Intelligent prediction and early warning method for life cycle of electric energy meter

Abstract

The invention provides an intelligent prediction and early warning method for the life cycle of an electric energy meter, and the method comprises the steps: installing a voltage transformer at a charging station, monitoring the voltage data of a power grid connected with the charging station in real time, and indicating that a large load of the power grid is connected or disconnected when the voltage data of the power grid connected with the charging station is monitored to have instantaneous overvoltage. Operating parameters of the target charging pile are collected in real time; performing data processing on the historical operation parameter data of the target charging pile to obtain an operation parameter normal range of the target charging pile, when the real-time operation parameter of the target charging pile exceeds the normal range, judging that the operation of the charging pile is abnormal, and identifying abnormal data in the operation parameters; and carrying out fusion calculation on the expected life of each device of the electric energy meter to obtain the comprehensive expected life of the electric energy meter, and if the comprehensive expected life of the electric energy meter is lower than a preset threshold, sending out an early warning.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to an intelligent prediction and early warning method for the life cycle of an electric energy meter. Background Art

[0002] At an electric bus station, a large number of electric buses return to the station for charging at night. The concentrated charging will cause a significant load peak at night, or a regional power grid failure will cause some lines to trip, and the charging station connected to the line will be forced to stop operating. In these cases, the power grid connected to the charging station will have a large load connected or disconnected phenomenon. When the large load is connected or disconnected, an instantaneous overvoltage will be generated. This overvoltage will propagate along the line to the electric energy meter of the charging pile, forming a surge. In the power grid connected to the charging station, when a large load is connected or disconnected, it will cause a sudden change in the grid voltage and current, resulting in a surge phenomenon. This surge phenomenon will have a damaging effect on the electric energy meter on the charging pile. First, the surge voltage may break down the voltage transformer of the electric energy meter, causing the insulation failure of the voltage transformer and unable to work normally. Secondly, the surge current may burn the current transformer of the electric energy meter, causing the current transformer to open circuit and unable to continue to be used. Furthermore, the surge may also damage the metering chip of the electric energy meter, resulting in incorrect metering data or inability to measure. Finally, surges may also accelerate the aging of the internal components of the energy meter and shorten the service life of the energy meter. Therefore, how to predict the service life of the energy meter through surge phenomena is a technical problem that needs to be solved urgently. Summary of the invention

[0003] The present invention provides an intelligent prediction and early warning method for the life cycle of an electric energy meter, which mainly includes:

[0004] Install voltage transformers at charging stations to monitor the voltage data of the power grid to which the charging station is connected in real time. When a transient overvoltage is detected in the voltage data of the power grid to which the charging station is connected, it indicates that a large load on the power grid is connected or disconnected, and the operating parameters of the target charging pile are collected in real time.

[0005] Process the historical operating parameter data of the target charging pile to obtain the normal range of the operating parameters of the target charging pile. When the real-time operating parameters of the target charging pile exceed the normal range, it is determined that the charging pile operation is abnormal, and the abnormal data in the operating parameters are identified;

[0006] The wavelet analysis method is used to analyze the abnormal data in the operating parameters, and the characteristics of the surge voltage amplitude, surge current amplitude, and surge duration of the charging pile are extracted when the large load of the power grid is connected or disconnected, and the characteristic mode of the charging pile surge phenomenon under the interference of the large load of the power grid being connected or disconnected is obtained. When the characteristic mode of the surge phenomenon is detected, it is judged that the charging pile has a surge phenomenon;

[0007] Build a simulation test platform for connecting or disconnecting large loads in the power grid, analyze the impact of the surge phenomenon generated by the charging pile on the voltage impact amplitude, current impact amplitude and impact duration of the mutual inductor, metering chip and relay devices in the charging pile when the large load of the power grid is connected or disconnected, determine the damage degree of each device, and establish a mapping relationship between the device damage degree and the surge phenomenon;

[0008] When a surge is detected at the charging pile, the damage to the transformer, metering chip, and relay components inside the electric energy meter is determined based on the mapping relationship between the damage degree of the components and the surge phenomenon. The expected life of each component is input into the pre-built electric energy meter life prediction model to obtain the expected life of each component.

[0009] The expected life of each component of the electricity meter is integrated and calculated to obtain the comprehensive expected life of the electricity meter. If the comprehensive expected life of the electricity meter is lower than the preset threshold, an early warning is issued.

[0010] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0011] The present invention discloses an intelligent prediction and early warning method for the life cycle of an electric energy meter. A voltage transformer is installed at a charging station to monitor the grid voltage in real time. When a transient overvoltage is detected, the operating parameters of the charging pile are collected and compared with historical data to determine whether an abnormality occurs. The surge characteristics are extracted using a wavelet analysis algorithm, and a characteristic pattern of the surge phenomenon is established. The impact on the insulation of the equipment is evaluated based on the surge amplitude and duration, and the corresponding protection process is triggered. The impact of the surge on the internal components of the charging pile is analyzed through simulation experiments, and a damage degree mapping model is established. Combined with the device life expectancy and damage degree discrimination model, the comprehensive life of the electric energy meter is predicted, and an early warning is issued when it is below the threshold. The present invention realizes real-time monitoring of power grid surges in charging stations, damage assessment of electric energy meter components and life prediction, effectively improving the reliability and safety of charging equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 The present invention is a flow chart of an intelligent prediction and early warning method for the life cycle of an electric energy meter.

[0013] Figure 2 It is a schematic diagram of an intelligent prediction and early warning method for the life cycle of an electric energy meter according to the present invention.

[0014] Figure 3 It is another schematic diagram of an intelligent prediction and early warning method for the life cycle of an electric energy meter according to the present invention. DETAILED DESCRIPTION

[0015] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0016] like Figure 1-3 In this embodiment, a method for intelligent prediction and early warning of the life cycle of an electric energy meter may specifically include:

[0017] S101. Install a voltage transformer at the charging station to monitor the voltage data of the power grid to which the charging station is connected in real time. When a transient overvoltage is detected in the voltage data of the power grid to which the charging station is connected, it indicates that a large load of the power grid is connected or disconnected, and the operating parameters of the target charging pile are collected in real time.

[0018] For the voltage sampling sequence collected by the voltage transformer, the fundamental component of the sampling sequence is extracted by Fourier transform, and the voltage amplitude curve is obtained from the fundamental component; the instantaneous voltage value is extracted according to the voltage amplitude curve, and the instantaneous voltage value is compared with the preset standard voltage threshold. If the instantaneous voltage value exceeds the standard voltage threshold for a preset judgment time, it is determined that a load change event occurs in the power grid; the charging pile operation parameter collection instruction is triggered for the load change event, and the output voltage sequence and output current sequence of the charging pile are obtained by the data collection unit, and the Kalman filter state space equation is established to obtain the noise smoothing data; the voltage change rate and the current change rate are calculated based on the noise smoothing data, and a real-time data recording form is established according to the voltage change rate and the current change rate.

