Data line charging protection method and system based on machine learning

Abstract

The embodiment of the invention provides a data line charging protection method and system based on machine learning, and the method comprises the steps: obtaining the insulation resistance change and environment humidity fluctuation data during the operation of a data line through a miniature multi-dimensional sensor array, and carrying out the time domain denoising to generate an initial state data set; extracting coupling characteristics of the two, and determining humidity abrupt change and resistance drop rate as key influence factors; monitoring a running state according to the key factor, and generating an abnormal mark if a threshold value is exceeded; predicting an electric leakage risk trend through a time sequence smoothing technology in combination with the abnormal mark and historical data, and outputting a risk probability distribution curve; generating a threat level label according to the curve, and calling a rule set to determine a potential threat degree; a charging strategy is dynamically adjusted based on the threat degree, and optimal configuration is generated in combination with equipment compatibility. According to the invention, intelligent identification, prediction and prevention and control of the electric leakage risk of the data line are realized, the charging safety and reliability are improved, and stable operation of equipment is guaranteed.

Description

TECHNICAL FIELD

[0001] The embodiment of the application relates to the technical field of electronic equipment charging safety, and particularly relates to a data line charging protection method and system based on machine learning. BACKGROUND

[0002] In the intelligent era driven by industrial big data, data line charging protection has become a core field for ensuring the safe operation of modern electronic equipment, and its reliability is directly related to user safety and equipment life cycle management. With the widespread penetration of intelligent terminal devices in industrial scenarios and consumer markets, the use frequency of charging data lines has increased exponentially, and problems such as insulation performance degradation, physical wear accumulation, and environmental adaptability degradation have become core factors threatening charging safety. The deep application of industrial big data provides new opportunities to overcome these technical bottlenecks.

[0003] Traditional charging protection technologies mostly rely on single-point monitoring mechanisms with fixed thresholds, such as overcurrent protection and temperature warning. These methods lack dynamic perception and global analysis of data line operating states, making it difficult to adapt to complex and variable industrial application scenarios. From the perspective of industrial big data, key parameters such as data line insulation resistance degradation, mechanical wear degree, and environmental humidity changes actually constitute a multi-source heterogeneous data set with strong correlation. However, existing solutions are limited by data collection and analysis capabilities, and cannot fully utilize the value of industrial big data, making it difficult to achieve comprehensive perception and deep mining of data line operating states, resulting in hidden safety hazards such as insulation layer aging-induced leakage risks that are difficult to capture and warn in a timely manner.

[0004] The core technical challenge in this field is essentially due to the insufficient processing and analysis capabilities of industrial big data. The dynamic changes of data line insulation resistance are influenced by the combined effects of physical wear and environmental humidity, forming complex data characteristics that are nonlinear and time-varying. However, existing technologies have significant shortcomings in real-time collection of multi-source data, heterogeneous data fusion analysis, and other aspects, and cannot effectively integrate the data line operating state information contained in industrial big data. This lack of perception capability directly leads to a lack of data support for leakage risk prediction models, making it difficult to meet the industrial-level application requirements. At the same time, due to the lack of risk trend analysis based on industrial big data, the charging control strategy cannot be adaptively adjusted according to the real-time state of the data line, and cannot optimize the charging current and time parameters to actively prevent and control the leakage risk in high-humidity environments or data line high-wear working conditions. The disconnection of the whole chain from data perception, risk prediction to control strategy makes it difficult for traditional charging protection solutions to achieve precise protection in industrial-level complex scenarios.

[0005] Therefore, how to realize efficient collection and deep fusion of multi-dimensional data such as insulation resistance, wear degree, and environmental humidity based on industrial big data technology, construct a risk trend prediction model based on big data analysis, and design an adaptive charging control strategy accordingly, has become a key breakthrough for realizing precise protection of data line leakage risk, which is also the core application direction of industrial big data in the field of charging protection. SUMMARY

[0006] The embodiment of the application provides a data line charging protection method and system based on machine learning to solve the problem that the existing method cannot balance the demand for high efficiency, high precision, and high concurrency in the consumer finance scene.

[0007] In a first aspect, the embodiment of the application provides a data line charging protection method based on machine learning, comprising: S101, acquiring insulation resistance change data and environmental humidity fluctuation data of a data line during operation through a micro multi-dimensional sensor array, and performing time domain denoising processing on the original signal to generate an initial state data set; S102, extracting coupling features of insulation resistance change and environmental humidity fluctuation based on the initial state data set, and determining a key influence factor combination containing humidity mutation and resistance decline rate; S103, monitoring operating state parameters according to the key influence factor combination, and generating an abnormal state marker if the parameters exceed a preset abnormal fluctuation threshold; S104, combining the abnormal state marker with historical running track data, predicting a data line leakage risk trend through time series smoothing technology, and outputting a risk probability distribution curve; S105, generating a threat level division label according to the risk probability distribution curve, calling a preset risk judgment rule set for different labels, and determining the potential threat degree in the current scene; S106, dynamically adjusting the current regulation amplitude and the charging time limit in the charging control strategy based on the potential threat degree, and generating an optimized current output range and charging time configuration in combination with device compatibility.

[0008] Preferably, the method further comprises: S107, updating the output control instruction of the charging device in real time according to the optimized current output range and charging time configuration, judging whether the output control instruction meets the device operation limit condition in combination with the safety threshold and real-time load feedback, and executing the adjusted charging parameter if it meets the condition.

[0009] Preferably, the method further comprises: S108, monitoring the operating state parameters of the data line based on the adjusted charging parameter, setting an acquisition interval according to the feedback data frequency, analyzing the deviation of the collected data from the expected state benchmark, and generating a deviation analysis report.

[0010] As preferred, the method further comprises: S109, according to the deviation analysis report, combining the deviation trend with the environmental interference filtering, updating the collection strategy of the micro multi-dimensional sensor array, adjusting the feedback data calibration method and the collection time interval, generating a new initial state data set, and cyclically executing the state tracking and parameter adjustment process.

[0011] As preferred, the S102 specifically comprises: Based on the initial state data set, cross-comparing the insulation resistance change data and the environmental humidity fluctuation data, obtaining a feature set containing coupling characteristics by calculating the correlation coefficient of the two; Based on the time series data of the initial state data set, detecting the humidity mutation point and the corresponding resistance drop point by a sliding window algorithm, classifying and labeling specific data points that meet the humidity mutation threshold and the resistance drop rate threshold, and determining the distribution range of the key influencing factors; wherein the key influencing factors include humidity mutation and resistance drop rate; If the key factors in the distribution range do not match the preset threshold, the screening conditions are re-set according to the deviation degree, the distribution range of the key factors is re-counted after removing the abnormal data points, and the corrected key factor combination is obtained.

[0012] As preferred, the S104 specifically comprises: Combined with the abnormal state label and the historical running track data, a time-stamped continuous data stream is extracted from the stored running data, the data stream is denoised and cleaned to obtain a basic data set; The moving average method is used to process the basic data set, and when it is detected that the data fluctuation amplitude in the interval exceeds the preset threshold, the smoothing operation is performed, and the data curve after smoothing is generated. If the data curve shows that the abnormal fluctuation exceeds the preset abnormal fluctuation threshold, the probability value of the future time point of the electric leakage risk is calculated according to the amplitude, frequency and duration of the abnormal fluctuation, combined with the historical running track data, and the risk probability distribution curve is output.

