A temperature rise compensation control method in a bench test of an electric drive system of an electric vehicle

Abstract

The application belongs to the technical field of automobile system testing, and provides a temperature rise compensation control method in electric drive system bench testing of electric vehicles, comprising: calibrating the relationship model of the conduction voltage drop of the power device changing with the junction temperature and the current; then using the measurable parameters of the bench, combining the loss model and the thermal network model to recursively propagate the dynamic junction temperature online; under the low-speed light-load working condition, using the clamping interval near the current zero-crossing point to construct the voltage error expansion model, using the recursive least square algorithm with the forgetting factor to simultaneously identify the total nonlinear error of the inverter and the zero-point drift of the current sensor, and separating the temperature drift component of the conduction voltage drop from the total nonlinear error; finally, fusing the reference value output by the relationship model and the temperature drift component to obtain the optimal compensation voltage value, and superimposing the optimal compensation voltage value on the inverter command voltage, and using the zero-point drift for zero correction of the current sampling channel. The application simultaneously realizes the temperature self-adaptation of the dead zone compensation and the online correction of the sensor zero drift under the same framework.

Description

Technical Field

[0001] This invention belongs to the field of automotive system testing technology, specifically a temperature rise compensation control method for bench testing of electric vehicle electric drive systems. Background Technology

[0002] In electric drive systems for electric vehicles, the dead-time effect of the inverter is one of the main causes of current waveform distortion. Currently, fixed-parameter compensation strategies are commonly used in engineering, which calculate the compensation voltage based on the preset value of the on-state voltage drop of the power devices. However, in actual operation, the on-state voltage drop of both insulated-gate bipolar transistors (IGBTs) and silicon carbide metal-oxide-semiconductor field-effect transistors (MOSFETs) will drift significantly with changes in junction temperature. For example, the on-state voltage drop of an IGBT will decrease significantly after the junction temperature rises, while the on-resistance of a MOSFET will increase significantly. During bench temperature rise testing, the junction temperature of the devices changes widely from cold to hot. Fixed dead-time compensation, because it does not take this temperature drift into account, often results in compensation deviations. Especially under low-speed, light-load conditions, this will additionally increase harmonic losses, creating a vicious cycle: increased temperature leads to poorer compensation, which in turn causes more heat generation, which further increases the temperature. On the other hand, the zero-point drift of current sensors is also affected by temperature. Hall or magnetoresistive current sensors may have a zero-point offset that reaches a certain proportion of the rated current over a wide temperature range. This error not only directly affects the torque accuracy under low speed and light load, but also interferes with the identification benchmark for dead zone compensation.

[0003] Therefore, the present invention provides a temperature rise compensation control method for bench testing of electric vehicle electric drive systems. Summary of the Invention

[0004] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0005] The technical solution adopted by this invention to solve its technical problem is: a method for temperature rise compensation control in bench testing of electric vehicle electric drive systems, comprising the following steps: Step S10: For the target device, measure the forward voltage drop at different junction temperatures using a constant temperature chamber, and fit a model of the relationship between the forward voltage drop and the junction temperature by combining the effects of different current levels. Step S20: Using the measurable operating parameters of the test bench, combined with the device loss and equivalent thermal network model, the current dynamic junction temperature is recursively calculated online, serving as the input to the relationship model in step S10 and the identification reference benchmark in step S30. Step S30: Under low-speed and light-load conditions, construct a voltage error extension model using the clamping interval near the current zero-crossing point, including two parameters to be identified: The first item is the total nonlinear error of the inverter, which consists of dead time and on-state voltage drop, where the on-state voltage drop varies with junction temperature. The second item is the zero-point drift of the current sensor; A recursive least squares algorithm with a forgetting factor is used to simultaneously identify two parameters to be identified. By analyzing the difference in the correlation between the error term and the current polarity, the temperature drift component of the on-state voltage drop is separated from the first term, and the value of the second term is obtained. Step S40: The relational model and the on-state voltage drop temperature drift component are fused through Kalman filtering to obtain the optimal compensation voltage value at the current junction temperature, which is then superimposed on the inverter command voltage in real time to correct the voltage loss in dead zone compensation; at the same time, the identified zero-point drift is fed back to the current sampling channel for online zeroing correction.

[0006] As a further aspect of the present invention: the process of measuring the on-state voltage drop at different junction temperatures in step S10 is as follows: The target device is placed in a programmable constant temperature chamber, and thermocouples are attached to the surface of the device housing to monitor the test temperature; multiple discrete junction temperature points are selected according to the operating temperature range of the device. Set the constant temperature chamber to the target temperature, and wait until the temperature inside the chamber stabilizes and the temperature deviation between the device housing and the temperature inside the chamber is within the allowable range. Using a transistor curve tracer or a high-current source meter, apply multiple conduction currents of different levels to the device, and use a pulse test method to read the saturation voltage drop between the collector and emitter within the pulse duration.

[0007] As a further aspect of the present invention: the process of fitting the relational model in step S10 is as follows: The on-state voltage drop at different junction temperatures and current levels is integrated into multiple sets of three-dimensional data. A model is adopted in which the on-state voltage drop varies linearly with junction temperature, and the slope of the change is different for different currents. For each fixed current level, linear fitting is performed independently, and the reference on-state voltage drop and temperature coefficient for the current level are obtained by least squares method. Store the reference on-state voltage drop and temperature coefficient corresponding to each current level. During online calculation, the reference on-state voltage drop value under any junction temperature and current can be obtained by looking up the table.

