring that the execution current trajectory is dynamically adjusted by the controller.

[0134] The above is only a preferred embodiment of the present invention, and it does not limit the present invention in other forms. Any person skilled in the relevant art may use the disclosed technical content to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical content of the technical solution of the present invention, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A permanent magnet motor current control system based on prediction compensation, characterized in that The system includes: The thermal field recognition module obtains the operating state parameters of the permanent magnet motor, constructs a spatial coordinate set at each point for the stator temperature rise curve slope, temperature change rate, and torque fluctuation value, determines whether the trajectory undergoes a curvature inversion change, extracts the coordinate points as magnetic performance reduction characteristic points, and generates a magnetic reduction characteristic trajectory; Based on the magnetic reduction characteristic trajectory, the response offset recognition module determines whether there is a reverse difference in the change phase between the target current trend curve and the actual sampling curve before and after the magnetic reduction characteristic point cycle. If there is a reverse change, the relative drift amount is calculated, and a predicted response drift window is established; The response offset recognition module includes: Based on the magnetic reduction characteristic trajectory, the current trend determination sub-module obtains the target current vector amplitude, the actual sampling current change trend, and the time position of the characteristic points in the magnetic reduction characteristic trajectory, calculates the change phase difference between the target current and the actual current in the two cycles before and after the characteristic point respectively, determines whether the change directions of the two curves before and after the characteristic point cycle show a reverse trend. If there is a reverse relationship, the position of the phase turning point is recorded, and a reverse trend recognition position sequence is generated; The deviation calculation sub-module calls the reverse trend recognition position sequence, collects the target current amplitude curve and the actual sampling current amplitude curve in the corresponding cycle, calculates the amplitude difference and the occurrence time difference between each corresponding point along the time axis, constructs a set of offset distance vectors in the same cycle, combines the rotor angular velocity change curve, uses the fluctuation amplitude of the angular velocity at the corresponding deviation point as a scaling factor to adjust the amplitude difference, and uses the formula: ； Calculate the response relative drift amount through operation; Among them, represents the response relative drift amount, represents the point target current amplitude, represents the point actual sampled current amplitude, represents the point time difference, represents the point corresponding rotor angular velocity, is the total number of paired points participating in the calculation; Based on the response relative drift amount, the time window construction sub-module combines the speed volatility rate in the corresponding section of the rotor angular velocity change curve, sets the angular velocity fluctuation threshold as twice the standard deviation of the speed median, filters out the sections where the response relative drift amount is higher than the threshold, extracts the corresponding time segments, and arranges and combines them in chronological order to establish a predicted response drift window; The correction trajectory generation module calls the predicted response drift window, determines whether there are continuous trend offset segments. If so, it performs point-by-point difference on the relative vector between the actual sampling current value and the target value, obtains the offset amount superposition trajectory of the corresponding cycle correction path, and generates a predicted correction current trajectory; Based on the predicted correction current trajectory, the amplitude limit interval decision module combines the amplitude upper limit point in the current correction trajectory and the inductance change rate in the corresponding time interval to constrain and clip the maximum amplitude of the correction trajectory, and generates a dynamic amplitude-limited current trajectory.

2. The permanent magnet motor current control system based on prediction compensation according to claim 1, characterized in that The magnetic reduction characteristic trajectory includes the magnetic performance change critical point time series, the peak path coordinates of the thermal load mapping surface, and the spatial distribution of the curvature inversion points. The predicted response drift window includes the current amplitude deviation data set, the current trend phase difference section, and the response drift duration range. The predicted correction current trajectory includes the current vector offset path, the amplitude superposition sequence, and the trend shift nodes. The dynamic amplitude-limited current trajectory includes the amplitude limit upper boundary, the edge points of the clipped section, and the inductance amplitude change control calibration value.

3. The permanent magnet motor current control system based on prediction compensation according to claim 2, characterized in that, The thermal field recognition module includes: The thermal parameter extraction sub-module obtains the operating state parameters of the permanent magnet motor, including the stator winding temperature rise curve, the surface temperature change rate of the permanent magnet, and the driving torque fluctuation range per unit time. It synchronizes and pairs the points of the three data items according to the time period, screens the time segments with inflection points of the temperature rise slope, sudden increases in the temperature change rate, and sudden jumps in the torque fluctuation amplitude, and performs joint normalization processing on the fluctuation amplitude to generate the periodic thermal response rate value; The thermal field construction sub-module constructs a spatial point set with the slope of the stator temperature rise curve as the X-axis, the temperature change rate of the permanent magnet as the Y-axis, and the torque fluctuation amplitude as the Z-axis based on the periodic thermal response rate value, calculates the surface connection angle and gradient change rate between each point within each period, identifies the continuously changing range of the gradient curvature and the extreme value distribution range, and establishes the thermal load mapping surface trend value; The critical trajectory identification sub-module extracts the coordinate sequence of the intersection points of the principal curvature poles and the time axis within a unit period according to the thermal load mapping surface trend value, determines whether there is a trajectory segment with a direction reversal and an amplitude mutation within two consecutive periods in the sequence, screens the spatial coordinates with prominent changes as the magnetic property change identification points, and generates the magnetic reduction characteristic trajectory.