[0013] As preferred, the S105 specifically comprises: According to the risk probability distribution curve, a preset level division standard is used to segment the curve to generate a threat level division label; The time-stamped running record is obtained from the data stream, and the initial time series data is generated by arranging in chronological order, and if the abnormal fluctuation exceeds the threshold, the frequency and amplitude of the abnormal occurrence are analyzed to generate risk distribution data; For different threat level labels, a preset risk judgment rule set is called to determine the potential threat degree in combination with current scene features, wherein the current scene features include environmental humidity and device load.

[0014] As preferred, the S106 specifically includes: Based on the potential threat degree, an adaptive parameter adjustment mechanism is constructed to dynamically adjust the current regulation amplitude and the charging time limit in the charging control strategy according to the threat level division; Real-time device running data is acquired to judge whether the adjusted current and charging time length are within the device rated working range, if compatible, the optimized current output range and charging time length configuration are calculated and generated based on the potential threat degree; The charging scene is simulated in combination with the adjusted parameters, the occurrence probability of the electric leakage risk under the parameters is counted, and the parameters are verified whether meeting the safety requirements by comparing with a preset safety threshold.

[0015] As preferred, the S109 specifically includes: According to the deviation analysis report, deviation trend data is extracted, high-frequency noise is separated through Fourier transform and other environmental interference filtering technologies to obtain a filtered deviation trend data set; The collection strategy of the micro multi-dimensional sensor array is clustered and classified according to data features, and if the classification result exceeds a preset strategy threshold, the collection frequency and time interval are updated; The feedback data calibration mode is adjusted, real-time data is acquired according to the new collection strategy, a new initial state data set is generated, and the state tracking of S103 and the parameter adjustment process of S106 are cyclically executed.

[0016] In a second aspect, an embodiment of the present application is a data line charging protection system based on machine learning, which includes: A data acquisition module is configured to acquire insulation resistance change data and environmental humidity fluctuation data of the data line during operation through a micro multi-dimensional sensor array, and perform time domain denoising processing on the original signal to generate an initial state data set; A coupling feature extraction module is configured to extract coupling features of insulation resistance change and environmental humidity fluctuation based on the initial state data set, and determine a key influence factor combination including humidity mutation and resistance decline rate; An abnormality monitoring module is configured to monitor running state parameters according to the key influence factor combination, and generate an abnormal state marker if the parameters exceed a preset abnormal fluctuation threshold; A risk prediction module is configured to predict a data line electric leakage risk trend through time series smoothing technology in combination with the abnormal state marker and historical running trajectory data, and output a risk probability distribution curve; The threat assessment module is configured to generate a threat level division label according to a risk probability distribution curve, and to determine a potential threat degree in the current scene by calling a preset risk judgment rule set for different labels. The charging strategy adjustment module is configured to dynamically adjust a current regulation amplitude and a charging time limit in the charging control strategy based on the potential threat degree, and to generate an optimized current output range and charging time length configuration in combination with device compatibility.

[0017] Compared with the prior art, the method and system for data line charging protection based on machine learning provided by the embodiments of the present application have the following beneficial effects: (1) The micro multi-dimensional sensor array is used to collect insulation resistance change and environmental humidity fluctuation data in real time, and the coupling characteristics of humidity mutation and resistance drop rate are extracted by combining time domain denoising and correlation analysis. The problem that the traditional fixed threshold monitoring cannot capture nonlinear correlation is solved, dynamic perception of multi-dimensional factors such as physical wear and environmental humidity is realized, and implicit risks such as insulation layer aging are accurately identified.

[0018] (2) The method combines abnormal markers and historical data, uses time series smoothing technology to predict leakage risk trends and generate probability distribution curves, dynamically adjusts current amplitude and charging time length according to threat levels, and verifies the feasibility of parameters through device compatibility. The reliability of data line charging safety protection is improved.

[0019] (3) The adaptive parameter adjustment mechanism is constructed based on the threat level, the charging parameters are executed in combination with real-time load feedback and safety threshold verification, and the decision is continuously optimized through reinforcement learning. The rigid defects of the traditional static strategy are overcome, the charging efficiency is dynamically balanced under the premise of ensuring safety, and the device adaptability and stability in the industrial complex environment are enhanced.

[0020] (4) The sensor acquisition strategy is updated through bias analysis report, environmental noise is filtered by Fourier transform, new data sets are generated in a cycle and the state tracking and parameter adjustment process is iteratively optimized. The data acquisition accuracy and risk response sensitivity are continuously evolved, the data line life cycle is prolonged, and the long-term maintenance cost is reduced. BRIEF DESCRIPTION OF DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0022] Figure 1 The flow chart of the method for data line charging protection based on machine learning provided by an embodiment of the present application is shown in the figure. Figure 2 A machine learning-based data line charging protection method flowchart provided for another embodiment of the present application; Figure 3 A structural block diagram of a machine learning-based data line charging protection system provided for an embodiment of the present application.

[0023] Figure 4 A structural schematic diagram of an electronic device provided for an embodiment of the present application. DETAILED DESCRIPTION

[0024] To make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0025] The existing data line charging protection technology mainly relies on the monitoring of fixed thresholds, such as overcurrent protection or temperature detection. These technologies are difficult to simultaneously obtain and integrate multi-dimensional dynamic data such as insulation resistance changes, wear and tear, and environmental humidity, and lack real-time perception of the data line state. On the one hand, the change of insulation resistance is jointly affected by wear and tear and environmental humidity, and the existing technology cannot accurately predict and respond to the hidden problems such as the risk of electric leakage caused thereby; on the other hand, there is a lack of dynamic analysis of risk trends, and the charging control strategy cannot be adaptively adjusted according to the data line state, breaking the chain from perception to prediction to control, making it difficult for the charging protection scheme to achieve precise protection in complex scenarios.

[0026] The present application aims to provide a machine learning-based data line charging protection method and system, which collects multi-dimensional data by deploying a micro multi-dimensional sensor array, extracts key influencing factors, constructs a real-time state tracking system, predicts the risk trend of electric leakage in combination with historical data and generates a threat level label, dynamically adjusts the charging control strategy according to different threat levels, continuously monitors the operating state and updates the sensor collection strategy, forming a closed-loop management. In this way, intelligent identification, prediction and prevention and control of data line electric leakage risk are achieved, the safety and reliability of the charging process are improved, and the stable operation of data transmission and charging equipment is ensured. The following will be described and introduced in detail through multiple embodiments with reference to the accompanying drawings.

[0027] Figure 1 is a machine learning-based data line charging protection method flowchart according to an embodiment of the present application, referring to Figure 1 , the method comprises: In step S101, the insulation resistance change data and the environmental humidity fluctuation data of the data line during operation are acquired by the micro multi-dimensional sensor array, and the original signal is subjected to time domain denoising processing to generate an initial state data set.

[0028] Specifically, by deploying a multi-dimensional array composed of micro sensors, insulation resistance change and environmental humidity fluctuation data are continuously collected during the operation of the data line, and the collected original signals are preliminarily stored to form an unprocessed signal data set. The time domain denoising method is used to process the data set, filter out the noise interference in the original signal, and determine whether the signal meets the requirements through a preset threshold. If the signal value exceeds the threshold, adjust the denoising parameters for secondary processing to obtain the filtered signal data set. Subsequently, the insulation resistance change and environmental humidity fluctuation data in the filtered data set are classified and arranged, and the two types of data are time axis aligned through data comparison to form a resistance-humidity correlation data set. Key characteristic values of resistance change rate and humidity fluctuation amplitude can be extracted from the correlation data set. The fluctuation trend of the characteristic value is determined by statistical analysis method, and if the characteristic value fluctuation exceeds the preset threshold, it is marked as an abnormal data point, and finally an initial state data set is generated.