[0008] As a further aspect of the present invention: the measurable operating parameters of the test bench and the method of obtaining them in step S20 are as follows: Operating parameters include three-phase current, DC bus voltage, switching frequency, and coolant inlet temperature; The DC bus voltage is acquired by a voltage Hall sensor installed between the positive and negative terminals of the DC bus; the three-phase current is acquired by a Hall current sensor installed on the three-phase output lines of the inverter. The switching frequency is read directly from the controller's pulse width modulation module; the coolant inlet temperature is collected by a temperature sensor located at the inlet of the cooling circuit.

[0009] As a further aspect of the present invention: the process of online recursion of the current dynamic junction temperature in step S20 is as follows: The average current and effective current during the switching cycle are calculated based on the three-phase current. The average current is used to calculate the portion of the conduction loss related to constant voltage drop, and the effective current is used to calculate the portion of the conduction loss related to dynamic resistance. The instantaneous amplitude is extracted from the three-phase current, and the switching loss is estimated by combining the DC bus voltage and the switching frequency. The total power loss is obtained by adding the conduction loss and the switching loss. Using a Cauer-type thermal network model, the power loss is injected into the first thermal capacity node. The heat flow is calculated based on the temperature difference across each thermal resistance. The temperature increment is obtained by dividing the net heat within the current step by the thermal capacity value. The temperature of each node is updated recursively, and the current dynamic junction temperature is output.

[0010] As a further aspect of the present invention: the process of low-speed, light-load operating conditions and constructing the voltage error extension model in step S30 is as follows: In three-phase current, the region centered on the current zero-crossing point and where the current amplitude is less than a fixed proportion of the rated current is defined as the clamping zone. Within the clamping range, the difference between the command voltage issued by the controller and the voltage actually applied to the motor terminal is taken as the original voltage error, and expressed as the total nonlinear error of the inverter multiplied by the current polarity plus the zero-point drift of the current sensor.

[0011] As a further aspect of the present invention: the process of identifying parameters and separating temperature drift components in step S30 is as follows: Multiple sampling points are continuously collected within the clamping interval. The polarity of the current and the measured voltage error are recorded at each sampling point. The initial estimated values ​​of the total nonlinear error of the inverter and the zero-point drift of the current sensor are set to zero. For each new data point, the voltage error is predicted and the difference is calculated with the measured value. The estimated values ​​of the total nonlinear error and the zero-point drift of the current sensor are adjusted according to the predicted error. When the parameter update amount of multiple consecutive sampling points is less than the preset threshold, convergence is determined and the current identification result is output. The dead zone reference value is obtained by performing identification once at room temperature. The temperature drift component of the on-state voltage drop is obtained by subtracting the dead zone reference value from the total nonlinear error of the current identification.

[0012] As a further aspect of the present invention: the process of obtaining the optimal compensation voltage value in step S40 is as follows: Based on the relational model, the reference on-state voltage drop value is output according to the current junction temperature and current level, and added to the on-state voltage drop temperature drift component to obtain the preliminary correction value; the optimal compensation voltage value of the previous moment is read, and the preliminary correction value is fused with the optimal compensation voltage value of the previous moment through Kalman filtering to obtain the optimal compensation voltage value of the current moment.

[0013] As a further aspect of the present invention: the process of superimposing the compensation voltage in step S40 is as follows: The original command voltage output from the current loop is read, and the base compensation voltage output from the fixed dead zone compensation module is read. The original command voltage, the base compensation voltage, and the optimal compensation voltage value are superimposed to obtain the total command voltage. The total command voltage is used to generate the actual switching signal to drive the inverter, thereby correcting the voltage loss caused by temperature changes in the dead zone compensation.

[0014] As a further aspect of the present invention: the online zeroing correction process in step S40 is as follows: At the start of each current sampling, the original current sample value is read, and the zero-point drift value is subtracted from the original current sample value to obtain the corrected current value.

[0015] The beneficial effects of this invention are as follows: This invention overcomes the temperature drift inaccuracy caused by treating the on-state voltage drop as a constant value in traditional dead-zone compensation by combining offline calibration and online identification. It uses a recursive least squares algorithm to extract the component of the on-state voltage drop that changes with the junction temperature in real time and integrates it with the offline model, so that the compensation voltage can be dynamically adjusted with temperature. This avoids overcompensation or undercompensation caused by device temperature rise, thereby cutting off the positive feedback loop of "temperature rise leading to compensation failure, failure leading to additional heat generation, and heat generation further aggravating temperature rise". This invention, within the same low-speed, light-load identification framework, simultaneously addresses two types of problems: temperature drift of inverter nonlinearity error and zero-point temperature drift of current sensor. By constructing a voltage error extension model including current polarity-related and constant terms, and using a recursive least squares algorithm to simultaneously identify the two parameters, and separating the temperature drift component of the on-state voltage drop based on a room-temperature reference value, it achieves both adaptive updates for dead-zone compensation and accurate estimation of sensor zero drift. Unlike existing technologies that handle the two problems separately, this invention requires no additional test conditions or dedicated hardware, utilizing only the existing low-speed, light-load operating conditions of the test bench. This invention enables joint identification by fusing the model with online identification results. This ensures a rapid response of the compensation value to temperature changes while avoiding severe fluctuations caused by noise. The optimal compensation voltage obtained by fusion is directly superimposed on the inverter command voltage, making it suitable for embedded real-time implementation. At the same time, the identified zero-point drift is directly used for zero-point correction of the current sampling channel, eliminating the impact of sensor temperature drift on current control accuracy. In summary, this invention can effectively suppress dead-zone compensation inaccuracy and current sensor drift caused by temperature changes in bench testing, improving the control consistency of the electric drive system over a wide temperature range. Attached Figure Description

[0016] The invention will now be further described with reference to the accompanying drawings.

