Electric energy metering box performance test system based on data analysis

Abstract

The invention belongs to the technical field of electric energy metering box performance testing, and particularly relates to an electric energy metering box performance testing system based on data analysis, which comprises a multi-dimensional data acquisition module, a self-adaptive preprocessing module, a multi-modal feature fusion module, a dynamic health assessment module, an intelligent decision output module and an operation supervision center. The multi-dimensional data acquisition module outputs an original signal, the original signal is corrected by the self-adaptive preprocessing module and then flows into the multi-modal feature fusion module to generate a high-order feature, the dynamic health assessment module generates a health index based on a feature vector, and a data acquisition strategy is reversely regulated according to an assessment result. A complete test and analysis link is formed from original data collection to final maintenance decision generation, quantitative characterization and trend pre-judgment of the performance of the electric energy metering box are achieved, the real-time performance is guaranteed, meanwhile, the long-term state pre-estimation precision is improved, and the intelligent level and the automation level are high.

Description

Technical Field

[0001] This invention relates to the field of performance testing technology for electricity metering boxes, specifically a performance testing system for electricity metering boxes based on data analysis. Background Technology

[0002] An electricity metering box is a device that houses one or more electricity meters, transformers, and related secondary circuits and terminals in a closed or semi-closed enclosure. It provides a safe, reliable, and easy-to-manage and maintain working environment for electricity metering equipment and plays a vital role in the power system.

[0003] Traditional power metering box performance testing relies on manual inspection or single parameter monitoring, which has the drawbacks of high data bias, delayed anomaly identification, and weak ability to suppress environmental interference. Although relevant data acquisition schemes have been proposed, they do not involve multi-parameter coupling analysis and dynamic evaluation mechanisms. Furthermore, the testing process and data analysis are separated, which cannot meet the real-time performance testing and evaluation needs of power metering boxes under complex operating conditions, and the level of intelligence and automation is low.

[0004] To address the aforementioned technical shortcomings, a solution is proposed. Summary of the Invention

[0005] The purpose of this invention is to provide a data analysis-based performance testing system for electricity metering boxes. This system solves the problems of existing technologies that do not involve multi-parameter coupling analysis and dynamic evaluation mechanisms, and whose testing process is separated from the data analysis process, thus failing to meet the real-time performance testing and evaluation needs of electricity metering boxes under complex operating conditions, and having low levels of intelligence and automation.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] The data analysis-based power metering box performance testing system includes a multi-dimensional data acquisition module, an adaptive preprocessing module, a multi-modal feature fusion module, a dynamic health assessment module, an intelligent decision output module, and an operation monitoring center. The multi-dimensional data acquisition module synchronously collects electrical parameters, environmental parameters, and electromagnetic compatibility parameters through a composite sensor array deployed inside and outside the metering box, and sends the collected raw data to the adaptive preprocessing module.

[0008] After receiving the raw data, the adaptive preprocessing module preprocesses it and assigns quality labels to the preprocessed data, which are then synchronously transmitted to the multimodal feature fusion module. Suspicious data is marked but still participates in subsequent calculations, while invalid data is discarded directly. After receiving the preprocessed data, the multimodal feature fusion module constructs a time-frequency joint feature set and transmits the fused feature vector to the dynamic health assessment module through analysis and processing.

[0009] The dynamic health assessment module loads a pre-trained fuzzy inference engine. After inputting the feature vector, the fuzzy inference engine performs state mapping through multi-level fuzzy rules. First, it determines whether a threshold alarm has been triggered. If not, it enters the trend prediction branch, uses a long short-term memory network to predict the future performance degradation trajectory, and performs a weighted calculation based on the current state value and the predicted degradation rate to generate a dynamic health index. The health assessment results are then transmitted to the intelligent decision output module. The intelligent decision output module automatically matches maintenance strategies based on the health assessment results and sends them to the operation monitoring center.

[0010] Furthermore, the sensors involved in the multi-dimensional data acquisition module adopt a non-contact installation design, which is fixed to the surface of the box by a magnetic base without physical intervention on the original circuit. The data sampling frequency is dynamically adjusted according to the parameter type, with electrical parameters using 1kHz high-frequency sampling and environmental parameters using 10Hz low-frequency sampling.