[0019] Specifically, data is collected at a fixed sampling frequency for the voltage transformer in the charging station, a voltage sampling sequence is obtained from the voltage transformer, the fundamental component of the sampling sequence is extracted using Fourier transform, and a voltage amplitude curve is obtained based on the fundamental component. The instantaneous voltage value is extracted from the voltage amplitude curve, and the instantaneous voltage value is compared using a preset standard voltage threshold. When the instantaneous voltage value exceeds the standard voltage threshold for a preset judgment time, it is determined that a load change event occurs in the power grid. The charging pile operation parameter collection instruction is triggered according to the load change event, and the output voltage sequence and output current sequence of the charging pile are obtained using a data acquisition unit. For the output voltage sequence and output current sequence, a Kalman filter is used to establish a state space equation, and the noise of the collected data is smoothed by alternating measurement updates and time updates. The steady-state operation parameters of the charging pile are obtained after the noise smoothing process, and the voltage change rate and current change rate are calculated using a variational algorithm, and the operation status of the charging pile is quantitatively evaluated based on the change rate value. A real-time data recording form is established for the evaluation result of the operation status of the charging pile, and the operation parameter fluctuation information is extracted from the data recording form, and the working status level of the charging pile is determined based on the fluctuation range. When the voltage transformer in the charging station collects data, a fixed sampling frequency of 512 Hz is used to monitor the voltage data. 512 sampling points are collected every second to form a voltage sampling sequence. The fundamental component is extracted by Fourier transforming the sampling sequence to obtain the voltage amplitude curve. The discrete Fourier transform algorithm is used to extract the fundamental component. After converting the time domain signal to the frequency domain, the amplitude and phase information corresponding to the power frequency component are extracted. In the voltage monitoring process, the standard voltage threshold is set to 380 volts. When the instantaneous voltage value exceeds this threshold and the duration reaches 10 milliseconds, it can be determined that a load change event has occurred in the power grid. At this time, the operation status of the charging pile is monitored. During the voltage sampling process, the sampling data is processed in real time by the digital signal processing unit, and the sampling value is converted into a digital quantity by the analog-to-digital conversion chip and stored in the data cache area. During the charging pile operation parameter collection process, the output voltage sequence and output current sequence of the charging pile are collected by the data acquisition unit. The sampling frequency is set to 1024 Hz and the sampling accuracy is 16 bits. The collected operating parameter data is processed using the Kalman filter algorithm, the noise covariance matrix in the state space equation is set, and the optimal estimate is obtained through iterative calculation. After noise smoothing, the steady-state operating parameters of the charging pile are obtained. The voltage change rate and current change rate are calculated using a variational algorithm for the operating parameters. The change rate threshold is set to 1% per second. When the voltage or current change rate exceeds the threshold, the charging pile operation state is determined to be abnormal. The operating state evaluation results are recorded in a real-time data form, which contains fields such as timestamp, operating parameter value, and change rate value.In the process of monitoring the operation status of the charging pile, the real-time data recording form is analyzed to extract the operating parameter fluctuation information, and the fluctuation range threshold is set to plus or minus 5% of the rated value. According to the parameter fluctuation amplitude, the operation status is divided into three levels: stable, fluctuating, and abnormal. The operation status level determination result is used to guide the power output control of the charging pile. When it is determined to be an abnormal state, the charging pile protection mechanism is triggered. The data recording form adopts a circular storage method to save the operation data of the last 24 hours, and the historical data is archived regularly. The record fields in the form include sampling time, voltage amplitude, current amplitude, power factor, voltage change rate, current change rate, operation status level and other information.

[0020] S102. Process the historical operating parameter data of the target charging pile to obtain the normal range of the operating parameters of the target charging pile. When the real-time operating parameters of the target charging pile exceed the normal range, it is determined that the operation of the charging pile is abnormal, and the abnormal data in the operating parameters is identified.

[0021] The original data of historical operation parameters is obtained from the charging pile operation log storage unit, the original data is standardized through the data preprocessing module, and the standardized data is denoised by the median filtering method to obtain the historical processing data; the historical processing data is segmented according to the time window, and the mean and standard deviation of the operation parameters in the time window are calculated by the Gaussian distribution fitting method, and the operation parameter fluctuation range is generated according to the mean and standard deviation, and the operation parameter abnormality judgment threshold is extracted from the boundary value of the fluctuation range; the real-time operation parameter sequence is obtained from the charging pile data acquisition unit, and the state of the real-time operation parameter sequence is estimated by the Kalman filtering algorithm to obtain the operation parameter estimation value sequence; if the parameter value in the operation parameter estimation value sequence exceeds the abnormal judgment threshold, abnormal identification information is generated, and an operation status evaluation form is established for the abnormal identification information, and a clustering algorithm is used to classify the abnormal data in the evaluation form to obtain the abnormal type result.

[0022] Specifically, the original data of historical operation parameters is obtained from the charging pile operation log storage unit, the original data is standardized by the data preprocessing module, and the standardized data is subjected to noise reduction processing by the median filtering method to obtain the historical processing data. The historical processing data is segmented according to the time window, and the mean and standard deviation of the operation parameters in each time window are calculated by the Gaussian distribution fitting method, and the operation parameter fluctuation range is generated according to the mean and standard deviation. The upper and lower limit thresholds of the normal range are calculated according to the operation parameter fluctuation range, and the operation parameter abnormality judgment threshold is extracted from the fluctuation range boundary value, and the real-time monitoring rule is established by the sliding time window method. The real-time operation parameter sequence is obtained from the charging pile data acquisition unit, and the state of the real-time operation parameter sequence is estimated by the Kalman filtering algorithm to obtain the operation parameter estimated value sequence. Abnormal detection is performed according to the operation parameter estimated value sequence and the real-time monitoring rules, and abnormal identification information is generated when the parameter value in the estimated value sequence exceeds the normal range threshold. An operation status evaluation form is established for the abnormal identification information, abnormal data records are extracted from the evaluation form, and the abnormal data is classified by the clustering algorithm to obtain the abnormal type result. The historical operating parameters recorded in the charging pile operation log storage unit include key information such as charging voltage, charging current, output power, and charging time. The data sampling frequency is 1024 Hz, and the generated raw data is mapped to the range of 0 to 1 after standardization. The standardized data is subjected to 5-point median filtering for noise reduction. The filter window corresponds to a time interval of 5 milliseconds in the time domain, which eliminates sudden noise interference during the sampling process. The historical processing data is divided into time windows in units of 10 minutes. In each time window, Gaussian distribution is used to perform statistical modeling on the operating parameters. The calculated charging voltage mean is 220 volts, the standard deviation is 4.4 volts, and the charging current mean is 32 amperes, with a standard deviation of 0.64 amperes. Based on these statistical characteristics, the operating parameter fluctuation range is set to the range of plus or minus 3 times the standard deviation of the mean, so that the voltage fluctuation range is 206.8 to 233.2 volts and the current fluctuation range is 30.08 to 33.92 amperes. In the real-time monitoring process, a 60-second sliding time window is used, and the window slides forward 1 second each time to perform real-time analysis on the newly collected operating parameters. In the Kalman filter algorithm, the charging voltage and charging current are used as state variables, and the constant speed model is used to describe the parameter change process. The measurement noise standard deviation is set to 0.1, and the process noise standard deviation is set to 0.01. The state estimate is obtained through iterative calculation. When an abnormality occurs in the estimated value of the operating parameter, the timestamp of the abnormality, the type of abnormal parameter, the duration of the abnormality, the amplitude of the abnormality and other information are recorded. The abnormal identification information record shows that the charging voltage abnormality was detected at 9:30, and the voltage value rose to 240 volts, exceeding the upper limit of the normal range, and the duration of the abnormality reached 300 milliseconds.The abnormal data recorded in the operation status evaluation form is classified using a density-based clustering algorithm, and the abnormal data is divided into three types: instantaneous fluctuations, continuous deviations, and periodic fluctuations. Instantaneous fluctuations account for 75% of the total number of abnormalities, continuous deviations account for 15%, and periodic fluctuations account for 10%. In daily operation, the fluctuations in the operating parameters of charging piles are mainly due to factors such as grid voltage fluctuations, load changes, and ambient temperature changes. Through statistical analysis of historical data to establish a normal range benchmark, timely identification and classification of abnormal states are achieved. The clustering analysis results of abnormal data reflect the main problem types and distribution characteristics faced by charging piles during operation. The abnormal records in the evaluation form are stored in chronological order, and the abnormal data of the last 7 days are retained for operation status trend analysis.

