Method for identifying noise sources of an automotive domain controller power supply network

Abstract

The application provides a noise source identification method for a power supply network of an automotive domain controller. The method applies a predefined test excitation to the power supply network of the automotive domain controller, synchronously collects response signals of the power supply network to the test excitation at at least two power supply nodes, obtains a measured frequency response of the power supply network at the current moment based on the response signals and the test excitation, updates equivalent model parameters of the power supply network under preset physical constraint conditions based on the measured frequency response and the current equivalent model parameters of the power supply network, so that the model can dynamically track and approximate the time-varying characteristics of the real network, and finally, the updated equivalent model parameters of the power supply network are used to perform online noise source identification for the power supply network, so that the network parameter drift and the newly added noise source can be effectively distinguished, and the long-term accuracy and reliability of noise source positioning are improved, and a reliable basis is provided for subsequent filtering strategy making and abnormal diagnosis.

Description

Technical Field

[0001] This application relates to data processing technology, and more particularly to a method for noise source identification in an automotive domain controller power supply network. Background Technology

[0002] In the field of automotive electronics, the stability and cleanliness of the power supply network of domain controllers, as high-performance computing units, are crucial to the functional safety of the system. Online noise source identification is an effective means of diagnosing and suppressing power supply network interference, which typically relies on a pre-established equivalent model of the power supply network. The common practice in existing technologies is to obtain the impedance characteristics of the power supply network during the development phase through offline measurements (such as using a network analyzer), and fit a set of initial equivalent model parameters (used to characterize the resistance, inductance, and capacitance characteristics of the power supply network). These parameters are then embedded into the controller software for use throughout its entire lifecycle.

[0003] However, during long-term use, the parameters of components such as capacitors, inductors, and resistors in the power supply network of a vehicle drift due to temperature changes, bias voltage fluctuations, and aging effects. Since the model parameters used in existing technologies are fixed, the characteristics of the equivalent power supply network model gradually become mismatched with those of the actual network. This mismatch causes a systematic deviation between the model's predicted frequency response (such as resonant frequency, impedance amplitude, and phase) and the actual situation. The accuracy of online noise source identification based on this mismatched model will significantly decrease over time, failing to effectively distinguish between network parameter drift and newly added noise sources, potentially leading to misjudgments and ineffective corrective measures. Summary of the Invention

[0004] This application provides a noise source identification method for automotive domain controller power supply networks to solve the technical problem in the prior art where the dynamic drift of power supply network component parameters with temperature and aging leads to a mismatch of the fixed equivalent model, which in turn causes a long-term decrease in the accuracy of online noise source identification. This method achieves the technical effects of model self-calibration and accurate noise source identification.

[0005] This application provides a method for noise source identification in an automotive domain controller power supply network, including: Apply predefined test stimuli to the power supply network of the vehicle domain controller; At at least two power supply nodes, the response signals of the power supply network to the test stimulus are collected synchronously. Based on the response signal and the test stimulus, the measured frequency response of the power supply network at the current moment is obtained; Based on the measured frequency response and the current power supply network equivalent model parameters, the power supply network equivalent model parameters are updated under preset physical constraints; wherein, the power supply network equivalent model parameters are used to characterize the resistance, inductance, and capacitance characteristics of the power supply network. Using the updated power supply network equivalent model parameters, perform online noise source identification for the power supply network.

[0006] The noise source identification method for automotive domain controller power supply networks provided in this application applies a predefined test stimulus to the power supply network of the automotive domain controller, and then synchronously collects the response signals of the power supply network to the test stimulus at at least two power supply nodes. This directly captures the actual impact of component parameter drift caused by temperature and aging on network characteristics. Then, based on the response signal and the test stimulus, the measured frequency response of the power supply network at the current moment is obtained. Next, based on the measured frequency response and the current equivalent model parameters of the power supply network, the equivalent model parameters of the power supply network are updated under preset physical constraints, so that the model can dynamically track and approximate the time-varying characteristics of the real network. Finally, the updated equivalent model parameters of the power supply network are used to perform online noise source identification for the power supply network, thereby effectively distinguishing between network parameter drift and newly added noise sources, thus improving the long-term accuracy and reliability of noise source localization, and providing a reliable basis for subsequent filtering strategy formulation and anomaly diagnosis. Attached Figure Description

[0007] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0008] Figure 1 This is an application scenario diagram illustrating a noise source identification method for an automotive domain controller power supply network according to an example embodiment of this application; Figure 2 This is a flowchart illustrating a noise source identification method for an automotive domain controller power supply network according to an example embodiment of this application; Figure 3 This is a flowchart illustrating a specific implementation of S140 according to an example embodiment of this application; Figure 4 This is a flowchart illustrating a target parameter subset update method according to an example embodiment of this application; Figure 5 This is a flowchart illustrating a specific implementation of S140 according to another example embodiment of this application; Figure 6 This is a flowchart illustrating a specific implementation of S140 according to another example embodiment of this application; Figure 7This is a flowchart illustrating the process of determining a specific implementation method based on a discrete parameter table according to an example embodiment of this application; Figure 8 This is a schematic diagram of the structure of an electronic device according to an example embodiment of this application.

[0009] The accompanying drawings have illustrated specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to specific embodiments. Detailed Implementation

[0010] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0011] Figure 1 This is an application scenario diagram illustrating a noise source identification method for an automotive domain controller power supply network according to an example embodiment of this application. For example... Figure 1 As shown, First, a predefined test stimulus is applied to the power supply network of the vehicle domain controller. This is achieved by injecting a pre-set test signal into a designated injection point of the power supply network through an internal or external stimulus generation module of the domain controller. The test signal is then superimposed on the DC power supply voltage of the power supply network and propagates through the network.

[0012] Subsequently, at least two power supply nodes are selected as measurement points in the power supply network, and the voltage and / or current response signals of each power supply node during the application of the test excitation are synchronously acquired using a synchronous sampling device. Based on the acquired response signals and the known test excitation signals, the measured frequency response of the power supply network from the test excitation injection point to each power supply node at the current moment is determined using frequency domain analysis methods.

[0013] Then, the measured frequency response is compared with the theoretical frequency response determined by the current power supply network equivalent model parameters. Under preset physical constraints, the power supply network equivalent model parameters are updated. The power supply network equivalent model parameters are used to characterize the resistance, inductance, and capacitance characteristics of wire harnesses, connectors, and filter elements in the power supply network. By updating the parameters, the equivalent model is made closer to the actual electromagnetic characteristics of the power supply network.