[0029] As an optional implementation, the deployment of the micro sensor array is the key to realizing high-precision data acquisition. The micro sensor is arranged at the key nodes of the data line, such as cable joints or weak insulation layers, to form a multi-dimensional array to cover signal changes in different directions. For example, a resistance micro probe with a size of 2mm x 3mm is used to collect data in real time at a frequency of 10 times per second, ensuring the capture of transient fluctuations and ensuring data integrity.

[0030] The filtered signal data set is classified and arranged, and the insulation resistance and humidity data are stored separately. Time axis alignment can be achieved through data comparison, such as using a timestamp matching algorithm to align the two types of data to the millisecond level. Assuming that a certain section of the data line records a resistance value from 10kΩ to 8kΩ in 10 minutes, and the humidity rises from 40% to 60%, the aligned data clearly shows the trend of resistance change with humidity, which lays a foundation for subsequent feature extraction and helps to find potential fault points.

[0031] Specifically, the extraction of key feature values mainly focuses on the resistance change rate and the humidity fluctuation amplitude. For example, the resistance drop rate and the humidity change slope are extracted from the resistance-humidity correlation data set every minute, and 0.1 kΩ / min and 2% / min are set as the normal range, respectively. Statistical analysis methods such as mean-variance analysis are used to determine whether the feature value fluctuation is abnormal. If the resistance drop rate of a certain period of time reaches 0.3 kΩ / min, which exceeds the threshold, it is marked as an abnormal point. Preferably, the abnormal point triggers an alarm to prompt the operation and maintenance personnel to check the insulation state of the data line to avoid potential short circuit risk.

[0032] As an optional implementation, the final initial state data set can be used to build a health model of the data line. For example, by analyzing the distribution of abnormal points, it is found that the resistance of a certain section of data line drops significantly when the humidity exceeds 55%, which indicates that the insulation material may have aging problems. Such a data set not only supports real-time monitoring, but also provides a basis for predictive maintenance, which helps to extend the service life of the data line. In summary, the above method significantly improves the reliability and safety of the data line operation through multi-dimensional data acquisition, accurate denoising and feature analysis.

[0033] The embodiment of the application acquires the state information of the data line by deploying a micro multi-dimensional sensor array to collect insulation resistance change and environmental humidity fluctuation data during the operation of the data line, and generates an initial state data set through time domain denoising processing. High-frequency acquisition and accurate denoising ensure data integrity and accuracy, time axis alignment and key feature value extraction help to find potential fault points, abnormal point marking and alarm mechanism can early warning risk, the health model built supports real-time monitoring and predictive maintenance, significantly improves the reliability and safety of the data line operation, and prolongs its service life.

[0034] Step S102, based on the initial state data set, extracting the coupling features of insulation resistance change and environmental humidity fluctuation, determining the key influence factor combination containing humidity mutation and resistance drop rate.

[0035] Step S102 specifically includes the following steps S201-S203: S201, based on the initial state data set, cross-comparing insulation resistance change data and environmental humidity fluctuation data, and obtaining a feature set containing coupling features by calculating the correlation coefficient of the two.

[0036] In this embodiment, the correlation between insulation resistance change and environmental humidity fluctuation is quantified to generate a feature set. The correlation analysis is realized by constructing a correlation matrix, and the Pearson correlation coefficient is used to calculate the linear correlation degree r of resistance (R) and humidity (H): wherein r is a Pearson correlation coefficient, r ∈ [-1, 1], and the closer the absolute value is to 1, the stronger the correlation is. R i is the insulation resistance measurement value of the i th sampling point, H i is the environmental humidity measurement value of the i th sampling point, is the average value of the insulation resistance measurement values of all samples, is the average value of the environmental humidity measurement values of all samples, and n is the number of samples.

[0037] For example, the resistance of a certain data line decreases from 12 kΩ to 9 kΩ and the humidity increases from 30% to 65% within one day. Substituting these values into the Pearson correlation coefficient formula, r =-0.85 is obtained, indicating that the increase in humidity and the decrease in resistance are strongly negatively correlated, thus forming a feature set containing coupling characteristics.

[0038] S202, based on the time series data of the initial state data set, detecting the humidity mutation point and the corresponding resistance decrease point through a sliding window algorithm, classifying and labeling specific data points that meet the humidity mutation threshold and the resistance decrease rate threshold, and determining the distribution range of the key influencing factors; wherein the key influencing factors include humidity mutation and resistance decrease rate.

[0039] Specifically, time series comparison can be performed through a sliding window algorithm to detect the humidity mutation point and the corresponding resistance decrease point. For example, the humidity increases from 40% to 60% within 5 minutes and the resistance decreases from 10 kΩ to 8 kΩ. These points are labeled as “mutation pairs” and their distribution range is determined. Assuming that the distribution range shows that the mutation pairs are concentrated in the humidity range of 50%-60% and the resistance decrease range of 1-2 kΩ, if they do not match the preset threshold (e.g., the humidity range of 45%-55% and the resistance decrease range of 0.5-1.5 kΩ), they need to be adjusted.

[0040] S203, if the key factors in the distribution range do not match the preset threshold, the screening conditions are re-set according to the degree of deviation, the distribution range of the key factors is re-calculated after removing the abnormal data points, and the modified key factor combination is obtained.

[0041] In this embodiment, a density-based screening method is used to remove abnormal mutation pairs. For example, the humidity is 90% and the resistance decreases by 5 kΩ. The distribution range is recalculated to obtain the modified key factor combination, such as the humidity range of 48%-58% and the resistance decrease range of 0.8-1.8 kΩ. This combination is more consistent with the actual operating scenario, ensuring the reliability of the analysis.

[0042] The embodiment of the application quantifies the correlation between the insulation resistance and the humidity by the Pearson correlation coefficient, can accurately extract the coupling characteristics of the two, and labels the "mutation pair" of the humidity mutation and the resistance drop by using the sliding window algorithm, can determine the distribution range of the key factor, and when the distribution range does not match the preset threshold, a more actual key factor combination can be obtained through data screening. The scheme realizes the extraction and analysis of the coupling characteristics of the insulation resistance change and the environmental humidity fluctuation, provides a reliable basis for subsequent data line state evaluation and abnormality judgment, and improves the accuracy and reliability of the data line operation state analysis.

[0043] In step S103, the running state parameters are monitored according to the key factor combination, and if the parameters exceed the preset abnormal fluctuation threshold, an abnormal state marker is generated.

[0044] Specifically, according to the key factor combination, real-time data streams of voltage fluctuation and current stability are obtained from the running environment, the state parameters are continuously recorded, and an initial data set containing the fluctuation range is obtained. For example, the voltage and current data are collected by a high-precision sensor at a frequency of 10 times per second to form an initial data set with a time stamp, such as the voltage of a certain data line fluctuating between 220-230 volts and the current fluctuating between 1-1.2 amperes within 1 hour, to ensure data accuracy and avoid misjudgment.

[0045] Based on the initial data set, the real-time collected voltage fluctuation and current stability values are dynamically compared with the preset normal fluctuation range, the value changes are continuously monitored, and it is judged whether an abnormality occurs. The preset voltage fluctuation threshold is ±5 volts, and the current stability threshold is ±0.1 amperes. If the value exceeds the threshold, an abnormality recognition signal is generated and the abnormality is preliminarily located. For example, the signal is triggered when the voltage suddenly increases to 235 volts or the current decreases to 0.9 amperes. In a certain monitoring, the voltage increases from 225 volts to 235 volts within 5 minutes, and the current decreases from 1.1 amperes to 0.9 amperes, that is, an abnormal signal is generated.