[0023] S103. Use wavelet analysis method to analyze abnormal data in operating parameters, extract the characteristics of charging pile surge voltage amplitude, surge current amplitude, and surge duration when large loads of the power grid are connected or disconnected, and obtain the characteristic pattern of charging pile surge phenomenon under the interference of large loads of the power grid being connected or disconnected. When the characteristic pattern of surge phenomenon is detected, it is determined that a surge phenomenon occurs in the charging pile.

[0024] Wavelet transform is used to decompose the waveform data in the charging pile anomaly database, and the surge signal characteristic component is obtained according to the reconstruction coefficient; wavelet packet decomposition is performed on the surge signal characteristic component, and the surge voltage amplitude characteristic and the surge current amplitude characteristic are obtained by calculating the energy coefficient of each frequency band; the surge amplitude interval is judged according to the surge voltage amplitude characteristic and the surge current amplitude characteristic, and the time-frequency distribution characteristics of the surge amplitude interval are calculated by short-time Fourier transform; a feature recognition rule base is established for the time-frequency distribution characteristics, and a surge feature recognition model is obtained through support vector machine training, and feature matching is performed on the waveform data of real-time monitoring of the charging pile. If the feature matching degree exceeds the matching degree threshold, surge phenomenon identification information is generated.

[0025] Specifically, the surge waveform data is obtained from the charging pile abnormality database, and the waveform data is decomposed into four layers using wavelet transform. The high-frequency component of the surge signal is extracted according to the reconstruction coefficient to obtain the surge characteristic component. The surge characteristic component is decomposed by wavelet packet, and the surge voltage amplitude characteristic and surge current amplitude characteristic are generated by calculating the energy coefficient of each frequency band, and the surge amplitude interval is determined according to the size of the energy coefficient. The surge duration characteristic is extracted from the surge amplitude interval, and the time-frequency distribution characteristic is calculated using short-time Fourier transform, and the surge duration interval is obtained through the characteristic distribution spectrum. A feature recognition rule base is established based on the surge amplitude characteristic and duration characteristic, and the feature combination is trained using a support vector machine to obtain a surge feature recognition model. The waveform data of the current monitoring point is extracted from the real-time monitoring data of the charging pile, and the waveform data is decomposed by wavelet transform to obtain the real-time surge feature. The surge feature recognition model is used to perform feature matching for the real-time surge feature, and surge phenomenon identification information is generated when the feature matching degree exceeds the preset threshold. The surge waveform data stored in the charging pile abnormality database is collected at a sampling frequency of 1024 Hz, recording the voltage and current waveforms of the charging pile at the time when the grid load changes. When performing wavelet transform on the waveform data, the db4 wavelet basis function is selected, and the four-layer wavelet decomposition can decompose the signal into different frequency bands, where the high-frequency component corresponds to the surge characteristics. The first layer of decomposition corresponds to the 256-512 Hz band, the second layer corresponds to the 128-256 Hz band, the third layer corresponds to the 64-128 Hz band, and the fourth layer corresponds to the 32-64 Hz band. Wavelet packet decomposition divides the signal into finer frequency bands, analyzes the energy distribution of the surge characteristic components on the time-frequency plane, and calculates the energy coefficient of each frequency band. In actual operation, the surge voltage amplitude often appears in the range of 1.5-2.0 times the rated value, corresponding to an energy coefficient between 0.6-0.8, and the surge current amplitude is in the range of 2.0-3.0 times the rated value, corresponding to an energy coefficient between 0.7-0.9. When using short-time Fourier transform to perform time-frequency analysis on surge signals, the window length is set to 64 sampling points, and the overlap rate of adjacent windows is 50%. By analyzing the time-frequency distribution feature map, it is found that the surge duration is usually in the range of 50-200 milliseconds, and the time-frequency map shows the characteristics of energy density rapidly rising over time and then decaying. The energy of the high-frequency component reaches a peak at the beginning of the surge, and then gradually decays. The feature recognition rule library contains a variety of typical surge feature combinations, and the support vector machine uses radial basis kernel function for feature mapping. The training samples include 500 sets of normal working state data and 200 sets of surge state data. The feature vector contains characteristic components such as voltage amplitude, current amplitude, duration, and high-frequency energy ratio. The classification accuracy of the trained model reaches 95%. During the real-time monitoring process, feature extraction is performed on the collected waveform data every 100 milliseconds, and the extracted features are matched with the pre-trained model.The characteristic matching degree sets a threshold of 0.85, and the surge alarm is triggered when the matching degree exceeds the threshold. The actual operation shows that the surge phenomenon caused by large load switching has obvious time-frequency characteristics, the voltage surge amplitude reaches 1.8 times the rated value, the current surge amplitude reaches 2.5 times the rated value, the duration is 120 milliseconds, and the high-frequency energy ratio is 0.75, which has a high degree of matching with the characteristic patterns stored in the rule base. When large load switching occurs in the power grid, the output parameters of the charging pile will fluctuate to varying degrees. By establishing a surge feature recognition model, accurate identification of the surge phenomenon is achieved. The feature recognition results show that the surge phenomenon is completed within 0.2 seconds, during which the voltage and current fluctuation amplitudes are large, but then they will return to normal working conditions. The surge feature data is recorded in the database for subsequent statistical analysis.

[0026] According to the amplitude and duration of surge voltage and current, the impact of surge on the insulation of power equipment is evaluated. If it exceeds the insulation level threshold of the equipment, it is judged that there is a risk of equipment insulation breakdown. According to the time, location and impact degree of the surge, a surge alarm event is generated to trigger the preset corresponding protection process.