[0014] After updating the equivalent model parameters, based on the updated power supply network equivalent model and combined with the actual power supply noise data measured during operation, online noise source identification and analysis of the power supply network is performed. By solving the inversion relationship between the noise source and the noise response of each measurement node, the spatial location and characteristics of the noise source in the power supply network are determined, thereby realizing online identification and location of noise sources in the power supply network of the vehicle domain controller under actual vehicle operating conditions.

[0015] Figure 2 This is a flowchart illustrating a noise source identification method for an automotive domain controller power supply network according to an example embodiment of this application. Figure 2 As shown, the method provided in this embodiment includes: S110. Apply a predefined test stimulus to the power supply network of the vehicle domain controller.

[0016] Optionally, an excitation signal with an amplitude lower than a preset threshold and containing multiple discrete frequency components can be generated, and then the excitation signal can be injected into a designated injection point of the power supply network.

[0017] Specifically, within the signal generation module of the domain controller, based on a target frequency list stored in read-only memory, multiple discrete frequencies that are prone to resonance in the domain controller's power supply network are selected from the target frequency list at preset frequency intervals. For each discrete frequency, a corresponding sinusoidal waveform with an amplitude normalized to 1 is generated. These sinusoidal waveforms are then linearly superimposed after phase alignment to obtain a multi-sinusoidal composite signal. Based on the permissible disturbance range of the power supply network, the overall amplitude of the multi-sinusoidal composite signal is scaled down to a level below a preset threshold. The scaled multi-sinusoidal composite signal is then output as a mixed digital-analog excitation signal at a fixed sampling rate via a digital-to-analog converter, thus forming an excitation signal with an amplitude below the preset threshold and containing multiple discrete frequency components.

[0018] Then, on the automotive domain controller, an excitation injection pad is reserved between the power supply network input filter circuit and the internal load of the domain controller as a designated injection point, and a controllable small-signal injection resistor is connected in series between the designated injection point and the power supply network bus. Before performing the test, the output of the signal generation module is connected to one end of the controllable small-signal injection resistor by controlling the analog switching device inside the domain controller, so that the excitation signal is coupled to the power supply network bus through the controllable small-signal injection resistor. At the same time, in order to prevent the excitation signal from being injected back into the vehicle battery or upstream power source, a unidirectional isolation diode or power switching device can be set between the power supply network bus and the vehicle power supply, and the power switching device can be controlled to be in a turned-off or current-limiting state during the application of excitation, so as to realize that the excitation signal is injected into the designated injection point of the power supply network without affecting the overall vehicle power supply safety.

[0019] Furthermore, the application of predefined test stimuli to the power supply network is periodically triggered when the vehicle domain controller is in an idle state or when the load fluctuation of the power supply network is below a stable threshold. It is worth noting that the aforementioned idle state of the vehicle domain controller could mean the vehicle is undergoing maintenance or repair.

[0020] S120. At at least two power supply nodes, the response signals of the power supply network to the test excitation are collected synchronously.

[0021] Specifically, in the power supply network of the automotive domain controller, at least two power supply nodes can be pre-selected based on the power supply bus topology. These power supply nodes include a bus node near the power supply inlet and a branch node near the critical load. A voltage sampling circuit is set between each power supply node and the reference ground. This voltage sampling circuit includes a resistor divider to attenuate the supply voltage to the input range of the analog-to-digital converter, and an RF suppression capacitor connected in parallel with the resistor divider to suppress high-frequency noise folding. The total resistance and voltage division ratio of the resistor divider are selected based on the maximum operating voltage of the power supply network and the target sampling bandwidth. Furthermore, this embodiment does not limit the specific form of the voltage sampling circuit; it only needs to be able to implement the voltage sampling function.

[0022] The outputs of each voltage sampling circuit are connected to different analog input channels of the analog-to-digital converter module inside the domain controller. A unified sampling clock source and a common sampling trigger signal are configured for the analog-to-digital converter module. The sampling trigger signal is generated by a hardware timing unit shared with the signal generation module of the predefined test stimulus, so that the output start time of the test stimulus has a definite time alignment relationship with the sampling start time of each channel.

[0023] During the acquisition process, the analog-to-digital conversion module is controlled to perform simultaneous sampling or quasi-simultaneous sampling with a time deviation lower than the preset synchronization error threshold on each analog input channel at a sampling frequency no less than twice that of the highest frequency component of the test excitation, so as to obtain a multi-channel response voltage discrete time sequence under a unified time base.

[0024] After completing the data acquisition within the predetermined sampling window, the response signal data of each channel, together with the timestamp generated by the hardware timing unit, are stored in the storage unit of the domain controller. In subsequent processing, the response signals of each power supply node are time-axis aligned and interpolated based on the timestamps to ensure that the requirement of synchronously acquiring the response signals of the power supply network to the test stimulus at at least two power supply nodes is met.

[0025] S130. Based on the response signal and test excitation, obtain the measured frequency response of the power supply network at the current moment.

[0026] While applying a predefined test stimulus to the power supply network, a multi-channel discrete-time series of signals characterizing the power supply network's response to the test stimulus is simultaneously acquired at at least two power supply nodes.

[0027] In the processing unit of the domain controller, a test stimulus reference sequence matching the sampling duration and sampling frequency is constructed based on the predefined test stimulus generation parameters. The test stimulus reference sequence includes discrete-time waveforms of the injection point voltage or current.

[0028] Then, window function weighting and discrete Fourier transform are performed on the test excitation reference sequence and the response signal time series of each power supply node to obtain the excitation spectrum and the corresponding response spectrum at multiple discrete frequency points.

[0029] At each discrete frequency point, the complex value of the corresponding response spectrum is compared with the complex value of the excitation spectrum to obtain the complex transfer function value that characterizes the gain and phase characteristics of the power supply network at that frequency point.

[0030] The complex transfer function values ​​are arranged in frequency order to form a set of frequency response points covering each discrete frequency component of the test excitation. Based on this, according to a preset frequency interpolation strategy, the transfer function of the missing frequency points is interpolated or preserved in terms of amplitude and phase, thereby obtaining the measured frequency response curve of the power supply network from the test excitation injection point to each power supply node within the time interval corresponding to the current sampling window. This curve serves as the measured frequency response of the power supply network at the current moment.

[0031] S140. Based on the measured frequency response and the current power supply network equivalent model parameters, update the power supply network equivalent model parameters under preset physical constraints.

