Insulation detection method of electric automobile

Abstract

The invention discloses an insulation detection method for an electric automobile, belongs to the technical field of high-voltage safety monitoring of the electric automobile, and solves the problem that the insulation detection precision is reduced due to electromagnetic interference, impedance fluctuation and capacitive effect coupling in a dynamic load switching process. According to the scheme, a test sequence of discrete frequency points is applied to a high-voltage system; measuring a response current value through a current sensor and calculating a ground distributed capacitance value; connecting a test voltage source between positive and negative electrodes of the high-voltage system through a coupling circuit; applying a first pulse signal during no-load, dynamically setting the width according to the capacitance value, measuring a first response current value, and calculating an initial insulation resistance value; a second pulse signal with the same parameter is applied in the load state window, a second response current value is measured, and differential filtering is carried out in combination with the initial value to eliminate load noise; and comparing the filtered insulation resistance value with a threshold value, and outputting a fault signal when the insulation resistance value is lower than the threshold value. The method is used for monitoring the insulation state of the high-voltage system in real time and accurately early warning faults.

Description

Technical Field

[0001] The present invention belongs to the technical field of high-voltage safety monitoring of electric vehicles, and in particular relates to an insulation resistance detection method for a high-voltage system of an electric vehicle based on dynamic pulse response and differential filtering. Background Art

[0002] Monitoring the insulation condition of electric vehicles' high-voltage electrical systems is crucial for ensuring safe vehicle operation. Under dynamic load switching conditions, the accuracy of online testing of high-voltage system insulation performance decreases significantly. This problem is primarily due to strong electromagnetic interference introduced by load transients and the time-varying nature of system parameters. Existing testing methods maintain basic accuracy under steady-state conditions, but during dynamic processes such as motor startup and shutdown and sudden power changes, measured values ​​can deviate significantly, potentially leading to missed or false fault reports.

[0003] During dynamic load switching, the switching action of the motor controller's power devices generates high-frequency common-mode interference, with peak values ​​reaching several amperes. This interference couples to the insulation detection circuit through the parasitic capacitance of the high-voltage cable to ground, where it aliases with the actual insulation leakage current in the time domain. Because the interference spectrum overlaps with the detection signal's frequency band (typically concentrated in the 2kHz-20MHz and DC-100Hz bands), conventional filtering methods struggle to effectively isolate the characteristic signal. Using high-order filters to suppress interference introduces millisecond-level phase delays, making it impossible to track rapid load switching. Furthermore, when the high-voltage system switches between no-load and loaded states, the equivalent impedance of the positive and negative poles to ground exhibits nonlinear jumps. For example, when the motor winding is connected, the AC-side impedance is connected in parallel to the DC bus, causing the system's total impedance to ground to drop by 30%-50%. This sudden change invalidates detection models based on static calibration parameters, causing the calculated insulation resistance to deviate from the actual value by over 200%.

[0004] The transient response of distributed capacitance further exacerbates measurement errors. The distributed capacitance of the high-voltage system to the ground (typical value 0.1-1μF) generates charging and discharging currents at the moment of load switching, and this current is superimposed on the leakage current within the detection pulse period. When the pulse width is fixed: the capacitance does not reach a steady state under a short pulse, and the capacitive current component is mistakenly included in the leakage current, resulting in a low calculated insulation resistance value; although long pulses can suppress the capacitive influence, the detection period is extended to hundreds of milliseconds, which cannot meet the real-time monitoring requirements of dynamic processes. Experiments show that under a test voltage of 500V, it takes more than 15ms for the capacitive current of a 10μF distributed capacitor to decay to a stable state. During this period, the superimposed transient current can drown out the microampere leakage current signal.

[0005] Existing technologies for addressing these issues have limitations. The unbalanced bridge method relies on switching standard resistors, but bus voltage fluctuations (±10% of rated value) caused by load switching can increase the error in the voltage divider ratio calculation to over 25%. The balanced bridge method cannot detect bipolar synchronous degradation faults. The low-frequency AC injection method requires multi-frequency scanning (e.g., 5Hz / 50Hz / 100Hz), and a single test takes over 200ms, far longer than the typical load switching time (50-100ms). While the DC high-voltage injection method improves immunity to capacitive interference, the injected power supply and drive system have poor electromagnetic compatibility, and inverter switching noise can degrade the detection signal-to-noise ratio to below -10dB. Hall-effect leakage current sensor solutions are affected by temperature gradients and electromagnetic fields. When the cabin temperature change rate exceeds 5°C / min, the sensor zero-drift error can reach 20% of the full-scale range. Furthermore, under dynamic operating conditions, it cannot distinguish between capacitive and resistive leakage currents.

[0006] The essence of these problems is that insulation detection requires fast and accurate extraction of microampere-level current in a strongly nonlinear system, while electromagnetic interference, impedance fluctuations, and capacitive effects under dynamic conditions form multiple coupled disturbances. Especially at the moment of load switching, the system has millisecond-level transient processes, and its characteristic signals are buried in the noise background. Due to principle limitations or insufficient dynamic adaptability, existing methods are difficult to complete effective measurements under conditions where the interference amplitude exceeds 100 times that of the signal. Its technical bottlenecks are concentrated in two points: first, there is a lack of a rapid separation mechanism for transient interference and characteristic signals; second, it is impossible to establish a real-time matching model between time-varying system parameters (such as distributed capacitance, equivalent impedance) and detection timing. This results in the reliability of insulation monitoring under dynamic conditions being significantly lower than that under static conditions, which constitutes a weak link in the high-voltage safety protection system of electric vehicles. Summary of the Invention

[0007] An object of the present invention is to solve at least the above problems and to provide at least the advantages which will be described hereinafter.

[0008] The present invention at least solves the following technical problems:

[0009] Solve the problem of reduced insulation detection accuracy caused by electromagnetic interference, impedance fluctuation and capacitive effect coupling during dynamic load switching of electric vehicles, and avoid missed or false fault reports.

[0010] Solve the problem of measurement error caused by the pulse response not reaching steady state due to the difference in distributed capacitance of the high-voltage system, and ensure that the capacitance parameters dynamically adapt to the detection timing.

[0011] By associating the capacitance with the lower limit of insulation resistance through the time constant, the pulse width can be accurately set to suppress transient capacitive current interference.

[0012] Solve the problem of single-point measurement being interfered with due to the long scanning time of multiple frequency points in the test frequency sequence in a complex electromagnetic environment, and improve the efficiency and anti-interference performance of capacitance detection.

[0013] Solve the problem of insulation resistance eigenvalue drift caused by temperature gradient during high-voltage system operation in differential filtering processing, and avoid fault misjudgment caused by temperature changes.

[0014] Solve the problem of compensation inaccuracy caused by the temperature sensor being only arranged in an external accessible part and unable to reflect the actual temperature of the key insulation area inside the high-voltage system.

[0015] The measured temperature and theoretical temperature are integrated through weight distribution to improve the accuracy of temperature estimation of insulation weak points.

[0016] Clarify the physical parameters of key heat conduction paths in the thermal network topology model to ensure that the model matches the actual heat conduction characteristics.

[0017] This solves the problem of being unable to accurately locate the faulty pole when asymmetric insulation degradation occurs simultaneously at the positive and negative poles of the high-voltage system, and enables independent detection of each pole.

[0018] Through polarity differential filtering and independent threshold judgment, it outputs fault polarity location signals to support precise maintenance intervention.

[0019] In order to achieve these objects and other advantages of the present invention, a method for detecting insulation of an electric vehicle is provided, comprising the following steps:

[0020] S0: Applying a predefined test frequency sequence to the high-voltage system, the sequence comprising at least three discrete sinusoidal wave frequency points between 50 Hz and 200 Hz, measuring the response current value at each frequency point using a current sensor, and transmitting the measured value to a microprocessor; the microprocessor calculates the distributed capacitance of the high-voltage system to ground based on the response current value;

[0021] S1: Connect the test voltage source between the positive and negative poles of the high-voltage system of the electric vehicle through a coupling circuit. The coupling circuit includes an isolation transformer and a current-limiting resistor. The output voltage range of the test voltage source is 50V to 500V.

[0022] S2: When the high-voltage system is in a no-load state, applying a first pulse signal of a test voltage source, the duration of which is dynamically determined by the microprocessor based on the distributed capacitance to ground calculated in step S0; measuring a first response current value under the first pulse signal by a current sensor, and transmitting the first response current value to the microprocessor; and calculating an initial insulation resistance value of the high-voltage system based on the first response current value and the test voltage value by the microprocessor;

[0023] S3: within a set time window after the high-voltage system switches to a load operation state, applying a second pulse signal of the test voltage source, wherein the duration of the second pulse signal is the same as that of the first pulse signal and is dynamically determined by the microprocessor, and the amplitude range of the second pulse signal is the same as that of the first pulse signal; measuring a second response current value under the second pulse signal by a current sensor, and transmitting the second response current value to the microprocessor; the microprocessor calculates the insulation resistance value under the load state based on the second response current value and the test voltage value, and performs differential filtering processing in combination with the initial insulation resistance value to eliminate noise interference caused by load fluctuations;

[0024] S4: The microprocessor compares the insulation resistance value after differential filtering with a preset threshold range of 100kΩ to 10MΩ. When the insulation resistance value is lower than the preset threshold range, a fault signal is output to the vehicle control unit.