[0011] Furthermore, the adaptive preprocessing module operates as follows:

[0012] After receiving the raw data, a spatiotemporal alignment operation is first performed to keep the data from multiple sensors synchronized in the time dimension. Then, trend analysis is performed on environmental parameters including temperature and humidity. When a sudden change is detected, the data validity verification process is triggered, suspicious data points are automatically marked, and wavelet packet decomposition is performed on the electromagnetic interference data to extract pulse group features in a specific frequency band. An interference attenuation model is established in combination with the dielectric constant of the enclosure material to correct the coupling effect of electromagnetic interference on the measurement of electrical parameters.

[0013] Furthermore, the analysis and processing procedure of the multimodal feature fusion module is as follows:

[0014] Short-time Fourier transform is performed on voltage and current waveforms to extract time-frequency features including the proportion of harmonic components and waveform distortion rate. Envelope demodulation is performed on vibration signals to identify the characteristic frequencies of loose mechanical parts. A parameter correlation matrix is ​​established to analyze the coupling relationship between electrical parameters and environmental parameters. An improved principal component analysis method is used to reduce the dimensionality of the feature set, retaining principal components with high variance contribution rates to generate feature vectors.

[0015] Furthermore, the specific operation process of the intelligent decision output module is as follows:

[0016] When the dynamic health index is at an unqualified level, an emergency maintenance process is generated and the specific fault location and handling plan are output. When the dynamic health index is at a medium level, the enhanced monitoring mode is activated and the sampling frequency of key parameters is adjusted. When the dynamic health index is at a good level, a routine inspection plan is generated. A visual test report is also generated, including a three-dimensional heat map, a spectrum waterfall chart and a maintenance priority ranking table.

[0017] Furthermore, the operation monitoring center communicates with the operation judgment module to obtain all sensors involved in the multi-dimensional data acquisition module, and marks the corresponding sensor as i, where i is a natural number greater than 1; through sensor evaluation and analysis, sensor i is marked as a superior operation object or a rebellious operation object, and the marking information of sensor i is sent to the operation monitoring center.

[0018] Furthermore, the specific analysis process for sensor evaluation is as follows:

[0019] The system acquires all the times when sensor i collects data within a unit time period, calculates the time difference between two adjacent sets of data collection times to obtain the sampling difference value, and calculates the difference between the sampling difference value and the set standard collection interval duration and takes the absolute value to obtain the collection anomaly value.

[0020] The number of times an abnormal value exceeds the corresponding preset abnormal value within a unit of time is obtained and marked as an abnormal frequency value. The average value of all abnormal values ​​within a unit of time is calculated to obtain the abnormal table value. The abnormal value with the largest value within a unit of time is marked as the abnormal amplitude value.

[0021] The acquisition feature value is obtained by weighted summation of the acquired frequency value, acquired meter value, and acquired amplitude value. The acquired feature value is compared with the corresponding preset acquisition feature threshold. If the acquired feature value exceeds the preset acquisition feature threshold, sensor i is marked as a blocking operation object; if the acquired feature value does not exceed the preset acquisition feature threshold, sensor i is marked as a high-performance operation object.

[0022] Furthermore, the acquisition and operation judgment module communicates with the acquisition and coordination analysis module. The acquisition and operation judgment module sends the tag information of all sensors to the acquisition and coordination analysis module. If the acquisition and coordination analysis module receives an obstruction object, it generates an acquisition and coordination failure signal.

[0023] If no obstruction object is received, a sampling failure signal or a sampling failure signal is generated through comprehensive analysis of the collected data. The sampling failure signal or the sampling failure signal is then sent to the operation monitoring center. When the operation monitoring center receives the sampling failure signal, it issues a corresponding warning.

[0024] Furthermore, the specific analysis process for data collection and comprehensive analysis is as follows:

[0025] Set a detection period, and set several analysis periods within the detection period. If a sensor malfunctions during the corresponding analysis period, the corresponding analysis period is marked as a fault period; if no sensor malfunctions during the corresponding analysis period, the corresponding analysis period is marked as a stable period. Obtain the number of fault periods within the detection period and mark them as fault detection values. Compare the fault detection values ​​with a preset fault detection threshold. If the fault detection value exceeds the preset fault detection threshold, a sampling failure signal is generated.

[0026] If the fault detection value does not exceed the preset fault detection threshold, the number of fault periods between two adjacent stable periods is marked as the characteristic duration value, the number of characteristic duration values ​​exceeding the preset characteristic duration threshold during the detection period is marked as the characteristic frequency value, and the characteristic duration value with the largest value during the detection period is marked as the characteristic amplitude value.