[0032] In this step, the equivalent model parameters of the power supply network can be updated based on the measured frequency response and the current equivalent model parameters of the power supply network, under preset physical constraints. These equivalent model parameters characterize the resistance, inductance, and capacitance properties of the power supply network.

[0033] Optionally, the aforementioned preset physical constraints include parameter monotonicity constraints, which are used to limit the parameters characterizing the aging effect of the component to only be allowed to change along a preset monotonic direction.

[0034] Specifically, during the initialization phase of the power supply network equivalent model parameters, for each parameter in the target parameter subset characterizing the aging effect of components, the changing trend of the parameter with time or service life can be determined in advance based on the component type and failure mechanism. The capacitance value of the electrolytic capacitor can be set to only allow monotonically decreasing with increasing aging, the equivalent series resistance value of the electrolytic capacitor can be set to only allow monotonically increasing with increasing aging, the inductance of the power inductor can be set to only allow monotonically decreasing, and the DC resistance of the power inductor can be set to only allow monotonically increasing. The above monotonic directions are stored in the parameter constraint configuration table by marking the parameter attributes.

[0035] In the process of determining the update value of the target parameter subset under the constraint of parameter variation range, with the goal of minimizing the error between the measured frequency response and the theoretical frequency response determined based on the parameters of the equivalent model of the current power supply network, the update value of the target parameter subset obtained in each iteration is checked for monotonicity. Specifically, for each parameter characterizing the aging effect of the component, the update value obtained in this iteration is compared with the parameter value of the previous update cycle. When the update value violates the preset monotonic direction, the update value of the parameter is forcibly truncated to the same value as the parameter value of the previous cycle or the boundary value that is closest to the parameter value of the previous cycle in the monotonic direction, so as to ensure that the parameter does not change in the opposite direction.

[0036] After completing the monotonicity verification and necessary truncation of all target parameters, the updated parameter values ​​filtered by the monotonicity constraints are written into the equivalent model parameters of the power supply network to ensure that, during multiple online update iterations, all parameters characterizing the aging effect of components always change in a monotonic direction that conforms to the physical aging law.

[0037] It is worth noting that in practical engineering applications, due to factors such as strong electromagnetic interference in the vehicle environment, limited sensor accuracy, model simplification errors, and changes in operating conditions, the equivalent model parameters obtained often exhibit significant random fluctuations or even directional changes. For example, the resistance of the wiring harness and the contact resistance of the connector show a periodic decreasing trend after long-term vehicle service, which is clearly inconsistent with the actual physical aging law of the components. This results in a lack of physical reliability in the online diagnostic results, which in turn affects the reliability of noise source localization and fault warning.

[0038] Therefore, in the above-mentioned method for identifying noise sources in the power supply network of automotive domain controllers based on online updating of equivalent model parameters, it is still necessary to further address how to suppress non-physical fluctuations in parameter estimation caused by measurement noise, model incompleteness, and operating condition disturbances. In particular, it is necessary to suppress the situation where parameters such as wiring harness resistance, connector contact resistance, solder joint resistance, and grounding loop resistance, which should only monotonically deteriorate with aging, decrease or fluctuate drastically during multiple update iterations. This will enable the equivalent power supply network model obtained through online identification to stably reflect the true aging trend of components throughout the entire vehicle lifecycle, providing a reliable parameter basis for subsequent noise source identification, hazard warning, and maintenance decisions.

[0039] In response, the aforementioned parameter monotonicity constraints may further include: pre-classifying the parameters characterizing harness resistance, connector contact resistance, solder joint resistance, and grounding loop resistance in the power supply network equivalent model parameters as aging monotonic parameters, classifying the parameters characterizing capacitance and inductance as non-monotonic parameters, and for each aging monotonic parameter, recording its initial parameter value during the factory calibration stage, and limiting the updated value of the aging monotonic parameter in each update cycle during the online update stage to be no less than the parameter value of the previous update cycle, and not exceeding the upper limit corresponding to the parameter value of the previous update cycle and the maximum relative change rate under the preset maximum relative change rate constraint, so that the aging monotonic parameters only change slowly in a monotonic direction consistent with the component aging law throughout the vehicle's entire life cycle.

[0040] In the above specific implementation method, the equivalent parameters are first classified by physical properties at the model level. The wire harness resistance, connector contact resistance, solder joint resistance and grounding loop resistance, which reflect the conductivity of conductors and contact paths, are clearly marked as aging monotonic parameters. The capacitance and inductance, which are mainly affected by operating conditions and environmental transients, are marked as non-monotonic parameters. The initial reference value of the aging monotonic parameters is recorded during the factory calibration stage.

[0041] Subsequently, during each online parameter update, the method for determining the candidate parameter values ​​obtained by the original recognition algorithm is not changed. Instead, a projection constraint on monotonicity and maximum rate of change is added to the parameter update write-back step: For each aging monotonic parameter, the candidate new value is compared with the parameter value of the previous period. When the candidate new value is less than the parameter value of the previous period, the parameter value of the previous period is directly used to replace the candidate value, and updating in physically unreasonable directions is prohibited. When the candidate new value is greater than the parameter value of the previous period, it is then restricted to between the parameter value of the previous period and the corresponding upper limit according to the preset maximum relative change rate. Thus, the parameter update result is mathematically projected into the feasible region that satisfies monotonicity and slow change.

[0042] By combining the above parameter classification, initial recording, and projection update with monotonic constraints, the monotonic physical laws of component aging are embedded into the online identification process in the form of explicit inequality constraints without adding additional sensors and complex control logic. This enables automatic suppression and filtering of non-physical parameter updates, thereby ensuring the physical rationality and engineering credibility of the evolution of equivalent model parameters over time.

[0043] Figure 3 This is a flowchart illustrating a specific implementation of S140 according to an example embodiment of this application. For example... Figure 3 As shown, the above-mentioned S140 includes: S210. Determine the subset of target parameters that can be updated in the equivalent model parameters of the power supply network.

[0044] Specifically, in the topology of the power supply network equivalent model, the equivalent resistance, inductance, and capacitance parameters are pre-grouped and labeled according to their corresponding physical component categories, circuit locations, and sensitivity to the frequency response of the power supply network, forming a parameter grouping table.

[0045] During the update, based on the current operating conditions and historical noise problem records, parameter groups that have a significant impact on the frequency response of the current target frequency band and are physically prone to drift are selected from the parameter grouping table as candidate parameter sets.