4. The permanent magnet motor current control system based on prediction compensation according to claim 3, characterized in that, The correction trajectory generation module includes: The slope extraction sub-module obtains the predicted response drift window, extracts the component slope of each data point on the time axis, the amplitude difference between consecutive points, and the corresponding drift time value according to the target current vector trajectory within the corresponding time period, and calculates the change amplitude value of the component slope within a unit period based on the amplitude change rate and time offset ratio between adjacent points to generate the current slope change amplitude; The trend identification sub-module calls the current slope change amplitude to detect whether there are three consecutive monotonic change sequences within the slope change section. If so, it determines whether the fluctuation direction has a double reversal, extracts the corresponding start and end drift time intervals according to the position index of the reversal section, and establishes the time range of the offset section to obtain the trend offset time period interval; The trajectory generation sub-module calls the target current vector trajectory and the actual sampled current value within the interval according to the trend offset time period interval, performs point-by-point vector difference on the two sets of current data at the corresponding time sequence points, and sequentially superimposes the difference results to establish a correction path vector sequence and supplement the head and tail smoothing sections to generate the predicted correction current trajectory.

5. The permanent magnet motor current control system based on prediction compensation according to claim 4, characterized in that, The amplitude limit interval decision module includes: The inductance fluctuation identification sub-module obtains the stator inductance trend change curve within the current period based on the predicted correction current trajectory, calculates the inductance change slope value and the range difference of the change interval within three consecutive periods, and determines whether the maximum slope value exceeds the inductance stability bandwidth threshold. If the judgment is established, it records the section as the high-variation section to generate the inductance abnormal change interval; The current peak positioning sub-module calls the inductance abnormal change interval, obtains the amplitude distribution sequence of the predicted correction current trajectory, extracts the point with the maximum amplitude within the change interval, and calculates the amplitude amplification ratio by combining the inductance slope value at the time index corresponding to the point with the maximum amplitude and the average inductance value of the previous period to generate the current amplitude slope ratio; The cropping boundary generation sub-module sets the regulation boundary points and calibrates the section response time according to the current amplitude slope ratio, in combination with the maximum amplitude point and the corresponding inductance change rate in the predicted corrected current trajectory, using the formula: ； Calculate to obtain the cropping boundary value, compress the amplitude of the peak section in the corrected trajectory, and establish a dynamically limited current trajectory; Among them, represents the clipping boundary value, represents the maximum amplitude point in the predicted corrected current trajectory, represents the target limit setting value, represents the inductance change rate at the current moment, represents the response duration at which the amplitude peak is located, represents the difference between the current corrected amplitudes before and after.

6. The permanent magnet motor current control system based on prediction compensation according to claim 5, characterized in that, The system further includes: The compensation instruction output module calls the dynamically limited current trajectory, collects the sampling error offset value in the current cycle and the current response delay value generated by the PWM dead time, judges whether the error offset value has an amplification trend within the time corresponding to the peak of the limited trajectory wave. If it holds, accumulate and offset-correct the offset value and the delay response value, adjust the instruction vector, and generate a predicted compensation current execution instruction; The predicted compensation current execution instruction includes an end instruction vector sequence, a response delay correction amount, and an execution output offset structure.

7. The permanent magnet motor current control system based on prediction compensation according to claim 6, characterized in that The compensation instruction output module includes: The offset trend recognition sub-module calls the dynamically limited current trajectory, obtains the time period corresponding to the peak in the dynamically limited current trajectory, collects the sampling error offset value sequence in the current cycle, performs a sliding mean process on the sequence, recognizes the mean difference change trend of three consecutive points, and judges whether there is a continuously increasing section. If the judgment holds, extract the maximum increment value of the corresponding section and generate an offset amplification trend value; The delay correction amount generation sub-module calls the offset amplification trend value, collects the PWM dead time in the current cycle, the current response delay change value within the time period, and the sampling current change rate at the end of the limited trajectory, calculates the relative amplification amount of the current response delay and the PWM dead time compensation ratio respectively, using the formula: ； Calculate to obtain the correction offset amount; wherein, is the correction offset, represents the amplification increment of the sampling bias, represents the PWM dead time duration, represents the current response slope, represents the current change amplitude, represents the response peak duration; The vector instruction adjustment sub-module performs amplitude augmentation correction on each point in the data sequence at the end of the limited current trajectory according to the correction offset amount, uses the augmented current trajectory as the corrected execution current input, combines the vector angle distribution interval and the instruction angle of the previous cycle to make a difference, calculates the corrected vector angle sequence, and establishes a predicted compensation current execution instruction.

Images