[0025] Preferably, the microprocessor calculates the distributed capacitance value of the high-voltage system to the ground based on the response current value, the applied sinusoidal voltage amplitude and the frequency value at each frequency point; based on the calculated distributed capacitance value to the ground, the microprocessor dynamically adjusts the duration of all pulse signals in subsequent insulation detection, including the duration of the first pulse signal and the second pulse signal.

[0026] Preferably, the rule for dynamic adjustment is: set the duration of the pulse signal to the calculated capacitance value and R min The value corresponding to the time constant range of 1.5 to 3 times the product; in the step of measuring the response current value under the first pulse signal by the current sensor, the measurement occurs after the first pulse signal is applied for a duration dynamically adjusted by the microprocessor; in the step of measuring the response current value under the second pulse signal by the current sensor, the steady-state response current value of the current sensor is collected at the end of the duration of the application of the first pulse signal; wherein, R min The lower limit of the insulation resistance of the high voltage system to ground is preset to 100kΩ.

[0027] Preferably, in the step of applying a set of predefined test frequency sequences to the high voltage system, the test frequency sequences are applied in the following manner:

[0028] 1) applying a step wave test voltage signal to the high voltage system through the coupling circuit, wherein the rise time of the step wave test voltage signal is less than 100 μs and the step voltage jumps from 0 V to a target voltage value, wherein the target voltage value range is set to 10 V to 30 V;

[0029] 2) After the step wave test voltage signal reaches the target voltage value, maintaining the constant voltage for 5 ms to 20 ms, and collecting the transient current response waveform during the duration by the current sensor;

[0030] 3) The collected transient current response waveform is transmitted to the microprocessor, and the microprocessor performs exponential function fitting on the transient current response waveform, extracts the current decay time constant τ, and calculates the current decay time constant τ according to the formula:

[0031] C=τ / R min ,

[0032] Calculate the distributed capacitance of the high voltage system to ground, where τ is the current decay time constant, R min A lower limit of 100 kΩ is preset for the insulation resistance of the high-voltage system to ground, which is used to simplify the equivalent RC circuit model.

[0033] In the dynamic adjustment rule, the duration of the pulse signal is set to the calculated distributed capacitance to ground and R min 1.5 to 3 times the product of .

[0034] Preferably, after the step of calculating the insulation resistance value under load based on the second response current value and the test voltage value by the microprocessor, the following steps are added:

[0035] 1) Synchronously collect the surface temperature values ​​of the positive busbar, negative busbar, and power battery casing of the high-voltage system through at least two temperature sensors distributed at key nodes of the high-voltage system;

[0036] 2) The collected temperature value is transmitted to the microprocessor; the microprocessor calls a pre-stored insulation material resistance-temperature relationship model, which is expressed as:

[0037]

[0038] Among them, R T Indicates the theoretical value of insulation resistance at temperature T, unit: Ω; R0 represents the insulation resistance value at reference temperature T0, set to 298K; β is the thermal sensitivity coefficient of the insulation material, ranging from 3000K to 5000K;

[0039] 3) The microprocessor calculates the weighted average temperature T of each measuring part of the high-voltage system based on the currently collected temperature value avg , and combining the insulation material resistance-temperature relationship model, temperature compensation is performed on the positive electrode insulation resistance value and the negative electrode insulation resistance value under the load state to obtain the standardized positive / negative electrode insulation resistance value at a reference temperature of 298K;

[0040] In the differential filtering processing step, the standardized positive / negative insulation resistance values ​​are used in the calculation; in the step of comparing the insulation resistance value with the preset threshold range, the standardized insulation resistance value is used for judgment.

[0041] Preferably, in the step of synchronously collecting the surface temperature values ​​of the positive busbar, the negative busbar, and the power battery housing of the high-voltage system through at least two temperature sensors distributed at key nodes of the high-voltage system, the following operation is added:

[0042] 1) Based on the physical structural parameters of the high-voltage system, a thermal network topology model is constructed in the microprocessor;

[0043] 2) The synchronously collected temperature values ​​are input as boundary conditions into the thermal network topology model. The microprocessor solves the thermal network steady-state equations to calculate the theoretical temperature values ​​of the insulation weak points within the high-voltage system. The insulation weak points include the central area of ​​the insulating partition between the power battery modules and the joint surface of the high-voltage connector.

[0044] Preferably, in the step of calculating the weighted average temperature of each measuring location of the high-voltage system, the theoretical temperature value of the insulation weak point is included in the weighted calculation, wherein the temperature weight distribution range of the insulation weak point is 30% to 40%, the total weight of the measured temperature of the positive busbar, the negative busbar and the power battery casing is 60% to 70%, and the sum of all weight coefficients is 1;

[0045] The calculation formula of the temperature weighted average value is:

[0046]

[0047] Among them, T sensor,i is the measured value of the i-th temperature sensor, K i is its weight coefficient; T virtual,j is the theoretical temperature value of the jth insulation weak point, k j is its weight coefficient; and satisfies ∑k i +∑k j =1.

[0048] Preferably, the model includes heat conduction paths between the high-voltage system positive busbar, negative busbar, power battery module insulating partition and battery box, and the thermal conductivity of each path is set according to the material properties, wherein the thermal conductivity range of aluminum is set to 200W / (m·K) to 240W / (m·K), and the thermal conductivity range of epoxy resin insulation board is set to 0.2W / (m·K) to 0.5W / (m·K).

[0049] Preferably, in the step of calculating the insulation resistance value under load by the microprocessor based on the second response current value and the test voltage value, the insulation resistance value is divided into the insulation resistance value R of the positive electrode to the ground of the high voltage system and the insulation resistance value R of the positive electrode to the ground of the high voltage system. pos And the insulation resistance value R neg Calculate separately, including the following operations:

[0050] 1) During the application of the first pulse signal, the first positive electrode response current value I between the positive terminal and the ground terminal of the high voltage system is synchronously measured by the current sensor pos1 and the first negative electrode response current value I between the negative terminal and the ground terminal neg1 , and I pos1 and I neg1 Transmit to microprocessor;

[0051] 2) The microprocessor calculates the insulation resistance of the positive and negative electrodes under no-load conditions according to the formula: test is the test voltage value;

[0052] During the application of the second pulse signal, the second positive electrode response current value I between the positive terminal and the ground terminal is synchronously measured by the current sensor. pos2 And the second negative electrode response current value I between the negative terminal and the ground terminal neg2 , and transmit it to the microprocessor, which calculates the positive and negative insulation resistance values ​​under load according to the formula.

[0053] Preferably, in the differential filtering step, the following steps are performed independently for the positive and negative electrodes:

[0054] The positive insulation resistance value R' under load pos The positive insulation resistance value R under no-load condition pos Perform differential filtering to obtain the final insulation resistance value R of the positive electrode pos-final ,

[0055] The negative electrode insulation resistance value R' under load neg The negative insulation resistance value R under no-load condition neg Perform differential filtering to obtain the final insulation resistance value R of the negative electrode neg-final ,

[0056] The functional expression of the differential filtering process is:

[0057] R pos-final =R pos +K·(R′ pos -R pos );

[0058] R neg-final =R neg +K·(R′ neg -R neg );

[0059] Among them, the value range of the filter coefficient K is 0.2 to 0.8;

[0060] In the step of comparing the insulation resistance value with the preset threshold range, Rpos-final and R neg-final Judge separately: when R pos-final or R neg-final When the voltage is lower than the preset threshold value range of 100kΩ to 10MΩ, a fault location signal of the corresponding polarity is output to the vehicle control unit.

[0061] The present invention has at least the following beneficial effects:

[0062] Through dual-pulse coordinated detection in both no-load and loaded states and differential filtering, the system effectively suppresses the impact of common-mode interference and equivalent impedance fluctuations generated during dynamic load switching on measurement results. Furthermore, the system dynamically adjusts the pulse width based on pre-calculated distributed capacitance to ensure that the measurement point is within the steady-state region after the capacitive current has decayed, thus avoiding misjudgment of leakage current due to transient response superposition.

[0063] A multi-frequency sine wave sequence is used to calculate the system's distributed capacitance to ground in real time, ensuring that the insulation detection pulse width is strictly matched with the current capacitance parameters. This eliminates the pulse timing mismatch caused by capacitance drift and improves the capacitive current suppression capability under dynamic working conditions.