[0046] Further, based on the magnitude and trend of change of each candidate parameter in the previous update cycle, parameters whose changes tend to be stable and whose contribution to the current error is less than the preset threshold are eliminated, while parameters whose changes are active and whose contribution to the error between the measured frequency response and the theoretical frequency response is not less than the preset threshold are retained. These parameters are combined to form a subset of target parameters, which can be used as the set of equivalent model parameters that can be adjusted in subsequent update operations.

[0047] S220. Based on the physical characteristics of each parameter in the target parameter subset, set corresponding parameter variation range constraints.

[0048] For each equivalent resistance, inductance, and capacitance parameter in the target parameter subset, obtain its nominal value, tolerance range, and long-term aging characteristic data of the corresponding physical component recorded during the circuit design phase.

[0049] By combining the maximum allowable temperature rise of the power supply network, the estimated service life, and the currently detected ambient temperature, the minimum and maximum values ​​that each parameter may theoretically reach at the current moment are determined using a pre-stored temperature-lifetime-drift relationship model.

[0050] Based on this, a unidirectional variation constraint is applied to the equivalent resistance parameter, allowing only monotonically increasing changes with aging; a variation constraint is applied to the equivalent capacitance parameter, allowing only monotonically decreasing changes with aging or fluctuations within a limited range; and a symmetrical or asymmetrical variation constraint is applied to the equivalent inductance parameter based on core saturation and coil temperature rise. The variation range of all parameters is numerically clipped to ensure that it does not exceed the intersection of the tolerance range and the aging prediction range. This generates a parameter variation range constraint containing upper and lower limits for each target parameter, which is then used for amplitude limiting constraints during subsequent updates.

[0051] S230. With the goal of minimizing the error between the measured frequency response and the theoretical frequency response determined based on the parameters of the equivalent model of the current power supply network, the updated values ​​of the target parameter subset are determined under the constraint of parameter variation range.

[0052] Figure 4 This is a flowchart illustrating a target parameter subset update method according to an example embodiment of this application. For example... Figure 4 As shown, this can be achieved by extracting the measured resonant frequency and measured quality factor corresponding to at least one resonance peak from the measured frequency response. Then, based on the ratio of the measured resonant frequency to the theoretical resonant frequency of the model, the change in the product of the equivalent inductance and capacitance is determined. Next, based on the changes in the measured quality factor and the theoretical quality factor of the model, combined with the change in the product of the equivalent inductance and capacitance, the change in the equivalent resistance is determined. According to a preset component drift-dominant rule, the change in the product of the equivalent inductance and capacitance is allocated to the updated values ​​of the target inductance parameters and / or target capacitance parameters, and the target resistance parameters are updated in conjunction with the change in the equivalent resistance.

[0053] Specifically, the extraction of at least one resonance peak corresponding to the measured resonance frequency and measured quality factor can be performed on the target frequency band where historical noise problems in the power supply network are frequent. The resonance peak with the largest amplitude is identified from the measured frequency response. Then, the center frequency corresponding to the resonance peak is determined as the measured resonance frequency, and the measured quality factor is determined based on the amplitude and bandwidth of the resonance peak.

[0054] Furthermore, after acquiring the measured frequency response of the power supply network at the current moment, bandpass smoothing and baseline correction can be performed on the amplitude-frequency curve of the measured frequency response to remove high-frequency jitter and overall gain shift caused by measurement noise. Then, within a pre-configured target frequency band, a peak search algorithm is used to scan the amplitude-frequency curve point by point, identifying local maxima whose amplitudes are greater than the amplitudes of several neighboring frequency points and exceed a preset detection threshold. The frequency points corresponding to each local maxima are taken as candidate resonant frequencies. For each candidate resonant, two cutoff frequency points are searched along the frequency axis on its left and right sides, where the amplitude drops to a predetermined proportion of the peak amplitude, and the frequency interval between the two cutoff frequency points is determined. The frequency at the candidate resonant is taken as the center frequency of the candidate resonant, and the corresponding quality factor is further determined. The quality factor is the ratio of the center frequency to the frequency interval. Among all candidate resonance peaks, sort them from largest to smallest according to peak amplitude or quality factor, select at least one resonance peak that ranks first and meets the preset amplitude and quality factor thresholds, take its center frequency as the measured resonance frequency, and output its corresponding quality factor as the measured quality factor for subsequent equivalent model parameter updates.

[0055] Furthermore, to determine the change in the equivalent inductance-capacitance product based on the ratio of the measured resonant frequency to the theoretical resonant frequency of the model, the parameter values ​​of the current equivalent inductance and equivalent capacitance can be obtained in the equivalent model of the power supply network for the model resonant mode corresponding to the measured resonant peak. Based on the correspondence between the frequency of the series or parallel resonant circuit and the inductance-capacitance product, the theoretical resonant frequency of the model can be determined. The ratio of the measured resonant frequency to the theoretical resonant frequency of the model is further determined, and based on the correspondence between the measured and theoretical resonant frequencies, the proportional relationship between the updated equivalent inductance-capacitance product and the current equivalent inductance-capacitance product is derived, thereby determining the target value of the equivalent inductance-capacitance product. The current equivalent inductance-capacitance product is taken as the existing value, and the difference between it and the target value is used to obtain the change in the equivalent inductance-capacitance product. This change is then used in subsequent parameter update steps to adjust the target inductance parameter and / or target capacitance parameter.

[0056] To determine the change in equivalent resistance based on the changes in the measured quality factor and the theoretical quality factor of the model, combined with the change in the product of equivalent inductance and capacitance, the following steps can be taken in the equivalent power supply network model: for the resonant branch corresponding to the measured resonance peak, obtain the current equivalent resistance and the aforementioned equivalent inductance and equivalent capacitance parameter values. Utilize the relationship between the theoretical resonant frequency and the equivalent resistance, and determine the theoretical quality factor of the model according to the preset functional relationship between the resonant network quality factor and resistance, inductance, and capacitance. Compare the measured quality factor with the theoretical quality factor of the model to determine the ratio of the quality factors. Considering the change in the product of equivalent inductance and capacitance, and based on the constraints between the updated product of equivalent inductance and capacitance and the definition of the quality factor, replace the updated equivalent inductance and equivalent capacitance with combined parameters that satisfy the updated product. Under this combination, derive the target value of the equivalent resistance. The target value of the equivalent resistance can be obtained by scaling the current equivalent resistance proportionally to the quality factor ratio, or by numerical iteration. The change in equivalent resistance is further determined as the difference between the target value of equivalent resistance and the current equivalent resistance. This change in equivalent resistance is used as the change in equivalent resistance to represent the magnitude of the equivalent resistance update, and is then used for subsequent updates to the target resistance parameters.