[0017] Figure 1 This is a flowchart illustrating the steps of a temperature rise compensation control method for bench testing of an electric vehicle electric drive system according to an embodiment of the present invention. Figure 2 This is a flowchart illustrating the steps of obtaining a temperature rise compensation control method for bench testing of an electric vehicle drive system according to an embodiment of the present invention. Detailed Implementation

[0018] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0019] Example: Please refer to Figure 1-2 As shown in the embodiment of the present invention, a temperature rise compensation control method for bench testing of an electric vehicle electric drive system includes the following steps: Step S10: For the target device, measure the forward voltage drop at different junction temperatures using a constant temperature chamber, and fit a model of the relationship between the forward voltage drop and the junction temperature by combining the effects of different current levels. In step S10, the target device, the power switching transistor used in the inverter of the electric vehicle electric drive system, such as the insulated gate bipolar transistor, is the core target of the temperature rise compensation scheme. Its on-state voltage drop temperature characteristics directly affect the accuracy of dead zone compensation. In bench testing, the target device is usually installed inside the inverter module and has a clear model, package type and rated parameters. In step S10, the procedure for measuring the on-state voltage drop at different junction temperatures using a constant temperature chamber is as follows: Place the target device (a single power transistor or a power module containing the device) in a programmable temperature chamber. The device needs to be pre-soldered or crimped onto the test fixture, and lead out the collector, emitter (or drain, source) and gate leads. Attach thermocouples or thermistors to the device housing or substrate surface to monitor the test temperature. Based on the device's operating temperature range (e.g., -40°C to 150°C), select several discrete junction temperature points, typically spaced at 25°C or 20°C intervals, including room temperature (25°C) and the maximum allowable junction temperature; Set the constant temperature chamber to the target temperature. After the temperature inside the chamber stabilizes and the temperature difference between the device housing temperature and the temperature inside the chamber is less than ±1℃, keep it at the temperature for at least 30 minutes to ensure that the device junction temperature is consistent with the housing temperature. Using a transistor curve tracer or a high-current source meter, different levels of on-state current (e.g., 10%, 25%, 50%, 75%, 100% of the rated current) are applied to the device. During measurement, the gate is driven to fully turn on the device, and the saturation voltage drop between the collector and emitter is recorded. To avoid the influence of self-heating on the junction temperature, a pulse test method is used (pulse width less than 100 μs, duty cycle less than 1%). The voltage drop is read within the pulse duration. At each junction temperature point, the on-state voltage drop corresponding to different on-state currents is measured and recorded. In step S10, the process of fitting a model showing the relationship between the on-state voltage drop and the junction temperature, taking into account the influence of different current levels, is as follows: The on-state voltage drop corresponding to different on-state currents at different junction temperatures is integrated into multiple sets of three-dimensional data: ,in, For the first The temperature of the junction. This represents the number of junction temperature points. For the first One current level, For the current rating, For the first The first junction temperature The on-state voltage drop corresponding to each current level; For an insulated-gate bipolar transistor, the on-state voltage drop varies linearly with the junction temperature, and the slope of this variation differs slightly for different currents. The model used is as follows: ;

[0020] in, For reference temperature (e.g., 25℃). For the on-state voltage drop at the reference temperature, Temperature coefficient (unit: V / °C); Among them, temperature coefficient The results were obtained by fitting the data at different current levels, specifically as follows: Taking an insulated-gate bipolar transistor as an example, for each fixed current level Perform linear fitting independently: For 3D data: Fitting a straight line using the least squares method: ;

[0021] in, For reference temperature On-state voltage drop at current rating For the first Temperature coefficient at current rating; The specific calculation is as follows: , ;

[0022] in, For the first The average on-state voltage drop at various junction temperatures for different current ratings. This is the average junction temperature; Finally, the reference on-state voltage drop corresponding to each current level was obtained. and temperature coefficient ; The reference on-state voltage drop and temperature coefficient for each current level are stored and obtained by looking up a table during calculation, ultimately yielding a model showing the relationship between on-state voltage drop and junction temperature. ;