[0064] By constraining the pulse duration with a time constant, the measurement window is forced to cover the steady-state operating range of the RC circuit, significantly reducing the flooding effect of the transient current of the distributed capacitance on the microampere-level leakage signal and improving the accuracy of resistive component extraction.

[0065] The step wave step response combined with exponential fitting is used to directly extract the decay time constant, significantly shortening the distributed capacitance detection cycle, avoiding the electromagnetic interference-sensitive period of multi-frequency scanning, and improving the detection robustness in complex noise environments.

[0066] A real-time compensation mechanism based on the insulation material resistance-temperature model eliminates the insulation resistance eigenvalue drift caused by temperature gradients, ensuring that fault threshold judgment is not affected by ambient temperature changes.

[0067] The theoretical temperature of insulation weak points is inverted using the thermal network topology model to solve the temperature measurement blind spot problem where sensors cannot be directly deployed in key areas, and to improve the spatial coverage accuracy of temperature compensation.

[0068] The weight distribution mechanism prioritizes reflecting the temperature rise risk of weak points in insulation, while taking into account the reliability of external measured temperature, and enhances the early warning capability of insulation degradation caused by internal heat accumulation.

[0069] The precise setting of the thermal conductivity of aluminum and epoxy resin ensures that the thermal network model is consistent with the actual physical heat transfer characteristics, supporting cross-platform model reuse and rapid deployment.

[0070] The synchronous and independent detection mechanism for positive and negative pole leakage currents accurately captures single-pole insulation degradation events, overcoming the limitations of traditional methods in identifying bipolar asymmetric faults.

[0071] Polarity differential filtering combined with independent threshold judgment enables precise fault polarity location and improves the targeted maintenance intervention; the adaptive design of the filter coefficient enhances output stability under load disturbance.

[0072] Other advantages, objectives and features of the present invention will be reflected in part through the following description, and in part will be understood by those skilled in the art through study and practice of the present invention. DETAILED DESCRIPTION

[0073] The present invention is described in further detail below so that those skilled in the art can implement the invention with reference to the description.

[0074] It should be understood that terms such as “having”, “including” and “comprising” used herein do not preclude the existence or addition of one or more other elements or combinations thereof.

[0075] An insulation detection method for an electric vehicle comprises the following steps:

[0076] S0: Applying a predefined test frequency sequence to the high-voltage system, the sequence comprising at least three discrete sinusoidal wave frequency points between 50 Hz and 200 Hz, measuring the response current value at each frequency point using a current sensor, and transmitting the measured value to a microprocessor; the microprocessor calculates the distributed capacitance of the high-voltage system to ground based on the response current value;

[0077] S1: Connect the test voltage source between the positive and negative poles of the high-voltage system of the electric vehicle through a coupling circuit. The coupling circuit includes an isolation transformer and a current-limiting resistor. The output voltage range of the test voltage source is 50V to 500V.

[0078] S2: When the high-voltage system is in a no-load state, applying a first pulse signal of a test voltage source, the duration of which is dynamically determined by the microprocessor based on the distributed capacitance to ground calculated in step S0; measuring a first response current value under the first pulse signal by a current sensor, and transmitting the first response current value to the microprocessor; and calculating an initial insulation resistance value of the high-voltage system based on the first response current value and the test voltage value by the microprocessor;

[0079] S3: within a set time window after the high-voltage system switches to a load operation state, applying a second pulse signal of the test voltage source, wherein the duration of the second pulse signal is the same as that of the first pulse signal and is dynamically determined by the microprocessor, and the amplitude range of the second pulse signal is the same as that of the first pulse signal; measuring a second response current value under the second pulse signal by a current sensor, and transmitting the second response current value to the microprocessor; the microprocessor calculates the insulation resistance value under the load state based on the second response current value and the test voltage value, and performs differential filtering processing in combination with the initial insulation resistance value to eliminate noise interference caused by load fluctuations;

[0080] S4: The microprocessor compares the insulation resistance value after differential filtering with a preset threshold range of 100kΩ to 10MΩ. When the insulation resistance value is lower than the preset threshold range, a fault signal is output to the vehicle control unit.

[0081] This insulation testing method for electric vehicle high-voltage systems achieves accurate measurements under dynamic load conditions through the following steps. First, a test sequence consisting of at least three discrete frequency points ranging from 50 Hz to 200 Hz is applied to the high-voltage system. A current sensor collects the response current at each frequency point and transmits it to a microprocessor, which uses this current to calculate the distributed capacitance to ground in real time. A 50V to 500V test voltage source is then connected between the positive and negative terminals of the high-voltage system via an isolation transformer and a current-limiting resistor. When the high-voltage system is in an unloaded state, a first pulse signal is applied, the duration of which is dynamically determined by the distributed capacitance. The first response current is simultaneously collected, and the initial insulation resistance value is calculated. Within a set time window after the high-voltage system is switched to a loaded state, a second pulse signal with the same parameters as the first pulse is applied. The second response current is collected, and the insulation resistance value under load is calculated. A key improvement lies in differential filtering the load-state resistance value and the initial resistance value to eliminate common-mode interference and impedance fluctuation noise introduced by load switching. Finally, the filtered resistance value is compared with a threshold value between 100 kΩ and 10 MΩ. A fault signal is triggered if the value falls below the threshold.

[0082] Compared to the closest existing technology, the traditional unbalanced bridge method, this solution overcomes two major technical bottlenecks: First, the traditional method relies on switching standard resistors. Bus voltage fluctuations during load transients can cause voltage divider calculation errors to increase by more than 25%. This solution, however, uses coordinated no-load / load dual pulse detection to directly offset the systematic deviations caused by voltage fluctuations. Second, the traditional method uses a fixed pulse width, sampling the capacitive current before the distributed capacitance changes, resulting in the leakage current measurement value containing a capacitive component. This solution dynamically adjusts the pulse width based on the real-time calculated distributed capacitance, ensuring that the sampling point is within the time window when the capacitive current decays to a stable state, avoiding transient interference.

[0083] Further comparison with the low-frequency AC injection method shows that the existing technology requires the execution of 5Hz-100Hz multi-frequency scanning, and a single detection takes more than 200ms, which is much longer than the 50-100ms time scale of load switching and cannot track dynamic processes. This solution decouples capacitance detection from insulation measurement: a step wave step response is used in the capacitance pre-calculation stage, and a single detection can be completed within 20ms; the total duration of the double pulses in the insulation measurement stage is controlled within 50ms, realizing real-time monitoring synchronized with load switching. In addition, the signal-to-noise ratio of the traditional injection method deteriorates to below -10dB under inverter switching noise, while the differential filtering mechanism of this solution can remove common-mode noise and maintain effective signal integrity under motor start-stop conditions.

[0084] To address the temperature drift limitations of Hall-effect leakage current sensors, this implementation incorporates an optional integrated temperature compensation module. Traditionally, sensors are affected by the cabin's temperature gradient, resulting in zero drift reaching 20% ​​of full scale when the temperature change rate exceeds 5°C / min, and the leakage current component cannot be distinguished. This solution eliminates impedance fluctuation interference through differential filtering, and combined with temperature compensation, further eliminates the temperature drift component, achieving stable measurement under all operating conditions.

[0085] Operational considerations: The voltage source step response test must be performed during the high-voltage system's dormant period to avoid conflicts with the drive system. The differential filter coefficient defaults to 0.5 and can be adaptively adjusted based on the load disturbance intensity. An automotive-grade MCU with floating-point support is preferred for the microprocessor, ensuring millisecond-level completion of capacitance calculations and pulse control. Implementation cases have demonstrated that this method effectively suppresses abnormal jumps in insulation resistance values ​​during rapid motor acceleration / braking, resolving the issue of false faults caused by dynamic interference with traditional methods.

[0086] In another technical solution, the microprocessor calculates the distributed capacitance value of the high-voltage system to the ground based on the response current value, the applied sinusoidal voltage amplitude and the frequency value at each frequency point; based on the calculated distributed capacitance value to the ground, the microprocessor dynamically adjusts the duration of all pulse signals in subsequent insulation detection, including the duration of the first pulse signal and the second pulse signal.

[0087] The real-time measurement of distributed capacitance and dynamic adjustment of pulse parameters are achieved through the following steps. First, a predefined sinusoidal wave test frequency sequence is applied to the high-voltage system. The sequence contains at least three discrete frequency points (selected in the range of 50Hz to 200Hz). At each frequency point, the response current value of the high-voltage system is synchronously collected by the current sensor, and the current measurement values ​​of all frequency points are transmitted to the microprocessor. Based on Ohm's law and the capacitive reactance model, the microprocessor combines the sinusoidal voltage amplitude and frequency value applied at each frequency point to calculate the distributed capacitance value of the high-voltage system to the ground in real time. Subsequently, the duration of all pulse signals in subsequent insulation detection is dynamically adjusted according to this capacitance value, including the first pulse signal in the no-load state and the second pulse signal in the load state, to ensure that the pulse width strictly matches the current system distributed capacitance.