[0046] For the abnormality recognition signal, the abnormality occurrence time point and the running state are associated and matched in chronological order, the specific distribution interval of the abnormality marker is obtained, and it is judged whether it is consistent with the fluctuation range. For example, when the voltage suddenly increases to 235 volts, the current decreases to 0.9 amperes, and the device is in high-load operation, the abnormality marker distribution interval is labeled as voltage 230-235 volts and current 0.9-1.0 amperes. If it does not match the preset range, it is prompted that the device is overloaded or the line is aged.

[0047] According to the distribution interval, the abnormality mark is classified and archived with the state parameter, the real-time tracking update record is obtained, and the matching degree with the dynamic monitoring result is determined. For example, the abnormality mark of voltage 230-235 volts and current 0.9-1.0 ampere is associated and stored with the 'high load' state, and if the matching degree of the abnormality mark appearing in the high load period is more than 90%, it can be judged that the abnormality is caused by too high load, and the time or state filtering is supported to trace back and analyze.

[0048] The embodiment of the application realizes accurate monitoring and abnormality marking of the data line running state. Through high-precision data acquisition, the monitoring basis is ensured to be accurate, the preset threshold dynamic comparison can capture the abnormality in time, the time sequence comparison can determine the abnormality distribution interval, and the classification and archiving can provide convenience for subsequent analysis. The overall technical scheme improves the timeliness and accuracy of abnormality detection, can quickly locate potential risks such as equipment overload or line aging, provides a strong guarantee for the safe and stable operation of the data line, and effectively reduces the probability of safety accidents such as electric leakage.

[0049] Step S104, in combination with the abnormal state mark and the historical running track data, the data line electric leakage risk trend is predicted through the time sequence smoothing technology, and a risk probability distribution curve is output.

[0050] As an optional implementation, step S104 specifically includes: S401, in combination with the abnormal state mark and the historical running track data, a continuous data stream with a time stamp is extracted from the stored running data, the data stream is denoised and cleaned, and a basic data set is obtained.

[0051] Specifically, first, according to the abnormal state mark and the historical running track data, a continuous data stream with a time stamp is extracted from the stored running data. For example, a certain data line records voltage and current values at a frequency of once per minute, and lasts for 24 hours, which will form a data stream containing 1440 time stamps. This way completely preserves the time sequence characteristics of the data, laying a foundation for subsequent analysis.

[0052] Then, the extracted data stream is preliminarily cleaned. This step mainly deals with the noise in the data stream, such as abnormal value points caused by sensor jitter. The median filtering method is adopted to remove the values that deviate from the normal range, so as to obtain a relatively clean basic data set.

[0053] Then, the time correspondence between the basic data set and the abnormality mark is determined. Through time stamp alignment, the abnormality mark points and the cleaned data points are matched one by one. Even if there is a slight difference between the cleaned data points and the abnormality mark points, such as the abnormality mark showing that the voltage suddenly increases to 240 volts, and the basic data set corresponding to the time point voltage is 238 volts, but the correspondence between the two can still be confirmed, providing a basis for subsequent analysis.

[0054] S402, adopt moving average method to process the basic data set, and perform smoothing operation when it is detected that the fluctuation amplitude of the data in the interval exceeds the preset threshold value, perform smoothing operation on the interval with larger fluctuation, and generate a smoothed data curve; In this embodiment, the smoothing operation is performed on the interval with large data fluctuation. For example, the voltage frequently jumps between 210 volts and 230 volts within 1 hour of a certain data, and after being processed by the moving average method, the fluctuation range is reduced to 215 volts to 225 volts, and the smoothed data curve is obtained. Whether there is a potential leakage risk trend is determined by observing the curve. If the curve shows that the voltage is continuously high and exceeds the preset threshold value, it may imply a risk.

[0055] S403, if the data curve shows abnormal fluctuation exceeding the preset abnormal fluctuation threshold value, the probability value of leakage risk at a future time point is calculated according to the amplitude, frequency and duration of the abnormal fluctuation, in combination with the historical running track data, and a risk probability distribution curve is output.

[0056] Specifically, if the smoothed data curve abnormally fluctuates beyond the preset threshold range, the amplitude (such as voltage mutation), frequency (fluctuation times per unit time) and duration characteristics of the abnormal fluctuation are extracted, these characteristics are compared with the leakage records corresponding to the same type of fluctuation in the historical running track data, the leakage occurrence rate of the characteristic matching cases is counted, the probability value at each future time point is calculated according to the time sequence, and a time-probability distribution curve is output, so as to obtain the risk probability distribution curve.

[0057] According to the probability distribution data, the running data is dynamically updated, and the potential change interval of the leakage risk is continuously tracked, so as to obtain the real-time updated risk probability distribution curve. For example, the system records that the voltage gradually rises from 220 volts to 228 volts, and the risk probability distribution curve shows that the risk probability rises from 50% to 75%, which indicates that the current state of the data line may deteriorate, and real-time tracking helps to find problems in time.

[0058] Finally, whether the smoothed data curve reflects the current state of the data line is judged in combination with the historical track and real-time data comparison. If the similar fluctuation mode in the historical track is often accompanied by leakage events, and the current curve also presents a similar trend, it can be inferred that the data line has potential problems. This multi-dimensional analysis ensures reliable judgment and provides a clear direction for subsequent maintenance.

[0059] The application can obtain accurate basic data sets by combining abnormal state markers with historical running track data, extracting and cleaning time-stamped continuous data streams. Time series smoothing processing is performed on the data to generate a smooth curve to determine the potential leakage risk trend. When the curve abnormal fluctuation exceeds the threshold, the risk is quantified and the real-time updated risk probability distribution curve is output to realize dynamic prediction of data line leakage risk. The scheme improves the accuracy and real-time performance of data line leakage risk prediction, provides a reliable basis for data line maintenance, and ensures safe operation.

[0060] In step S105, threat level division labels are generated according to the risk probability distribution curve, and preset risk determination rule sets are called for different labels to determine the potential threat degree in the current scene.

[0061] As an optional implementation, S105 specifically includes: S501, according to the risk probability distribution curve, the curve is segmented by using a preset level division standard to generate threat level division labels.

[0062] According to the shape of the risk probability distribution curve, the curve is segmented by using a preset level division standard (for example, low, medium and high risk intervals). For example, if the risk probability value of a certain segment of the curve exceeds 80%, it is marked as a "high risk" level, and the corresponding threat level division label is generated to realize quantitative classification of risk.

[0063] S502, time-stamped running records are obtained from the data stream, and initial time series data is generated by arranging them in chronological order. If abnormal fluctuations exceeding the threshold are detected, the frequency and amplitude of the abnormality are analyzed to generate risk distribution data.

[0064] The time-stamped running records (such as voltage, current and other parameters) are extracted from the data stream, and the initial time series data is generated by arranging them in chronological order. When the time-stamped continuous running records are obtained from the data stream, the running state of a data line is recorded at a frequency of every 5 minutes, covering key parameters such as voltage and current, with a time span of 48 hours, forming a continuous data set containing 576 time stamps. This ensures the integrity and time sequence of the data, laying a foundation for subsequent processing.

[0065] When arranging the running records, the above data set is rearranged in chronological order to ensure the accuracy of the data point position corresponding to each time stamp. If some time stamp data is missing during the rearrangement process, the missing values are estimated by linear interpolation, such as filling in the missing voltage value at a certain time point by taking the average of the voltage values at the previous and next time points to generate the initial time series data and maintain the continuity of the data.

