Automobile on-line current detection method and terminal equipment

Abstract

The invention relates to the field of vehicle current detection, in particular to an automobile online current detection method and terminal equipment. The method comprises the following steps: acquiring real-time current of a vehicle to obtain a multi-time-sequence sampling synchronous current waveform, performing microsecond offset identification and multi-dimensional sampling feature relevance mining, and constructing a synchronous current response matrix; a vehicle-mounted environment monitoring unit is called to collect real-time environment temperature information of the vehicle, temperature trend change mining and temperature drift factor mining are carried out, and temperature drift influence factors are extracted; current detection reference calibration analysis is carried out based on the temperature drift influence factor and the synchronous current response matrix, temperature deviation compensation intelligent optimization is carried out, and a temperature deviation equivalent regulation and control engine is constructed; the method comprises the following steps: monitoring full-band operation electromagnetic parameters of a plurality of subsystems of a vehicle, carrying out frequency-band electromagnetic fluctuation identification one by one, carrying out multi-scale decomposition, and extracting an electromagnetic interference disturbance spectral line set; according to the invention, real-time and accurate vehicle on-line current detection is realized.

Description

Technical Field

[0001] The present invention relates to the field of vehicle current detection, and in particular to an automobile online current detection method and terminal equipment. Background Art

[0002] With the rapid development of new energy vehicles and smart car technologies, the complexity of vehicle electronic systems continues to increase, and the degree of electrification continues to rise, placing higher demands on vehicle current monitoring and management. This is especially true in key electrical components such as power batteries, motor controllers, and onboard chargers. Current fluctuations are directly related to the vehicle's operating status, energy efficiency management, and safety assurance. Therefore, real-time online monitoring of vehicle current is crucial for improving vehicle performance, extending equipment life, and ensuring driving safety.

[0003] Traditional automotive current detection methods often rely on offline or fixed-point testing. These methods, such as assessing current status by connecting a test instrument during regular maintenance, or deploying current sensors at key locations for single-point monitoring, suffer from poor detection timeliness, limited coverage, and slow response speeds. These methods struggle to meet the global, continuous, and high-precision monitoring requirements of modern intelligent vehicles. Furthermore, during vehicle operation, current fluctuations exhibit significant nonlinearity and suddenness due to environmental factors, electromagnetic interference, and dynamic load changes. Failure to monitor current fluctuations in real time can easily lead to safety risks such as electrical system anomalies, equipment overloads, and even fires.

[0004] While some high-end vehicles have introduced advanced monitoring systems in recent years, achieving preliminary dynamic current sensing, most systems still rely on collecting data and uploading it to a control unit for analysis and judgment. This leads to response delays, low processing efficiency, and a lack of timely feedback on abnormal conditions. Furthermore, most current current detection solutions fail to intelligently identify current fluctuation patterns, making it difficult to detect potential faults and abnormal behavior in a timely manner, severely restricting the intelligent management of onboard electrical systems. Summary of the Invention

[0005] In order to solve the above technical problems, the present invention proposes an automobile online current detection method and terminal equipment to solve at least one of the above technical problems.

[0006] To achieve the above object, the present invention provides a method for detecting an online current of an automobile, comprising the following steps:

[0007] Step S1: Collect the real-time current of the vehicle to obtain the multi-time sampling synchronous current waveform, perform microsecond offset identification and multi-dimensional sampling feature correlation mining, and construct a synchronous current response matrix;

[0008] Step S2: calling the vehicle environment monitoring unit to collect the real-time ambient temperature information of the vehicle, and performing temperature trend change mining and temperature drift factor mining to extract the temperature drift influencing factors;

[0009] Step S3: performing current detection benchmark calibration analysis based on the temperature drift influencing factor and the synchronous current response matrix, and performing intelligent optimization of temperature deviation compensation to build a temperature deviation equivalent control engine;

[0010] Step S4: monitoring the full-band operating electromagnetic parameters of multiple subsystems of the vehicle, identifying electromagnetic fluctuations in each frequency band, and performing multi-scale decomposition to extract electromagnetic interference disturbance spectrum lines;

[0011] Step S5: clustering the disturbance patterns of the electromagnetic interference disturbance spectrum line set, performing dynamic electromagnetic disturbance suppression, and performing waveform distortion correction to construct an electromagnetic disturbance suppression engine;

[0012] Step S6: Based on the temperature deviation equivalent control engine and the electromagnetic disturbance suppression engine, intelligent collaborative current detection re-optimization is performed, and post-intelligent precision optimization is performed to build a self-evolving current detection precision maintenance engine.

[0013] This specification provides a terminal device, comprising a main control circuit board, an LCD screen, and a single-circuit module. The main control circuit board communicates data with the six single-circuit modules via six SPI interfaces; the single-circuit modules are powered by the main control board; the LCD screen displays the current time, Wi-Fi connection status, and the set and real-time values ​​of each vehicle-mounted circuit; the main control circuit board transmits real-time data to a host computer for analysis, processing, and display via a USB 2.0 interface; and the terminal device is used to execute the above-described vehicle online current detection method, comprising:

[0014] The synchronous sampling module is used to collect real-time vehicle current, obtain multi-time sampling synchronous current waveforms, identify microsecond offsets, and mine the correlation of multi-dimensional sampling features to construct a synchronous current response matrix.

[0015] The temperature drift mining module is used to call the vehicle environmental monitoring unit to collect the vehicle's real-time ambient temperature information, and conduct temperature trend change mining and temperature drift factor mining to extract the temperature drift influencing factors;

[0016] The deviation compensation module is used to perform current detection benchmark calibration analysis based on the temperature drift influencing factor and the synchronous current response matrix, and to perform intelligent optimization of temperature deviation compensation to build a temperature deviation equivalent control engine;

[0017] The electromagnetic interference module is used to monitor the full-band operating electromagnetic parameters of multiple subsystems of the vehicle, identify electromagnetic fluctuations band by band, and perform multi-scale decomposition to extract the electromagnetic interference disturbance spectrum line set;

[0018] The disturbance suppression module is used to cluster the disturbance patterns of the electromagnetic interference disturbance spectrum line set, perform dynamic electromagnetic disturbance suppression, and perform waveform distortion correction to build an electromagnetic disturbance suppression engine;

[0019] The intelligent precision optimization module is used to perform intelligent collaborative current detection re-optimization based on the temperature deviation equivalent control engine and the electromagnetic disturbance suppression engine, and perform post-intelligent precision optimization to build a self-evolving current detection precision maintenance engine.

[0020] The beneficial effects of the present invention are specifically as follows: utilizing a high-frequency sampling mechanism (such as the μs level) to achieve accurate capture of rapid dynamic current changes, ensuring that signal changes are not missed by traditional sampling frequencies; microsecond offset identification can accurately analyze the jitter or synchronization error of sensor signals in a very short time, which is helpful for subsequent error modeling; mining the correlation between multi-dimensional sampling features (such as peak values, envelopes, and zero crossings) enables the system to identify complex dynamic features and improve the understanding of nonlinear current fluctuations; the construction of a synchronous current response matrix provides a unified data structure interface for subsequent temperature and electromagnetic interference compensation, thereby enhancing system consistency and adaptability. Through real-time temperature trend mining, the system can quickly identify the dynamic impact of sudden changes in ambient temperature (such as cold start or exposure to the sun) on the current detection results; temperature drift factor mining can identify sensor deviation behavior in different temperature zones, and achieve modeling and tracking of drift trends; the extracted temperature drift influencing factors can be used for subsequent calibration optimization to avoid false alarms, missed alarms or power anomaly judgments caused by temperature changes; achieve adaptive reconstruction of the current detection benchmark to avoid the long-term error accumulation caused by the use of fixed calibration curves; introduce an intelligent temperature deviation compensation mechanism to achieve dynamic adjustment and correction of the sensor output signal, effectively eliminating thermal sensitivity. The system can eliminate measurement offsets caused by interference. The temperature deviation equivalent control engine has self-learning and self-correction capabilities, adapting to different vehicle conditions, seasons, and regional environments. Full-band monitoring enables full coverage identification of electromagnetic emissions from modules such as high-voltage motors, battery management systems, and wireless communications. Multi-scale decomposition technology improves the system's ability to classify strong, weak, and sudden interference. Extracting electromagnetic interference spectrum lines helps identify the patterned characteristics and propagation patterns of interference sources, laying the foundation for interference elimination. The system can adapt to and adaptively identify complex electromagnetic environments, making it particularly suitable for interference-stricken scenarios such as smart electric vehicles and hybrid vehicles. Disturbance pattern clustering can classify and manage complex electromagnetic interference, enabling behavioral modeling of interference waveforms. The dynamic interference suppression mechanism selectively suppresses or filters electromagnetic disturbances, rather than traditional static low-pass / band-pass filtering, to avoid useful signal loss. Waveform distortion correction restores the true current curve before interference superposition, improving the signal integrity and authenticity of the measurement. The constructed electromagnetic disturbance suppression engine can be ported as an independent system module to different vehicle models or platforms, demonstrating strong versatility and engineering value. It achieves collaborative anti-interference capabilities under dual interference of temperature and electromagnetic interference, solving the technical bottleneck that makes it difficult for the current industry to deal with these two types of problems at the same time; the post-intelligent precision optimization mechanism can continuously correct system parameters through historical error reverse analysis and online learning strategies; the constructed self-evolution engine has long-term adaptive capabilities, which is especially suitable for maintaining accuracy after long-term operation of vehicles; it significantly improves the system's stability and measurement accuracy in complex operating environments, and reduces problems such as system misjudgment, power scheduling imbalance, and false fault alarm triggering caused by errors. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a schematic flow chart of the steps of a method for detecting online current of an automobile according to the present invention;

[0022] Figure 2 Detailed implementation flow chart of step S1;

[0023] Figure 3 Detailed implementation flow chart of step S2;

[0024] Figure 4 This is a schematic diagram of the structure of the main control circuit board of the terminal device. DETAILED DESCRIPTION

[0025] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0026] This application provides an example of an automotive online current detection method and terminal device. The execution entities of the method and terminal device include, but are not limited to, the following: mechanical equipment, data processing platforms, cloud server nodes, network upload devices, etc., which can be considered as general computing nodes of this application. The data processing platform includes, but is not limited to, at least one of an audio and image management system, an information management system, and a cloud data management system.