[0027] The surge voltage amplitude and surge current amplitude are obtained according to the charging pile monitoring unit, the surge occurrence time and duration are extracted from the data acquisition unit, and the surge basic data is obtained by recording the surge occurrence location through the monitoring point information; the surge basic data is standardized, the surge number and interval time data are extracted from the surge monitoring record, and the impact intensity data is generated according to the surge amplitude sequence; the impact intensity data is quantitatively evaluated by using a neural network, the equipment insulation standard parameters are obtained from the preset insulation level parameter library, and the insulation impact level is obtained according to the evaluation value; the peak point and duration interval information are extracted according to the insulation impact level, the equipment insulation aging record is read from the historical database, the insulation breakdown risk value is obtained by comparative calculation, and the corresponding protection plan is triggered according to the risk value.

[0028] Specifically, the surge voltage amplitude and surge current amplitude are obtained according to the charging pile monitoring unit, the surge occurrence time and duration are extracted from the data acquisition unit, and the surge basic data is obtained by recording the surge occurrence location through the monitoring point information. The surge basic data is standardized, the surge number and interval time are extracted from the surge monitoring record, and the impact strength data is generated according to the surge amplitude sequence. The equipment insulation standard parameters are obtained from the preset insulation level parameter library, and the impact strength data is quantitatively evaluated using a neural network, and the insulation impact level is obtained based on the evaluation value. According to the insulation impact level, the peak point and continuous interval information are extracted, the equipment insulation aging record is read from the historical database, and the insulation breakdown risk value is obtained by comparative calculation. A real-time monitoring record is established for the insulation breakdown risk value, and an alarm trigger signal is generated when the risk value exceeds the preset threshold, and the random forest algorithm is used to determine the alarm level. The alarm event description is constructed from the alarm level information, the surge occurrence time, location and impact degree information are recorded, and the corresponding protection plan is matched according to the alarm level. Generate a protection control instruction sequence for the protection plan, obtain the standard protection process from the instruction sequence database, and trigger the corresponding protection mechanism according to the protection process. The basic surge data collected by the charging pile monitoring unit includes information such as voltage amplitude, current amplitude, occurrence time and duration. The voltage amplitude record shows that at the moment of large load switching, the voltage rises to 2.5 times the rated value instantaneously, and the duration reaches 150 milliseconds. The surge occurs at the incoming line end of the distribution cabinet of the charging station. The monitoring record shows that three surges occurred continuously within 24 hours, and the surge intervals were 4 hours and 6 hours respectively. The surge data standardization process uses the maximum and minimum value normalization method to map the voltage amplitude and current amplitude to the 0-1 interval. The standardized value is combined with the number of surge occurrences, interval time and other features to form an impact strength feature vector. The insulation withstand voltage level recorded in the equipment insulation standard parameter library is 1000 volts, and the impact tolerance time is 200 milliseconds. The neural network uses a three-layer structure to evaluate the impact strength. The input layer contains 6 feature nodes, the hidden layer uses 12 nodes, and the output layer is a single node to characterize the impact level. The insulation shock level is divided into three levels: mild, moderate, and severe. When the shock level is severe, the risk of insulation breakdown needs to be further evaluated. Historical data shows that the insulation resistance value in the equipment insulation aging record has dropped to 85% of the initial value. The breakdown risk value calculated in combination with the current shock intensity is 0.75, which exceeds the preset 0.7 threshold and triggers an alarm signal. The alarm level is determined by the random forest algorithm. The input features include information such as surge amplitude, duration, frequency of occurrence, degree of insulation aging, etc. The alarm level is determined by the voting results of 500 decision trees. The information recorded in the alarm event description shows that a severe surge shock was detected. The time of occurrence was 9:30 and the location was the incoming line end of the No. 2 charging pile. The impact level reached the tolerance limit of the equipment.The protection plan selects the corresponding control strategy according to the alarm level. The protection plan corresponding to the current alarm level includes: reducing the charging power to 50%, starting the surge suppressor, and triggering overvoltage protection. The protection control instructions are executed in sequence according to the standard process. First, the power reduction instruction is executed, then the surge suppression device is activated, and finally the overvoltage protection circuit is triggered to cut off the power supply in extreme cases. The actual operation shows that after the power reduction and surge suppression are executed, the surge amplitude is reduced to within 1.5 times the rated value, the duration is shortened to 80 milliseconds, and the insulation of the equipment has not been broken down. The monitoring data further shows that after the protection measures take effect, the output parameters of the charging pile gradually return to normal, the voltage amplitude returns to the range of plus or minus 5% of the rated value, the insulation breakdown risk value is reduced to 0.3, and the system enters normal operation. The protection process takes no more than 500 milliseconds, effectively protecting the insulation of the equipment.

[0029] S104. Build a simulation test platform for connecting or disconnecting large loads in the power grid, analyze the impact of the surge phenomenon generated by the charging pile on the voltage shock amplitude, current shock amplitude and shock duration of the mutual inductor, metering chip and relay devices in the charging pile when the large load of the power grid is connected or disconnected, determine the damage degree of each device, and establish a mapping relationship between the device damage degree and the surge phenomenon.

[0030] For the load switching test signal generated by the load switching test unit, the voltage and current data at the load switching time are obtained from the power grid monitoring device, and the voltage and current data are used to record the real-time response waveform of the key device port of the charging pile; according to the real-time response waveform, the input-output voltage ratio and phase difference are obtained from the transformer sampling end, the signal distortion is obtained from the metering chip sampling end, and the contact action time is obtained from the relay sampling end; for the voltage ratio, signal distortion and contact action time, wavelet transform is used to perform noise reduction processing, and the device response feature sequence is obtained through data reconstruction; according to the device response feature sequence, damage parameter indicators are extracted, and the damage parameter indicators include voltage deviation rate, metering error rate, and contact resistance, and a device damage assessment criterion is established using a deep learning method.