[0027] The initial analysis value is obtained by weighted summation of fault detection value, characteristic frequency value and characteristic amplitude value. The initial analysis value is then compared with the preset initial analysis threshold. If the initial analysis value exceeds the preset initial analysis threshold, a poor analysis signal is generated.

[0028] Furthermore, if the initial sampling and analysis value does not exceed the preset initial sampling and analysis threshold, the interval between the production time of sensor i and the current time is obtained and marked as the sensor production value, and the total duration of sensor i in the running state in the historical stage is marked as the sensor running value.

[0029] And obtain various environmental parameters of the environment where sensor i is located. If there are environmental parameters that do not meet the corresponding preset requirements, it is determined that sensor i is in a state of quality damage. Obtain the total duration of sensor i in a state of quality damage in the historical period and mark it as the sensor damage value.

[0030] The sensor detection value is obtained by weighted summation of sensor production value, sensor operation value and sensor damage value. The sensor detection value is then compared with the corresponding preset sensor detection threshold. If the sensor detection value exceeds the corresponding preset sensor detection threshold, sensor i is marked as an alarm object.

[0031] The system acquires the percentage of alarm objects involved in the multi-dimensional data acquisition module and marks it as the alarm status value. It then calculates the sensor status value by comparing the sensor detection value with the corresponding preset sensor detection threshold, and calculates the average of the sensor status values ​​of all sensors to obtain the sensor comprehensive measurement value. The alarm status value and the sensor comprehensive measurement value are compared with the preset alarm status threshold and the preset sensor comprehensive measurement threshold, respectively. If the alarm status value or the sensor comprehensive measurement value exceeds the corresponding preset threshold, a poor acquisition signal is generated. If neither the alarm status value nor the sensor comprehensive measurement value exceeds the corresponding preset threshold, a no-acquisition signal is generated.

[0032] Compared with the prior art, the beneficial effects of the present invention are:

[0033] 1. In this invention, the original signal is output by the multi-dimensional data acquisition module, corrected by the adaptive preprocessing module, and then flows into the multi-modal feature fusion module to generate high-order features. A health index is generated based on the feature vector, and the data acquisition strategy is adjusted in reverse according to the evaluation results. From the original data acquisition to the final maintenance decision generation, a complete test and analysis link is formed, realizing the quantitative characterization and trend prediction of the performance of the power metering box, ensuring real-time performance while improving the accuracy of long-term state prediction, with a high level of intelligence and automation.

[0034] 2. In this invention, all sensors involved in the multi-dimensional data acquisition module are obtained through the acquisition and operation judgment module. Based on sensor evaluation and analysis, the corresponding sensors are marked as high-performance or low-performance objects to remind managers to check and adjust the corresponding sensors to ensure their acquisition frequency performance. Furthermore, the acquisition and matching analysis module analyzes the data to generate acquisition and matching failure signals or acquisition and matching failure signals. When acquisition and matching failure signals are generated, sensor monitoring is strengthened and the corresponding sensors are replaced or repaired in a timely manner to ensure sensor matching performance, which is beneficial to the accuracy and stability of subsequent data acquisition. Attached Figure Description

[0035] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings;

[0036] Figure 1 This is a system block diagram of Embodiment 1 of the present invention;

[0037] Figure 2 This is a system block diagram of Embodiments 2 and 3 of the present invention. Detailed Implementation

[0038] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0039] Example 1: As Figure 1 As shown, the data analysis-based power metering box performance testing system proposed in this invention includes a multi-dimensional data acquisition module, an adaptive preprocessing module, a multi-modal feature fusion module, a dynamic health assessment module, an intelligent decision output module, and an operation monitoring center.

[0040] The multidimensional data acquisition module uses a composite sensor array deployed inside and outside the metering box to simultaneously collect electrical parameters (including voltage, current, power factor, etc.), environmental parameters (including temperature, humidity, vibration acceleration, etc.), and electromagnetic compatibility parameters (including power frequency magnetic field strength, electrostatic discharge amplitude, etc.). After assigning a timestamp to the collected raw data, it sends it to the adaptive preprocessing module.