[0027] See also Figures 1 to 4 The present invention provides a method for detecting an online current of an automobile, comprising the following steps:

[0028] Step S1: Collect the real-time current of the vehicle to obtain the multi-time sampling synchronous current waveform, perform microsecond offset identification and multi-dimensional sampling feature correlation mining, and construct a synchronous current response matrix;

[0029] Step S2: calling the vehicle environment monitoring unit to collect the real-time ambient temperature information of the vehicle, and performing temperature trend change mining and temperature drift factor mining to extract the temperature drift influencing factors;

[0030] Step S3: performing current detection benchmark calibration analysis based on the temperature drift influencing factor and the synchronous current response matrix, and performing intelligent optimization of temperature deviation compensation to build a temperature deviation equivalent control engine;

[0031] Step S4: monitoring the full-band operating electromagnetic parameters of multiple subsystems of the vehicle, identifying electromagnetic fluctuations in each frequency band, and performing multi-scale decomposition to extract electromagnetic interference disturbance spectrum lines;

[0032] Step S5: clustering the disturbance patterns of the electromagnetic interference disturbance spectrum line set, performing dynamic electromagnetic disturbance suppression, and performing waveform distortion correction to construct an electromagnetic disturbance suppression engine;

[0033] Step S6: Based on the temperature deviation equivalent control engine and the electromagnetic disturbance suppression engine, intelligent collaborative current detection re-optimization is performed, and post-intelligent precision optimization is performed to build a self-evolving current detection precision maintenance engine.

[0034] In the embodiment of the present invention, see Figure 1 , is a schematic flow chart of the steps of a method for detecting an online current of an automobile according to the present invention. In this example, the steps of the method for detecting an online current of an automobile include:

[0035] Step S1: Collect the real-time current of the vehicle to obtain the multi-time sampling synchronous current waveform, perform microsecond offset identification and multi-dimensional sampling feature correlation mining, and construct a synchronous current response matrix;

[0036] In this embodiment, high-time-precision current signals are collected from multiple key current channels of the entire vehicle, and the data at different sampling points are time-aligned and feature-cross-analyzed to ultimately construct a "synchronous current response matrix" that can support electromagnetic disturbance analysis. To ensure data integrity and accuracy, a multi-channel current sensing system with a sampling frequency of 1MHz and a resolution of 16 bits was selected in the experiment. Hall effect sensors (such as the LEM series) were deployed on multiple power paths, including the main drive motor phase current, DC / DC converter output current, on-board charging module input and output current, and auxiliary load power supply branches. During the acquisition process, the vehicle completed multiple typical operating conditions (such as constant speed driving, rapid acceleration, energy feedback deceleration, etc.) under simulated driving conditions. The acquisition time for each operating condition was no less than 20 seconds to ensure that the full change cycle of the current waveform was captured. In terms of time synchronization, due to the microsecond-level hardware delay and data write lag in different sensing channels, the system introduced a high-precision GPS timestamp to uniformly mark each sampling channel. At the same time, combined with the FPGA-controlled synchronization trigger module, nanosecond-level time alignment of the data frame was achieved. After data acquisition is complete, the first step is to identify microsecond offsets in the data of each current channel. Using time-domain peak matching and sliding window alignment algorithms, the temporal positional differences of response events (such as rising edges and breakpoints) in each channel are detected, compensating for micro-offsets caused by sampling channel differences. In experiments, after the drive motor control signal was triggered, current channels A, B, and C were found to have offsets of up to 3.6 microseconds. Dynamic interpolation and resampling techniques were used to synchronize all of these channels. Next, the system enters the multi-dimensional sampling feature correlation mining phase. Specifically, the current values ​​of each channel at each time point are combined into a feature vector, and the trajectory pattern formed by these vectors in the time series is calculated. By calculating metrics such as mutual information between channels, covariance trends, and correlation gradients, the degree of current feature coupling between different channels under different vehicle conditions is analyzed. During hill climbing conditions, the changes in the drive motor phase current and the DC / DC output current exhibit a dynamic correlation of 0.92. However, during braking energy regeneration, the DC / DC output current undergoes a reverse abrupt change, indicating a significant decoupling from the main motor channel. This stage also uses cluster analysis to model the current coupling structure under specific behavior modes, and classifies it into three types of structures: high correlation (such as between phase currents), medium correlation (such as power and auxiliary power supply), and weak correlation (such as system no-load current fluctuations in standby state). Ultimately, the system uses a unified time axis as a benchmark and constructs a "synchronous current response matrix" according to time windows (such as one window every 1 millisecond). Each row of the matrix corresponds to a time node, and each column represents the sampling value of a current channel. The metadata attached to the matrix also records the physical components corresponding to the channel, sampling accuracy, and synchronization offset compensation values. This matrix provides a high-quality raw data foundation for subsequent electromagnetic disturbance detection, spectrum analysis, and disturbance source tracing analysis, and is the first layer of structured data results for the entire current detection system.

[0037] Step S2: calling the vehicle environment monitoring unit to collect the real-time ambient temperature information of the vehicle, and performing temperature trend change mining and temperature drift factor mining to extract the temperature drift influencing factors;

[0038] In this embodiment, the vehicle's onboard environmental monitoring module is used to collect real-time temperature information and explore the patterns of temperature changes over time and operating conditions to identify temperature drift factors that may cause current detection errors. Because the various sensors in the on-board current detection system (especially the Hall effect sensor, current amplifier, and ADC module) are sensitive to temperature changes, the dynamic changes in ambient temperature are a key variable affecting the quality of the sampled data. In the experiment, multiple temperature monitoring units were deployed on the vehicle platform. The collection points included the motor compartment, the exterior of the BMS box, the electronic controller housing, the vicinity of the on-board power module, and the ambient temperature of the cockpit. These temperature sensors communicated with the central control unit using an I2C or CAN interface with an accuracy of ±0.5°C and a refresh rate of 1Hz. During a typical test, the vehicle was operated continuously for 30 minutes under different operating conditions (such as urban congestion, constant speed driving, and high-speed cruising). The system recorded the temperature values ​​and timestamps of five temperature points every second. After data collection was completed, the temperature trend change mining phase began. During this phase, a moving window analysis method (for example, with a 5-minute cycle) is used to identify the slope, amplitude, and frequency of temperature variations and construct a time-temperature evolution model. For example, during a long ramp, the motor compartment temperature rose from 45°C to 79°C in less than 8 minutes, demonstrating a typical "heat accumulation trend." Simultaneously, the temperature of the electronic controller housing increased at a slower rate, reflecting passive thermal conductivity. This trend information is valuable for predicting drift in certain current sampling modules. The system then conducts temperature drift factor mining, identifying the specific ways in which temperature changes affect the current sensing system. This step uses correlation analysis to compare temperature data with the current response matrix to identify the synchronous relationship between current offset and temperature anomalies. Experiments revealed that when the temperature near the electronic controller exceeded 65°C, some channel current signals exhibited systematic zero-bias drift, with the drift reaching 1.2% to 2.5% of the original signal. Further analysis revealed that this drift was related to the gain change of the amplifier circuit under high temperature conditions. In addition, it was observed that under low-temperature startup conditions (<10°C), the battery circuit current response speed slowed down by about 40ms, indicating that temperature also affects the dynamic response of the signal. Finally, the system extracts key temperature drift influencing factors from the above analysis. These factors include but are not limited to: the temperature rise rate exceeds a specific threshold (such as 3°C / min), the local temperature reaches an absolute value threshold (such as the shell temperature >65°C), and the temperature gradient exceeds the device operating specification (such as temperature difference >20°C / 5min). Each influencing factor is assigned a priority label and incorporated into the parameter compensation module or disturbance identification model of the subsequent current detection system.

[0039] Step S3: performing current detection benchmark calibration analysis based on the temperature drift influencing factor and the synchronous current response matrix, and performing intelligent optimization of temperature deviation compensation to build a temperature deviation equivalent control engine;

[0040] In this embodiment, based on the temperature drift influencing factors acquired in the previous step, the system focuses on several key temperature zones, such as the engine compartment upper edge, the power battery interface, the onboard air conditioning module, and the sensor cavity. Temperature fluctuations in these areas significantly impact the current signal. By aligning the timestamps with the synchronized current response matrix, the system establishes a direct correlation mapping between temperature changes and current responses. This process utilizes a multidimensional data synchronization calibration method, superimposing the current changes and temperature curves within a 1-second sampling window at a 50ms sampling frequency to extract the specific current drift characteristics exhibited by temperature changes. In a typical test, when the engine compartment temperature increased from 35°C to 55°C, the peak current at the sampling point decreased by an average of 0.4A, corresponding to a standard current deviation of approximately 7%. The system records this high-temperature-induced current response slowdown and classifies it as a "temperature-induced negative drift factor." After determining the temperature-current relationship curve, the system enters the current baseline calibration phase. This step compares historical standard current waveforms (such as the low-temperature starting current template) with the current temperature-affected waveform to identify the specific drift direction, offset magnitude, and recovery trend. This comparison goes beyond simple mean shifts and also includes changes in waveform structure similarity, displacement of the main spectral peak, and energy variations in high-frequency components. By constructing a "temperature-induced current drift rate table," the system establishes current baseline correction factors for different temperature ranges. For example, within a 10°C range, the startup peak current is typically 2.5% higher, while within a 60°C range, the charging current is typically 5% lower. Based on this, the system generates calibration vectors for each temperature zone, which serve as input to the subsequent control engine. During compensation optimization, the system inputs the baseline calibration results into the adaptive optimization module, which uses a multivariate regression tree model to predict the current response in different temperature zones. The compensation strategy goes beyond linear stretching or compression and considers the synergistic effects of the temperature change rate (dT / dt), the duration of the temperature rise, and the frequency of current fluctuations. For example, in a measured operating condition, when the temperature rises at a rate of 2°C / min for more than 10 minutes, the system activates the high thermal stress model. Compared to the traditional short-term fluctuation model, its deviation compensation factor is automatically increased to ensure a stable current response output. The optimization process was verified with the help of an existing drift sample set. In the experimental data, the average current error after correction dropped to within ±0.12A, which is significantly better than the ±0.45A range in the uncompensated state. The above-mentioned temperature drift calibration factor, compensation optimization model and historical correction trajectory are integrated to construct a temperature deviation equivalent control engine. The engine has the capabilities of online adaptation, dynamic switching and error learning, and can actively adjust the current response processing strategy according to real-time temperature data. The engine contains a temperature perception layer, an offset decision layer, a compensation execution layer and a feedback learning layer to form a closed-loop dynamic control. In actual vehicle deployment, the engine communicates with the on-board temperature monitoring system in real time and refreshes the offset judgment logic every 5 seconds.When the temperature suddenly changes (such as a cold engine heating up quickly), the engine automatically increases its response sensitivity to compensate in advance. During continuous driving conditions, this control engine can stably maintain a current detection error of less than ±2%, ensuring accurate and consistent vehicle status perception.