[0088] The closest existing technology is the traditional multi-frequency sine wave scanning method. This technique similarly calculates distributed capacitance by injecting sinusoidal signals at multiple frequencies and sampling the resulting currents. However, its implementation has significant drawbacks: it requires switching between at least three frequencies for independent measurement, and a single scan takes over 200ms. Because load switching transients typically complete within 50–100ms, traditional methods are unable to simultaneously track capacitance changes under dynamic operating conditions. More critically, during the multi-frequency scanning process, the high-frequency switching noise (2kHz–20MHz) of the inverter's power devices interferes with the single-frequency measurement signal, distorting the current sampling value. Experiments have shown that at the moment of motor startup, common-mode interference peaks can reach ampere levels, far exceeding the microampere leakage signal. At this time, the capacitance calculation error of traditional methods can reach over 30%. Furthermore, the fixed-time scanning mechanism cannot adapt to the time-varying capacitance (such as ±30% capacitance drift caused by cable aging), resulting in a mismatch between the pulse width and the actual capacitance.

[0089] In contrast, this embodiment breaks through the above limitations through two core improvements: First, the capacitance detection and insulation resistance measurement timing are decoupled, and the capacitance pre-calculation is independent of the dynamic load process to avoid the scanning process being disturbed by load switching. Second, the pulse width is reversely controlled in real time by the distributed capacitance value to ensure that the sampling window always covers the decay period of the capacitive current. For example, when the distributed capacitance is detected to be 1μF, the pulse width is automatically set to 150ms (set at 1.5 times the time constant), and the capacitive current is forced to decay to a steady state before the leakage current is collected to eliminate the superposition of transient components. The traditional method uses a fixed pulse width (usually 100ms), and samples before reaching a steady state when the capacitance increases, resulting in the leakage current measurement value containing up to 40% capacitive components.

[0090] Attention should be paid to the following aspects at the operational level: the test frequency sequence is injected during the dormant period of the high-voltage system to avoid the working period of the drive system; the microprocessor uses an automotive-grade MCU that supports floating-point operations to ensure that the multi-frequency point data is solved within 20ms; the pulse width is dynamically adjusted in steps of 10ms to adapt to the slow-changing characteristics of the capacitor. For capacitor mutation scenarios (such as high-voltage plug-in events), the system automatically triggers capacitor re-detection to prevent historical data from becoming invalid. Implementation cases have shown that this method can still maintain the matching of pulse timing and capacitor parameters under conditions of capacitor drift caused by aging of battery pack cables.

[0091] In another technical solution, the rule for dynamic adjustment is: set the duration of the pulse signal to the calculated capacitance value and R min The value corresponding to the time constant range of 1.5 to 3 times the product; in the step of measuring the response current value under the first pulse signal by the current sensor, the measurement occurs after the first pulse signal is applied for a duration dynamically adjusted by the microprocessor; in the step of measuring the response current value under the second pulse signal by the current sensor, the steady-state response current value of the current sensor is collected at the end of the duration of the application of the first pulse signal; wherein, R min The lower limit of the insulation resistance of the high voltage system to ground is preset to 100kΩ.

[0092] Dynamic adaptation of the pulse width is achieved through the following steps. First, based on the pre-calculated distributed capacitance value and the preset insulation resistance lower limit of 100kΩ, the microprocessor calculates the time constant of the RC circuit. Then, the duration of the first pulse signal and the second pulse signal is set to a fixed multiple in the range of 1.5 to 3 times the time constant. When the first pulse signal is applied in the no-load state of the high-voltage system, the current sensor only collects the response current value after the pulse duration ends; similarly, when the second pulse signal is applied in the load state, the current sensor synchronously collects the steady-state response current value at the end of the pulse. By forcing the measurement window to cover the steady-state stage where the capacitive current decay is completed, it is ensured that the resistive component in the collected current signal is dominant.

[0093] The closest existing technology is the insulation detection method with a fixed pulse width. This method uses a preset fixed pulse width (usually 100ms) for current sampling, which has significant limitations: when the distributed capacitance of the high-voltage system drifts due to cable aging or changes in ambient humidity (such as from 0.5μF to 1μF), the original pulse width cannot meet the attenuation requirements of the new capacitance. For example, in an RC circuit consisting of a 1μF capacitor and 100kΩ, it takes more than 150ms for the capacitive current to decay to a steady state, and the fixed 100ms pulse is sampled before reaching a steady state, resulting in the collected current value containing a capacitive component of up to 40%. These capacitive currents are mistakenly counted as leakage current, resulting in a low calculated insulation resistance value, which may trigger a false fault alarm under conditions where the capacitance increases.

[0094] In contrast, this embodiment associates the capacitance with the lower limit of the insulation resistance through a time constant, so that the pulse width is automatically adjusted as the capacitance changes. For example, when the capacitance is detected to be 1μF, the pulse width is automatically set to 225ms (set according to 1.5 times the time constant) to ensure that the capacitive current has decayed to less than 5% of the leakage current during sampling. At the same time, the strict synchronous sampling mechanism at the end of the pulse avoids the transient interference introduced by the sampling timing deviation of the traditional method. In the prior art, some improvement schemes attempt to predict capacitance based on temperature or historical data, but the prediction model fails in the capacitance mutation scenario caused by load switching (such as high-voltage plug-in events). This scheme directly controls the pulse timing dynamically through the measured capacitance, and achieves full working condition coverage without the need for a prediction model.

[0095] Operational considerations: Floating-point arithmetic is used in time constant calculation to ensure accuracy error of less than 1%. Pulse width adjustment uses a minimum step size of 10ms to accommodate the slow-changing characteristics of the capacitor. The microprocessor has a built-in hardware timer for synchronous sampling triggering at the end of the pulse. The insulation resistance lower limit, Rmin, is set to the industry safety standard lower limit of 100kΩ. If adjustment is required, the pulse multiplier factor must be adjusted simultaneously.

[0096] In another technical solution, in the step of applying a set of predefined test frequency sequences to the high voltage system, the test frequency sequences are applied in the following manner:

[0097] 1) applying a step wave test voltage signal to the high voltage system through the coupling circuit, wherein the rise time of the step wave test voltage signal is less than 100 μs and the step voltage jumps from 0 V to a target voltage value, wherein the target voltage value range is set to 10 V to 30 V;

[0098] 2) After the step wave test voltage signal reaches the target voltage value, maintaining the constant voltage for 5 ms to 20 ms, and collecting the transient current response waveform during the duration by the current sensor;

[0099] 3) The collected transient current response waveform is transmitted to the microprocessor, and the microprocessor performs exponential function fitting on the transient current response waveform, extracts the current decay time constant τ, and calculates the current decay time constant τ according to the formula:

[0100] C=τ / R min ,

[0101] Calculate the distributed capacitance of the high voltage system to ground, where τ is the current decay time constant, R min A lower limit of 100 kΩ is preset for the insulation resistance of the high-voltage system to ground, which is used to simplify the equivalent RC circuit model.

[0102] In the dynamic adjustment rule, the duration of the pulse signal is set to the calculated distributed capacitance to ground and R min 1.5 to 3 times the product of .

[0103] Rapid measurement of distributed capacitance is achieved through step wave step response. First, a step wave test voltage signal with a rise time of less than 100 microseconds is applied to the high-voltage system through a coupling circuit. The signal jumps from 0 volts to a target voltage value of 10-30 volts and remains constant for 5-20 milliseconds. During the voltage holding stage, the current sensor synchronously collects the transient current response waveform and transmits the waveform data to the microprocessor. The microprocessor performs exponential function fitting on the current decay curve, extracts the current decay time constant τ, and then calculates the distributed capacitance to ground based on the formula C=τ / Rmin (Rmin is the preset insulation resistance lower limit of 100kΩ). Finally, the duration of the subsequent insulation detection pulse signal is dynamically set according to 1.5-3 times the product of the capacitance value and Rmin.

[0104] The closest existing technology is the traditional multi-frequency point sinusoidal wave scanning method. This method requires sequentially injecting sinusoidal wave signals of multiple discrete frequencies such as 50Hz, 100Hz, and 150Hz, and independently collecting the response current at each frequency point, and calculating the distributed capacitance through frequency domain analysis. However, this technology has two major bottlenecks: First, the multi-frequency point switching scan takes too long, and a single detection requires more than 200 milliseconds, which cannot match the transient process of load switching (usually 50-100 milliseconds). Under dynamic conditions such as motor start-up and stop, the high-frequency switching noise (2kHz-20MHz) generated by the inverter will interfere with the single-frequency point measurement signal, resulting in distortion of the current sampling value. Experiments show that the common-mode interference peak at the moment of motor startup can reach the ampere level, far exceeding the microampere leakage signal. At this time, the capacitance calculation error of the traditional method is as high as 30%. Second, the fixed timing scanning mechanism cannot adapt to the time-varying nature of capacitance. For example, cable aging can cause the distributed capacitance to drift by ±30%. Traditional methods use a preset pulse width (usually 100 milliseconds) and sample before the capacitance increases and reaches a steady state, resulting in the leakage current measurement value containing a capacitive component of up to 40%.