[0041] The sensors involved in the multi-dimensional data acquisition module adopt a non-contact installation design, which is fixed to the surface of the box by a magnetic base to avoid physical interference with the original circuit. The data sampling frequency is dynamically adjusted according to the parameter type. Electrical parameters are sampled at a high frequency of 1kHz (that is, electrical parameters are sampled at a high frequency to ensure the capture of waveform details), while environmental parameters are sampled at a low frequency of 10Hz (environmental parameters are sampled at a low frequency to balance data volume and real-time performance).

[0042] The adaptive preprocessing module receives the raw data, preprocesses it, assigns quality labels (valid / suspicious / invalid) to the preprocessed data, and synchronously transmits it to the multimodal feature fusion module. It also feeds back to the data acquisition module to adjust the sampling strategy. Suspicious data, after being labeled, still participates in subsequent calculations, while invalid data is discarded. The operation process of the adaptive preprocessing module is as follows:

[0043] After receiving the raw data, a spatiotemporal alignment operation is first performed to keep the data from multiple sensors synchronized in the time dimension. Then, trend analysis is performed on environmental parameters such as temperature and humidity. When a sudden change is detected, the data validity verification process is triggered, suspicious data points are automatically marked, and wavelet packet decomposition is performed on electromagnetic interference data to extract pulse group features in a specific frequency band. An interference attenuation model is established in combination with the dielectric constant of the enclosure material to correct the coupling effect of electromagnetic interference on electrical parameter measurement.

[0044] After receiving the preprocessed data, the multimodal feature fusion module constructs a time-frequency joint feature set, and transmits the fused feature vector to the dynamic health assessment module through analysis and processing. The analysis and processing process of the multimodal feature fusion module is as follows:

[0045] Short-time Fourier transform is performed on voltage and current waveforms to extract time-frequency features including the proportion of harmonic components and waveform distortion rate. Envelope demodulation is performed on vibration signals to identify the characteristic frequencies of loose mechanical parts. A parameter correlation matrix is ​​established to analyze the coupling relationship between electrical parameters and environmental parameters. An improved principal component analysis method is used to reduce the dimensionality of the feature set, retaining principal components with high variance contribution rates to generate feature vectors.

[0046] The dynamic health assessment module loads a pre-trained fuzzy inference engine. After inputting the feature vector, the fuzzy inference engine performs state mapping through multi-level fuzzy rules. First, it determines whether a threshold alarm is triggered. If not, it enters the trend prediction branch and uses a long short-term memory network to predict the future performance degradation trajectory. It then performs a weighted calculation by combining the current state value and the predicted degradation rate to generate a dynamic health index (preferredly, the dynamic health index (DHI) = current state value × 0.7 + predicted degradation rate × 0.3). The health assessment results are then transmitted to the intelligent decision output module.

[0047] The intelligent decision-making output module automatically matches maintenance strategies based on health assessment results and sends them to the operation monitoring center. The specific operation process is as follows: when the dynamic health index is at an unqualified level, an emergency maintenance process is generated and the specific fault location and handling plan are output. When the dynamic health index is at a medium level, the enhanced monitoring mode is activated and the sampling frequency of key parameters is adjusted. When the dynamic health index is at a good level, a routine inspection plan is generated. A visual test report is also generated, which includes a three-dimensional heat map, a spectrum waterfall chart, and a maintenance priority ranking table.

[0048] The technical solution of this invention outputs raw signals through a multi-dimensional data acquisition module, which are then corrected by an adaptive preprocessing module and fed into a multi-modal feature fusion module to generate high-order features. A dynamic health assessment module generates a health index based on the feature vectors, and an intelligent decision output module adjusts the data acquisition strategy in reverse according to the assessment results, forming a closed-loop control. The modules communicate asynchronously through message queues to ensure data integrity in high-concurrency scenarios. From raw data acquisition to final maintenance decision generation, a complete testing and analysis link is formed, realizing the quantitative characterization and trend prediction of meter box performance. While ensuring real-time performance, it improves the accuracy of long-term state prediction, filling the gap in dynamic health management technology in this field.

[0049] Example 2: Figure 2As shown, the difference between this embodiment and Embodiment 1 is that the operation monitoring center communicates with the operation judgment module to collect data. The operation judgment module obtains all the sensors involved in the multi-dimensional data acquisition module, marks the corresponding sensor as i, and i is a natural number greater than 1; through sensor evaluation and analysis, sensor i is marked as a superior operation object or a rebellious operation object.