[0031] Specifically, a load switching test unit is used to generate a large load switching test signal, and the voltage and current data at the load switching time are obtained from the power grid monitoring device. The real-time response waveform of the key device port of the charging pile is recorded through the data acquisition module. The response waveform data is characterized by decomposition, and the input-output voltage ratio and phase difference are obtained from the sampling end of the transformer, the signal distortion is obtained from the sampling end of the metering chip, and the contact action time is obtained from the sampling end of the relay. The device performance measurement record is established according to the device sampling data, and the measured data is subjected to noise reduction processing by wavelet transform, and the device response feature sequence is obtained by data reconstruction. The damage parameter index is extracted from the device response feature sequence, including the voltage deviation rate, the metering error rate, and the contact resistance, and the device damage assessment criterion is established by using the deep learning method. A damage degree quantification standard is constructed for the device damage assessment criterion, and the surge impact data is read from the data buffer area. The device performance attenuation curve is obtained by feature matching calculation. The device damage degree data is generated according to the performance attenuation curve, and the random forest algorithm is used to extract the features of the damage degree data, and the mapping relationship between surge features and damage degree is established. A device damage assessment model is constructed from the mapping relationship, and device damage prediction is performed for the real-time collected surge data to generate device damage risk warning information. The load switching test unit simulates the load change of the power grid by controlling the on-off state of the high-power switch. The switching power is set to 1000 kilowatts, the switching time interval is 1 second, and the response characteristics of each device of the charging pile are recorded at a sampling frequency of 1024 Hz. The monitoring data shows that at the moment of load switching, the voltage at the input end of the transformer has a peak of 2.5 times the rated value, the peak voltage at the output end reaches 1.8 times the rated value, and the phase difference increases from 2 degrees in normal operation to 12 degrees. After characteristic decomposition, the device response waveform shows that the input-output voltage ratio of the transformer deviates from the calibration value by 15%, the total harmonic distortion of the sampling data of the metering chip reaches 8%, and the relay contact action time is extended from the rated 15 milliseconds to 25 milliseconds. These parameter changes reflect the performance degradation of the device under surge impact. The measured data is denoised by wavelet transform, and the reconstructed characteristic sequence clearly shows the change trend of various performance parameters. The extraction results of damage parameter indicators show that the voltage deviation rate of the transformer reaches 15%, which exceeds the allowable range of 10%; the measurement error rate of the metering chip is 3%, close to the limit value of 5%; the contact resistance of the relay contact increases to 200 milliohms, which is twice the initial value. The deep learning network adopts a 5-layer structure. The input layer contains 12 feature nodes. Through the nonlinear mapping of 3 hidden layers, the output layer gives the device damage assessment value. The performance attenuation curve records the performance change process of the device under continuous surge impact, and the slope of the curve represents the performance attenuation rate.The test data shows that after 50 surge shocks, the measurement error of the transformer increases to 1.5 times the initial value; after 80 shocks, the signal distortion of the metering chip increases to 2 times the initial value; after 30 shocks, the contact resistance of the relay increases to 2.5 times the initial value. The random forest algorithm uses 200 decision trees to build a damage mapping model. The input features include parameters such as surge amplitude, duration, and frequency of occurrence. The output result is the predicted value of the device damage degree. Model verification shows that when the surge amplitude exceeds 2 times the rated value and the duration exceeds 100 milliseconds, the device damage risk reaches 0.8, and protective measures need to be taken in time. Actual operation data shows that after taking measures such as reducing charging power and starting surge suppression, the device performance parameters are basically maintained within the allowable range, and the damage risk is reduced to below 0.3. The generation of early warning information is based on real-time monitoring data. When the surge amplitude is detected to exceed the threshold, the device damage risk value is calculated immediately. The early warning level is divided into three levels: low risk, medium risk, and high risk, and the corresponding damage risk values ​​are 0.3, 0.6, and 0.9 respectively. After the risk warning is triggered, the corresponding level of protection measures will be automatically activated to achieve active protection of the key components of the charging pile. Monitoring records show that the response time of the warning mechanism does not exceed 50 milliseconds, which can effectively prevent component damage.

[0032] S105. When it is determined that a surge occurs in the charging pile, the extent of damage to the internal mutual inductor, metering chip, and relay components of the electric energy meter is determined based on the mapping relationship between the degree of component damage and the surge phenomenon. The expected life of each component is input into a pre-built electric energy meter life prediction model to obtain the expected life of each component.

[0033] Obtain surge characteristic parameters from the charging pile surge detection unit, and extract performance influencing parameters from the device damage mapping database according to the surge characteristic parameters; generate device performance evaluation data using a deep neural network model for the performance influencing parameters, and establish a device status record table according to the performance evaluation data; use a life prediction algorithm to analyze the device status record table to obtain the device performance decay rate, and obtain the device aging trend through the performance parameter change curve according to the performance decay rate; construct a life prediction data set according to the device aging trend, and use a random forest algorithm to train the life prediction data set to obtain the device expected life prediction result;

[0034] Specifically, the surge characteristic parameters, including surge amplitude, duration and occurrence frequency, are obtained from the charging pile surge detection unit, and the corresponding performance impact parameters are extracted according to the device damage mapping database. The device damage index is calculated for the performance impact parameters, the mutual inductor measurement error, metering chip accuracy, and relay contact parameters are read from the device performance database, and the device performance evaluation data is generated using a deep neural network. A device state record table is established based on the performance evaluation data, and the device performance decay rate is extracted using a life prediction algorithm, and the device aging trend is obtained through the performance parameter change curve. The performance limit threshold is obtained from the preset device standard library, and the performance decay fitting is performed for the device aging trend, and the remaining life value of the device is calculated through the performance degradation curve. A life prediction data set is constructed based on the remaining life value of the device, and the life prediction parameters are trained using a random forest algorithm to generate the device expected life prediction result. A device update warning mechanism is established for the expected life prediction result, and the device reference life value is extracted from the standard life database. The device life warning information is obtained through life comparison calculation. A device health file is established based on the warning information, and the cumulative damage value, performance decay rate, and expected life value of the device are recorded to generate a device status evaluation report. The charging pile surge detection record shows that when a large load switching occurs in the power grid, the surge voltage amplitude generated reaches 2.5 times the rated value, the duration is 150 milliseconds, and an average of 3 surges occur every 24 hours. These surge characteristic parameters are converted into device performance influencing parameters through damage mapping relationships, among which the measurement accuracy of the transformer decreases by 15%, the sampling accuracy of the metering chip decreases by 8%, and the action time of the relay contact is extended by 60%. In the device performance evaluation stage, the deep neural network uses a 5-layer structure to model the performance parameters. The input layer contains 12 feature nodes such as surge characteristics, ambient temperature, and load level. Through the nonlinear mapping of 3 hidden layers, the output layer gives the degree of device performance degradation. The evaluation results show that the measurement error of the transformer increases to level 0.5 after 1,000 surge shocks, the signal distortion of the metering chip increases to 5%, and the contact resistance of the relay contact increases to 200 milliohms. The device status record table shows the performance decay rate data. The transformer measurement error increases by 0.2 levels for every 1,000 surges, the metering chip accuracy decreases by 2% for every 1,000 surges, and the relay contact resistance increases by 50 milliohms for every 1,000 surges. The device performance limit thresholds are set as 1.0 level transformer measurement error, 8% metering chip accuracy error, and 300 milliohms relay contact resistance. Through performance degradation curve fitting calculation, it is predicted that the remaining life of the transformer is 4,000 surge shocks, the remaining life of the metering chip is 3,500 surge shocks, and the remaining life of the relay is 2,500 surge shocks. The random forest algorithm uses 200 decision trees to build a life prediction model, and the training data contains 500 sets of historical device performance records.The prediction results show that according to the current surge frequency, the expected life of the transformer is 16 months, the expected life of the metering chip is 14 months, and the expected life of the relay is 10 months. The reference life of the device recorded in the standard life database is 24 months for the transformer, 20 months for the metering chip, and 15 months for the relay. Comparative analysis shows that surge impact shortens the life of the device by 30% to 40%. The device health file records the entire process of performance degradation. The measurement error of the transformer has reached level 0.5, and the annual attenuation rate is 0.3; the accuracy error of the metering chip has reached 5%, and the annual attenuation rate is 3%; the contact resistance of the relay has reached 200 milliohms, and the annual attenuation rate is 100 milliohms. The cumulative damage assessment shows that the damage degree of the transformer is 45%, the damage degree of the metering chip is 50%, and the damage degree of the relay is 65%. When the damage degree of the device exceeds 80% or the remaining life is less than 3 months, the update warning signal is triggered. Combined with historical operation data, it is expected that the relay will reach the update warning conditions within 3 months, and replacement should be given priority.