[0041] Step S4: monitoring the full-band operating electromagnetic parameters of multiple subsystems of the vehicle, identifying electromagnetic fluctuations in each frequency band, and performing multi-scale decomposition to extract electromagnetic interference disturbance spectrum lines;

[0042] In this embodiment, a high-precision electromagnetic sensor array is deployed to cover the main subsystems of the vehicle such as the engine control unit, power battery management module, on-board communication system, wiring harness nodes and sensor interfaces. The acquisition equipment covers a frequency range generally from tens of kHz to several GHz, ensuring that all electromagnetic bands from low-frequency motor interference to high-frequency radio waves are covered. The sampling rate is set above 1MHz, which can accurately capture rapidly changing electromagnetic signals. In specific operation, the system continuously collects electromagnetic parameters of each subsystem in real time, including electromagnetic field intensity, electromagnetic wave frequency and phase information. The sampling time window is set to several short periods per second (such as 10ms) to ensure sensitive capture of sudden and periodic interference. During the acquisition process, the sensor uses a clock synchronization mechanism to ensure data timing consistency to avoid misjudgment caused by time deviation. The collected full-band electromagnetic data needs to be processed by frequency band division. Common methods include bandpass filters and short-time Fourier transform (STFT) technology. Through bandpass filtering, the electromagnetic signal is divided into multiple preset frequency bands, such as low frequency (tens of kHz to hundreds of kHz), mid frequency (hundreds of kHz to several MHz), and high frequency (several MHz to several GHz). The temporal variation trend of the electromagnetic intensity within each frequency band is individually extracted, generating two-dimensional time-frequency fluctuation data. Based on this segmented data, the system automatically identifies electromagnetic fluctuations using an energy envelope detection algorithm. Specifically, within each frequency band, fluctuations are considered when the electromagnetic intensity exceeds a threshold above the baseline noise level. This threshold is typically set at the noise mean plus two standard deviations to ensure sensitivity to weak but persistent interference while avoiding misinterpretation of ambient background noise. The time points and amplitudes of the fluctuations in each frequency band are recorded, providing a detailed fluctuation trajectory for subsequent analysis. Electromagnetic signal fluctuations inherently contain multiple time-frequency characteristics that cannot be fully captured using a single scale. To this end, multi-scale decomposition methods, such as the wavelet transform, are used to analyze the fluctuation signal in each frequency band. The wavelet transform localizes the signal in both time and frequency dimensions, accurately decomposing interference features at different scales. Taking the experimental setup as an example, the Daubechies wavelet transform is applied to the collected electromagnetic waveform, decomposing it into five scale levels, each representing signal components of different frequency bandwidths. Low-scale levels correspond to high-frequency burst interference and can capture transient jumps; high-scale levels correspond to low-frequency periodic fluctuations, which facilitate analysis of the changing patterns of periodic electromagnetic interference. Through multi-scale decomposition, the system transforms complex electromagnetic fluctuation signals into a set of fine-grained spectral lines across multiple frequency bands. After completing the multi-scale decomposition, the system quantitatively extracts the energy distribution, frequency concentration, and temporal variation trends of each scale level, forming a set of electromagnetic interference disturbance spectral lines. This involves calculating the energy peak position, frequency bandwidth, and duration of each component and comparing it with the baseline noise spectrum to eliminate low-energy, ineffective components. In actual testing, common interference types such as high-frequency switching power supply noise, periodic fluctuations of wiring harness resonance, and wireless communication frequency hopping signals can be clearly identified in the spectral lines.Based on these characteristic spectral lines, the system further generates an electromagnetic interference database to provide support for subsequent interference identification and suppression strategies.

[0043] Step S5: clustering the disturbance patterns of the electromagnetic interference disturbance spectrum line set, performing dynamic electromagnetic disturbance suppression, and performing waveform distortion correction to construct an electromagnetic disturbance suppression engine;

[0044] In this embodiment, starting from the constructed electromagnetic interference disturbance spectrum line set, these disturbance data are first clustered to identify their patterns and regularities, and a dynamic suppression mechanism is developed based on the identification results to intervene and repair the affected current waveform, thereby building a set of intelligent electromagnetic disturbance suppression engines. The first stage is the clustering of disturbance patterns. In the previous steps, the electromagnetic disturbance spectrum line set already contains multi-dimensional features such as disturbance frequency band, energy distribution, duration, and modulation characteristics. The clustering process uses unsupervised learning methods (such as DBSCAN or Gaussian Mixture Model) to aggregate and classify the spectrum lines. The system divides the disturbances into several pattern categories by analyzing the Euclidean distance, temporal locality, and spectral similarity between the spectrum lines. In actual tests, three types of high-frequency disturbances (concentrated above 2.4GHz), two types of periodic disturbances (typically in the 300-600kHz range), and one type of burst interference (mostly appearing in the 50-150kHz frequency band) were clustered from more than 3,000 spectrum lines. Each type of pattern is assigned a unique identifier, which is subsequently used for interference identification and control calls. The second stage is dynamic electromagnetic disturbance suppression. The system matches the timestamps and characteristic indicators of the disturbance spectrum with the currently acquired current waveform to identify the time period in the current waveform affected by the disturbance. For example, in experiments, during periods of high motor load operation, disturbance characteristics matching the high-frequency interference spectrum were detected in the 0.85- to 1.12-second window of the phase B current waveform. The system activates the dynamic suppression module, determines the type of interference, identifies its source (such as the communication module or power switch), and selects a corresponding suppression strategy, including sliding window filtering, disturbance stripping algorithms, or adaptive modulation cancellation. The disturbance stripping algorithm is particularly suitable for periodic interference with a narrow spectral range but high energy density, filtering out disturbances in the target frequency band while preserving the main current signal. The third stage is waveform distortion correction. Because electromagnetic interference can cause nonlinear distortions in the slope, amplitude, and period of the current waveform, the corrected waveform must be reconstructed and corrected. To this end, the system uses a method based on historical waveform morphology modeling to extract an approximate "healthy" waveform template outside the disturbance segment, and uses interpolation reconstruction, edge patching, and dynamic trend alignment techniques to integrate and correct the waveform within the disturbance segment. For a current segment where the energy is significantly reduced after disturbance suppression, the system compensates its slope structure through positive and negative peak alignment and periodic trend analysis to ensure consistency with the context waveform in terms of change trend and energy distribution. Finally, all the above processing modules are integrated into an electromagnetic disturbance suppression engine. This engine has multi-channel input, multi-mode recognition, and multi-strategy suppression capabilities, and can intelligently select corresponding strategies to operate under different operating conditions. In experimental verification, the engine's detection accuracy for three types of interference waveforms exceeded 93.7%, and the root mean square error of the current waveform after interference suppression decreased by more than 65%, verifying its real-time and robustness in dynamic environments.

[0045] Step S6: Based on the temperature deviation equivalent control engine and the electromagnetic disturbance suppression engine, intelligent collaborative current detection re-optimization is performed, and post-intelligent precision optimization is performed to build a self-evolving current detection precision maintenance engine.

[0046] In this embodiment, the data outputs of the two engines are integrated with the control signals. By fusing their calibration parameters and suppression effects, a collaborative working mechanism is formed. Specifically, the system employs data fusion technology, using weighted averaging and confidence assessment methods to dynamically adjust the contribution weights of the two engines, thereby improving the overall current detection accuracy. During the experiments, the temperature range was set to -20°C to 85°C, and the electromagnetic interference intensity covered a real-world vehicle environment of 0.1 to 10 V / m. This phase focused on testing the noise suppression effect and temperature drift compensation improvement of the current signal after the coordinated adjustment of the two engines. A current sampling system with a sampling frequency of 1 MHz was used to ensure high-resolution data input. After intelligent collaborative operation, the system collects current data optimized by the control engines in real time, while also simultaneously collecting ambient temperature, electromagnetic intensity, and other auxiliary parameters. The acquisition window is typically set to 10 seconds, with a frequency range from DC to 500 kHz, ensuring the capture of subtle fluctuations and sudden anomalies. This data serves as the basis for subsequent accuracy assessment and feedback. The system uses multi-point data acquisition devices to ensure the integrity and consistency of multi-location and multi-channel signals. Data is transmitted to the central processing unit in real time and synchronized with the system's timing to ensure accurate archiving of optimization process information. Statistical analysis of the collected optimized current measurement data calculates metrics such as measurement error, noise level, and drift trends, generating a long-term accuracy assessment report. This report not only includes the average measurement error and standard deviation but also incorporates correlation analysis with temperature and electromagnetic environment changes to assess the system's robustness in complex environments. The analysis utilizes sliding window statistical techniques (e.g., a window every 60 seconds) to test the stability and continuity of the measurement data. Experimental parameters, such as 72 hours of continuous current sampling data, help capture the impact of both daily operating conditions and extreme environments on measurement performance. Based on the errors and drift trends identified in the assessment report, the system initiates a self-correcting reinforcement learning mechanism. Using reinforcement learning algorithms (such as deep Q-learning or policy gradient methods), the parameters of the temperature deviation equivalent control and electromagnetic disturbance suppression models are dynamically adjusted to optimize the calibration strategy. The specific process involves defining a feedback reward function and cyclically updating the model parameters with the goal of reducing error and improving stability. In the experimental setup, the learning cycle is measured in hours, and the online update algorithm ensures that the system continuously adapts to environmental changes, enabling learning and progress from experience. Combining reinforcement learning results with real-time feedback, a closed-loop optimization architecture is formed to achieve self-evolutionary accuracy maintenance of current detection. The engine can not only automatically identify environmental change trends and adjust detection parameters in a timely manner, but also quickly restore detection accuracy after sudden interference. To ensure stability, the engine is equipped with a redundant calibration mechanism and anomaly detection module to ensure that new errors caused by excessive parameter adjustments are avoided during the optimization process. In the experiment, after one week of continuous operation, the monitoring accuracy increased by approximately 15%-20%, and the error fluctuation was significantly reduced, demonstrating the practical effect of the self-evolution engine.