[0057] Furthermore, updating the target resistance parameters can be achieved by pre-establishing a component drift dominance rule base based on the device type, operating voltage, operating current, temperature level, and historical failure statistics of the actual physical components in the power supply network. This rule base records the typical drift direction and amplitude ratio of inductors, capacitors, and related damping resistors under different operating conditions and life stages. During online updates, based on the currently detected ambient temperature, load current, and frequency band of the resonance peak, the corresponding component drift dominance rule is read from the rule base, and the classification result—whether inductor drift is dominant, capacitor drift is dominant, or both inductor and capacitor drift together—is determined for that resonance mode. When inductor drift is determined to be dominant, the change in the product of equivalent inductance and capacitance is fully converted into a change in equivalent inductance. While keeping the equivalent capacitance parameter unchanged, the updated equivalent inductance parameter is determined.

[0058] When it is determined that capacitance drift is dominant, the change in the product of equivalent inductance and capacitance is converted into the change in equivalent capacitance. Under the condition of keeping the equivalent inductance parameter unchanged, the updated equivalent capacitance parameter is determined.

[0059] When it is determined that both inductance and capacitance are drifting, according to the preset inductance and capacitance change ratio coefficients in the rule base, the change in the product of equivalent inductance and capacitance is allocated as the change in inductance and the change in capacitance, respectively. This ensures that the product of the updated equivalent inductance and the updated equivalent capacitance matches the product of the original equivalent inductance and capacitance plus the change, and the updated equivalent inductance parameters and the updated equivalent capacitance parameters are determined respectively.

[0060] Simultaneously, the aforementioned equivalent resistance change is added to the current target resistance parameter to obtain the updated equivalent resistance parameter. After obtaining the equivalent inductance, equivalent capacitance, and equivalent resistance through the above update, each updated parameter is further limited and truncated according to the parameter change range constraints set above. Parameter values ​​exceeding the allowable range are truncated to the corresponding upper or lower limit, thereby forming updated values ​​of the target inductance parameter, target capacitance parameter, and target resistance parameter that meet the preset physical constraints.

[0061] Figure 5 This is a flowchart illustrating a specific implementation of S140 according to another example embodiment of this application. For example... Figure 5 As shown, the above-mentioned S140 includes: S310. At multiple discrete frequency points, determine the complex difference vector between the measured frequency response and the theoretical frequency response.

[0062] Specifically, at least twenty discrete frequency points within the target frequency band, distributed at equal or quasi-equal intervals along the frequency axis, can be pre-selected as a set of defined frequency points for parameter updates. During online updates, the amplitude and phase information of the measured response at each defined frequency point are read from the measured frequency response, and the amplitude and phase information are converted into complex numbers represented by real and imaginary parts. Simultaneously, based on the current power supply network equivalent model parameters, the amplitude and phase information of the theoretical frequency response at each frequency corresponding to the set of defined frequency points are calculated, and similarly converted into complex numbers.

[0063] Subsequently, for each determined frequency point, the measured complex value of the response at that frequency point is subtracted from the theoretical complex value to determine the real part difference and the imaginary part difference, respectively. These are then arranged in a predetermined order, and the real part difference and the imaginary part difference at all determined frequency points are combined to form a complex difference vector, which is used to construct a system of linear equations.

[0064] S320. Construct the Jacobian matrix of the theoretical frequency response with respect to a subset of the target parameters.

[0065] Optionally, based on the structure of the power supply network equivalent model, the analytical expression or numerical template of the Jacobian matrix can be determined and stored offline in advance. During online updates, the current power supply network equivalent model parameters are substituted into the analytical expression or numerical template to generate the current Jacobian matrix.

[0066] Optionally, the sensitivity of the theoretical frequency response to the parameter can be determined at multiple discrete frequency points using a finite difference method. Specifically, for any target parameter, positive and negative small perturbations are applied to the current parameter value. While keeping other parameters constant, the complex values ​​of the theoretical frequency response at each determined frequency point after the perturbation are determined. The difference between the theoretical frequency responses obtained from the two perturbations is divided by the parameter change between the two perturbations to obtain the approximate values ​​of the complex derivatives of the target parameter at each determined frequency point. Then, the approximate values ​​of the real and imaginary derivatives of each target parameter at all determined frequency points are arranged in a predetermined order and used as elements of the column vector corresponding to that parameter in the Jacobian matrix, thereby constructing the Jacobian matrix of the theoretical frequency response for a subset of the target parameters.

[0067] S330. Construct a system of linear equations using the complex difference vector as the observation vector and the Jacobian matrix as the coefficient matrix.

[0068] First, the real and imaginary differences at each frequency point in the complex difference vector are arranged sequentially from low to high frequency, forming a one-dimensional real observation vector. Simultaneously, the real and imaginary derivatives of each target parameter at each frequency point in the Jacobian matrix are arranged in the same order, such that each row corresponds to the real or imaginary part at a specific frequency point, and each column corresponds to the sensitivity of the specific target parameter at all frequency points, thus obtaining a real coefficient matrix for linearizing the error approximation. Based on this, using the increments of each target parameter in the parameter increment vector as unknowns, the product of the coefficient matrix and the parameter increment vector is approximated to the observation vector, resulting in a set of overdetermined linear equations with the goal of minimizing the error. This provides a mathematical model for subsequently solving for the incremental adjustment values ​​of the target parameter subset.

[0069] S340. Solve the system of linear equations to obtain the incremental adjustment values ​​of the target parameter subset.

[0070] Based on the constructed real coefficient matrix and observation vector, the linear equation system is numerically solved using the least squares method with regularization terms.

[0071] Specifically, the coefficient matrix is ​​first normalized column by column to scale the numerical range of each column to a preset standard range, so as to reduce the impact of differences in the dimensions and orders of magnitude of different parameters on the stability of the solution.

[0072] Subsequently, a regularization constraint related to the norm of the parameter increment vector is introduced into the least squares objective function to suppress the problem of excessive parameter increments in the presence of noise or ill-conditioned matrices.

[0073] Based on this, singular value decomposition or Gaussian elimination with column pivoting is used to obtain the incremental adjustment value corresponding to each target parameter.

[0074] After the solution is completed, the scaling factors introduced for each parameter in the normalization step are restored to obtain the incremental adjustment value vector that characterizes the correction magnitude of each target parameter at the original parameter scale.