[0066] If the data anomaly fluctuation is detected to exceed the preset threshold (such as the voltage suddenly increases to 235 volts), the frequency and amplitude of the abnormality are analyzed to generate risk probability distribution data. For example, 10 times of voltage exceeding the threshold within 24 hours, 5 times concentrated within 1 hour, can be quantified as a higher risk probability and form a curve.

[0067] S503, for different threat level labels, call the preset risk judgment rule set, combine the current scene characteristics, determine the potential threat degree; wherein, the current scene characteristics include environmental humidity and device load.

[0068] For different threat level labels, call the preset risk judgment rule set, and combine the current scene characteristics to comprehensively judge the potential threat degree. For example, if the threat level is "high risk" and the environmental humidity is high, the rule set may determine the threat degree as "immediate treatment" is required, realizing the linkage of risk assessment and scene characteristics.

[0069] As an optional implementation, when the curve representation of the risk probability distribution data is segmented, the risk probability is divided into low, medium and high levels according to the preset level division standard. If the curve shows that the probability value of a certain segment is more than 80%, it is marked as a high risk segment and the corresponding threat level label is generated, which converts complex data into intuitive risk levels.

[0070] When judging whether the threat level label matches the current scene characteristics, the environment where the data line is located is analyzed. If the data line is running in a high temperature and high humidity environment, and the threat level label is high risk, it is determined to be consistent with the scene characteristics, and the environmental factors may exacerbate the risk.

[0071] When the threat level label is classified by calling the judgment rule set, the corresponding rule is selected according to the scene characteristics. If the rule set specifies that the high risk label in high temperature environment needs to be paid attention to first, the system will classify this label as a potential threat degree that needs to be paid attention to immediately, realizing the scene-based customization processing.

[0072] As an optional implementation, the traditional risk threshold setting is often static and cannot be dynamically adjusted according to environmental changes or device status. This scheme introduces a dynamic risk threshold adjustment mechanism, which dynamically adjusts the segment threshold of the risk probability distribution curve according to historical data, environmental humidity changes and device load conditions, so that the risk assessment is more flexible and accurate. The expression of the dynamic risk threshold adjustment mechanism is: T dynamic (t)=μ hist +β·σ hist ·f(t) In the formula, T dynamic (t) is the dynamic risk threshold, which is a function of time t, indicating the risk threshold calculated at a specific time point according to the dynamic adjustment mechanism. μhist is the average value of historical risk probability, and β is an adjustment coefficient used to adjust the sensitivity or amplitude of the dynamic threshold. By adjusting β, the degree of fluctuation of the dynamic threshold relative to the historical standard deviation can be controlled. hist is the standard deviation of historical risk probability. f(t) is a dimensionless normalization adjustment factor whose value is calculated by scaling the time or environmental parameters with a reference value.

[0073] In this embodiment, the dynamic threshold T dynamic (t) can be dynamically adjusted according to the average value and standard deviation of historical risk probability, as well as the current time or environmental conditions, making the risk assessment more flexible and accurate.

[0074] As an optional implementation, the present application uses machine learning algorithms to train historical risk data and build a threat level prediction model to predict the threat level in future time periods and take preventive measures in advance. Historical risk data and their corresponding threat levels are collected as a training set, and machine learning algorithms are used for training. After training is completed, real-time risk data is input into the model, and the predicted threat level is output. In this embodiment, a random forest algorithm can be used, and the prediction formula is: wherein, is the predicted threat level, R features is the input risk feature vector, and RandomForest is the trained random forest model.

[0075] This embodiment uses machine learning algorithms to train historical risk data and build a threat level prediction model, which can predict future threat levels in advance to take preventive measures in advance, improve the forward-looking prevention and control capability of data line electric shock risk, and improve charging safety and equipment operation reliability.

[0076] The present application embodiment generates threat level labels by segmenting the risk probability distribution curve, realizes risk quantization and grading, calls a preset risk judgment rule set, and combines with environmental humidity, device load and other scene characteristics to realize risk assessment and scene linkage, and accurately determines the potential threat level. This scheme improves the intuitiveness, accuracy and scene adaptability of risk assessment, and provides strong support for data line risk prevention and control.

[0077] Step S106, dynamically adjusting the current regulation amplitude and charging time limit in the charging control strategy based on the potential threat level, and generating an optimized current output range and charging time configuration combined with device compatibility.

[0078] In this embodiment, S106 specifically includes: S601, based on the potential threat degree, an adaptive parameter adjustment mechanism is constructed, and the current regulation amplitude and the charging time limit in the charging control strategy are dynamically adjusted according to the threat level division.

[0079] Specifically, based on the determined potential threat degree, an adaptive parameter adjustment mechanism is constructed. The key parameters in the charging control strategy can be dynamically adjusted according to the threat level division. The adjustment mode of the current regulation amplitude and the charging time limit will be different for different threat levels.

[0080] For example, when the threat level is high, the current regulation amplitude may be reduced and the charging time limit may be shortened to reduce the risk; when the threat level is low, the current regulation amplitude may be appropriately increased and the charging time limit may be extended to improve the charging efficiency.

[0081] S602, real-time device running data is obtained, and it is judged whether the adjusted current and the charging time are within the rated working range of the device. If compatible, the optimized current output range and the charging time configuration are calculated and generated based on the potential threat degree.

[0082] Specifically, real-time device running data containing scene characteristics are obtained, which reflect the current actual running state of the device. Then, through device compatibility detection, the parameters and performance of the device are evaluated to determine whether the device can withstand the current regulation amplitude and the charging time limit adjusted according to the threat level. If the device is compatible, based on the potential threat degree, the optimized configuration is generated according to the preset grading calculation rule: when the threat is low, the upper limit of the current output is moderately relaxed and the charging time is extended within the rated range; when the threat is medium, the current output is controlled in the middle of the rated range, and the charging time is reduced according to the reference value; when the threat is high, the current output is strictly limited to the lower limit of the rated range, and the charging time is greatly shortened. Through this quantitative calculation method combined with the threat degree, the optimized current output range and the charging time configuration are obtained.

[0083] S603, the charging scene is simulated combined with the adjusted parameters, the occurrence probability of the electric leakage risk under the parameters is counted, and the parameters are verified whether they meet the safety requirements by comparing with the preset safety threshold.

[0084] In this embodiment, the adjusted parameters are quantitatively verified, the risk probability possibly caused by the adjusted parameters is analyzed, and it is ensured that the optimized current output range and the charging time configuration meet the safety threshold requirements. Only when the configuration is proved to be within the safety range by the quantitative verification, the configuration can be finally determined as the applicable charging control strategy parameter.

[0085] For example, when acquiring real-time device operation data, taking a charging pile as an example, the device records parameters such as current, voltage, and temperature every minute, generating a 24-hour data set containing 1440 timestamps. The data includes a current value range of 20 to 50 amperes and a temperature range of 25 to 45 degrees Celsius. This high-frequency collection method ensures the time sequence integrity of the data, laying a foundation for subsequent analysis.

[0086] In a possible implementation, when arranging data, a device state sequence is generated in timestamp order. If it is found that current data is missing at a certain time point, it is completed by averaging the current values of the previous and next two minutes. For example, if the missing point is 30 amperes and 32 amperes respectively, it is completed as 31 amperes. This completion method maintains the continuity of the data.

[0087] Specifically, when screening the device state sequence by a preset threshold, the normal range of current is set to 25 to 45 amperes. If the current is detected to suddenly drop to 15 amperes in a certain time period, it is marked as an abnormal state, and the device compatibility identifier is generated as “unstable”. This screening method quickly locates potential problems.