[0023] in, and It is obtained by looking up a table during calculation; It is understandable that the significance of step S10 is as follows: Step S10 measures the on-state voltage drop at different junction temperatures and different current levels offline using a constant temperature chamber, and fits a model of the relationship between the on-state voltage drop and the junction temperature. This model provides a priori reference for the system, so that subsequent online compensation no longer depends on a fixed constant, but can obtain the reference on-state voltage drop value by looking up a table based on the actual junction temperature and current. The method of fitting the temperature coefficient by current level solves the model mismatch problem caused by treating the on-state voltage drop as a constant value in traditional dead zone compensation. The output of this model serves as the basic data source for the fusion calculation of the compensation voltage in step S40. Step S20: Using the measurable operating parameters of the test bench, combined with the device loss and equivalent thermal network model, the current dynamic junction temperature is recursively calculated online, serving as the input to the relationship model in step S10 and the identification reference benchmark in step S30. In step S20, the measurable operating parameters of the test bench include, but are not limited to, three-phase current. DC bus voltage Switching frequency Coolant inlet temperature ; The operating parameters can be obtained through methods such as DC bus voltage. The three-phase current is acquired by a voltage Hall sensor installed between the positive and negative terminals of the DC bus at a sampling frequency of 1kHz to 10kHz. The current is collected by Hall current sensors installed on the three-phase output lines of the inverter; switching frequency. The coolant inlet temperature is read directly from the controller's PWM module. Temperature data is collected via a temperature sensor at the inlet of the cooling circuit. In step S20, the process of recursively estimating the current dynamic junction temperature online, combining device losses and the equivalent thermal network model, is as follows: Two intermediate quantities are calculated based on the three-phase current: the average current value during the switching cycle and the effective current value. The average current is used to calculate the portion of conduction loss related to constant voltage drop, while the effective current value is used to calculate the portion of conduction loss related to dynamic resistance. Simultaneously, the instantaneous amplitude of the current (i.e., the current magnitude at the current moment) is extracted from the three-phase current. Combined with the DC bus voltage and switching frequency, the energy consumed by each switching action is estimated, and then multiplied by the number of switching operations per second to obtain the switching loss. The conduction loss and switching loss are added together to obtain the total power loss generated by the power device at the current moment (i.e., power loss). For example, for an insulated-gate bipolar transistor, the formula is: Conduction loss: ;

[0024] in, It is the on-state voltage drop related to junction temperature and current (from the relationship model in step S10). Average current value during the switching cycle It is the square of the effective value of the current. For dynamic resistance; in, The output characteristic curve of the power device is obtained through the dynamic resistance. - The slope is obtained by fitting the slope at different junction temperatures, or the standard temperature coefficient model can be used: ; in, This is the temperature coefficient of resistance, which is 0.003 to 0.005 / ℃ for IGBTs and 0.002 to 0.0035 / ℃ for SiC MOSFETs. Switching losses: ;

[0025] in, , This refers to the switching energy, derived from the family of switching energy curves in the power device datasheet. If the datasheet provides this information, a two-dimensional lookup table plus bilinear interpolation is used to obtain the switching energy under the current operating condition. If the datasheet does not provide this information, it is obtained through actual measurement using a dual-pulse test before bench testing. The dual-pulse test standard follows IEC 60747-9. The switching energy is modeled as follows: ; ; in, , The temperature coefficient of the switching energy is obtained from the power device datasheet. The total loss is: ;

[0026] The equivalent thermal network model employs a structure with multiple thermal resistances and thermal capacities connected in series to simulate the process of heat transfer from the chip junction to the coolant. In the model, the first thermal capacity corresponds to the thermal energy storage of the chip itself, and the first thermal resistance corresponds to the heat conduction path from the chip to the casing. Subsequent stages include casing to heat sink, heat sink to coolant (multiple stages), each with corresponding thermal resistance and thermal capacity. The coolant inlet temperature, as the outermost boundary condition of the entire network, is a known real-time measured value. For example, the specific calculation is as follows: A Cauer-type thermal network (lumped parameter model) is used to discretize the heat dissipation path of the power module into several thermal resistances. and heat capacity The concatenated network, model form: Forming into a shell: ; The thermal resistance between the chip junction and the case of power devices. The heat capacity of the chip itself; From shell to radiator: ; Thermal resistance between the casing and the heat sink Equivalent heat capacity of the outer shell and connecting layer; From radiator to coolant: ; Thermal resistance between the radiator and the coolant The heat capacity of the radiator itself; The model is as follows: ;

[0027] ;

[0028] in, For the junction temperature, This refers to the case temperature or radiator temperature. This is the coolant temperature (real-time measurement). , Each corresponds to a shell: ; Corresponding to the heat sink ; It should be noted that the parameters of the thermal network model ( , The method to obtain the transient thermal impedance curve is as follows: the power device datasheet provides the transient thermal impedance curve. The curve is decomposed into multi-order RC parameters using Foster or Cauer network fitting methods. For the required accuracy in this embodiment, a second-order RC network is sufficient, and the fitting method is the least squares method to fit the exponential function. ; This embodiment uses the transient thermal impedance curve provided in the device datasheet. Curve fitting yielded: ; At the start of each calculation step (e.g., every 10 milliseconds or every 100 milliseconds), the junction temperature, case temperature, and node temperatures of the previous moment are known. Power loss is used as a heat source and injected into the first thermally capacitive node. Then, based on the temperature difference across each thermal resistor, the heat flow through each thermal resistor at the current moment is calculated. The net heat within the current step is divided by the thermal capacity value to obtain the temperature increment within the step. This increment is then added to the temperature of the previous moment to obtain the new temperature at the current moment. This recursive update of node temperatures is repeated to obtain the junction temperature at each moment. The junction temperature reflects the true thermal state inside the power device. The recursively calculated junction temperature is fed back to the relational model in step S10 for table lookup or calculation of the on-state voltage drop at temperature. It also serves as a reference benchmark in the identification process of step S30. For example, the calculation process is as follows: The lumped parameter model is discretized into difference equations, and then recursively applied online using the first-order backward Euler method: ;