[0047] In this embodiment, refer to Figure 2 , is a flowchart of the detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include:

[0048] Based on the multi-time series displacement compensation sampling mechanism, a displacement dislocation sampling window is constructed;

[0049] The high-frequency vehicle-mounted current signal is collected under different time offsets according to the displacement staggered sampling window, and a multi-time-series sampling synchronous current waveform is constructed;

[0050] Reconstructing the time axis of the multi-time-series sampled synchronous current waveform to generate a standardized current waveform timestamp curve;

[0051] Perform microsecond offset identification on the standardized current waveform timestamp curve, calculate the response amplitude and response phase of the offset interval, and generate the microsecond offset characteristics of the current waveform;

[0052] Multi-dimensional sampling feature correlation mining is performed based on the microsecond offset characteristics of the current waveform to construct a synchronous current response matrix.

[0053] In this embodiment, a time-staggered sampling mechanism is employed to construct multiple sampling windows with time-shifted differences through sampling delays at minute intervals, enabling multi-time series parallel observation of the same current signal. Compared to traditional single-time series sampling methods, this mechanism significantly improves the ability to resolve subtle changes in high-frequency current signals. The system sampling controller configures multiple sampling trigger delay channels, each with a delay value ranging from 2 to 10 microseconds. Sampling windows are set at positions such as 0μs, 2μs, 4μs, 6μs, and 8μs, forming independent sampling channels for each time point. Each sampling window uses an analog-to-digital converter (ADC) with a high sampling rate exceeding 10kHz to acquire the target vehicle current signal point by point. In an experimental vehicle, this method was applied to engine ignition current monitoring, synchronously acquiring data across multiple vehicle ignition cycles to verify the sampling overlap and ability to detect subtle waveform drift using the staggered windows. Results show that the staggered sampling mechanism effectively compensates for signal loss during single-point sampling, providing a high-precision raw data foundation for subsequent timeline reconstruction. After establishing the offset sampling windows, the system synchronously initiates high-frequency sampling within each sampling window. Using parallel channel acquisition technology, the system acquires multiple sets of on-board current signal data at adjacent microsecond intervals. These data together form a multi-time-series sampled synchronous current waveform. During a vehicle start-up and no-load test, the system set five time-offset windows to sample current signal fluctuations above 30A. Results showed that the waveforms between adjacent sampling windows exhibited clear time-offset relationships, and the overall signal curve maintained good continuity. This provided solid data support for the subsequent construction of a timeline reconstruction model and the achievement of unified current waveform timing matching. The key to this multi-channel parallel sampling approach is maintaining the consistency and stability of time offsets. The system utilizes a high-precision local oscillator clock phase-locking mechanism within the sampling control chip to ensure that offset accuracy within 0.1μs for each channel, ensuring that even minor waveform variations are faithfully reproduced and not due to systematic errors. After completing multiple sets of sampling with different time offsets, the system enters the timeline reconstruction phase. This phase aims to integrate and align the multiple waveform segments obtained from the time-offset sampling and generate a standardized current waveform timestamp curve using a unified time reference system. During the implementation process, the system first extracts the signal amplitude and rising / falling edge change characteristics in each sampling channel, and aligns the edges using a sliding matching algorithm. It then uses the time offset value to perform data interpolation and fusion, reorganizing the original sampling points into a current curve with a continuous time axis. Taking the electric steering module of a vehicle equipped with a 12V power supply system as an example, when collecting high-frequency current signals within a period of approximately 3ms, the system successfully reconstructed a standardized waveform with a sampling time axis accuracy of 0.5μs. Compared with single-channel sampling, this multi-time series fusion waveform shows clearer pulse fluctuation details, especially in high-speed change sections, with higher signal-to-noise ratio and edge sharpness.Based on the standardized timeline waveform, the system further analyzes the dynamic excursions of the current signal at the microsecond level, extracting microsecond response features for each waveform segment. This process primarily identifies minute positional shifts in each rising / falling edge or waveform peak, calculating the signal's response amplitude and phase shift at different time intervals. During the driving process of a high-frequency motor on a vehicle, the system detected a 2μs shift in the peak position of a 40μs current waveform compared to the previous cycle, corresponding to an amplitude change of approximately 1.5A. The system recorded this as a typical microsecond excursion response. To improve feature extraction accuracy, the system uses a differential matching method to extract the difference between adjacent cycles and combines it with fast Fourier transform spectrum analysis to locate frequency components and annotate phase responses. This feature extraction method transforms the transient response behavior of the current signal into a quantifiable statistical description, facilitating subsequent feature association and model training. After acquiring a large number of microsecond excursion features, the system mines correlations between these multidimensional features to construct a current response matrix reflecting the responses at different time points and across different channels. This matrix reveals the spatiotemporal mapping between sampling behaviors and the tendency of a small perturbation to propagate synchronously across multiple channels. Taking the bus current monitoring of the high-voltage power supply unit of a new energy vehicle as an example, after 100 periodic samplings, the response matrix constructed by the system shows information such as the maximum amplitude response position, average phase difference and frequency drift intensity of each channel at different time points. The principal component analysis (PCA) and correlation clustering algorithm were further applied to classify the current response pattern, and it was found that the normal working state and two types of potential fault precursor patterns (such as contactor aging and instantaneous cable breakdown) can be distinguished. Through this matrix, the system not only improves the ability to understand the high-frequency current fluctuation behavior, but also provides basic support for achieving "non-disturbance" predictive maintenance. The matrix structure is scalable and can access more sampling dimensions, such as synchronous voltage sampling, temperature sensor information, etc., to further enhance the prediction depth of the model.

[0054] In this embodiment, refer to Figure 3 , is a flowchart of the detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include:

[0055] Call the vehicle environmental monitoring unit to collect real-time ambient temperature information in the engine compartment, wiring harness branch nodes, and sensor package cavity temperature zone;

[0056] Performing multi-point temperature distribution identification on the real-time ambient temperature information to construct a vehicle-mounted multi-point temperature distribution map;

[0057] Perform temperature trend mining on the multi-point temperature distribution map on the vehicle, render the change evolution, and generate a temperature trend change distribution map;

[0058] The temperature drift factors are mined on the temperature trend change distribution map to extract the temperature drift influencing factors.

[0059] In this embodiment, the multi-channel temperature sensing module in the onboard environmental monitoring unit (EMU) is deployed to collect real-time thermal field information from key vehicle areas. These areas include the high-temperature zone near the engine block within the engine compartment, the branch connection nodes of the main wiring harness (particularly near the electronic control module), and the interior of the current sensor package cavity. Each sampling point is equipped with a high-precision digital temperature sensor, such as an NTC thermistor or digital thermocouple, with an accuracy of ±0.5°C. The sampling period is set to 1 second. During the actual acquisition process, a hybrid SUV was used as a test platform, and monitoring was carried out under conditions of high-speed engine operation, unloaded standby mode, and traffic congestion. Experimental data showed that the temperature range within the engine compartment fluctuated from 45°C to 105°C, while the temperature at the wiring harness node fluctuated between 35°C and 70°C with changes in engine load. Due to the thermal insulation design of the sensor package cavity, the temperature changes are relatively stable, but a slow upward trend can also be observed under continuous high load, demonstrating the system's sensitive and real-time response to environmental thermal conditions. After collecting raw temperature data, the system maps the information from all sensor points onto a 3D virtual vehicle model, constructing a multi-point temperature distribution map of the vehicle environment. This map visually displays the relative temperature differences between monitoring points, the clustering characteristics of hot spots, and the spatial location of temperature anomalies. Specifically, the system uses the vehicle's CAN bus to uniformly report temperature information and incorporates a temperature spatial mapping engine to match different acquisition points to the specific coordinates of the vehicle's 3D model. A thermal imaging rendering algorithm then assigns color scales to different temperature zones, visualizing the temperature distribution map. For example, after 30 minutes of driving in urban conditions on the same test platform, the system's temperature distribution map shows a distinct red high-temperature cluster (temperature > 95°C) around the engine block, while wiring harness branch nodes appear orange-yellow, indicating medium-to-high temperatures. This map clearly reveals the thermal environment of the vehicle's electrical system, providing accurate thermal interference background support for current signal anomaly detection. Dynamic modeling is performed to analyze the changing trends of multi-point temperature data. The system uses a sliding window statistical mechanism and a time series fitting method to identify the evolution trajectory of the temperature at each point over time, and then generates a temperature trend change distribution map. During the implementation process, the system establishes a continuous temperature change sequence for each temperature sampling point, and sets three analysis windows of 5 minutes, 15 minutes, and 30 minutes to perform trend analysis on the data. In the data visualization process, the system uses color-changing dynamic heat map technology to display different temperature growth / decline rates in a color gradient. The transition from red to purple indicates a sharp increase in temperature, and the transition from blue to cyan indicates that the temperature tends to stabilize or decrease. During the experiment, under the condition of continuous idling, the temperature of a monitoring point in the engine compartment rose rapidly from 60°C to 92°C within 15 minutes, and the system displayed it as a deep red dramatic change trend.The temperature inside the sensor cavity remained stable within the same timeframe, fluctuating slightly around 45°C, indicated by a stable light blue area. This trend evolution diagram effectively reveals the intensity of thermal fluctuations in specific vehicle regions, providing a dynamic basis for fault prediction. Drift mechanisms are identified for significant temperature fluctuations within the temperature trend distribution diagram, identifying the primary thermal disturbances causing the current detection signal to destabilize. The system uses characteristic factor matching and dynamic drift curve regression analysis to extract typical temperature drift influencing factors, such as engine thermal load changes, air flow and ventilation obstruction, and heat accumulation caused by electrical insulation aging. In real-vehicle analysis, by statistically correlating outliers with vehicle status, the system identified certain temperature fluctuation peaks that were highly correlated with fan startup and shutdown, as well as battery management system loading frequency. For example, a retrospective analysis of an extreme temperature rise event revealed that the event occurred during a prolonged traffic jam, with the cooling fan disabled. This caused the local temperature to spike to 102°C, impacting the output stability of adjacent current sensors. The system ultimately constructs a matrix of temperature drift influencing factors, with each factor corresponding to its impact, triggering conditions, and duration. This provides a basis for decision-making regarding thermal compensation and intelligent alarms. This mechanism enhances the environmental adaptability of automotive current detection systems, enabling them to proactively respond to environmental changes and improving overall detection reliability.