[0105] This implementation overcomes the above limitations by using a step wave step response mechanism:

[0106] 1. Single-step instead of multi-frequency sweep: A staircase waveform, stepping from 0V to the target voltage, generates a wide-spectrum excitation covering the equivalent frequency band of 50-200Hz. Capacitor parameters can be extracted through a single transient response, shortening the detection cycle from >200ms with traditional methods to <20ms. This avoids the sensitive period of electromagnetic interference between multi-frequency switching points, significantly improving the signal-to-noise ratio in the inverter switching noise environment.

[0107] 2. Time-domain fitting replaces frequency-domain calculations: Directly exponentially fits the current decay waveform, avoiding the cumulative errors associated with frequency-domain impedance calculations in traditional methods. For example, when the distributed capacitance is 1μF, the traditional sine wave method requires calculating the capacitive reactance at three frequencies and solving a system of equations. However, the step response directly derives the time constant by fitting the slope of the decay curve, reducing algorithm complexity by 70%.

[0108] 3. Optimized transient synchronization: The step wave rise time is <100μs, much faster than the millisecond transient of load switching, ensuring that capacitance detection is completed before load disturbances occur. Traditional methods, however, often overlap with dynamic processes due to their lengthy scanning times, resulting in the insulation detection phase not yet being updated.

[0109] Compared with the low-frequency AC injection method, this solution further addresses its inherent flaws: the low-frequency injection method requires applying a 5-100Hz sweep frequency signal during the vehicle's dormant period, which is susceptible to background harmonic interference from equipment such as on-board chargers; the step wave test voltage amplitude is only 10-30V, which is 1 / 20 of the traditional injection voltage, and the duration is short, which has no impact on the working state of the high-voltage system and can be performed intermittently while the vehicle is driving.

[0110] Operational considerations include: The target voltage for the step wave is preferably 20V to balance signal-to-noise ratio and safety; the current sampling rate is no less than 1MHz to ensure accurate attenuation curve fitting; the exponential fit uses the Levenberg-Marquardt algorithm to avoid the initial value sensitivity of the conventional least squares method; and the step hold time is set based on the estimated capacitance range (e.g., 15ms for a 1μF capacitor). Implementation examples have shown that this method can reduce the overlap between capacitor detection and load switching by over 90% under rapid motor acceleration conditions, fundamentally preventing modeling mismatch caused by dynamic interference.

[0111] In another technical solution, after the step of calculating the insulation resistance value under load based on the second response current value and the test voltage value by the microprocessor, the following steps are added:

[0112] 1) Synchronously collect the surface temperature values ​​of the positive busbar, negative busbar, and power battery casing of the high-voltage system through at least two temperature sensors distributed at key nodes of the high-voltage system;

[0113] 2) The collected temperature value is transmitted to the microprocessor; the microprocessor calls a pre-stored insulation material resistance-temperature relationship model, which is expressed as:

[0114]

[0115] Among them, R TIndicates the theoretical value of insulation resistance at temperature T, unit: Ω; R0 represents the insulation resistance value at reference temperature T0, set to 298K; β is the thermal sensitivity coefficient of the insulation material, ranging from 3000K to 5000K;

[0116] 3) The microprocessor calculates the weighted average temperature T of each measuring part of the high-voltage system based on the currently collected temperature value avg , and combining the insulation material resistance-temperature relationship model, temperature compensation is performed on the positive electrode insulation resistance value and the negative electrode insulation resistance value under the load state to obtain the standardized positive / negative electrode insulation resistance value at a reference temperature of 298K;

[0117] In the differential filtering processing step, the standardized positive / negative insulation resistance values ​​are used in the calculation; in the step of comparing the insulation resistance value with the preset threshold range, the standardized insulation resistance value is used for judgment.

[0118] The temperature compensation method is implemented by the following steps: After completing the load state insulation resistance calculation, the temperature data is collected synchronously by at least two temperature sensors distributed on the positive busbar, negative busbar and power battery housing of the high voltage system. These temperature values ​​are transmitted to the microprocessor in real time, and the pre-stored insulation material resistance-temperature relationship model (expressed as

[0119] Where R0 is the resistance value at the reference temperature of 298K, and β is the thermal sensitivity coefficient of the insulating material. The microprocessor calculates the weighted average temperature T of each measuring part. avg The positive and negative insulation resistance values ​​under load are compensated accordingly, and the standardized resistance value at a reference temperature of 298 K is output. Differential filtering and fault threshold determination are both performed based on the standardized resistance value.

[0120] The closest existing technology is the direct compensation solution using a single-point temperature sensor. This solution only installs a single temperature sensor on the battery pack casing or busbar, and directly substitutes the collected temperature into an Arrhenius formula for compensation. However, there are significant defects: First, single-point temperature measurement cannot reflect the temperature gradient inside the high-voltage system. For example, the temperature of weak insulation points such as the center area of ​​the battery module partition may be more than 20°C higher than the casing, and traditional solutions completely ignore this; second, the heat conduction lag effect during dynamic temperature changes is not taken into account. For example, under rapid acceleration conditions of the motor, the busbar temperature may rise by 15°C within 30 seconds, but the temperature response delay inside the insulation material can reach several minutes, resulting in a mismatch between the compensation value and the actual temperature field.

[0121] This implementation overcomes the above limitations through multi-node synchronous temperature measurement and weighted fusion mechanism:

[0122] Spatial Coverage Optimization: Temperature sensor clusters are placed simultaneously on the positive and negative busbars and the battery casing to capture temperature differences at key nodes. For example, when charging, the negative busbar temperature is typically higher than the positive busbar temperature. Traditional solutions cannot distinguish such differences, but this solution avoids polarity deviation by independently compensating for these differences.

[0123] Internal temperature inversion using thermal network topology: A thermal network model is constructed based on the physical structural parameters of the high-voltage system (such as battery module spacing and wiring harness layout). Sensor measurements are used as boundary conditions to determine the theoretical temperature at insulation weaknesses (such as the center of the battery module partition). This process utilizes material parameters such as the thermal conductivity of aluminum (200-240 W / (m·K)) and epoxy resin board (0.2-0.5 W / (m·K)) to ensure that the model's heat transfer paths align with actual physical properties.

[0124] Dynamic Weighting: In the temperature-weighted average calculation, insulation weaknesses are weighted 30%-40% higher than external sensors (60%-70% combined). This prioritizes internal heat buildup risks. For example, if the battery separator's theoretical temperature rises to 85°C, the weighted result can still accurately trigger high-temperature compensation even if the outer casing temperature is 60°C.

[0125] Compared with the traditional thermal model solution, this solution abandons the static model that relies on fixed heat capacity parameters and introduces fluid-heat conduction coupling calculation. The traditional model will cause the prediction deviation to exceed 100% due to the change of material contact thermal resistance in the vehicle driving vibration environment.

[0126] ±5℃, while this solution controls the internal temperature prediction error within ±3℃ through a real-time synchronous multi-sensor data closed-loop correction model.

[0127] Operational considerations: PT1000 platinum resistance thermometers are preferred, with a sampling interval of ≤1 second. The β coefficient is preset based on the insulation material type (3000K for epoxy resin and 4200K for silicone). The compensation cycle is triggered synchronously with the insulation detection pulse to avoid processor overload. If a sensor fails, the system automatically switches to the remaining sensor data and reduces the weight coefficient to maintain basic compensation functionality.

[0128] In another technical solution, in the step of synchronously collecting the surface temperature values ​​of the positive busbar, the negative busbar, and the power battery housing of the high-voltage system through at least two temperature sensors distributed at key nodes of the high-voltage system, the following operation is added:

[0129] 1) Based on the physical structural parameters of the high-voltage system, a thermal network topology model is constructed in the microprocessor;

[0130] 2) The synchronously collected temperature values ​​are input as boundary conditions into the thermal network topology model. The microprocessor solves the thermal network steady-state equations to calculate the theoretical temperature values ​​of the insulation weak points within the high-voltage system. The insulation weak points include the central area of ​​the insulating partition between the power battery modules and the joint surface of the high-voltage connector.