[0088] For example, when extracting threat level division criteria, it is assumed that the rule set stipulates that a sudden drop in current may cause device overload and needs to be marked as medium risk. If the device compatibility identifier is “unstable” and the scene feature is a high-temperature environment, it matches the medium risk standard. This matching method improves the pertinence of judgment in combination with the scene.

[0089] In a possible implementation, when processing the threat level division criteria, the upper limit of the current output is preliminarily adjusted to 40 amperes, and the charging time is shortened by 10%. This adjustment parameter set provides a basis for subsequent optimization.

[0090] Specifically, when quantifying the preliminary adjustment parameter set, the frequency of current sudden drop is analyzed in combination with risk probability. It is assumed that 5 sudden drops occur within 24 hours, concentrated in the high-temperature period, and then probability distribution data is generated, showing that the medium risk probability is 60%. Processing the probability distribution data optimizes the current output range to 30 to 40 amperes and the charging time to 2 hours. This optimization configuration balances efficiency and safety.

[0091] For example, when verifying the optimized configuration, the charging process after adjustment is monitored in real time. If the current is stable at 35 amperes and the temperature does not exceed 40 degrees Celsius, the verification result is consistent with the “unstable” identifier, indicating that the configuration is effective. Subsequently, the configuration is applied to the charging control process to determine the final charging strategy as current 35 amperes and charging time 2 hours. This verification and application method ensures the reliability of the strategy.

[0092] The embodiment of the present application can dynamically adjust the charging parameters according to the threat level by constructing an adaptive parameter adjustment mechanism, thereby improving the pertinence of the charging strategy. In combination with the device compatibility detection and risk probability quantification verification, it is ensured that the adjusted parameters are safe and feasible, and device failure or safety risks caused by improper parameters are avoided.

[0093] As an optional implementation, in step S106, a parameter optimization algorithm based on reinforcement learning can be used, and specifically, a reinforcement learning algorithm is used to dynamically optimize the charging parameter configuration according to the real-time feedback charging process data and device state. Reinforcement learning can learn the optimal strategy through trial and error, thereby improving the charging efficiency and safety.

[0094] The charging process is modeled as a Markov decision process (MDP), the state space is defined to include threat levels and device states, the action space is defined to include current adjustment amplitudes and charging time limits, and the reward function includes charging efficiency and safety.

[0095] The reinforcement learning algorithm is used to learn the optimal strategy, and the optimal action is selected in each state to maximize the long-term reward. In this embodiment, the Q-Learning algorithm is used to learn the optimal strategy by updating the Q value table. The Q value represents the expected long-term reward of taking a specific action in a specific state. The update formula is: In the formula, Q' (s, a) represents the expected long-term reward of taking action a in state s after this update; Q (s, a) represents the historical Q value before the update; s is the current state, a is the current action, is the immediate reward, s' is the next state, and a' is the next action, is the learning rate, is the discount factor, which measures the degree of influence of future rewards on the current Q value update.

[0096] In this embodiment, the charging process is modeled as a Markov decision process, and reinforcement learning (such as Q-Learning) is used to dynamically optimize the parameters. According to the real-time data and device state, the optimal action is selected in the state space to balance efficiency and safety, and the strategy is learned through trial and error iteration, so that the charging parameters adapt to variable scenarios, improve the system adaptability and control accuracy, and ensure the charging stability.

[0097] Figure 2 The flowchart of the data line charging protection method based on machine learning provided by another embodiment of the present application is shown in Figure 2 On the basis of the above embodiment, the data line charging protection method based on machine learning further comprises: At step S107, the output control instruction of the charging device is updated in real time according to the optimized current output range and charging duration configuration, and it is judged whether the output control instruction meets the device operation limit condition in combination with the safety threshold and real-time load feedback. If yes, the adjusted charging parameter is executed.

[0098] Specifically, device operation data containing scene characteristics are acquired, device state sequences are generated in time sequence, compatibility detection is performed, it is judged whether the preset operation limit condition is met, and device state identifiers are obtained. Real-time load data are acquired according to the identifiers, risk probability distribution data are generated by processing in combination with the load data and scene characteristics, and a safety threshold range is determined. If the threshold range is met, the current output and charging duration are quantitatively processed, the optimized configuration is obtained by processing the adjusted parameters, real-time output control instructions are generated based on this, it is verified whether the instructions meet the operation limit condition, and the final charging control instruction is determined.

[0099] For example, when acquiring electric vehicle charging pile operation data, the device records current, voltage, environmental humidity and other parameters every 5 minutes, generates a data set containing 288 time stamps within 24 hours, the current value ranges from 15 amperes to 60 amperes, and the humidity range is 30% to 80%, ensuring the time sequence integrity of the data. The device state sequences are generated in time stamp order. If the humidity data is missing at a certain time point, it is completed by the average value of the humidity values of the previous and next 5 minutes, and the data continuity is maintained.

[0100] When performing compatibility analysis on the device state sequences, the operation limit condition is set as a current range of 20 amperes to 55 amperes and a humidity lower than 75%. If the humidity rises to 78% in a certain period, a “restricted” device state identifier is generated to quickly identify environmental abnormalities. Real-time load data are acquired, for example, the charging pile power suddenly increases from 10 kilowatts to 15 kilowatts. Risk probability is calculated in combination with the load and humidity characteristics, the frequency of load sudden increase under high humidity is analyzed, for example, 4 times of sudden increase within 24 hours and concentrated in the period when the humidity is greater than 70%; risk probability distribution data (high risk probability 50%) are generated, and compared with the safety threshold (<40%). When the risk probability exceeds the threshold, the parameter optimization process is triggered.

[0101] If the safety threshold is not met, the current output upper limit is initially adjusted to 50 amperes and the charging duration is shortened by 15% through parameter adjustment. In combination with the load sudden increase frequency, the current is optimized to 25 amperes to 45 amperes and the charging duration is 1.5 hours. Real-time control instructions (for example, current 40 amperes and duration 1.5 hours) are generated accordingly. When verified, if the current is stable at 40 amperes and the humidity is reduced to below 70%, it is confirmed that the operation limit condition is met, and the instructions are finally executed.

[0102] The application ensures that the charging parameters are safe and feasible by updating the charging device output control instruction in real time, combining the safety threshold and load feedback to determine whether the operating conditions are met. The scheme realizes dynamic optimization and adjustment of the charging parameters, ensures the safety of the charging process, and improves the reliability and adaptability of the charging system.

[0103] On the basis of the above-mentioned embodiments, the data line charging protection method based on machine learning further comprises: Step S108, based on the adjusted charging parameters, monitoring the operating state parameters of the data line, setting the collection interval according to the feedback data frequency, analyzing the deviation of the collected data from the expected state reference, and generating a deviation analysis report.

[0104] Specifically, the operating state data including voltage fluctuation and current stability are obtained through data line monitoring, the operating state data are arranged in time sequence, the state sequence arranged in time sequence is generated, the state sequence is processed to determine whether the preset stability threshold is met, and the state stability identifier is obtained. According to the state stability identifier, the collection interval of the feedback data is quantitatively adjusted, the real-time operating state data are obtained according to the adjusted collection interval, the real-time operating state data are preprocessed, and the cleaned operating state data set is obtained. The cleaned operating state data set is compared with the preset state reference, the deviation between the two is calculated by using the mean square deviation calculation method, and the deviation analysis data is obtained. If the deviation analysis data exceeds the preset deviation threshold, the charging parameters are reconfigured, and the final charging parameter adjustment instruction is determined.