[0075] S350. Based on the parameter change range constraint, the incremental adjustment value is limited, and the limited incremental adjustment value is superimposed on the current parameter value to obtain the updated value.

[0076] Before performing an online update, for each parameter in the target parameter subset, a corresponding maximum and minimum allowable increment are pre-set to represent the upper limit that each parameter can increase or decrease during a single update.

[0077] After solving the linear equations and obtaining the incremental adjustment values ​​for each target parameter, the incremental adjustment value of each parameter is compared with its corresponding maximum and minimum allowable increments. When the incremental adjustment value of a parameter is greater than the maximum allowable increment, it is truncated to the maximum allowable increment. When the incremental adjustment value of a parameter is less than the minimum allowable increment, it is truncated to the minimum allowable increment. Incremental adjustment values ​​within the allowable range remain unchanged.

[0078] Subsequently, the incremental adjustment values ​​after amplitude limiting are added to the current parameter values ​​of the corresponding target parameters to obtain the updated values ​​of each target parameter. The updated values ​​are then used to replace the original parameters, forming a new set of parameters for the power supply network equivalent model, which is used for frequency response determination and noise source identification in the next cycle.

[0079] Figure 6 This is a flowchart illustrating a specific implementation of S140 according to another example embodiment of this application. For example... Figure 6 As shown, the above-mentioned S140 includes: S410. Obtain the pre-stored discrete parameter table, which stores multiple parameter combinations and their corresponding theoretical frequency response fingerprints.

[0080] Optionally, the aforementioned discrete parameter table can be obtained based on offline measurements and fitting under various temperature and load conditions. Furthermore, when updating parameters, the current temperature and load information of the power supply network can be obtained. Based on the current temperature and load information, a subset of parameter combinations for the corresponding operating conditions can be selected from the discrete parameter table, and a lookup and interpolation determination can be performed within the subset.

[0081] Figure 7 This is a flowchart illustrating the process of determining a specific implementation method based on a discrete parameter table shown in an example embodiment of this application. For example... Figure 7 As shown, the determination of the above discrete parameter table can be carried out in the vehicle prototype or bench test stage by pre-selecting multiple sets of resistance, inductance and capacitance parameter combinations covering the working range of the power supply network. For each set of parameter combinations, the corresponding theoretical frequency response data is determined by simulation and / or obtained by physical testing at multiple discrete frequency points. The amplitude and phase information at each frequency point are uniformly converted into real and imaginary parts and arranged in a predetermined frequency order to form a theoretical frequency response fingerprint vector of fixed length.

[0082] Based on this, a one-to-one correspondence is established between the specific values ​​of each set of resistance, inductance, and capacitance parameters and their corresponding theoretical frequency response fingerprint vectors. These parameters are then stored as discrete parameter tables in the form of tables or database records. Each record in the discrete parameter table contains at least a parameter combination identifier, the value of each parameter, the theoretical frequency response fingerprint vector, and optional generation condition identifier information. This allows the discrete parameter table to be directly read from the storage medium for matching and interpolation during online updates.

[0083] S420 converts the measured frequency response into a measured response fingerprint.

[0084] After acquiring the measured frequency response, select the same set of frequency points used when establishing the discrete parameter table from the measured data, and read the amplitude and phase information of the measured frequency response at each frequency point one by one.

[0085] Optionally, the measured amplitude and phase are preprocessed, including removing DC bias, performing amplitude normalization, and phase expansion, to reduce the impact of test noise and range differences. Subsequently, the amplitude and phase at each frequency point are converted into complex numbers in real and imaginary form according to a predetermined mathematical relationship. Then, the real and imaginary parts of all frequency points are sequentially arranged and combined into a real number vector of fixed length according to the frequency from low to high. This real number vector is defined as the measured response fingerprint, so that the measured response fingerprint is consistent with the theoretical frequency response fingerprint stored in the discrete parameter table in terms of data structure and dimension, which facilitates the subsequent determination of the matching degree.

[0086] S430. Find at least two candidate parameter combinations in the discrete parameter table that have the highest matching degree with the measured response fingerprint.

[0087] For each record in the discrete parameter table, extract its corresponding theoretical frequency response fingerprint vector and compare it with the measured response fingerprint in terms of element dimension. Determine the similarity or distance index between the two according to the preset matching degree measurement method. The matching degree measurement method includes, but is not limited to, the Euclidean distance obtained by summing the squares of the differences between the two vectors, or the correlation coefficient obtained by determining the inner product of the normalized two vectors.

[0088] After determining the matching degree of all records in the discrete parameter table, all parameter combinations are sorted according to similarity or distance index. At least two parameter combinations with the highest matching degree or the smallest distance are selected as candidate parameter combinations. The parameter values ​​of the candidate parameter combinations and the corresponding matching degree or distance index are temporarily stored for subsequent interpolation determination steps.

[0089] S440. Based on the matching weight of each candidate parameter combination, interpolate at least two candidate parameter combinations to determine the updated values ​​of the target parameter subset.

[0090] First, based on the matching degree or distance index of the candidate parameter combinations, they are converted into non-negative weight values. When using the distance index, the reciprocal of the distance or a linear mapping is preferred to ensure that candidate parameter combinations with smaller distances correspond to larger weights. Then, the weights of all candidate parameter combinations are normalized so that the sum of all weights equals one.

[0091] Based on this, for each target parameter in the subset of target parameters, the value of the parameter is read from each candidate parameter combination as an interpolation node, and then weighted and summed according to the corresponding normalized weights to obtain the weighted average value of the target parameter. The weighted average value is then used as the updated value of the target parameter.

[0092] When there are more than two candidate parameter combinations and they form a hypercube or polyhedron in the parameter space, restrictions on parameter distribution uniformity and extrapolation risk can be added on the basis of the above weighted summation to avoid the interpolation results from exceeding the preset parameter variation range, thereby forming updated values ​​of the target parameter subset that satisfy physical constraints.

[0093] S150. Using the updated power supply network equivalent model parameters, perform online noise source identification for the power supply network.

[0094] Optionally, uncertainty intervals can be associated with at least some parameters in the updated power supply network equivalent model parameters. Noise source inversion analysis is then performed within these uncertainty intervals to obtain confidence intervals or probability distributions of the noise source locations.