[0029] ;

[0030] in, For each recursive step size (e.g., every 10 milliseconds or every 100 milliseconds), the real-time current, bus voltage, switching frequency, and coolant temperature are read during each control cycle (e.g., 10 ms). Based on the relational model in step S10, the on-state voltage drop under the current junction temperature estimate is calculated, and then the conduction loss and switching loss are calculated. The loss values ​​and coolant temperature are substituted into the difference equation to update the junction temperature and intermediate node temperature, and the current dynamic junction temperature is output. ; Understandably, the significance of step S20 lies in the following: Step S20 utilizes the measurable three-phase current, DC bus voltage, switching frequency, and coolant inlet temperature from the test bench, combined with a loss model and a Cauer-type thermal network model, to recursively estimate the dynamic junction temperature of the power device online. This junction temperature value is directly used as the input to the relationship model in step S10, enabling the reference on-state voltage drop to be updated in real time with temperature. Simultaneously, it serves as a reference benchmark for the identification process in step S30, used to determine the temperature rise stage. This step solves the engineering problem that junction temperature cannot be directly measured, providing a key state variable for subsequent temperature-related compensation. Step S30: Under low-speed and light-load conditions, construct a voltage error extension model using the clamping interval near the current zero-crossing point, including two parameters to be identified: The first item is the total nonlinear error of the inverter, which consists of dead time and on-state voltage drop, where the on-state voltage drop varies with junction temperature. The second item is the zero-point drift of the current sensor; A recursive least squares algorithm with a forgetting factor is used to simultaneously identify two parameters to be identified. By analyzing the difference in the correlation between the error term and the current polarity, the temperature drift component of the on-state voltage drop is separated from the first term, and the value of the second term is obtained. In step S30, the low-speed, light-load operating condition specifically refers to: the motor speed being less than 10% of the rated speed and the motor output torque being less than 10% of the rated torque, and the corresponding phase current amplitude being less than 10% of the rated current; In step S30, the process of constructing the voltage error extension model using the clamping interval near the current zero-crossing point is as follows: Determine the clamping interval: In three-phase current, whenever a phase current changes from positive to negative or from negative to positive, it will pass through the zero point. A time window (e.g., the area where the current amplitude is less than 2% of the rated current) is set with the zero point as the center, which is called the clamping interval. It should be noted that the detection method for the clamping range is as follows: set the rated current. The detection threshold is 3%. When the absolute value of the current is less than the detection threshold Determined to enter the clamping zone, rated current. It is the peak value of the rated phase current output by the inverter, given by the motor parameters; it enters the threshold. Exit threshold That is, the current flows from large to small through the path. Enter at that time, pass through from small to large Exit at the specified time; the minimum dwell time is set to 50 microseconds to ensure that the clamping interval remains at least this duration before data processing, avoiding instantaneous noise interference and affecting voltage error signals. A moving average with a window length of 5 is performed to obtain the filtered voltage error signal. Only when the current polarity of three consecutive sampling points is consistent is the polarity used for parameter identification; otherwise, no update is performed. If the number of valid sampling points (satisfying the above filtering and having stable polarity) in a clamping interval is less than 10, all data in that interval is discarded and no recursive least squares algorithm update is performed (i.e., the A3 step update operation in subsequent step S30). It should be noted that within the clamping range, the switching state of the upper and lower bridge arms of the inverter is most significantly affected by the dead time, and the deviation between the output voltage and the ideal value is the largest. Voltage error extraction: Within the clamping range, the motor back electromotive force is approximately zero (due to the low speed) and the stator resistance voltage drop is very small. Therefore, the difference between the command voltage issued by the controller and the voltage actually applied to the motor terminals mainly comes from the nonlinear effects of the inverter (dead time and conduction voltage drop) and the zero-point drift of the current sensor. The original voltage error signal is obtained by subtracting the command voltage from the phase voltage of this range in real time (or by reconstructing the DC bus voltage and switch state). Establish a voltage error extension model: represent the original voltage error as the sum of two terms: First item The total nonlinear error of the inverter is related to the polarity of the current (positive compensation for positive current and negative compensation for negative current). Its magnitude is affected by the dead time and the on-state voltage drop, with the on-state voltage drop varying with the junction temperature. Second item The equivalent voltage error caused by the zero-point drift of the current sensor is manifested as a constant offset (approximately constant within the clamping interval) that is independent of the current polarity. The constructed voltage error extension model is as follows: ;