[0131] Internal temperature inversion is achieved by constructing a thermal network topology model. First, at least two PT1000 platinum resistance temperature sensors are placed on the positive and negative busbars of the high-voltage system and on the surface of the power battery housing. These sensors synchronously collect temperature data at one-second intervals and transmit it to a microprocessor. Based on the physical structural parameters of the high-voltage system—including battery module spacing, wiring layout, insulation separator thickness, and connector interface geometry—a thermal network topology model is established within the microprocessor. This model includes the heat conduction paths from the positive and negative busbars to the battery housing, the lateral heat conduction paths through the insulation separators between modules, and the interfacial thermal resistance between the connector metal terminals and the insulation housing. The thermal conductivity of each path is set based on the material properties: 220 W / (m·K) ± 5% for the aluminum path and 0.35 W / (m·K) ± 10% for the epoxy resin insulation board path. The real-time sensor temperatures are input into the model as boundary conditions. The thermal network steady-state equations are solved to output theoretical temperatures at insulation weaknesses (such as the center of the battery module separator and the high-voltage connector interface).

[0132] The closest existing technology is the single-point temperature extrapolation compensation solution. This technology only installs a single temperature sensor in the center of the battery pack shell and directly infers the internal temperature through an empirical formula. Its core defects are two points: First, single-point measurement cannot perceive the spatial temperature gradient inside the high-voltage system. For example, when the motor accelerates suddenly, the temperature of the negative busbar can be 15°C higher than the positive busbar, while the center area of ​​the battery module partition plate is more than 20°C higher than the shell temperature due to the heat accumulation effect. The existing technology completely ignores this; second, the heat conduction hysteresis under dynamic conditions is not taken into account. For example, when the vehicle decelerates suddenly, the busbar temperature can drop by 10°C within 30 seconds, but the internal insulation material has a response delay of several minutes due to the heat capacity effect, resulting in a mismatch between the estimated temperature and the actual temperature field of more than ±8°C.

[0133] In contrast, this implementation overcomes the above limitations by coupling multi-node temperature measurement with a thermal network:

[0134] Three-dimensional heat conduction path modeling: A physical parameterized model is constructed based on the measured thermal conductivity of aluminum and epoxy resin, accurately depicting the heat flow path from the busbar to the battery box. Traditional static models use fixed heat capacity parameters and cannot adapt to changes in material contact thermal resistance caused by vehicle vibration. However, this solution uses a closed-loop correction model using real-time sensor data to control the predicted temperature error of insulation weak points to within ±3°C.

[0135] Dynamic Capture of Internal Hotspots: The thermal network model discretizes unmeasurable areas, such as the center of the insulating separator, into virtual nodes. The theoretical temperature is inverted by solving the Fourier heat conduction equation between adjacent nodes. For example, during fast charging, even though the outer casing sensor only reads 45°C, the model can infer the separator center temperature to be 68°C based on the measured value of the negative busbar at 60°C and the thermal resistance parameter, triggering an alert. Traditional solutions, lacking internal node modeling, are unable to identify this risk.

[0136] Operational considerations include: the thermal network model is pre-installed on the microprocessor at the production line end, supporting OTA remote updates of material parameters; the solver utilizes a Newton-Raphson iterative algorithm, with the calculation cycle synchronized with the insulation detection pulse; and if a sensor fails, the remaining sensor data is automatically activated and the corresponding path weight coefficient is reduced to 0.1 to maintain basic compensation functionality. Implementation cases have demonstrated that this method accurately inverts the actual temperature trend at the center of the battery module separator under -30°C cold start conditions, resolving the low-temperature insulation misjudgment caused by thermal lag in traditional solutions.

[0137] In another technical solution, in the step of calculating the weighted average temperature of each measuring location of the high-voltage system, the theoretical temperature value of the insulation weak point is included in the weighted calculation, wherein the temperature weight distribution range of the insulation weak point is 30% to 40%, and the total weight of the measured temperature of the positive busbar, negative busbar and power battery casing is 60% to 70%, and the sum of all weight coefficients is 1;

[0138] The calculation formula of the temperature weighted average value is:

[0139]

[0140] Among them, T sensor,i is the measured value of the i-th temperature sensor, K i is its weight coefficient; T virtual,j is the theoretical temperature value of the jth insulation weak point, k j is its weight coefficient; and satisfies ∑k i +∑k j =1.

[0141] Multi-node temperature fusion is used to achieve early warning of insulation weakness risks. First, a PT1000 platinum resistance temperature sensor is installed on each of the positive and negative busbars of the high-voltage system. Four sensors of the same model are symmetrically arranged at the four corners of the power battery housing. All sensors synchronously collect temperature data at one-second intervals and transmit it to the microprocessor. A thermal network model is constructed based on the battery module spacing, insulation partition thickness, and high-voltage connector geometric parameters. The thermal conductivity of aluminum is set to 220W / (m·K), and the thermal conductivity of epoxy resin board is set to 0.35W / (m·K). Real-time sensor data is input into the model as boundary conditions to solve the theoretical temperature values ​​of the center area of ​​the insulation partition and the joint surface of the high-voltage connector. In the temperature weighted calculation, the theoretical temperature of the insulation weakness is assigned a 35% weight, and the measured busbar and housing temperatures are assigned the remaining weights according to their spatial location (20% for the negative busbar, 15% for the positive busbar, and 30% for the average of the four corners of the housing). The total weight is strictly 1. When the battery is fast charging, if the model inversion shows that the center temperature of the partition reaches 75°C while the outer shell sensor only shows 50°C, the weighted result will be significantly biased towards the high-risk area, triggering an early warning.

[0142] The closest existing technology is the single-point extrapolation temperature compensation solution. This technology only installs a single temperature sensor in the center of the battery pack shell and directly calculates the internal temperature using a linear formula:

[0143] Assuming the shell temperature is 50°C, the center temperature of the partition is calculated to be 60°C using a fixed coefficient of 1.2.

[0144] When the temperature rise of the negative busbar is 15°C higher than that of the positive busbar during motor operation, the solution cannot detect this due to the lack of independent monitoring.

[0145] This type of method has two fundamental flaws: First, it does not take into account the temperature gradient in the internal space of the high-voltage system. Experiments show that during fast charging, the actual temperature at the center of the battery module partition can be 25°C higher than that of the outer shell, and the single-point estimation error is as high as ±8°C. Second, it ignores the heat conduction lag under dynamic conditions. When the vehicle decelerates suddenly, the busbar temperature drops by 10°C within 30 seconds, but the response of the insulation material is delayed by several minutes due to the heat capacity effect inside, resulting in a mismatch between the compensation value and the actual temperature field.

[0146] In contrast, this implementation breaks through traditional limitations by using a dynamic weight allocation mechanism:

[0147] Risk-Based Weighting: Insulation weaknesses are assigned a 30%-40% decision weight, significantly higher than measured points (e.g., a single point on the housing is weighted only 8%). When model inversion indicates that the connector mating surface temperature exceeds the 85°C safety threshold, even if the busbar temperature is measured at 60°C, the weighted result still amplifies the internal risk signal with a 37% weighting. Traditional solutions, which lack internal temperature information, completely ignore this risk.

[0148] Real-time closed-loop correction of the thermal network: Traditional static models experience a ±5°C deviation due to changes in material contact thermal resistance when the vehicle vibrates. This solution refreshes model parameters with new sensor data every 5 seconds, compressing the internal temperature prediction error to within ±3°C.

[0149] In low-temperature battery startup scenarios, traditional solutions mistakenly extrapolate a -20°C partition center temperature to -10°C due to thermal hysteresis, leading to a false alarm of low-temperature insulation. This solution, however, inverts the partition temperature in real time and combines it with a 35% weighting to output a weighted value of -19.2°C, preventing false insulation alarms. Important notes: The default weighting coefficients are configured as 35% for weak points, 35% for busbars, and 30% for the casing. If a sensor fails, its weight is automatically distributed proportionally to other nodes. The thermal network model supports OTA remote updates of material parameters, adapting to different vehicle platforms. Implementation cases have shown that this approach reduces the lead time for battery thermal runaway warning by 80% compared to traditional single-point solutions.

[0150] In another technical solution, the model includes heat conduction paths between the high-voltage system positive busbar, negative busbar, power battery module insulating partition, and battery box. The thermal conductivity of each path is set according to the material properties, where the thermal conductivity range of aluminum is set to 200W / (m·K) to 240W / (m·K), and the thermal conductivity range of epoxy resin insulation board is set to 0.2W / (m·K) to 0.5W / (m·K).