[0095] It is worth noting that in most engineering practices, equivalent model parameters are either simplified to fixed nominal values ​​or only roughly adjusted within empirical ranges, without systematically modeling and constraining the value range of each parameter from an uncertainty perspective. This leads to a situation where, during noise source inversion and online parameter identification, parameter updates often rely on the numerical convergence characteristics of the algorithm itself, lacking effective constraints from device manufacturing tolerances, prototype measurement statistical characteristics, and physical environmental factors such as temperature and aging. This easily results in parameter drift or distortion, thereby weakening the credibility and interpretability of power supply network noise source identification results.

[0096] Therefore, initial parameter ranges can be determined for each target parameter characterizing the resistance, inductance, and capacitance of the power supply network, based on the nominal values ​​in the corresponding component's datasheet and manufacturing tolerances. Multi-condition frequency response measurements are performed on multiple representative hardware samples under prototype or bench conditions. The frequency response is then fitted based on the parameters of the current power supply network equivalent model, and the distribution range of the fitting results and the distribution of the fitting residuals for each target parameter across different samples are statistically analyzed. Based on the relationship between the fitting result distribution and the initial parameter range, the initial parameter range is adjusted by widening or narrowing. Combined with temperature and aging mileage information, the adjusted parameter range is constructed as a dynamic uncertainty range that varies with temperature and service time, used to constrain the value range of each target parameter during noise source inversion analysis.

[0097] Specifically, for each target parameter that characterizes the resistance, inductance, and capacitance of the power supply network, an initial parameter range reflecting the upper and lower bounds of manufacturing error is defined using the nominal values ​​and manufacturing tolerances of the corresponding component datasheets.

[0098] Subsequently, multiple representative hardware samples were selected under prototype or bench conditions, and the actual frequency response was measured under various working conditions. The measured frequency response was then fitted based on the equivalent model parameters of the current power supply network. By statistically analyzing the distribution range of each target parameter obtained from fitting multiple samples and the distribution of fitting residuals, the actual impact of model structure error and sample dispersion on the parameters was quantified.

[0099] Then, the statistical distribution is compared with the initial parameter range. When the fitting results are concentrated in a narrow range within the initial range, the range is appropriately narrowed. When the fitting results frequently approach or exceed the range boundary, the range is appropriately widened to form a parameter range that is more consistent with the characteristics of the real hardware group in a statistical sense.

[0100] Based on this, and by combining temperature information and aging mileage information, and according to the device thermal characteristic curve and aging law, the adjusted parameter range is constructed into a dynamic uncertainty range model that varies with temperature and service time, so that the allowable fluctuation range of parameters is related to the actual operating conditions.

[0101] Ultimately, during the noise source inversion analysis, the updated values ​​of all target parameters are constrained by the aforementioned dynamic uncertainty interval, thereby achieving regularization and physical limitation of the inverse problem solution. This ensures stable numerical solution, maintains physical rationality, and directly yields stable and reliable noise source identification results.

[0102] After updating the parameters of the equivalent model of the power supply network, for the subset of target parameters in the equivalent model parameters, the parameter update values ​​and corresponding measured frequency response fitting errors in several historical update iterations are read. Based on the relationship between the fitting error and the parameter update values, a preset statistical analysis method or empirical rule is used to determine the upper and lower offsets corresponding to the current parameter value for each target parameter. The upper and lower offsets are then added to the current parameter value to form the lower and upper limits of the target parameter, thereby defining the uncertainty interval of the target parameter.

[0103] The statistical analysis methods include: determining the standard deviation or absolute deviation of the target parameter in the most recent update records, and amplifying the standard deviation or absolute deviation by a preset factor to determine the upper and lower offsets.

[0104] The empirical rules include: when the fitting error between the measured frequency response and the theoretical frequency response is lower than the error threshold, reduce the uncertainty interval width; when the fitting error is close to or exceeds the error threshold, increase the uncertainty interval width.

[0105] After determining the uncertainty interval for each parameter in the target parameter subset, the current value of each parameter and its corresponding uncertainty interval boundary are written into the uncertainty parameter configuration table in the form of parameter identifier, current value, lower limit value and upper limit value, for subsequent noise source inversion analysis within the uncertainty interval.

[0106] Based on the uncertainty parameter configuration table, multiple sets of parameter samples are generated within the uncertainty interval of each target parameter according to a preset sampling strategy. The sampling strategy includes uniform grid sampling, Latin hypercube sampling, or Monte Carlo random sampling within the parameter interval.

[0107] For each set of parameter samples, the set of parameter samples is written into the equivalent model of the power supply network. The updated equivalent model is used to perform forward simulation to determine the possible candidate locations of multiple noise sources in the power supply network one by one. The theoretical frequency response or characteristic index corresponding to each candidate noise source location under the set of parameter samples is obtained, and the determination results are compared with the current measured frequency response to measure the error.

[0108] After completing the forward simulation and error measurement of all parameter samples and all candidate noise source locations, the number of times the combination of parameter samples and candidate noise source locations that can match the measured frequency response within the preset error threshold is counted. The number of times each candidate noise source location is successfully matched is divided by the total number of simulations to obtain the probability value corresponding to the noise source location. The noise source locations are then sorted in physical space according to the probability values.

[0109] Based on this, the set of candidate noise source locations whose cumulative probability reaches the preset confidence level is determined as the confidence interval of the noise source location, or a discrete spatial probability distribution is formed directly using the probability values ​​corresponding to each candidate noise source location, which is used to guide fault location and hardware rectification.

[0110] In this embodiment, by applying a predefined test stimulus to the power supply network of the automotive domain controller, and then simultaneously acquiring the response signal of the power supply network to the test stimulus at at least two power supply nodes, the actual impact of component parameter drift caused by temperature and aging on network characteristics is directly captured. Then, based on the response signal and the test stimulus, the measured frequency response of the power supply network at the current moment is obtained. Next, based on the measured frequency response and the current equivalent model parameters of the power supply network, the equivalent model parameters of the power supply network are updated under preset physical constraints, so that the model can dynamically track and approximate the time-varying characteristics of the real network. Finally, the updated equivalent model parameters of the power supply network are used to perform online noise source identification for the power supply network, thereby effectively distinguishing between network parameter drift and new noise sources, thus improving the long-term accuracy and reliability of noise source localization, and providing a reliable basis for subsequent filtering strategy formulation and anomaly diagnosis.

[0111] Figure 8 This is a schematic diagram of the structure of an electronic device according to an example embodiment of this application. For example... Figure 8 As shown, the electronic device 500 provided in this embodiment includes: a processor 501 and a memory 502; wherein: Memory 502 is used to store computer programs, and the memory may also be flash memory.