[0031] in, Indicates the polarity of the current (positive is +1, negative is -1). In step S30, a recursive least squares algorithm with a forgetting factor is used to simultaneously identify two parameters to be identified. By analyzing the difference in correlation between the error term and the current polarity, the temperature drift component of the on-state voltage drop is separated from the first term, and the value of the second term is obtained as follows: A1, Data Acquisition: Under low-speed and light-load conditions, the system enters the clamping interval near the current zero-crossing point and continuously acquires multiple sampling points. Each sampling point records two pieces of information: the polarity of the current (positive or negative) and the measured voltage error value. A2, Initial settings for recursive least squares: Set the initial estimates of total nonlinear error and equivalent voltage error to 0; set a forgetting factor (0.98) to make the algorithm pay more attention to the most recently collected data, so as to track the slow changes of parameters with temperature; maintain an internal covariance parameter, which controls the step size of each adjustment of total nonlinear error and equivalent voltage error, and initially set the covariance to 1000. A3, point-by-point recursive update: For each new data point (current polarity, measured voltage error), perform the following operations: Using the current total nonlinear error and equivalent voltage error estimate, combined with the current current polarity, predict the voltage error value; subtract the measured voltage error from the predicted voltage error value to obtain the predicted error; calculate the adjustment range according to the magnitude and direction of the predicted error using the least squares criterion, where the adjustment range is proportional to the predicted error and the current covariance value; update the estimates of the total nonlinear error and equivalent voltage error: new estimate = old estimate + adjustment range; update the covariance parameter to gradually decay (determined by the least squares algorithm). The specific recursive steps are: For each new data point... Execute once, The number of the sampling point (an integer, starting from 1 and incrementing). No. The polarity of the current at each sampling point is taken as +1 when the current is positive and -1 when it is negative. No. The measured voltage error value at each sampling point is obtained by subtracting the command voltage from the actual voltage within the clamping interval; Calculate the predicted voltage error: ;in, The observation vector is a 2-dimensional column vector, defined as follows: , Indicates transpose; For the previous moment (the first moment) (Step) Estimated value of the parameter vector; Calculate the prediction error: ; The first time predicted based on the parameters of the previous time step Step voltage error value; Calculate the gain vector: ; For the first The covariance matrix of the step; The forgetting factor is set to 0.98 in this embodiment; Update parameter estimates: ;at this time, This is the current identification result; For the first The total nonlinearity error of the inverter and the zero-point drift of the current sensor were identified step by step; Update the covariance matrix: ;in It is a row vector. We obtain a 2×2 matrix; A4 continuously records the parameter update amounts of the last 10 sampling points. If the parameter update amount of the total nonlinear error is less than 0.1 volts and the parameter update amount of the equivalent voltage error is less than 0.01 volts, then convergence is determined, and the current identification result, i.e., the total nonlinear error and the equivalent voltage error, is output. Specifically, the calibration method for the preset threshold in the convergence criterion: A complete temperature rise test (from cold state to thermal steady state) was conducted on the test bench, and the parameter update at each sampling point was recorded. (Parameter update amount of total nonlinear error) and (Parameter update amount of equivalent voltage error), extract the parameter update amount data during the thermal steady-state stage (junction temperature change rate <1℃ / min, usually lasting more than 2 minutes), and calculate its standard deviation. and The convergence threshold is set to 3σ to ensure that the parameter update amount during the stable phase has a 99.7% probability of being less than the threshold. In this embodiment, the parameter update amount threshold for the total nonlinear error after calibration is 0.1V, and the parameter update amount threshold for the equivalent voltage error is 0.01V. Convergence is determined when all 10 consecutive sampling points are less than the threshold. If convergence fails within 5 seconds, the recursive least squares algorithm is forcibly reset (the covariance matrix is ​​reset to the initial value), and the identification is performed again. A5, when the test bench is at room temperature (e.g., 25 degrees Celsius), perform the same low-speed, light-load identification process once. At this time, the temperature drift of the on-state voltage drop is approximately zero. The total nonlinear error identified is the fixed voltage loss caused by the dead time. The total nonlinear error is then used as the dead-time reference value. At the current temperature, the total nonlinear error identified is subtracted from the dead-time reference value to obtain the temperature drift component of the on-state voltage drop. Finally, the temperature drift component of the on-state voltage drop and the zero-point drift of the current sensor (equivalent voltage error) are output. Understandably, the significance of step S30 is as follows: Under low-speed and light-load conditions, step S30 utilizes the clamping interval near the current zero-crossing point to construct a voltage error extended model that includes the total nonlinear error of the inverter and the zero-point drift of the current sensor. It uses a recursive least squares algorithm with a forgetting factor to simultaneously identify the two parameters, and separates the temperature drift component of the on-state voltage drop by subtracting the dead zone reference value calibrated at room temperature. This step decouples and identifies the two types of errors that drift with temperature (on-state voltage drop and sensor zero drift) under the same operating condition, providing accurate temperature drift components and zero drift correction amounts for step S40. Step S40: The relational model and the on-state voltage drop temperature drift component are fused through Kalman filtering to obtain the optimal compensation voltage value at the current junction temperature, which is then superimposed on the inverter command voltage in real time to correct the voltage loss in dead zone compensation; at the same time, the identified zero-point drift is fed back to the current sampling channel for online zeroing correction. In step S40, the process of fusing the relational model with the on-state voltage drop temperature drift component using Kalman filtering to obtain the optimal compensation voltage value at the current junction temperature is as follows: Define state variables: Given the optimal compensation voltage value at the current moment, the state transition model (assuming the optimal compensation voltage changes slowly in adjacent moments due to temperature drift): ; in, For process noise, For process noise covariance (scalar); Observed variables That is, the sum of the reference on-state voltage drop output by the relational model and the temperature drift component separated by the recursive least squares algorithm; Observation equation: ; in, To measure noise, To measure the noise covariance (scalar); Kalman filter recursive formula: Assume initial state estimation Initial estimation error covariance ; The prediction steps are as follows: ; ; The update steps are as follows: ; ; ; Output the optimal compensation voltage value at the current moment. ; It should be noted that the process noise covariance and measurement noise covariance The calibration method is as follows: Under steady-state conditions on the test bench (constant junction temperature), continuous data collection was performed. A total of 200 points were used to calculate the sample variance. The estimated value: , The average; During the temperature rise process, the actual change in the optimal compensation voltage at adjacent time points is recorded. Take the variance as The calibration results obtained in this embodiment are as follows: ; It should be noted that once the Kalman filter reaches steady state, the gain... tending to a constant At this point, the filtering equation degenerates into: ; That is, a weighted average with fixed coefficients, in this embodiment The value is 0.7; In step S40, the process of real-time superimposing onto the inverter command voltage to correct voltage loss in dead-time compensation is as follows: Read the original command voltage output from the current loop and the basic compensation voltage output from the fixed dead zone compensation module, and then add the optimal compensation voltage value, the original command voltage and the basic compensation voltage to obtain the total command voltage. In step S40, the process of feeding back the identified zero-point drift to the current sampling channel for online zero-point correction is as follows: At the beginning of each current sampling, the original current sampling value output by the current sampling channel is read, and the zero drift value is subtracted from the original current sampling value to obtain the corrected current value. It is understandable that the significance of step S40 is as follows: Step S40 adds the reference on-state voltage drop output by the relational model in step S10 to the on-state voltage drop temperature drift component separated in step S30 to obtain a preliminary correction value. Then, it merges the weighted average with the compensation value at the previous moment through a fixed weight coefficient to obtain the optimal compensation voltage value at the current junction temperature, and adds it to the inverter command voltage to correct the voltage loss in dead zone compensation. At the same time, the identified zero drift is directly used for the zero-point correction of the current sampling channel. This step completes the temperature adaptation of dead zone compensation and the online correction of sensor zero drift in the same framework.