[0035] According to the preset parameter thresholds of the transformer, metering chip and relay devices inside the electricity meter, a corresponding training sample set is established. The training sample set is studied and classified to obtain a discrimination model for the damage degree of each device. The surge characteristics are input into the trained discrimination model to obtain the damage degree assessment results of the transformer, metering chip and relay.

[0036] The transformer measurement error value, metering chip accuracy value, and relay contact resistance value are extracted from the device parameter database, and a device parameter training set is generated through a standard parameter file; the device parameter training set is decomposed using Fourier transform, and amplitude parameters, frequency parameters, and duration parameters are obtained from surge waveform data, and a feature vector matrix is ​​constructed according to the surge parameters; a deep neural network is used for training based on the feature vector matrix, and device performance attenuation data is obtained through network calculation, and a damage degree index is extracted from the performance attenuation data; a support vector machine is used for classification of the damage degree index, and a discrimination standard is obtained from a preset damage level threshold, and a device damage discrimination function is obtained through indicator mapping, and a device damage assessment result is calculated using the discriminant function.

[0037] Specifically, the mutual inductor measurement error value, metering chip accuracy value, and relay contact resistance value are extracted from the device parameter database, a standard parameter file is established according to the preset threshold interval, and a device parameter training set is generated through data labeling. Feature extraction is performed on the device parameter training set, surge waveform data is decomposed using Fourier transform, and amplitude parameters, frequency parameters, and duration parameters are obtained from the surge sequence. A feature vector matrix is ​​constructed according to the surge parameters, and the feature data is trained using a deep neural network, and device performance attenuation data is obtained through network calculation. A damage degree index is extracted from the performance attenuation data, a damage level standard is established for different devices, and a support vector machine is used to classify the damage degree index. A discrimination criterion library is established according to the damage classification result, and the discrimination criterion is obtained from the preset damage level threshold, and a device damage discrimination function is obtained through indicator mapping. A verification test is performed on the damage discrimination function, a surge feature sequence is extracted from the verification data set, and a device damage assessment result is obtained by using the discriminant function calculation. A device status file is constructed from the damage assessment result, and device performance parameters, damage degree values, and assessment timestamps are recorded to generate a device damage assessment report. The device parameter database records the standard parameter ranges of different devices, among which the standard value of the mutual inductor measurement error is 0.2 level, the allowable error range is plus or minus 0.2%, the standard value of the metering chip accuracy is 0.5 level, the allowable error range is plus or minus 0.5%, and the standard value of the relay contact resistance is 100 milliohms, and the allowable fluctuation range is plus or minus 20 milliohms. These parameters are divided into four levels of normal, slightly damaged, moderately damaged, and severely damaged by the data labeling method to construct a training sample set. The feature extraction of surge waveform data uses the Fourier transform method to convert the time domain waveform to the frequency domain for analysis. The extracted characteristic parameters include the surge peak amplitude reaching 2.5 times the rated value, the main frequency components distributed in the range of 50 Hz to 1000 Hz, and the duration is 150 milliseconds. These characteristic parameters are processed by data standardization to form feature vectors as input data of the deep neural network. The deep neural network adopts a 5-layer structure, the input layer contains 12 feature nodes, and the device performance attenuation is calculated through nonlinear mapping of 3 hidden layers. The training results show that the measurement error of the mutual inductor rises to level 0.5, which is equivalent to 2.5 times the standard value; the accuracy of the metering chip drops to level 1.2, which is 2.4 times the standard value; the contact resistance of the relay increases to 250 milliohms, which is 2.5 times the standard value. The support vector machine establishes a criterion for judging the degree of damage by calculating the optimal classification hyperplane between different damage levels. The classification features include information such as performance parameter deviation value, parameter change rate, and cumulative surge times. The judgment standard shows that when the performance parameter deviation exceeds 2 times the standard value and lasts for more than 100 milliseconds, it is judged as moderate damage; when the deviation exceeds 3 times or lasts for more than 200 milliseconds, it is judged as severe damage.The verification test used 200 sets of actual surge data for model verification, including 80 sets of normal data, 60 sets of lightly damaged data, 40 sets of moderately damaged data, and 20 sets of severely damaged data. The accuracy of the discrimination results reached 92%, of which the accuracy of normal state recognition was 96%, the accuracy of lightly damaged recognition was 94%, the accuracy of moderately damaged recognition was 90%, and the accuracy of severe damage recognition was 88%. The device status archive records show that after the transformer has experienced 1,000 surge shocks, the measurement error has increased to level 0.6, which has reached the standard of moderate damage; after 800 surge shocks, the accuracy of the metering chip has dropped to level 1.5, which is in a state of light damage; after 500 surge shocks, the contact resistance of the relay has increased to 280 milliohms, which has reached the standard of severe damage. The evaluation report recommends replacing the relay, focusing on monitoring the transformer, and regularly checking the metering chip.

[0038] S106. The expected lifespans of the various components of the electric energy meter are integrated and calculated to obtain a comprehensive expected lifespan of the electric energy meter. If the comprehensive expected lifespan of the electric energy meter is lower than a preset threshold, an early warning is issued.

[0039] The expected life data of the transformer, the expected life data of the metering chip, and the expected life data of the relay are obtained from the electric energy meter status database, and a device importance score is established for the expected life data to obtain a device weight coefficient; a weighted operation is performed based on the device weight coefficient and the expected life data, and a life assessment matrix is ​​constructed using a device performance attenuation curve, and a device life feature sequence is extracted from the life assessment matrix; a life association data table is established for the device life feature sequence, and a deep neural network is used to train the life association data table to obtain a device life association value; the device combination life is calculated based on the device life association value, and a support vector regression method is used to perform data fitting on the device combination life, a device state sequence is generated from the data fitting result, and a life prediction is performed for the device state sequence.