[0112] Processor 501 is used to execute the execution instructions stored in the memory to implement the various steps in the above method. For details, please refer to the relevant descriptions in the preceding method embodiments.

[0113] Alternatively, the memory 502 can be either standalone or integrated with the processor 501.

[0114] When the memory 502 is a device independent of the processor 501, the electronic device 500 may further include: Bus 503 is used to connect the memory 502 and the processor 501.

[0115] This embodiment also provides a readable storage medium storing a computer program, which, when executed by at least one processor of an electronic device, enables the electronic device to perform the methods provided in the various embodiments described above.

[0116] This embodiment also provides a program product including a computer program stored in a readable storage medium. At least one processor of an electronic device can read the computer program from the readable storage medium, and the at least one processor executes the computer program to cause the electronic device to perform the methods provided in the various embodiments described above.

[0117] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the claims.

[0118] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A method for noise source identification in an automotive domain controller power supply network, characterized in that, include: Apply predefined test stimuli to the power supply network of the vehicle domain controller; At at least two power supply nodes, the response signals of the power supply network to the test stimulus are collected synchronously. Based on the response signal and the test stimulus, the measured frequency response of the power supply network at the current moment is obtained; Based on the measured frequency response and the current power supply network equivalent model parameters, the power supply network equivalent model parameters are updated under preset physical constraints. Using the updated power supply network equivalent model parameters, perform online noise source identification for the power supply network.

2. The method according to claim 1, characterized in that, Applying a predefined test stimulus to the power supply network includes: Generate an excitation signal with an amplitude lower than a preset threshold and containing multiple discrete frequency components; The excitation signal is injected into a designated injection point in the power supply network.

3. The method according to claim 1, characterized in that, The step of updating the power supply network equivalent model parameters based on the measured frequency response and the current power supply network equivalent model parameters, under preset physical constraints, includes: Determine the subset of target parameters that are allowed to be updated in the equivalent model parameters of the power supply network; Based on the physical characteristics of each parameter in the target parameter subset, set corresponding parameter variation range constraints; With the goal of minimizing the error between the measured frequency response and the theoretical frequency response determined based on the parameters of the equivalent model of the current power supply network, the updated values ​​of the target parameter subset are determined under the constraint of the parameter variation range.

4. The method according to claim 3, characterized in that, The step of minimizing the error between the measured frequency response and the theoretical frequency response determined based on the equivalent model parameters of the current power supply network, and determining the updated values ​​of the target parameter subset under the constraint of the parameter variation range, includes: From the measured frequency response, extract the measured resonant frequency and measured quality factor corresponding to at least one resonant peak; Based on the ratio of the measured resonant frequency to the theoretical resonant frequency of the model, the change in the product of the equivalent inductance and capacitance is determined. Based on the changes in the measured quality factor and the theoretical quality factor of the model, and combined with the changes in the product of the equivalent inductance and capacitance, the changes in the equivalent resistance are determined. Based on the preset component drift-dominant rule, the change in the product of the equivalent inductance and capacitance is allocated to the updated values ​​of the target inductance parameter and / or target capacitance parameter, and the target resistance parameter is updated in conjunction with the change in the equivalent resistance.

5. The method according to claim 4, characterized in that, The extraction of the measured resonance frequency and measured quality factor corresponding to at least one resonance peak includes: For the target frequency band where historical noise problems in the power supply network are frequent, identify the resonance peak with the largest amplitude from the measured frequency response; The center frequency corresponding to the resonance peak is determined as the measured resonance frequency, and the measured quality factor is determined based on the amplitude and bandwidth of the resonance peak.

6. The method according to claim 3, characterized in that, The step of minimizing the error between the measured frequency response and the theoretical frequency response determined based on the equivalent model parameters of the current power supply network, and determining the updated values ​​of the target parameter subset under the constraint of the parameter variation range, includes: At multiple discrete frequency points, determine the complex difference vector between the measured frequency response and the theoretical frequency response; Construct the Jacobian matrix of the theoretical frequency response for the subset of the target parameters; Using the complex difference vector as the observation vector and the Jacobian matrix as the coefficient matrix, a system of linear equations is constructed. Solve the system of linear equations to obtain the incremental adjustment values ​​of the subset of target parameters; Based on the parameter variation range constraint, the incremental adjustment value is subjected to amplitude limiting, and the amplitude-limited incremental adjustment value is superimposed on the current parameter value to obtain the updated value.

7. The method according to claim 6, characterized in that, The construction of the Jacobian matrix of the theoretical frequency response for the subset of the target parameters includes: Based on the structure of the equivalent model of the power supply network, the analytical expression or numerical template of the Jacobian matrix is ​​determined and stored offline in advance. During online updates, the parameters of the current power supply network equivalent model are substituted into the analytical expression or numerical template to generate the current Jacobian matrix.

8. The method according to claim 3, characterized in that, The step of minimizing the error between the measured frequency response and the theoretical frequency response determined based on the equivalent model parameters of the current power supply network, and determining the updated values ​​of the target parameter subset under the constraint of the parameter variation range, includes: Obtain the pre-stored discrete parameter table, which stores multiple parameter combinations and their corresponding theoretical frequency response fingerprints; The measured frequency response is converted into a measured response fingerprint; Search the discrete parameter table for at least two candidate parameter combinations that have the highest matching degree with the measured response fingerprint; Based on the matching weight of each candidate parameter combination, interpolation is performed on the at least two candidate parameter combinations to obtain the updated values ​​of the target parameter subset.

9. The method according to claim 8, characterized in that, The discrete parameter table is obtained based on offline measurements and fitting under various temperature and load conditions; the method further includes: When updating parameters, obtain the current temperature and load information of the power supply network; Based on the current temperature and load information, a subset of parameter combinations for the corresponding operating conditions is selected from the discrete parameter table, and the search and interpolation determination is performed within the subset.

10. The method according to claim 1, characterized in that, The step of using the updated power supply network equivalent model parameters to perform online noise source identification for the power supply network includes: Associate uncertainty intervals with at least some parameters in the updated power supply network equivalent model parameters; Noise source inversion analysis is performed within the uncertainty interval to obtain the confidence interval or probability distribution of the noise source location.

11. The method according to claim 1, characterized in that, The application of a predefined test stimulus to the power supply network is periodically triggered when the vehicle domain controller is idle or when the load fluctuation of the power supply network is below a stable threshold.

Images