[0032] Example 2

[0033] Please see Figure 2 As shown in the embodiment of the present invention, a temperature rise compensation control system for bench testing of an electric vehicle electric drive system includes the following modules: Fitting Analysis Module: For the target device, the forward voltage drop is measured at different junction temperatures using a constant temperature chamber. Combined with the influence of different current levels, a model is fitted to fit the relationship between the forward voltage drop and the junction temperature. Online recursive module: Utilizing the measurable operating parameters of the test bench, combined with device losses and equivalent thermal network models, the current dynamic junction temperature is recursively calculated online, serving as the input to the relationship model in the fitting analysis module and the identification reference benchmark in the error expansion module; Error extension module: Under low-speed and light-load conditions, a voltage error extension model is constructed using the clamping interval near the current zero-crossing point, including two parameters to be identified: The first item is the total nonlinear error of the inverter, which consists of dead time and on-state voltage drop, where the on-state voltage drop varies with junction temperature. The second item is the zero-point drift of the current sensor; A recursive least squares algorithm with a forgetting factor is used to simultaneously identify two parameters to be identified. By analyzing the difference in the correlation between the error term and the current polarity, the temperature drift component of the on-state voltage drop is separated from the first term, and the value of the second term is obtained. Correction and compensation module: The relationship model and the on-state voltage drop temperature drift component are fused through Kalman filtering to obtain the optimal compensation voltage value at the current junction temperature, which is then superimposed on the inverter command voltage in real time to correct the voltage loss in dead zone compensation; at the same time, the identified zero-point drift is fed back to the current sampling channel for online zeroing correction.

[0034] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for temperature rise compensation control in bench testing of an electric vehicle's electric drive system, characterized in that: Includes the following steps: step S10: For the target device, the forward voltage drop at different junction temperatures is measured by a constant temperature chamber. Combined with the influence of different current levels, a model is fitted to fit the relationship between the forward voltage drop and the junction temperature. Step S20: Using the measurable operating parameters of the test bench, combined with the device loss and equivalent thermal network model, the current dynamic junction temperature is recursively calculated online, serving as the input to the relationship model in step S10 and the identification reference benchmark in step S30. Step S30: Under low-speed and light-load conditions, construct a voltage error extension model using the clamping interval near the current zero-crossing point, including two parameters to be identified: The first item is the total nonlinear error of the inverter, which consists of dead time and on-state voltage drop, where the on-state voltage drop varies with junction temperature. The second item is the zero-point drift of the current sensor; A recursive least squares algorithm with a forgetting factor is used to simultaneously identify two parameters to be identified. By analyzing the difference in the correlation between the error term and the current polarity, the temperature drift component of the on-state voltage drop is separated from the first term, and the value of the second term is obtained. Step S40: The relationship model and the on-state voltage drop temperature drift component are fused through Kalman filtering to obtain the optimal compensation voltage value at the current junction temperature, which is then superimposed on the inverter command voltage in real time to correct the voltage loss in dead zone compensation. Simultaneously, the identified zero-point drift is fed back to the current sampling channel for online zeroing correction.

2. The temperature rise compensation control method for bench testing of an electric vehicle electric drive system according to claim 1, characterized in that: The process of measuring the on-state voltage drop at different junction temperatures in step S10 is as follows: The target device is placed in a programmable constant temperature chamber, and thermocouples are attached to the surface of the device housing to monitor the test temperature; multiple discrete junction temperature points are selected according to the operating temperature range of the device. Set the constant temperature chamber to the target temperature, and wait until the temperature inside the chamber stabilizes and the temperature deviation between the device housing and the temperature inside the chamber is within the allowable range. Using a transistor curve tracer or a high-current source meter, apply multiple conduction currents of different levels to the device, and use a pulse test method to read the saturation voltage drop between the collector and emitter within the pulse duration.