[0060] In this embodiment, the detailed implementation steps of step S3 include:

[0061] Based on the temperature trend change distribution diagram, the synchronous current response matrix is ​​time-series correlated and matched, and the high-order drift rate of the current response with temperature fluctuation is calculated to obtain the current-temperature response correlation law;

[0062] Calculating the current deviation under temperature drift based on the current-temperature response correlation law to generate a temperature drift-current deviation value;

[0063] Perform current detection benchmark calibration analysis based on temperature drift influencing factors to generate a current drift calibration index;

[0064] Adaptive temperature deviation compensation is performed based on the current drift calibration index and the temperature drift-current deviation value to obtain a current detection value with temperature drift eliminated;

[0065] performing deep reinforcement learning on the current detection value to generate a temperature drift feedback sample;

[0066] Intelligently optimize temperature deviation compensation for temperature drift feedback samples and build a temperature deviation equivalent control engine.

[0067] In this embodiment, cross-dimensional time series data alignment is performed by calling upon previously constructed temperature trend change distribution maps and synchronized current response matrices. Specifically, the system introduces a sliding window analysis mechanism to match the temperature gradient change at each time node with the current response amplitude and phase behavior within the corresponding time period, forming a temperature-current correlation comparison table. In a test experiment, a plug-in hybrid vehicle was used as a platform. The system selected the temperature rise interval in the engine compartment area (from 55°C to 85°C, lasting approximately 18 minutes) as the key change area and extracted multiple node response sequences from the corresponding current response matrix within this time period. A comparison revealed that the current waveforms output by some sensors showed significant offsets, with their response amplitudes decreasing by an average of approximately 4.3% and the phase lag increasing by 0.7 milliseconds. Subsequently, the system used a high-order drift rate model based on time series cointegration analysis to calculate the second-order response drift rate of the current signal under temperature changes, ultimately forming a set of current-temperature response functions that describe the behavioral pattern of current signal distortion under changes in a specific temperature range. After establishing the correlation between temperature and current response, the system further uses this correlation to perform drift error retrospective analysis on each set of actual current data samples. The system introduces a historical sample error model and compares it point-by-point with the measured current signal, extracting the abnormal deviation components caused by temperature fluctuations and generating a set of "temperature drift-current deviation values." For example, during a long, low-speed vehicle driving test, as the engine compartment temperature slowly rose from 60°C to 90°C, the current value recorded by the current sensor at point B above the cylinder block exhibited a persistent average negative offset of approximately 2.8%. By comparing the sensor's standard response curve under historical normal temperature conditions, the system accurately identified the deviation points and constructed a drift table for the current temperature state. The final output of this process, the "temperature drift-current deviation value," includes not only the absolute value of the offset (e.g., the change in amperes), but also dynamic response characteristics such as the drift onset, peak offset duration, and buffer recovery rate, providing a data foundation for subsequent drift compensation. The deviation values ​​extracted in the previous step are then combined with the constructed temperature drift influencing factor matrix for analysis. The matrix contains typical interference patterns of different heat sources (such as engine heat waves, wiring harness load heat accumulation, cabin heat backflow, etc.) on the current sensor response under various working conditions. The system determines the current drift type by matching the cause, temperature zone, duration, frequency and other information of the current temperature drift, and deduce the current reference correction coefficient accordingly. In a thermal backflow condition test, the system identified that the main cause of temperature drift was not the temperature rise of the engine output, but the residual heat after the engine stopped diffusing to the sensor cavity. This type of drift characteristic is relatively slow, but lasts for a long time and has a significant impact on current accuracy. By comparing with this type of influencing factor model, the system generates a set of "current drift calibration indexes" to indicate the correction amplitude and response timing for signal offsets under current temperature conditions.Using the drift calibration index as the dominant factor and combined with the actual current deviation value, the system adaptively corrects the current data using a dynamic weighting method. Compensation not only adjusts the current value itself but also corrects the response phase error caused by drift, making the final current measurement closer to the actual load state. Measured data shows that without compensation, sensor C can have an error rate of up to 5.1% in a high-temperature environment. However, after introducing the compensation mechanism, the final output error is reduced to less than 1.2%. This improved accuracy is of great significance for on-board fault monitoring and current waveform analysis. To further enhance the generalization capability of the compensation mechanism, the system uses deep reinforcement learning to generate empirical samples of the corrected current measurement values. During this process, the system sets the objective function to "minimize the interference of temperature drift on current measurement accuracy." The system trains an intelligent agent model using a large amount of historical temperature-current data pairs. By interacting with the environment, the model learns how to select the optimal compensation strategy combination under various temperature variations. During training, the system samples data every 30 seconds and classifies all samples according to temperature gradients to construct a feedback sample pool. Each feedback sample contains key features such as temperature trajectory, drift occurrence point, and changes in current deviation before and after, for the model to continuously learn and optimize. The feedback sample pool generated in the previous step is used to train and build a "temperature deviation equivalent control engine." This engine has online adaptive adjustment capabilities. It can select the most suitable compensation strategy in real time according to the current temperature status during the actual operation of the vehicle, and evaluate and update the compensation results cycle by cycle. Under certain low-speed congestion conditions, the temperature in the engine compartment fluctuates violently, and traditional static models are difficult to respond in time. Based on the learned sample behavior, the intelligent engine quickly identifies similar overheating modes and activates the corresponding calibration strategy to effectively suppress current deviation. Ultimately, the intelligent compensation engine is integrated into the vehicle's current detection system, realizing a closed-loop optimization of "temperature perception-error correction-strategy iteration", which greatly improves the stability and accuracy of the system in complex thermal environments, and provides strong support for the intelligent health management of new energy vehicles.

[0068] In this embodiment, performing time-series correlation matching on the synchronous current response matrix based on the temperature trend change distribution diagram, calculating the high-order drift rate of the current response with temperature fluctuation, and thus obtaining the current-temperature response correlation law includes the following steps:

[0069] Calculate the local temperature rise rate and temperature fluctuation gradient based on the temperature trend change distribution diagram to obtain a local temperature rise rate diagram and a temperature fluctuation gradient diagram;

[0070] Define a frame of 5 seconds and divide the synchronous current response matrix into frames of equal length to obtain multiple current response time windows;

[0071] Perform feature extraction on the current response time window to obtain the maximum value, minimum value, mean value, fluctuation rate, FFT spectrum, and differential slope, and generate multimodal features of the current in each response time window;

[0072] Performing time stamp alignment on the temperature trend change distribution graph and the synchronous current response matrix to obtain a synchronous time relationship;

[0073] Based on the synchronous time relationship, the local temperature rise rate map, temperature fluctuation gradient map, and multi-modal characteristics of the response time window current are matched with sliding windows, and first-order drift rate modeling is performed to obtain the current-temperature sensitivity of different windows;

[0074] Performing a second-order fitting analysis on the current-temperature sensitivity to extract a nonlinear drift trend;

[0075] Differential correlation analysis is performed based on the nonlinear drift trend to obtain the current-temperature response correlation law.

[0076] In this embodiment, the system uses a temperature trend distribution map pre-collected and generated by the onboard environmental monitoring unit to extract time-series temperature data from the sensor layout location and its surrounding areas. With a sampling period of 1 second, the system calculates the local temperature change rate on the time axis, i.e., the rate of temperature rise per unit time. It also measures the gradient difference between different temperature zones within the spatial distribution range, i.e., the temperature fluctuation gradient. During a typical vehicle cold-start acceleration cycle, the engine compartment temperature rises from room temperature (approximately 27°C) to approximately 76°C within the first 10 minutes. The temperature rise rate in the cylinder area is the highest, reaching 1.5°C / second, while the central wiring harness area is only 0.3°C / second. The system performs differential analysis on each acquisition point and, based on the heat conduction path and fan startup time, annotates the temperature fluctuation trend with isothermal lines. These two maps construct a "local temperature rise rate map" and a "temperature fluctuation gradient map." These two maps provide detailed foundational data for subsequent analysis of the heat source background of current fluctuations. The synchronous current response matrix is ​​divided into equal-length frames along the time dimension. A frame length of 5 seconds ensures that the signal changes within each frame are relatively stable and representative for analysis. The sampling frequency of the current data is 1000 Hz, that is, each frame contains 5000 sampling points. The multiple time windows after division form a structured "current response time window" sequence.