[0151] In another technical solution, in the step of calculating the insulation resistance value under load by the microprocessor based on the second response current value and the test voltage value, the insulation resistance value is split into the insulation resistance value R of the high voltage system positive electrode to ground. pos And the insulation resistance value R neg Calculate separately, including the following operations:

[0152] 1) During the application of the first pulse signal, the first positive electrode response current value I between the positive terminal and the ground terminal of the high voltage system is synchronously measured by the current sensor pos1 and the first negative electrode response current value I between the negative terminal and the ground terminal neg1 , and I pos1 and I neg1 Transmit to microprocessor;

[0153] 2) The microprocessor calculates the insulation resistance of the positive and negative electrodes under no-load conditions according to the formula: test is the test voltage value;

[0154] During the application of the second pulse signal, the second positive electrode response current value I between the positive terminal and the ground terminal is synchronously measured by the current sensor. pos2 And the second negative electrode response current value I between the negative terminal and the ground terminalneg2 , and transmit it to the microprocessor, which calculates the positive and negative insulation resistance values ​​under load according to the formula.

[0155] The implementation of the thermal network topology model is achieved through the following process: Based on the physical structural parameters of the high-voltage system, a heat conduction path model is constructed in the microprocessor, including the positive and negative busbars, the insulating partitions of the power battery module, and the battery box. The thermal conductivity of each path is strictly set according to the material properties. The thermal conductivity of the aluminum material of the positive and negative busbars is set to 220W / (m·K)±5%, with an error band covering the range of 200-240W / (m·K); the thermal conductivity of the epoxy resin insulating plate between the battery modules is set to 0.35W / (m·K)±10%, covering the industry typical value of 0.2-0.5W / (m·K). When building the model, the busbar length (typical value 1.2m), insulating partition thickness (typical value 3mm) and joint surface dimensions in the three-dimensional drawing of the battery pack are first extracted, and the geometric data is converted into equivalent thermal resistance network nodes. Next, the materials were injected and their thermal conductivity parameters were measured: the average measured value for the extruded aluminum busbar was 220 W / (m·K), and the measured value for the vacuum-impregnated epoxy insulation board was 0.35 W / (m·K). Finally, a microprocessor solved the steady-state heat conduction equation and output the theoretical temperature field distribution at the insulation weak point.

[0156] The closest existing technology is the static heat capacity parameter model (such as the electric motor temperature estimation patent CN102654423A). This technology uses a fixed heat capacity value to simulate the system's heat transfer path (for example, the thermal conductivity of aluminum is uniformly set to 237W / (m·K)). This technology has two key flaws: First, it ignores thermal conductivity fluctuations caused by differences in material processing. For example, the actual thermal conductivity of cast aluminum busbars can be 15% lower than that of extruded aluminum due to differences in porosity, but traditional models do not distinguish between them. Second, it does not consider changes in contact thermal resistance caused by vehicle vibration. Experiments have shown that long-term vibration can increase the contact thermal resistance between the busbar and the insulation plate by 30%, causing the model's predicted temperature to deviate from the actual value by more than ±5°C.

[0157] Compared with static models, this solution breaks through limitations through physical parametric modeling and dynamic compensation mechanisms:

[0158] Material and process adaptation: Different thermal conductivity ranges are set based on the material processing techniques of different high-voltage system components. For example, the battery case uses cast aluminum (set at 200-210W / (m·K)), while the positive busbar uses high-conductivity extruded aluminum (set at 230-240W / (m·K)), which better aligns with actual physical properties. Epoxy resin insulation boards are graded by filler ratio—glass fiber reinforced types are rated at 0.45-0.5W / (m·K), while pure resin-based types are rated at 0.2-0.3W / (m·K). This avoids the missed detection of local hotspots caused by the uniform value used in traditional solutions.

[0159] Closed-loop correction of contact thermal resistance: Vibration acceleration sensor data is embedded in thermal network nodes. When continuous vibration exceeding the threshold is detected, the contact thermal resistance coefficient of the busbar-insulation plate interface is automatically increased by 25%-30%. A self-calibration mode is also triggered every six months: based on the deviation between the measured battery casing temperature and the model output value, the contact thermal resistance parameters are iteratively corrected in reverse order, ensuring that the long-term prediction error of the model is stable within ±3°C.

[0160] Operational considerations: The default thermal conductivity value is preset based on the vehicle's bill of materials and supports OTA updates; the vibration compensation module is only activated at speeds greater than 60 km / h; and the self-calibration mode requires the vehicle to remain stationary in a 25°C constant-temperature garage. Implementation examples demonstrate that this method improves the temperature prediction accuracy of the battery module's central partition by 40% compared to traditional fixed-parameter models, resolving the issue of thermal distribution distortion under vibration conditions.

[0161] In another technical solution, in the differential filtering process step, the following steps are performed independently for the positive and negative electrodes:

[0162] The positive insulation resistance value R' under load pos The positive insulation resistance value R under no-load condition pos Perform differential filtering to obtain the final insulation resistance value R of the positive electrode pos-final ,

[0163] The negative electrode insulation resistance value R' under load neg The negative insulation resistance value R under no-load condition neg Perform differential filtering to obtain the final insulation resistance value R of the negative electrode neg-final ,

[0164] The functional expression of the differential filtering process is:

[0165] R pos-final =R pos +K·(R′ pos -R pos );

[0166] R neg-final =R neg +K·(R′ neg -R neg );

[0167] Among them, the value range of the filter coefficient K is 0.2 to 0.8;

[0168] In the step of comparing the insulation resistance value with the preset threshold range, R pos-final and R neg-final Judge separately: when R pos-final or R neg-finalWhen the voltage is lower than the preset threshold value range of 100kΩ to 10MΩ, a fault location signal of the corresponding polarity is output to the vehicle control unit.

[0169] The implementation steps of the polarization insulation detection method are as follows: When the high-voltage system is in a no-load state, the first positive pole response current value between the positive terminal and the ground terminal, and the first negative pole response current value between the negative terminal and the ground terminal are synchronously measured by the current sensor. These two current values ​​are transmitted to the microprocessor in real time, and the positive pole insulation resistance value and the negative pole insulation resistance value under the no-load state are calculated in combination with the test voltage value. When the high-voltage system is switched to a load state, the second positive pole response current value and the second negative pole response current value are synchronously collected at the same position, and the microprocessor independently calculates the positive and negative pole insulation resistance values ​​under the load state. The entire process uses a hardware trigger mechanism to ensure that the positive and negative pole current sampling moments are strictly synchronized to avoid calculation errors introduced by timing deviations.

[0170] The closest existing technology is a single-channel leakage current detection solution (such as patent CN105738701A). This technology monitors the total leakage current of the high-voltage system through a single Hall sensor and can only calculate the overall insulation resistance value. Its core defect is that when asymmetric insulation degradation occurs on the positive and negative poles (for example, the positive pole drops to 50kΩ and the negative pole remains at 500kΩ), the total insulation resistance value may still be higher than the 100kΩ threshold, causing the system to be unable to identify positive pole faults. Experiments have shown that such asymmetric faults account for more than 30% of insulation failure cases, and traditional methods are completely ineffective for this. What's more serious is that if the degree of insulation degradation on the negative pole is higher than that on the positive pole, the system will misjudge that the fault occurs in the positive pole circuit, resulting in the wrong direction of maintenance.

[0171] This solution overcomes the above limitations through dual-channel synchronous sampling and independent calculation mechanism:

[0172] Precise fault polarity isolation: During rapid motor acceleration, if the negative electrode momentarily shorts to ground due to vibration and friction (resistance drops to 20kΩ), while the positive electrode maintains a normal 1MΩ, this solution can detect the abnormal increase in negative electrode current in real time and independently output a negative electrode fault signal. Traditional single-channel solutions may completely miss this fault because the total resistance is pulled up by the high-resistance positive electrode.

[0173] Dynamic interference suppression by polarization: Common-mode interference generated by load switching exhibits asymmetry between the positive and negative conduction paths (e.g., the negative circuit has greater parasitic capacitance). This solution's polarization calculations autonomously separate the interference components of each pole, avoiding the miscalculation caused by traditional methods that mistakenly include negative interference in the positive leakage current.

[0174] Operational considerations include: The positive and negative current sensors should preferably be identical models to eliminate gain errors; the microprocessor utilizes dual ADC channels for microsecond-level simultaneous sampling; and if a sensor fails, the system automatically switches to unipolar monitoring mode and issues a degradation warning. A case study demonstrated that this method successfully located the fault polarity and guided precise repairs in a scenario where a ruptured liquid cooling tube within a battery pack caused partial grounding of the positive electrode, avoiding the full system disassembly and inspection required by traditional solutions.

[0175] Although the embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the description and implementation methods. They can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to specific details.