3. The temperature rise compensation control method for bench testing of an electric vehicle electric drive system according to claim 2, characterized in that: The process of fitting the relational model in step S10 is as follows: The on-state voltage drop at different junction temperatures and current levels is integrated into multiple sets of three-dimensional data. A model is adopted in which the on-state voltage drop varies linearly with junction temperature, and the slope of the change is different for different currents. For each fixed current level, linear fitting is performed independently, and the reference on-state voltage drop and temperature coefficient for the current level are obtained by least squares method. Store the reference on-state voltage drop and temperature coefficient corresponding to each current level. During online calculation, the reference on-state voltage drop value under any junction temperature and current can be obtained by looking up the table.

4. The temperature rise compensation control method for bench testing of an electric vehicle electric drive system according to claim 1, characterized in that: The measurable operating parameters of the test bench and the methods for obtaining them in step S20 are as follows: Operating parameters include three-phase current, DC bus voltage, switching frequency, and coolant inlet temperature; The DC bus voltage is acquired by a voltage Hall sensor installed between the positive and negative terminals of the DC bus; the three-phase current is acquired by a Hall current sensor installed on the three-phase output lines of the inverter. The switching frequency is read directly from the controller's pulse width modulation module; the coolant inlet temperature is collected by a temperature sensor located at the inlet of the cooling circuit.

5. The temperature rise compensation control method for bench testing of an electric vehicle electric drive system according to claim 4, characterized in that: The process of online recursion of the current dynamic junction temperature in step S20 is as follows: The average current and effective current during the switching cycle are calculated based on the three-phase current. The average current is used to calculate the portion of the conduction loss related to constant voltage drop, and the effective current is used to calculate the portion of the conduction loss related to dynamic resistance. The instantaneous amplitude is extracted from the three-phase current, and the switching loss is estimated by combining the DC bus voltage and the switching frequency. The total power loss is obtained by adding the conduction loss and the switching loss. Using a Cauer-type thermal network model, the power loss is injected into the first thermal capacity node. The heat flow is calculated based on the temperature difference across each thermal resistance. The temperature increment is obtained by dividing the net heat within the current step by the thermal capacity value. The temperature of each node is updated recursively, and the current dynamic junction temperature is output.

6. The temperature rise compensation control method for bench testing of an electric vehicle electric drive system according to claim 1, characterized in that: Step S30 involves the process of handling low-speed, light-load conditions and constructing the voltage error extension model as follows: In three-phase current, the region centered on the current zero-crossing point and where the current amplitude is less than a fixed proportion of the rated current is defined as the clamping zone. Within the clamping range, the difference between the command voltage issued by the controller and the voltage actually applied to the motor terminal is taken as the original voltage error, and expressed as the total nonlinear error of the inverter multiplied by the current polarity plus the zero-point drift of the current sensor.

7. The temperature rise compensation control method for bench testing of an electric vehicle electric drive system according to claim 6, characterized in that: The process of identifying parameters and separating temperature drift components in step S30 is as follows: Multiple sampling points are continuously collected within the clamping interval. The polarity of the current and the measured voltage error are recorded at each sampling point. The initial estimated values ​​of the total nonlinear error of the inverter and the zero-point drift of the current sensor are set to zero. For each new data point, the voltage error is predicted and the difference is calculated with the measured value. The estimated values ​​of the total nonlinear error and the zero-point drift of the current sensor are adjusted according to the predicted error. When the parameter update amount of multiple consecutive sampling points is less than the preset threshold, convergence is determined and the current identification result is output; the dead zone reference value is obtained by performing identification once at room temperature, and the temperature drift component of the conduction voltage drop is obtained by subtracting the dead zone reference value from the total nonlinear error of the current identification.

8. The temperature rise compensation control method for bench testing of an electric vehicle electric drive system according to claim 1, characterized in that: The process of obtaining the optimal compensation voltage value in step S40 is as follows: Based on the relational model, the reference on-state voltage drop value is output according to the current junction temperature and current level, and added to the on-state voltage drop temperature drift component to obtain the preliminary correction value; the optimal compensation voltage value of the previous moment is read, and the preliminary correction value is fused with the optimal compensation voltage value of the previous moment through Kalman filtering to obtain the optimal compensation voltage value of the current moment.

9. The temperature rise compensation control method for bench testing of an electric vehicle electric drive system according to claim 8, characterized in that: The process of superimposing the compensation voltage in step S40 is as follows: Read the original command voltage output by the current loop, read the basic compensation voltage output by the fixed dead zone compensation module, and add the original command voltage, basic compensation voltage and optimal compensation voltage value together to obtain the total command voltage. The total command voltage is used to generate the actual switching signal to drive the inverter, thereby correcting the voltage loss caused by temperature changes in dead-zone compensation.

10. The temperature rise compensation control method for bench testing of an electric vehicle electric drive system according to claim 9, characterized in that: The online zeroing correction process in step S40 is as follows: At the start of each current sampling, the original current sample value is read, and the zero-point drift value is subtracted from the original current sample value to obtain the corrected current value.

Images