[0105] For example, in the electric vehicle charging pile operation scenario, voltage fluctuation and current stability data are collected every minute, 1440 data sets are formed at 24 time points, the voltage fluctuation range is 200-240 volts, and the current fluctuation amplitude is less than 2 amperes, fully reflecting the dynamic changes of the equipment. The data are arranged according to the time stamp, if the voltage data at a certain time point is missing, it is completed by the average value of the previous and next two minutes, and the sequence integrity is maintained.

[0106] In the stability detection process, the voltage fluctuation threshold is set to 5 volts and the current fluctuation threshold is set to 1 ampere. If the voltage fluctuation of a certain segment reaches 6 volts, it is determined to be unstable, and a "fluctuation abnormality" state stability identifier is generated. Accordingly, the collection interval is shortened from 1 minute to 30 seconds, and the equipment state is monitored in real time. After obtaining the data according to the new interval, abnormal values such as voltage surge to 300 volts are removed to obtain a reliable data set.

[0107] By comparing the cleaned data set with the preset reference, if the average voltage of a certain segment deviates from the reference by 10 volts, the deviation analysis data is obtained by mean square deviation calculation. If the deviation exceeds the preset threshold of 5 volts, the original charging current of 50 amperes is adjusted to 40 amperes, and the charging time of 2 hours is extended to 2.5 hours, an adjustment instruction is generated, and the safe and stable operation of the equipment is ensured.

[0108] The application can realize dynamic optimization of charging parameters and guarantee stable operation of the equipment by monitoring data line operation state parameters, dynamically adjusting collection intervals and analyzing data deviation, and discovering equipment operation abnormalities in time.

[0109] On the basis of the above-mentioned embodiments, the data line charging protection method based on machine learning further comprises: In step S109, the collection strategy of the micro multi-dimensional sensor array is updated according to the deviation analysis report, combined with deviation trend and environmental interference filtering, the feedback data calibration mode and the collection time interval are adjusted, a new initial state data set is generated, and the state tracking and parameter adjustment process is executed cyclically.

[0110] In this embodiment, step S109 specifically comprises: In S901, deviation trend data is extracted according to the deviation analysis report, high-frequency noise is separated by Fourier transform and other environmental interference filtering technologies, and filtered deviation trend data set is obtained.

[0111] In this embodiment, deviation trend data is extracted from the deviation analysis report, high-frequency noise caused by environmental interference is separated by Fourier transform and other technologies, for example, voltage jitter caused by electromagnetic interference, filtered deviation trend data set is obtained, and the data is ensured to be close to the real operation state of the equipment.

[0112] In S902, the collection strategy of the micro multi-dimensional sensor array is clustered and classified according to data characteristics, and if the classification result exceeds the preset strategy threshold, the collection frequency and time interval are updated. The collection strategy of the micro multi-dimensional sensor array is classified by cluster analysis, for example, classified into high-frequency, medium-frequency and low-frequency collection modes. If the classification result exceeds the preset threshold (such as high-frequency collection response delay), the collection frequency and time interval are updated (for example, the collection frequency is shortened from once every minute to once every 30 seconds), and the real-time performance of data collection is improved.

[0113] In S903, the feedback data calibration mode is adjusted, real-time data is obtained according to the new collection strategy, a new initial state data set is generated, and the state tracking of S103 and the parameter adjustment of S106 are executed cyclically.

[0114] The feedback data calibration mode is adjusted, for example, the calibration period can be shortened. Real-time data is obtained according to the new strategy, a new initial state data set is generated. The state tracking of S103 and the parameter adjustment of S106 are executed cyclically, forming a closed loop of "data collection-deviation analysis-strategy update-parameter adjustment", and continuously improving the adaptability and safety of the charging system.

[0115] In this embodiment, the deviation trend data is obtained from the deviation analysis report, the data filtering method is used for pre-processing the environmental interference, the high-frequency noise is separated by the Fourier transform method, and the filtered deviation trend data set is obtained. According to the data set, the clustering analysis method is used to classify the collection strategy of the sensor array, if the classification result exceeds the preset strategy threshold, the collection frequency is updated, and a new collection strategy configuration is formed. With the new configuration, the calibration mode of the feedback data is adjusted, the real-time data is obtained according to the adjusted time interval, and the initial state data set is obtained. Based on the data set, the state tracking is performed through time series analysis, the parameter adjustment instruction is generated by comparing the deviation calculation with the preset reference, and the final dynamic update configuration is determined.

[0116] For example, in the operation monitoring of the electric vehicle charging pile, the voltage deviation trend of a charging pile running 24 hours shows that there is an 8-volt deviation peak every 4 hours, which exceeds the preset reference of 3 volts. The data filtering method separates the high-frequency noise such as external electromagnetic interference by Fourier transform to obtain the real deviation data. When the clustering analysis method is used to classify the collection strategy of the charging pile equipped with multiple sensors, if the high-frequency collection group response time is too long and exceeds the threshold, the collection frequency is dynamically adjusted from 1 time / minute to 1 time / 30 seconds. Under the new configuration, the original 1-hour / period calibration mode is shortened to 30 minutes / period, and real-time data of voltage 210-230 volts and current about 45 amperes are collected. Through time series analysis, it is tracked that the voltage is continuously high to 235 volts, which is 5 volts higher than the preset reference of 220 volts, and an adjustment instruction is generated to reduce the charging current from 50 amperes to 42 amperes, forming a dynamic update configuration to ensure stable operation of the equipment.

[0117] The embodiment of the present application dynamically updates the sensor collection strategy and the calibration mode through deviation analysis and environmental interference filtering, generates a new data set and optimizes the charging parameters in a cycle, improves the data collection accuracy and system adaptability, forms a closed-loop management, realizes accurate monitoring of the data line state and dynamic adjustment of the charging parameters, and ensures charging safety and stable operation of the equipment.

[0118] Figure 3 The structure block diagram of the data line charging protection system based on machine learning provided by the present application is shown in Figure 3 The data line charging protection system based on machine learning comprises: The data acquisition module 301 is used to acquire the insulation resistance change data and environmental humidity fluctuation data of the data line during operation through a micro multi-dimensional sensor array, and to perform time domain denoising processing on the original signal to generate an initial state data set; The coupling feature extraction module 302 is used to extract the coupling features of insulation resistance change and environmental humidity fluctuation based on the initial state data set, and to determine the key influence factor combination containing humidity mutation and resistance decline rate; The anomaly monitoring module 303 is configured to monitor the operating state parameters according to the key influence factor combination, and generate an anomaly state label if the parameters exceed a preset anomaly fluctuation threshold; The risk prediction module 304 is configured to combine the anomaly state label with historical operation trajectory data, predict a data line electric leakage risk trend through a time series smoothing technique, and output a risk probability distribution curve; The threat assessment module 305 is configured to generate a threat level division label according to the risk probability distribution curve, call a preset risk judgment rule set for different labels, and determine a potential threat degree in a current scenario; The charging strategy adjustment module 306 is configured to dynamically adjust a current regulation amplitude and a charging time limit in a charging control strategy based on the potential threat degree, and generate an optimized current output range and charging time length configuration in combination with device compatibility.

[0119] The application provides a data line charging protection system based on machine learning, which is used for executing the machine learning-based data line charging protection method provided in the above embodiments.

[0120] Based on the same idea, the application also provides an electronic device, as shown in the figure. Figure 4 The electronic device can include a processor 410, a communications interface 420, a memory 430, and a communications bus 430, wherein the processor 410, the communications interface 420, and the memory 430 can communicate with each other through the communications bus 440. The processor 410 can call the logic instructions in the memory 430 to execute the steps of the machine learning-based data line charging protection method described in the above embodiments.