Claims

1. A method for detecting insulation of an electric vehicle, comprising the following steps: S0: Applying a predefined test frequency sequence to the high-voltage system, the sequence comprising at least three discrete sinusoidal wave frequency points between 50 Hz and 200 Hz, measuring the response current value at each frequency point using a current sensor, and transmitting the measured value to a microprocessor; the microprocessor calculates the distributed capacitance of the high-voltage system to ground based on the response current value; S1: Connect the test voltage source between the positive and negative poles of the high-voltage system of the electric vehicle through a coupling circuit. The coupling circuit includes an isolation transformer and a current-limiting resistor. The output voltage range of the test voltage source is 50V to 500V. S2: When the high-voltage system is in a no-load state, applying a first pulse signal of a test voltage source, the duration of which is dynamically determined by the microprocessor based on the distributed capacitance to ground calculated in step S0; measuring a first response current value under the first pulse signal by a current sensor, and transmitting the first response current value to the microprocessor; Based on the first response current value and the test voltage value, the microprocessor calculates the initial insulation resistance value of the high voltage system; S3: within a set time window after the high-voltage system switches to a load operation state, applying a second pulse signal of the test voltage source, wherein the duration of the second pulse signal is the same as that of the first pulse signal and is dynamically determined by the microprocessor, and the amplitude range of the second pulse signal is the same as that of the first pulse signal; measuring a second response current value under the second pulse signal by a current sensor, and transmitting the second response current value to the microprocessor; The microprocessor calculates the insulation resistance value under load based on the second response current value and the test voltage value, and performs differential filtering on the initial insulation resistance value to eliminate noise interference caused by load fluctuations. S4: The microprocessor compares the insulation resistance value after differential filtering with a preset threshold range of 100kΩ to 10MΩ. When the insulation resistance value is lower than the preset threshold range, a fault signal is output to the vehicle control unit.

2. The insulation detection method of an electric vehicle according to claim 1, characterized in that: The microprocessor calculates the distributed capacitance value of the high-voltage system to the ground based on the response current value, the applied sinusoidal voltage amplitude and the frequency value at each frequency point; based on the calculated distributed capacitance value to the ground, the microprocessor dynamically adjusts the duration of all pulse signals in subsequent insulation testing, including the duration of the first pulse signal and the second pulse signal.

3. The insulation detection method of an electric vehicle according to claim 2, characterized in that: The rule for dynamic adjustment is: set the duration of the pulse signal to the calculated capacitance value and R min The value corresponding to the time constant range of 1.5 to 3 times the product; in the step of measuring the response current value under the first pulse signal by the current sensor, the measurement occurs after the first pulse signal is applied for a duration dynamically adjusted by the microprocessor; in the step of measuring the response current value under the second pulse signal by the current sensor, the steady-state response current value of the current sensor is collected at the end of the duration of the application of the first pulse signal; wherein, R min The lower limit of the insulation resistance of the high voltage system to ground is preset to 100kΩ.

4. The insulation detection method for an electric vehicle according to claim 3, characterized in that: In the step of applying a set of predefined test frequency sequences to the high voltage system, the test frequency sequences are applied in the following manner: 1) applying a step wave test voltage signal to the high voltage system through the coupling circuit, wherein the rise time of the step wave test voltage signal is less than 100 μs and the step voltage jumps from 0 V to a target voltage value, wherein the target voltage value range is set to 10 V to 30 V; 2) After the step wave test voltage signal reaches the target voltage value, maintaining the constant voltage for 5 ms to 20 ms, and collecting the transient current response waveform during the duration by the current sensor; 3) The collected transient current response waveform is transmitted to the microprocessor, and the microprocessor performs exponential function fitting on the transient current response waveform, extracts the current decay time constant τ, and calculates the current decay time constant τ according to the formula: C=τ / R min , Calculate the distributed capacitance of the high voltage system to ground, where τ is the current decay time constant, R min A lower limit of 100 kΩ is preset for the insulation resistance of the high-voltage system to ground, which is used to simplify the equivalent RC circuit model. In the dynamic adjustment rule, the duration of the pulse signal is set to the calculated distributed capacitance to ground and R min 1.5 to 3 times the product of .

5. The insulation detection method of an electric vehicle according to claim 1, characterized in that: After the step of calculating the insulation resistance value under load based on the second response current value and the test voltage value by the microprocessor, the following steps are added: 1) Synchronously collect the surface temperature values ​​of the positive busbar, negative busbar, and power battery casing of the high-voltage system through at least two temperature sensors distributed at key nodes of the high-voltage system; 2) The collected temperature value is transmitted to the microprocessor; the microprocessor calls a pre-stored insulation material resistance-temperature relationship model, which is expressed as: Among them, R T Indicates the theoretical value of insulation resistance at temperature T, unit: Ω; R0 represents the insulation resistance value at reference temperature T0, set to 298K; β is the thermal sensitivity coefficient of the insulation material, ranging from 3000K to 5000K; 3) The microprocessor calculates the weighted average temperature T of each measuring part of the high-voltage system based on the currently collected temperature value avg , and combining the insulation material resistance-temperature relationship model, temperature compensation is performed on the positive electrode insulation resistance value and the negative electrode insulation resistance value under the load state to obtain the standardized positive / negative electrode insulation resistance value at a reference temperature of 298K; In the differential filtering processing step, the standardized positive / negative insulation resistance values ​​are used in the calculation; in the step of comparing the insulation resistance value with the preset threshold range, the standardized insulation resistance value is used for judgment.

6. The insulation detection method for an electric vehicle according to claim 5, characterized in that: In the step of synchronously collecting the surface temperature values ​​of the positive busbar, the negative busbar, and the power battery casing of the high-voltage system through at least two temperature sensors distributed at key nodes of the high-voltage system, the following operation is added: 1) Based on the physical structural parameters of the high-voltage system, a thermal network topology model is constructed in the microprocessor; 2) The synchronously collected temperature values ​​are input as boundary conditions into the thermal network topology model. The microprocessor solves the thermal network steady-state equations to calculate the theoretical temperature values ​​of the insulation weak points within the high-voltage system. The insulation weak points include the central area of ​​the insulating partition between the power battery modules and the joint surface of the high-voltage connector.

7. The insulation detection method of an electric vehicle according to claim 6, characterized in that: In the step of calculating the weighted average temperature of each measurement location of the high-voltage system, the theoretical temperature value of the insulation weak point is included in the weighted calculation, wherein the temperature weight distribution range of the insulation weak point is 30% to 40%, and the total weight of the measured temperature of the positive busbar, negative busbar, and power battery casing is 60% to 70%, and the sum of all weight coefficients is 1; The calculation formula of the temperature weighted average value is: Among them, T sensor,i is the measured value of the i-th temperature sensor, K i is its weight coefficient; T virtual,j is the theoretical temperature value of the jth insulation weak point, k j is its weight coefficient; and satisfies ∑k i +∑k j =1.

8. The insulation detection method for an electric vehicle according to claim 6, characterized in that: The model includes heat conduction paths between the high-voltage system positive busbar, negative busbar, power battery module insulating partition, and battery box. The thermal conductivity of each path is set according to the material properties. The thermal conductivity range of aluminum is set to 200W / (m·K) to 240W / (m·K), and the thermal conductivity range of epoxy resin insulation board is set to 0.2W / (m·K) to 0.5W / (m·K).

9. The insulation detection method for an electric vehicle according to claim 1, characterized in that: In the step of calculating the insulation resistance value under load by the microprocessor based on the second response current value and the test voltage value, the insulation resistance value is divided into the insulation resistance value R of the positive electrode to the ground of the high voltage system. pos And the insulation resistance value R neg Calculate separately, including the following operations: 1) During the application of the first pulse signal, the first positive electrode response current value I between the positive terminal and the ground terminal of the high voltage system is synchronously measured by the current sensor pos1 and the first negative electrode response current value I between the negative terminal and the ground terminal neg1 , and I pos1 and I neg1 Transmit to microprocessor; 2) The microprocessor calculates the insulation resistance of the positive and negative electrodes under no-load conditions according to the formula: test is the test voltage value; During the application of the second pulse signal, the second positive electrode response current value I between the positive terminal and the ground terminal is synchronously measured by the current sensor. pos2 And the second negative electrode response current value I between the negative terminal and the ground terminal neg2 , and transmit it to the microprocessor, which calculates the positive and negative insulation resistance values ​​under load according to the formula.

10. The insulation detection method of an electric vehicle according to claim 9, characterized in that: In the differential filtering step, the following steps are performed independently for the positive and negative electrodes: The positive insulation resistance value R' under load pos The positive insulation resistance value R under no-load condition pos Perform differential filtering to obtain the final insulation resistance value R of the positive electrode pos-final , The negative insulation resistance value R' under load neg The negative insulation resistance value R under no-load condition neg Perform differential filtering to obtain the final insulation resistance value R of the negative electrode neg-final , The functional expression of the differential filtering process is: R pos-final =R pos +K·(R' pos -R pos ); R neg-final =R neg +K·(R' neg -R neg ); Among them, the value range of the filter coefficient K is 0.2 to 0.8; In the step of comparing the insulation resistance value with the preset threshold range, R pos-final and R neg-final Judge separately: when R pos-final or R neg-final When the voltage is lower than the preset threshold value range of 100kΩ to 10MΩ, a fault location signal of the corresponding polarity is output to the vehicle control unit.

Images