[0040] Specifically, the expected life data of mutual inductors, metering chips, and relays are extracted from the electric energy meter status database, and the importance score is established according to the role of the device in electric energy metering, and the device weight coefficient is generated by the score value. The device weight coefficient and the expected life data are weighted and calculated, and the device life evaluation matrix is ​​constructed using the device performance attenuation curve, and the device life feature sequence is extracted from the evaluation matrix. A life association data table is established according to the device life feature sequence, and a deep neural network is used to analyze the life influence relationship between devices, and the device life association value is obtained through network training. The device combined life is calculated from the life association value, and the combined life data is fitted using the support vector regression method, and the device state sequence is generated according to the fitting result. The life prediction calculation is performed for the device state sequence, and the device standard life value is obtained from the preset life parameter library. The comprehensive life value of the electric energy meter is obtained by life comparison. The warning trigger condition is established according to the comprehensive life value, and the life warning threshold is read from the warning rule database. When the comprehensive life value is lower than the warning threshold, a warning signal is generated. A warning level determination table is constructed for the warning signal, and the device life state, warning time, and warning level are recorded to generate an electric energy meter warning notification. The expected life data of key components are recorded in the status database of the electric energy meter. The expected life of the transformer is 16 months, the expected life of the metering chip is 14 months, and the expected life of the relay is 10 months. The importance score is determined according to the role of the device in the metering process. The transformer is scored 0.4, which is responsible for basic power measurement; the metering chip is scored 0.35, which processes data calculation; the relay is scored 0.25, which controls the on and off of the circuit. The performance decay curve shows the life characteristics of the device. The measurement error of the transformer increases by 0.1 level every 2 months of operation, the accuracy of the metering chip decreases by 0.2 level every 2 months of operation, and the contact resistance of the relay increases by 50 milliohms every 2 months of operation. The life evaluation matrix contains three dimensions: performance parameters, operating time, and environmental factors. The comprehensive life characteristics of the device are obtained through weighted calculation. The deep neural network uses a 5-layer structure to analyze the life correlation between devices. The input layer contains 12 feature nodes, and the degree of influence of the device life is calculated through nonlinear mapping of 3 hidden layers. The results of correlation analysis show that the reduction of transformer life will lead to a decrease in the measurement accuracy of the metering chip, with an influence coefficient of 0.6; the reduction of metering chip life will cause an increase in data calculation errors, with an influence coefficient of 0.5; the reduction of relay life will cause circuit switching instability, with an influence coefficient of 0.4. The support vector regression method is used to model the combined life data, and the feature vector contains device performance parameters, life correlation values, and operating environment data. The fitting results show that when the transformer life is reduced to 12 months, the metering chip life will be reduced to 10 months accordingly; when the metering chip life is reduced to 8 months, the measurement data distortion increases to 8%; when the relay life is reduced to 6 months, the contact jitter time increases to 30 milliseconds. The weighted average method is used to predict the comprehensive life of the electric energy meter, taking into account the importance of the device and the correlation effect.The prediction results show that the comprehensive life of the energy meter is 9 months, which is lower than the preset threshold of 12 months. The warning rules are set at three levels. When the comprehensive life is less than 12 months, a low-level warning is triggered, when it is less than 9 months, a medium-level warning is triggered, and when it is less than 6 months, a high-level warning is triggered. The current warning judgment result is a medium-level warning. The warning record shows that the warning time is September 15. The device status shows that the relay has reached the replacement standard, the transformer is close to the replacement standard, and the metering chip is in a normal monitoring state. The warning notice recommends that the relay be replaced within 3 months, the operating status monitoring of the transformer be strengthened, and the working parameters of the metering chip be checked regularly. The comprehensive life assessment results show that the overall performance of the energy meter continues to decline, and maintenance measures need to be taken in a timely manner.

[0041] The description of the above embodiments is only used to help understand the technical solutions and core ideas of the present application. Ordinary technical personnel in this field should understand that they can still modify the technical solutions recorded in the aforementioned embodiments, or replace some of the technical features therein with equivalents. However, these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. An intelligent prediction and early warning method for the life cycle of an electric energy meter, characterized in that: The method comprises: Install voltage transformers at charging stations to monitor the voltage data of the power grid to which the charging station is connected in real time. When a transient overvoltage is detected in the voltage data of the power grid to which the charging station is connected, it indicates that a large load on the power grid is connected or disconnected, and the operating parameters of the target charging pile are collected in real time. Process the historical operating parameter data of the target charging pile to obtain the normal range of the operating parameters of the target charging pile. When the real-time operating parameters of the target charging pile exceed the normal range, it is determined that the charging pile operation is abnormal, and the abnormal data in the operating parameters are identified; The wavelet analysis method is used to analyze the abnormal data in the operating parameters, and the characteristics of the surge voltage amplitude, surge current amplitude, and surge duration of the charging pile are extracted when the large load of the power grid is connected or disconnected, and the characteristic mode of the charging pile surge phenomenon under the interference of the large load of the power grid being connected or disconnected is obtained. When the characteristic mode of the surge phenomenon is detected, it is judged that the charging pile has a surge phenomenon; Build a simulation test platform for connecting or disconnecting large loads in the power grid, analyze the impact of the surge phenomenon generated by the charging pile on the voltage impact amplitude, current impact amplitude and impact duration of the mutual inductor, metering chip and relay devices in the charging pile when the large load of the power grid is connected or disconnected, determine the damage degree of each device, and establish a mapping relationship between the device damage degree and the surge phenomenon; When a surge is detected at the charging pile, the damage to the transformer, metering chip, and relay components inside the electric energy meter is determined based on the mapping relationship between the damage degree of the components and the surge phenomenon. The expected life of each component is input into the pre-built electric energy meter life prediction model to obtain the expected life of each component. The expected life of each component of the electricity meter is integrated and calculated to obtain the comprehensive expected life of the electricity meter. If the comprehensive expected life of the electricity meter is lower than the preset threshold, an early warning is issued.

2. The method according to claim 1, characterized in that The voltage transformer is installed at the charging station to monitor the voltage data of the power grid connected to the charging station in real time. When the voltage data of the power grid connected to the charging station is monitored to have an instantaneous overvoltage, it indicates that a large load of the power grid is connected or disconnected, and the operating parameters of the target charging pile are collected in real time, including: For the voltage sampling sequence collected by the voltage transformer, the fundamental component of the sampling sequence is extracted by Fourier transform, and the voltage amplitude curve is obtained from the fundamental component; Extracting the instantaneous voltage value according to the voltage amplitude curve, comparing the instantaneous voltage value with a preset standard voltage threshold, and if the instantaneous voltage value exceeds the standard voltage threshold for a preset determination time, determining that a load change event occurs in the power grid; The load change event triggers a charging pile operation parameter collection instruction, uses a data collection unit to obtain the charging pile output voltage sequence and output current sequence, and establishes a Kalman filter state space equation to obtain noise smoothing processing data; The voltage change rate and the current change rate are calculated based on the noise smoothing data, and a real-time data recording form is established according to the voltage change rate and the current change rate.

3. The method according to claim 1, characterized in that The data processing of the historical operating parameter data of the target charging pile obtains the normal range of the operating parameters of the target charging pile. When the real-time operating parameters of the target charging pile exceed the normal range, it is determined that the charging pile operation is abnormal, and the abnormal data in the operating parameters are identified, including: Acquire the original data of historical operation parameters from the charging pile operation log storage unit, perform standardization processing on the original data through the data preprocessing module, and perform noise reduction processing on the standardized data by using the median filtering method to obtain historical processing data; The historical processing data is segmented according to time windows, a Gaussian distribution fitting method is used to calculate the mean and standard deviation of the operating parameters in the time window, an operating parameter fluctuation range is generated according to the mean and standard deviation, and an operating parameter abnormality determination threshold is extracted from the fluctuation range boundary value; Acquire a real-time operating parameter sequence from a charging pile data acquisition unit, and use a Kalman filter algorithm to perform state estimation on the real-time operating parameter sequence to obtain an operating parameter estimation value sequence; If the parameter value in the operating parameter estimation value sequence exceeds the abnormality determination threshold, abnormality identification information is generated, an operating status evaluation form is established for the abnormality identification information, and a clustering algorithm is used to classify the abnormal data in the evaluation form to obtain an abnormality type result.