[0077] In a sampling sequence during the vehicle acceleration phase, a total of 900 seconds of data was recorded, and 180 current response windows were obtained after framing. The system ensures that each frame is compatible with the corresponding temperature change frame on the time axis. The main goal of this step is to convert the continuous signal into a block time unit that is convenient for feature extraction and modeling, thereby enhancing the flexibility and contrast of the subsequent sliding matching process. The original signal in each 5-second current response time window is converted into a highly compressed and expressive multimodal feature set. The system first calculates the maximum, minimum, and mean values ​​of the current amplitude to characterize the overall power fluctuation trend; then the fluctuation rate is calculated through the standard deviation and deviation rate; then the fast Fourier transform (FFT) is performed to extract the frequency domain energy concentration feature, and the slope change of the current change is extracted through the first-order differential. In frame 42, current data from a section of wiring harness overheating exhibits a significant increase in low-frequency energy and volatility. Its FFT spectrum center of gravity shifts from 48Hz to approximately 36Hz, accompanied by a decrease in peak amplitude of approximately 7%. These detailed features are combined into a "current multimodal feature vector" for this frame, forming a high-dimensional input representation for subsequent thermal interference correlation modeling. By comparing the timestamps of the current and temperature sampling data, an accurate temporal synchronization relationship model is constructed. Because the temperature sampling frequency is typically low (e.g., 1Hz) while the current sampling frequency is high (1000Hz), the system uses interpolation to expand the temperature data time axis to ensure a one-to-one correspondence with the current frame structure. In specific implementation, the system defines a time reference point for each frame every 5 seconds and extracts temperature status data (temperature rise rate graph and temperature fluctuation gradient graph) for the same time period, ensuring high-precision temporal registration of the two data domains. This step lays a critical foundation for subsequent sliding window matching, preventing feature misinterpretation caused by signal misalignment. After completing time alignment, the system uses a sliding window approach to analyze the temperature signature for the current time period in conjunction with the current multimodal eigenvectors. Each sliding window is 10 seconds wide (i.e., two current frames). The system calculates the slope of the current parameter's response to temperature fluctuations within this time period and constructs a first-order drift rate model. Within a high-temperature fluctuation region, the system detects a significant positive slope between the current mean and the temperature gradient, indicating that the current signal is highly sensitive to thermal disturbances. The system then records the "current-temperature sensitivity" value for this window and plots it on the timeline to identify critical periods with a high risk of current error. Second-order curve fitting is used to analyze the current sensitivity curve's characteristics across different temperature gradients. The system employs sliding curvature assessment to identify turning points where sensitivity transitions from linear increase to nonlinear acceleration. For example, during high-temperature periods, the relationship between current fluctuation and temperature is no longer linear, potentially exhibiting an acceleration effect. In experiments, some sensors exhibited a significant amplification in current response after temperatures exceeded 70°C. Their sensitivity curve slope increased from an initial 0.4 to over 1.2, demonstrating a clear nonlinear drift trend.The system records this trend information to determine whether the sensor has entered a "high-sensitivity zone" or a "high-distortion zone." The system performs differential processing on the extracted nonlinear drift trend curve, analyzes its growth rate, turning point interval, response lag, and overlapping period, and forms a deep response model for current-temperature behavior. This model not only describes the extent to which temperature affects current, but also predicts dynamic properties such as when it occurs, how fast it responds, and how long the impact lasts. By analyzing the sensitivity change rate of the 90-frame current window, the system concludes that the rapid heating stage of the engine (approximately 60 to 120 seconds) is a high-sensitivity outbreak zone. The current-temperature nonlinear response is strongest during this period, requiring focused monitoring and real-time compensation. This law will directly guide the vehicle's online current detection module to implement a temperature control response adaptive mechanism.

[0078] In this embodiment, step S4 includes the following steps:

[0079] Monitor the full-band operating electromagnetic parameters of multiple subsystems of the vehicle;

[0080] Calculating the electromagnetic intensity of the electromagnetic parameters of the full-band operation, and identifying electromagnetic fluctuations in each frequency band to generate electromagnetic fluctuation detection maps for multiple frequency bands;

[0081] Performing electromagnetic mutation interference cross analysis on the electromagnetic fluctuation detection diagram based on the synchronous current response matrix to identify the current change time period of the electromagnetic mutation interference;

[0082] Extracting a time period of the electromagnetic fluctuation detection graph based on the current change time period to obtain an electromagnetic fluctuation curve of the corresponding time period;

[0083] The electromagnetic wave curve is subjected to Fourier transformation and multi-scale decomposition to extract the electromagnetic interference disturbance spectrum line set.

[0084] In this embodiment, full-band electromagnetic emissions from each major subsystem of the vehicle during actual operation are monitored to obtain its electromagnetic radiation characteristics under different operating conditions. To achieve full-band electromagnetic parameter acquisition, a multi-channel broadband antenna array and high-precision electromagnetic signal acquisition equipment were deployed on the experimental platform. The acquisition frequency range extends from 9kHz to 3GHz, encompassing electromagnetic leakage from low-frequency power electronics, radiation from high-frequency communication modules, and possible motor drive interference bands. The subsystems include but are not limited to the motor controller (MCU), battery management system (BMS), on-board charger (OBC), and intelligent driving module. In actual deployment, to reduce errors introduced by environmental interference, the test is conducted in a shielded environmental chamber, and vehicle operating condition data such as speed, load, voltage, and current are simultaneously collected. When the vehicle runs to different load and speed combinations (such as 1200rpm @ 30% load and 2500rpm @ 75% load), the electromagnetic fluctuation data of each subsystem is recorded in a time-synchronized manner to ensure that subsequent data can be correlated and analyzed. The acquisition system records electromagnetic signals at a rate of 50MS / s, and simultaneously records the channel power spectrum, time domain waveform and amplitude modulation characteristics, providing a basis for the next step of electromagnetic intensity analysis and fluctuation identification. The original electromagnetic data collected in step 1 is spectrally decoded, and a multi-window Hanning filter is used to improve the frequency domain resolution, and the signal is averaged to reduce background noise interference. Subsequently, by statistically analyzing the amplitude response in each frequency band, a frequency band intensity map is constructed, and the change trajectory of the electromagnetic intensity is calculated. In order to achieve frequency band identification, the spectrum is divided into multiple analysis segments (such as one frequency band every 100kHz), and fluctuation detection is performed on each frequency band to identify sudden peaks, periodic disturbances and modulation behaviors. Experiments revealed multiple periodic signals in the 400-500 MHz frequency range, associated with the high-frequency clock of the battery's BMS. Furthermore, strong transient fluctuations occurred in the 100-300 kHz frequency range during high-speed vehicle acceleration, consistent with the inverter switching frequency response. This frequency band information was converted into a two-dimensional electromagnetic fluctuation detection map. Each map reflects the electromagnetic fluctuation characteristics at different time points within a specific frequency band. The intensity and duration of the fluctuations were expressed as a heat map, enabling visual analysis of electromagnetic interference behavior and laying the foundation for identifying sudden interference. The current response of the vehicle's main power supply circuit was collected using a high-sensitivity Hall effect current sensor (with a bandwidth of up to 1 MHz) for real-time synchronous sampling at a sampling frequency of 100 kHz. This was processed into a current response matrix, recording the instantaneous current of multiple circuits (e.g., the drive motor circuit, auxiliary system circuit, and charging system circuit) at different time points. Subsequently, a time alignment algorithm was used to synchronize the current response matrix with the electromagnetic fluctuation detection map obtained in the previous step. The corresponding locations of the signal sudden changes on the time axis were compared and analyzed, with particular attention paid to regions of high gradient change.Using a sliding time window and cross-correlation calculations, we found that the sudden electromagnetic disturbances occurring in multiple frequency bands coincided with the current jumps during sudden acceleration of the drive motor, with particularly significant matching peaks at 750ms and 1290ms. Based on these analysis results, we extracted the time periods where the current mutations occurred, typically lasting between 50ms and 300ms. These time periods were marked as areas of concern for sudden electromagnetic interference, which served as the basis for subsequent curve extraction. To achieve high-precision extraction, the system used linear interpolation to align the data points of the detection image to the current change time window (e.g., 750ms to 800ms), ensuring the integrity and coherence of the signal curves for each frequency band. During the extraction process, a separate fluctuation curve was generated for each frequency band, with time plotted on the horizontal axis and signal strength plotted on the vertical axis. We used key frequency bands (e.g., 300kHz and 2.4GHz) as representative examples and output their electromagnetic response behavior. The extracted electromagnetic fluctuation curves clearly show abnormal variations in electromagnetic intensity corresponding to the rapid current rise, manifesting as spikes in some frequency bands and longer-lasting, high-intensity disturbances in others. Experiments revealed significant amplitude modulation signal drift within the 300–500 kHz frequency band during the motor startup acceleration phase (e.g., the 0–2 s interval). Simultaneously, the 2.4 GHz communication module experiences signal interference during the acceleration trigger period, leading to abnormal power gain in the original frequency-hopping channel. This precise fluctuation curve extraction provides fundamental data for subsequent spectrum analysis and interference source identification. In-depth frequency domain analysis of the extracted electromagnetic fluctuation curves was performed to identify key spectral features potentially related to interference behavior. First, the fast Fourier transform (FFT) algorithm was used to convert the time-domain fluctuation curves into frequency-domain spectrograms, allowing the identification of the main frequency components and modulation sidebands. To enhance the depth of analysis, wavelet packet decomposition (WPD) was introduced for multi-scale signal separation, decomposing the original curves into sub-signals with varying frequency bandwidths and time resolutions to capture details of short-term mutations and periodic disturbances. In the experiment, by performing wavelet packet decomposition on multiple fluctuation curves (using the db4 wavelet basis with four layers), the disturbance signal related to the inverter switching frequency in the low-frequency band (<500kHz) and the modulated interference spectrum caused by the working cycle of the wireless communication module in the medium- and high-frequency bands (1–3GHz) were extracted. The disturbance spectral line set was further classified and labeled, such as periodic carrier sidebands, non-periodic transient spikes, and harmonic linear structures. The resulting disturbance spectral line set can clearly indicate the electromagnetic behavior patterns of different subsystems under specific operating conditions, helping to establish interference source models, improve vehicle electromagnetic compatibility design, and provide high-value feature input for future AI-based interference prediction models.