[0121] Moreover, the logic instructions in the memory 430 described above can be implemented in the form of software functional units and sold or used as independent products, and can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the parts that make contributions to the prior art or parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0122] Based on the same concept, the embodiments of the present application also provide a non-transitory computer readable storage medium, which stores a computer program containing at least one code, the at least one code being executable by a host device to control the host device to implement the steps of the machine learning-based data line charging protection method as described in the above embodiments.

[0123] Based on the same technical concept, the embodiments of the present application also provide a computer program, which, when executed by a host device, is used to implement the above method embodiments.

[0124] The various embodiments of the present application can be combined arbitrarily to achieve different technical effects.

[0125] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by a computer program instructing relevant hardware, and the program can be stored in a computer readable storage medium. The program, when executed, can include the processes of the above-mentioned method embodiments. The aforementioned storage medium includes: ROM or random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0126] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements for some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A machine learning based data line charging protection method, characterized in that, The method comprises the following steps: S101, obtaining insulation resistance change data and environmental humidity fluctuation data of a data line in operation through a micro multi-dimensional sensor array, and performing time domain denoising processing on the original signal to generate an initial state data set; S102, based on the initial state data set, extracting the coupling characteristics of insulation resistance change and environmental humidity fluctuation, and determining the key factor combination containing humidity mutation and resistance drop rate; S103, monitoring the operating state parameters according to the key factor combination, and generating an abnormal state label if the parameters exceed the preset abnormal fluctuation threshold; S104, combining the abnormal state label and the historical operation trajectory data, predicting the data line leakage risk trend through time series smoothing technology, and outputting a risk probability distribution curve; S105, generating a threat level division label according to the risk probability distribution curve, calling a preset risk judgment rule set for different labels, and determining the potential threat degree in the current scene; S106, based on the potential threat degree, dynamically adjusting the current regulation amplitude and the charging time limit in the charging control strategy, and combining the device compatibility to generate an optimized current output range and charging time configuration. 2.The machine learning based data line charging protection method of claim 1, wherein, Further comprising: S107, updating the output control instruction of the charging device in real time according to the optimized current output range and charging time configuration, combining the safety threshold and the real-time load feedback to judge whether the output control instruction meets the device operation limit condition, and executing the adjusted charging parameters if it meets the condition. 3.The machine learning based data line charging protection method of claim 2, wherein, Further comprising: S108, based on the adjusted charging parameters, monitoring the operating state parameters of the data line, setting the collection interval according to the feedback data frequency, analyzing the deviation of the collected data from the expected state benchmark, and generating a deviation analysis report.

4. The machine learning based data line charging protection method of claim 3, wherein, Further comprising: S109, according to the deviation analysis report, combining the deviation trend and environmental interference filtering, updating the collection strategy of the micro multi-dimensional sensor array, adjusting the feedback data calibration method and the collection time interval, generating a new initial state data set, and executing the state tracking and parameter adjustment process cyclically. 5.The machine learning based data line charging protection method of claim 1, wherein, The S102 specifically comprises: Based on the initial state data set, cross-comparing the insulation resistance change data and the environmental humidity fluctuation data, and obtaining a feature set containing coupling characteristics by calculating the correlation coefficient of the two; Based on the time series data of the initial state data set, detecting the humidity mutation point and the corresponding resistance drop point through a sliding window algorithm, classifying and labeling specific data points that meet the humidity mutation threshold and the resistance drop rate threshold, and determining the distribution range of the key factors; wherein the key factors include humidity mutation and resistance drop rate; If the key factors in the distribution range do not match the preset threshold, the screening conditions are re-set according to the deviation degree, the distribution interval of the key factors is re-counted after removing the abnormal data points, and the corrected key factor combination is obtained. 6.The machine learning based data line charging protection method of claim 1, wherein, The S104 specifically comprises: Combining the abnormal state label and the historical operation trajectory data, extracting a time-stamped continuous data stream from the stored operation data, denoising and cleaning the data stream to obtain a basic data set; The moving average method is used to process the basic data set, and when it is detected that the fluctuation amplitude of the data in the interval exceeds the preset threshold, the smoothing operation is performed, the interval with large fluctuation is smoothed, and the smoothed data curve is generated. If the data curve shows abnormal fluctuation exceeding the preset abnormal fluctuation threshold, the probability value of the leakage risk at the future time point is calculated according to the amplitude, frequency and duration of the abnormal fluctuation, combined with the historical running track data, and the risk probability distribution curve is output.

7. The machine learning based data line charging protection method of claim 1, wherein, The S105 specifically includes: According to the risk probability distribution curve, a preset grade division standard is used to segment the curve, and a threat level division label is generated; From the data stream, the time-stamped running record is obtained, and the initial time sequence data is generated in time sequence. If it is detected that the abnormal fluctuation exceeds the threshold, the frequency and amplitude of the abnormality are analyzed, and the risk distribution data is generated; For different threat level labels, a preset risk judgment rule set is called, and the potential threat degree is determined combined with the current scene characteristics; wherein the current scene characteristics include environmental humidity and device load. 8.The machine learning based data line charging protection method of claim 1, wherein, The S106 specifically includes: Based on the potential threat degree, an adaptive parameter adjustment mechanism is constructed, and the current adjustment amplitude and the charging time limit in the charging control strategy are dynamically adjusted according to the threat level division; Real-time device running data is obtained, and it is judged whether the adjusted current and charging time are within the rated working range of the device. If compatible, the optimized current output range and charging time configuration are calculated and generated based on the potential threat degree; The charging scene is simulated combined with the adjusted parameters, the occurrence probability of the leakage risk under the parameters is counted, compared with the preset safety threshold, and it is verified whether the parameters meet the safety requirements. 9.The machine learning based data line charging protection method of claim 4, wherein, The S109 specifically includes: According to the deviation analysis report, the deviation trend data is extracted, the high-frequency noise is separated through Fourier transform and other environmental interference filtering technologies, and the filtered deviation trend data set is obtained; The collection strategy of the micro multi-dimensional sensor array is clustered and classified according to data characteristics. If the classification result exceeds the preset strategy threshold, the collection frequency and time interval are updated; The feedback data calibration mode is adjusted, the real-time data is obtained according to the new collection strategy, the new initial state data set is generated, and the state tracking of S103 and the parameter adjustment process of S106 are cyclically executed.

10. A machine learning based data line charging protection system, characterized in that, It includes: A data acquisition module is used to acquire insulation resistance change data and environmental humidity fluctuation data of a data line during operation through a micro multi-dimensional sensor array, and to perform time domain denoising processing on the original signal to generate an initial state data set; A coupling feature extraction module is used to extract the coupling features of insulation resistance change and environmental humidity fluctuation based on the initial state data set, and to determine the key influence factor combination including humidity mutation and resistance decline rate; An abnormality monitoring module is used to monitor the running state parameters according to the key influence factor combination. If the parameters exceed the preset abnormal fluctuation threshold, an abnormal state mark is generated. A risk prediction module is used to combine the abnormal state mark and the historical running track data, predict the data line leakage risk trend through time sequence smoothing technology, and output the risk probability distribution curve. The threat assessment module is configured to generate threat level division labels according to the risk probability distribution curve, and determine the potential threat degree in the current scene by calling a preset risk determination rule set for different labels. The charging strategy adjustment module is configured to dynamically adjust the current regulation amplitude and the charging time limit in the charging control strategy based on the potential threat degree, and generate an optimized current output range and charging time configuration in combination with device compatibility.

Images