4. The method according to claim 1, characterized in that: The wavelet analysis method is used to analyze the abnormal data in the operating parameters, extract the characteristics of the surge voltage amplitude, surge current amplitude, and surge duration of the charging pile when the large load of the power grid is connected or disconnected, and obtain the characteristic mode of the charging pile surge phenomenon under the interference of the large load of the power grid being connected or disconnected. When the characteristic mode of the surge phenomenon is detected, it is determined that the charging pile has a surge phenomenon, including: Wavelet transform is used to decompose the waveform data in the charging pile abnormality database, and the characteristic components of the surge signal are obtained according to the reconstruction coefficient; Performing wavelet packet decomposition on the characteristic components of the surge signal, and obtaining the surge voltage amplitude characteristics and the surge current amplitude characteristics by calculating the energy coefficient of each frequency band; Determine the surge amplitude interval according to the surge voltage amplitude characteristics and the surge current amplitude characteristics, and calculate the time-frequency distribution characteristics of the surge amplitude interval by short-time Fourier transform; A feature recognition rule base is established for the time-frequency distribution characteristics, a surge feature recognition model is obtained through support vector machine training, and feature matching is performed on the waveform data of real-time monitoring of the charging pile. If the feature matching degree exceeds the matching degree threshold, surge phenomenon identification information is generated; It also includes: evaluating the impact of surges on the insulation of power equipment based on the amplitude and duration of surge voltage and current. If the insulation level threshold of the equipment is exceeded, it is judged that there is a risk of equipment insulation breakdown. Based on the time, location and impact level of the surge, a surge alarm event is generated to trigger the preset corresponding protection process.

5. The method according to claim 4, characterized in that The impact degree of the surge on the insulation of the power equipment is evaluated according to the amplitude and duration of the surge voltage and current. If the equipment insulation level threshold is exceeded, it is determined that there is a risk of equipment insulation breakdown. According to the time, location and impact degree of the surge, a surge alarm event is generated to trigger the preset corresponding protection process, including: The surge voltage amplitude and surge current amplitude are obtained from the charging pile monitoring unit, the surge occurrence time and duration are extracted from the data acquisition unit, and the surge occurrence location is recorded through the monitoring point information to obtain the surge basic data; Standardizing the basic surge data, extracting surge frequency and interval time data from surge monitoring records, and generating impact intensity data based on surge amplitude sequences; A neural network is used to quantitatively evaluate the impact strength data, and the insulation standard parameters of the equipment are obtained from a preset insulation level parameter library, and the insulation impact level is obtained according to the evaluation value; The peak point and duration interval information are extracted according to the insulation impact level, the equipment insulation aging record is read from the historical database, the insulation breakdown risk value is obtained by comparison and calculation, and the corresponding protection plan is triggered according to the risk value.

6. The method according to claim 1, characterized in that The simulation test platform for connecting or disconnecting a large load of the power grid is constructed, and the impact of the surge phenomenon generated by the charging pile on the voltage impact amplitude, current impact amplitude and impact duration of the mutual inductor, metering chip and relay device in the charging pile when the large load of the power grid is connected or disconnected is analyzed, and the damage degree of each device is determined, and a mapping relationship between the device damage degree and the surge phenomenon is established, including: For the load switching test signal generated by the load switching test unit, the voltage and current data at the load switching time are obtained from the power grid monitoring device, and the voltage and current data are used to record the real-time response waveform of the key device port of the charging pile; According to the real-time response waveform, the input-output voltage ratio and phase difference are obtained from the transformer sampling end, the signal distortion is obtained from the metering chip sampling end, and the contact action time is obtained from the relay sampling end; According to the voltage ratio, signal distortion and contact action time, wavelet transform is used to perform noise reduction processing, and a device response characteristic sequence is obtained through data reconstruction; Damage parameter indicators are extracted according to the device response characteristic sequence, wherein the damage parameter indicators include voltage deviation rate, measurement error rate and contact resistance rate, and a device damage assessment criterion is established using a deep learning method.

7. The method according to claim 1, characterized in that When it is determined that a surge occurs in the charging pile, the extent of damage to the internal mutual inductor, metering chip, and relay components of the electric energy meter is determined based on the mapping relationship between the degree of component damage and the surge phenomenon, and the expected life of each component is input into a pre-built electric energy meter life prediction model to obtain the expected life of each component, including: Acquire surge characteristic parameters from a charging pile surge detection unit, and extract performance impact parameters from a device damage mapping database according to the surge characteristic parameters; Generate device performance evaluation data using a deep neural network model for the performance influencing parameters, and establish a device status record table based on the performance evaluation data; The device state record table is analyzed by using a life prediction algorithm to obtain a device performance decay rate, and the device aging trend is obtained through a performance parameter change curve according to the performance decay rate; Constructing a life prediction data set according to the aging trend of the device, and training the life prediction data set using a random forest algorithm to obtain a device life expectancy prediction result; It also includes: establishing a corresponding training sample set according to the preset parameter thresholds of the internal transformer, metering chip, and relay devices of the electric energy meter, learning and classifying the training sample set, obtaining a discrimination model for the degree of damage of each device, inputting the surge characteristics into the trained discrimination model, and obtaining the damage degree assessment results of the transformer, metering chip, and relay.

8. The method according to claim 7, characterized in that According to the preset parameter thresholds of the mutual inductor, metering chip, and relay components in the electric energy meter, a corresponding training sample set is established, the training sample set is studied and classified, a discrimination model of the damage degree of each component is obtained, and the surge feature is input into the trained discrimination model to obtain the damage degree assessment result of the mutual inductor, metering chip, and relay, including: Extract the transformer measurement error value, metering chip accuracy value, and relay contact resistance value from the device parameter database, and generate a device parameter training set through the standard parameter file; Decomposing the device parameter training set by Fourier transform, obtaining amplitude parameters, frequency parameters, and duration parameters from the surge waveform data, and constructing a feature vector matrix according to the surge parameters; Using a deep neural network to perform training according to the eigenvector matrix, obtaining device performance attenuation data through network calculation, and extracting a damage degree index from the performance attenuation data; A support vector machine is used to classify the damage degree index, a discrimination standard is obtained from a preset damage level threshold, a device damage discrimination function is obtained through index mapping, and a device damage assessment result is calculated using the discrimination function.

9. The method according to claim 1, characterized in that: The expected life of each component of the electric energy meter is integrated and calculated to obtain the comprehensive expected life of the electric energy meter. If the comprehensive expected life of the electric energy meter is lower than a preset threshold, an early warning is issued, including: Obtain the expected life data of the transformer, the expected life data of the metering chip and the expected life data of the relay from the electric energy meter status database, establish the component importance score for the expected life data, and obtain the component weight coefficient; Performing a weighted operation based on the device weight coefficient and the expected life data, constructing a life assessment matrix using the device performance attenuation curve, and extracting a device life feature sequence from the life assessment matrix; Establishing a life association data table for the device life feature sequence, and training the life association data table using a deep neural network to obtain a device life association value; The device combination lifetime is calculated according to the device lifetime correlation value, a support vector regression method is used to perform data fitting on the device combination lifetime, a device state sequence is generated from the data fitting result, and a lifetime prediction is performed on the device state sequence.

Images