[0085] In this embodiment, step S5 includes the following steps:

[0086] Perform disturbance pattern clustering on the electromagnetic interference disturbance spectrum set to extract high-frequency disturbance, periodic disturbance and sudden jump disturbance;

[0087] Constructing a spectrum of vehicle-borne electromagnetic interference sources based on the high-frequency disturbance, periodic disturbance and sudden jump disturbance;

[0088] Mining the interference source composition of the vehicle-borne electromagnetic interference source spectrum and classifying the types to obtain electromagnetic disturbance type characteristics;

[0089] Based on the electromagnetic disturbance type characteristics, dynamic electromagnetic disturbance suppression is performed on the standardized current waveform timestamp curve, and waveform distortion correction is performed to construct an electromagnetic disturbance suppression engine.

[0090] In this embodiment, a clustering algorithm based on time-frequency feature fusion is used to label different spectral lines. The specific processing process is divided into three categories: high-frequency disturbances, characterized by frequencies above 1 MHz, short durations, and high signal energy concentration; periodic disturbances, manifested as fluctuating signals with stable periods, commonly seen in PWM modulation secondary waves of vehicle controllers; and sudden jump disturbances, characterized by a sudden increase and rapid decrease in electromagnetic energy over a short period of time, commonly seen in relay closure, motor startup, or high-voltage discharge. A custom disturbance index system is introduced during the clustering process, including parameters such as frequency center, energy density change rate, disturbance period, duration, and sudden change slope. In actual measurements, disturbances from the engine ignition control unit were classified as high-frequency disturbances (center frequency 2.6 MHz), while periodic signals from the electronic power steering control unit were identified as periodic disturbances (stable frequency 780 kHz, period 15 ms). Meanwhile, pulse signals caused by sudden changes in current load from the AC compressor were clustered as sudden jump disturbances. After clustering disturbance patterns, the system maps each type of disturbance to its corresponding subsystem structure, creating a "spectral map" of vehicle-mounted electromagnetic interference sources. This map is constructed based on a three-layer structure: "disturbance type - source device - interference characteristics," categorizing and labeling each type of disturbance by device source. Periodic disturbances originating from vehicle-mounted DC-DC modules are identified as "energy conversion source nodes," which are further subdivided into frequency band characteristics, disturbance intensity, and synchronization duration with current anomalies. This construction process is driven by a large sample of data, using interference data collected by acquisition modules deployed on different vehicle models to form an automotive-grade interference knowledge graph. Each spectrum line is labeled with its device, disturbance type, typical waveform pattern, and frequency range, forming a visual spectral network structure diagram. The map shows that the high-voltage air-conditioning compressor is one of the main sources of high-frequency disturbances, with frequencies ranging from 1.8 to 3.0 MHz, causing short-term coupling effects on multiple current nodes. Based on the interference spectral map, the system further conducts structural mining to identify the mechanisms underlying the disturbance sources and classify them into categories. This process uses feature induction and logical deduction methods to abstractly classify interference sources. All interference sources are divided according to physical location, functional module, electrical topology, and signal characteristics, and several typical disturbance type characteristics are summarized. The system classifies interference sources into six categories: ① Switching power supply interference sources (such as DC-DC); ② High-frequency communication interference sources (such as LIN / CAN interfaces); ③ High-voltage discharge interference sources (such as ignition coils); ④ Mechanical start-stop interference sources (such as motor starting); ⑤ Sensor feedback interference; and ⑥ RF reflection coupling interference. Each interference type is assigned a typical characteristic parameter combination, including frequency range, impact duration, and impact current change rate. For example, high-voltage discharge is often accompanied by short pulses of 1–3 MHz, lasting less than 10 ms, causing simultaneous current changes in adjacent nodes. Based on the known electromagnetic disturbance type characteristics, the system performs disturbance cancellation processing on the standardized current waveform timestamp curve.The processing flow consists of three parts: dynamic suppression, waveform correction, and suppression engine construction. First, the system identifies the time period within the current waveform that falls within the disturbance window based on the matching relationship between timestamps and disturbance spectrum lines. It then invokes a pre-trained correction model for this type of disturbance. For high-frequency spike disturbances, a short-window sliding filter is used to remove instantaneous energy. For periodic disturbances, phase modulation techniques are used to reduce periodic error. For sudden jump signals, a fast slope compensation strategy is introduced to smooth the slope. After processing, the original distorted portion of the current waveform is suppressed or replaced, forming a corrected waveform with a continuous structure and stable response. The visual difference between the waveforms before and after correction is significant, especially in sudden disturbance segments, where the corrected signal is more easily interpreted. Finally, the system modularizes the above operational logic to construct an "electromagnetic disturbance suppression engine" based on disturbance type mapping. This engine supports dynamic input of new disturbance data and adaptively adjusts the suppression strategy, with online evolution and self-learning capabilities.

[0091] In this embodiment, step S6 includes the following steps:

[0092] Based on the temperature deviation equivalent control engine and electromagnetic disturbance suppression engine, intelligent collaborative current detection re-optimization is carried out, and current detection optimization process information is collected;

[0093] Perform long-term current detection accuracy calculation on the current detection optimization process information to obtain a current detection accuracy evaluation report;

[0094] Self-correction reinforcement learning is performed on the current detection accuracy evaluation report, and post-intelligent accuracy optimization is performed to build a self-evolving current detection accuracy maintenance engine.

[0095] In this embodiment, the "temperature deviation equivalent control engine" and the "electromagnetic disturbance suppression engine" are integrated into a unified intelligent collaborative processing framework, forming a dual-engine-driven current detection re-optimization mechanism. In actual operation, the system first retrieves the temperature compensation factor for the current vehicle environment from the temperature engine, including parameters such as temperature rise trends, multi-point heat source locations, and historical temperature drift behavior. Simultaneously, the electromagnetic disturbance suppression engine receives interference spectrum data from the spectrum analysis module in real time and synchronously locates key periods in the current waveform. The two engines perform collaborative optimization based on a shared timing synchronization mechanism. If both an ambient temperature increase (e.g., exceeding 65°C) and high-frequency pulse interference (e.g., a center frequency of 2.4MHz) are detected in a current waveform segment, the system combines the engine strategies to first perform a temperature calibration offset and then implement high-frequency signal reduction in the waveform distortion area. Each round of optimization is scheduled by the processing scheduler. After the optimization is completed, data such as parameter change records, quantitative indicators of waveform correction before and after correction, and interference source identification efficiency are automatically collected to form a sample of "current detection optimization process information" for subsequent analysis. After collecting a large amount of optimization process information, the system performs periodic accuracy assessment and analysis on this historical optimization data. This process first standardizes the data from each optimization period to ensure comparability across different acquisition periods for parameters such as temperature field background, electromagnetic environment intensity, and current waveform amplitude. Next, the system incorporates a multi-metric accuracy evaluation system, including short-term offset error, long-term mean square error, current waveform smoothness, and drift control success rate, forming a multi-dimensional indicator matrix. In practice, this evaluation mechanism performs a full cycle analysis every 48 hours and compares it with manually set actual current values ​​or manufacturer-provided experimental benchmark curves. During a 48-hour monitoring period, the target current of the vehicle's main power supply line should be maintained within 60A ± 0.5A. However, the original signal exhibits periodic offsets due to the combined effects of engine heat and line EMI. After dual-engine optimization, the current deviation in this section decreased from a maximum of 0.8A to 0.22A, the waveform became smooth, and response latency was reduced by 12.3%. The system will eventually summarize the above evaluation data into a current detection accuracy evaluation report, and present it graphically according to the usage scenario, including visualization results such as accuracy retention curve, abnormal interference density trend, and compensation success rate, for subsequent diagnosis and tuning. The system introduces a deep reinforcement learning network with a memory mechanism, uses the key indicators in the current detection accuracy evaluation report as environmental feedback signals (such as detection error rate, waveform recovery rate, disturbance response delay, etc.), and constructs a state-action-reward triple model to drive the engine to autonomously adjust the compensation strategy under different environmental variables. When the system identifies that the original temperature compensation strategy is insufficiently responsive to sudden slope distortion in multiple concurrent high-temperature + high-frequency interference detection cycles, the learning model will give the current strategy a lower reward value, driving the engine weight to shift to a new response structure.After repeated learning cycles, the engine adaptively updates the compensation strategy parameters to optimize its response behavior for complex interference environments. Simultaneously, based on the compensation behavior results from reinforcement learning, the system constructs an "intelligent precision optimizer" and embeds it into the pre-analysis module of the next monitoring cycle. This allows for early prediction of potential offsets and deployment of response strategies. Ultimately, this mechanism creates a "current detection accuracy maintenance engine" with self-learning, adaptive, and self-evolving capabilities, ensuring the long-term high reliability and low error rate of the detection system.

[0096] In this embodiment, a terminal device is provided, comprising a main control circuit board, an LCD screen, and a single-circuit module. The main control circuit board communicates data with the six single-circuit modules via six SPI interfaces; the single-circuit modules are powered by the main control board; the LCD screen displays the current time, Wi-Fi connection status, and the set and real-time values ​​of each vehicle-mounted circuit; the main control circuit board transmits real-time data to a host computer via a USB 2.0 interface for analysis, processing, and display; and the terminal device is used to execute the above-described vehicle online current detection method, comprising:

[0097] The synchronous sampling module is used to collect real-time vehicle current, obtain multi-time sampling synchronous current waveforms, identify microsecond offsets, and mine the correlation of multi-dimensional sampling features to construct a synchronous current response matrix.

[0098] The temperature drift mining module is used to call the vehicle environmental monitoring unit to collect the vehicle's real-time ambient temperature information, and conduct temperature trend change mining and temperature drift factor mining to extract the temperature drift influencing factors;

[0099] The deviation compensation module is used to perform current detection benchmark calibration analysis based on the temperature drift influencing factor and the synchronous current response matrix, and to perform intelligent optimization of temperature deviation compensation to build a temperature deviation equivalent control engine;

[0100] The electromagnetic interference module is used to monitor the full-band operating electromagnetic parameters of multiple subsystems of the vehicle, identify electromagnetic fluctuations band by band, and perform multi-scale decomposition to extract the electromagnetic interference disturbance spectrum line set;

[0101] The disturbance suppression module is used to cluster the disturbance patterns of the electromagnetic interference disturbance spectrum line set, perform dynamic electromagnetic disturbance suppression, and perform waveform distortion correction to build an electromagnetic disturbance suppression engine;

[0102] The intelligent precision optimization module is used to perform intelligent collaborative current detection re-optimization based on the temperature deviation equivalent control engine and the electromagnetic disturbance suppression engine, and perform post-intelligent precision optimization to build a self-evolving current detection precision maintenance engine.

[0103] The present invention is therefore intended to be illustrative and non-restrictive in all respects, with the scope of the invention being defined by the appended claims rather than the foregoing description, and all changes that come within the meaning and range of equivalents of the application documents are intended to be embraced therein.

[0104] The foregoing description is intended only to provide specific embodiments of the present invention, which are intended to enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is to be construed in the widest manner consistent with the principles and novel features disclosed herein.

Claims

1. A method for detecting online current of an automobile, characterized in that: The following steps are involved: Step S1: Collect the real-time current of the vehicle to obtain the multi-time sampling synchronous current waveform, perform microsecond offset identification and multi-dimensional sampling feature correlation mining, and construct a synchronous current response matrix; Step S2: calling the vehicle environment monitoring unit to collect the real-time ambient temperature information of the vehicle, and performing temperature trend change mining and temperature drift factor mining to extract the temperature drift influencing factors; Step S3: performing current detection benchmark calibration analysis based on the temperature drift influencing factor and the synchronous current response matrix, and performing intelligent optimization of temperature deviation compensation to build a temperature deviation equivalent control engine; Step S4: monitoring the full-band operating electromagnetic parameters of multiple subsystems of the vehicle, identifying electromagnetic fluctuations in each frequency band, and performing multi-scale decomposition to extract electromagnetic interference disturbance spectrum lines; Step S5: clustering the disturbance patterns of the electromagnetic interference disturbance spectrum line set, performing dynamic electromagnetic disturbance suppression, and performing waveform distortion correction to construct an electromagnetic disturbance suppression engine; Step S6: Based on the temperature deviation equivalent control engine and the electromagnetic disturbance suppression engine, intelligent collaborative current detection re-optimization is performed, and post-intelligent precision optimization is performed to build a self-evolving current detection precision maintenance engine.

2. The automobile online current detection method according to claim 1, characterized in that: The specific steps of step S1 are: Based on the multi-time series displacement compensation sampling mechanism, a displacement dislocation sampling window is constructed; The high-frequency vehicle-mounted current signal is collected under different time offsets according to the displacement staggered sampling window, and a multi-time-series sampling synchronous current waveform is constructed; Reconstructing the time axis of the multi-time-series sampled synchronous current waveform to generate a standardized current waveform timestamp curve; Perform microsecond offset identification on the standardized current waveform timestamp curve, calculate the response amplitude and response phase of the offset interval, and generate the microsecond offset characteristics of the current waveform; Multi-dimensional sampling feature correlation mining is performed based on the microsecond offset characteristics of the current waveform to construct a synchronous current response matrix.

3. The automobile online current detection method according to claim 1, characterized in that: The specific steps of step S2 are: Call the vehicle environmental monitoring unit to collect real-time ambient temperature information in the engine compartment, wiring harness branch nodes, and sensor package cavity temperature zone; Performing multi-point temperature distribution identification on the real-time ambient temperature information to construct a vehicle-mounted multi-point temperature distribution map; Perform temperature trend mining on the multi-point temperature distribution map on the vehicle, render the change evolution, and generate a temperature trend change distribution map; The temperature drift factors are mined on the temperature trend change distribution map to extract the temperature drift influencing factors.

4. The automobile online current detection method according to claim 1, characterized in that: The specific steps of step S3 are: Based on the temperature trend change distribution diagram, the synchronous current response matrix is ​​time-series correlated and matched, and the high-order drift rate of the current response with temperature fluctuation is calculated to obtain the current-temperature response correlation law; Calculating the current deviation under temperature drift based on the current-temperature response correlation law to generate a temperature drift-current deviation value; Perform current detection benchmark calibration analysis based on temperature drift influencing factors to generate a current drift calibration index; Adaptive temperature deviation compensation is performed based on the current drift calibration index and the temperature drift-current deviation value to obtain a current detection value with temperature drift eliminated; performing deep reinforcement learning on the current detection value to generate a temperature drift feedback sample; Intelligently optimize temperature deviation compensation for temperature drift feedback samples and build a temperature deviation equivalent control engine.

5. The automobile online current detection method according to claim 1, characterized in that: The specific steps of performing time-series correlation matching on the synchronous current response matrix based on the temperature trend change distribution diagram, calculating the high-order drift rate of the current response with temperature fluctuation, and thus obtaining the current-temperature response correlation law are as follows: Calculate the local temperature rise rate and temperature fluctuation gradient based on the temperature trend change distribution diagram to obtain a local temperature rise rate diagram and a temperature fluctuation gradient diagram; Define a frame of 5 seconds and divide the synchronous current response matrix into frames of equal length to obtain multiple current response time windows; Perform feature extraction on the current response time window to obtain the maximum value, minimum value, mean value, fluctuation rate, FFT spectrum, and differential slope, and generate multimodal features of the current in each response time window; Performing time stamp alignment on the temperature trend change distribution graph and the synchronous current response matrix to obtain a synchronous time relationship; Based on the synchronous time relationship, the local temperature rise rate map, temperature fluctuation gradient map, and the multimodal characteristics of the response time window current are matched with a sliding window, and first-order drift rate modeling is performed to obtain the current-temperature sensitivity of different windows. Performing a second-order fitting analysis on the current-temperature sensitivity to extract a nonlinear drift trend; Differential correlation analysis is performed based on the nonlinear drift trend to obtain the current-temperature response correlation law.

6. The automobile online current detection method according to claim 1, characterized in that: The specific steps of step S4 are: Monitor the full-band operating electromagnetic parameters of multiple subsystems of the vehicle; Calculating the electromagnetic intensity of the electromagnetic parameters of the full-band operation, and identifying electromagnetic fluctuations in each frequency band to generate electromagnetic fluctuation detection maps for multiple frequency bands; Performing electromagnetic mutation interference cross analysis on the electromagnetic fluctuation detection diagram based on the synchronous current response matrix to identify the current change time period of the electromagnetic mutation interference; Extracting a time period of the electromagnetic fluctuation detection graph based on the current change time period to obtain an electromagnetic fluctuation curve of the corresponding time period; The electromagnetic wave curve is subjected to Fourier transformation and multi-scale decomposition to extract the electromagnetic interference disturbance spectrum line set.

7. The automobile online current detection method according to claim 1, characterized in that: The specific steps of step S5 are: Perform disturbance pattern clustering on the electromagnetic interference disturbance spectrum set to extract high-frequency disturbance, periodic disturbance and sudden jump disturbance; Constructing a spectrum of vehicle-borne electromagnetic interference sources based on the high-frequency disturbance, periodic disturbance and sudden jump disturbance; Mining the interference source composition of the vehicle-borne electromagnetic interference source spectrum and classifying the types to obtain electromagnetic disturbance type characteristics; Based on the electromagnetic disturbance type characteristics, dynamic electromagnetic disturbance suppression is performed on the standardized current waveform timestamp curve, and waveform distortion correction is performed to construct an electromagnetic disturbance suppression engine.

8. The automobile online current detection method according to claim 1, characterized in that: The specific steps of step S6 are: Based on the temperature deviation equivalent control engine and electromagnetic disturbance suppression engine, intelligent collaborative current detection re-optimization is carried out, and current detection optimization process information is collected; Perform long-term current detection accuracy calculation on the current detection optimization process information to obtain a current detection accuracy evaluation report; Self-correction reinforcement learning is performed on the current detection accuracy evaluation report, and post-intelligent accuracy optimization is performed to build a self-evolving current detection accuracy maintenance engine.

9. A terminal device, characterized in that: The terminal device includes a main control circuit board, an LCD screen, and a single-circuit module. The main control circuit board communicates with the six single-circuit modules through six SPI interfaces. The single-circuit modules are powered by the main control board. The LCD screen displays the current time, Wi-Fi connection status, and the set and real-time values ​​of each vehicle-mounted circuit. The main control circuit board transmits real-time data to the host computer through the USB2.0 interface for analysis, processing, and display. The method for detecting an online current of an automobile according to claim 1 comprises: The synchronous sampling module is used to collect real-time vehicle current, obtain multi-time sampling synchronous current waveforms, identify microsecond offsets, and mine the correlation of multi-dimensional sampling features to construct a synchronous current response matrix. The temperature drift mining module is used to call the vehicle environmental monitoring unit to collect the vehicle's real-time ambient temperature information, and conduct temperature trend change mining and temperature drift factor mining to extract the temperature drift influencing factors; The deviation compensation module is used to perform current detection benchmark calibration analysis based on the temperature drift influencing factor and the synchronous current response matrix, and to perform intelligent optimization of temperature deviation compensation to build a temperature deviation equivalent control engine; The electromagnetic interference module is used to monitor the full-band operating electromagnetic parameters of multiple subsystems of the vehicle, identify electromagnetic fluctuations band by band, and perform multi-scale decomposition to extract the electromagnetic interference disturbance spectrum line set; The disturbance suppression module is used to cluster the disturbance patterns of the electromagnetic interference disturbance spectrum line set, perform dynamic electromagnetic disturbance suppression, and perform waveform distortion correction to build an electromagnetic disturbance suppression engine; The intelligent precision optimization module is used to perform intelligent collaborative current detection re-optimization based on the temperature deviation equivalent control engine and the electromagnetic disturbance suppression engine, and perform post-intelligent precision optimization to build a self-evolving current detection precision maintenance engine.

Images