[0050] Furthermore, the tagging information of sensor i is sent to the operation monitoring center to remind management personnel to check and adjust the corresponding sensor to ensure its data acquisition frequency performance and avoid affecting the subsequent analysis process; the specific analysis process of sensor evaluation and analysis is as follows:

[0051] The system acquires all data collection moments of sensor i within a unit time period, calculates the time difference between two adjacent data collection moments to obtain the sampling difference value, and calculates the difference between the sampling difference value and the set standard collection interval duration, taking the absolute value to obtain the collection anomaly value; where, the larger the value of the collection anomaly value, the less accurate the collection interval time of the corresponding collection process is.

[0052] The number of times an abnormal value exceeds the corresponding preset abnormal value within a unit of time is obtained and marked as an abnormal frequency value. The average value of all abnormal values ​​within a unit of time is calculated to obtain the abnormal table value. The abnormal value with the largest value within a unit of time is marked as the abnormal amplitude value.

[0053] The acquisition feature value is obtained by weighted summation of the acquired frequency value, acquired table value, and acquired amplitude value. Specifically, the acquired frequency value, acquired table value, and acquired amplitude value are each assigned a corresponding preset weight coefficient, and the acquired frequency value, acquired table value, and acquired amplitude value are each multiplied by the corresponding preset weight coefficient. The sum of the three sets of product results is marked as the acquisition feature value.

[0054] It should be noted that the larger the value of the acquired feature, the worse the overall performance of the acquisition frequency per unit time. The acquired feature value is compared with the corresponding preset acquisition feature threshold. If the acquired feature value exceeds the preset acquisition feature threshold, it indicates that the overall performance of the acquisition frequency per unit time is poor, and sensor i is marked as a blocked operation object. If the acquired feature value does not exceed the preset acquisition feature threshold, it indicates that the overall performance of the acquisition frequency per unit time is good, and sensor i is marked as a superior operation object.

[0055] Example 3: Figure 2As shown, the difference between this embodiment and Embodiment 1 and Embodiment 2 is that the acquisition and operation judgment module is connected to the acquisition and coordination analysis module. The acquisition and operation judgment module sends the marking information of all sensors to the acquisition and coordination analysis module. If the acquisition and coordination analysis module receives an obstruction object, it generates an acquisition and coordination failure signal.

[0056] If no obstruction object is received, a data acquisition and coordination analysis is performed to generate a data acquisition failure signal or a data acquisition failure signal. This signal is then sent to the operation monitoring center. Upon receiving the data acquisition failure signal, the operation monitoring center issues a corresponding warning to remind management personnel to strengthen sensor monitoring and promptly replace or repair the corresponding sensors to ensure sensor performance and improve the accuracy and stability of subsequent data acquisition. The specific analysis process of data acquisition and coordination analysis is as follows:

[0057] The detection period is set, preferably seven days. Several analysis periods are set within the detection period. If a sensor malfunction occurs within the corresponding analysis period, the corresponding analysis period is marked as a fault period. If no sensor malfunction occurs within the corresponding analysis period, the corresponding analysis period is marked as a stable period. The number of fault periods within the detection period is obtained and marked as fault detection values. The fault detection values ​​are compared with a preset fault detection threshold. If the fault detection value exceeds the preset fault detection threshold, a poor acquisition signal is generated.

[0058] If the fault detection value does not exceed the preset fault detection threshold, the number of fault periods between two adjacent stable periods is marked as the characteristic duration value, the number of characteristic duration values ​​exceeding the preset characteristic duration threshold during the detection period is marked as the characteristic frequency value, and the characteristic duration value with the largest value during the detection period is marked as the characteristic amplitude value.

[0059] The preliminary analysis value of the sampling and distribution is obtained by weighted summation of the fault condition value, characteristic frequency value, and characteristic amplitude value. Specifically, the fault condition value, characteristic frequency value, and characteristic amplitude value are assigned corresponding preset weight coefficients, and the fault condition value, characteristic frequency value, and characteristic amplitude value are multiplied by the corresponding preset weight coefficients, and the sum of the three sets of product results is marked as the preliminary analysis value of the sampling and distribution.

[0060] It should be noted that the larger the initial sampling analysis value, the worse the overall stability of sensor acquisition during the detection period. The initial sampling analysis value is compared with the preset initial sampling analysis threshold. If the initial sampling analysis value exceeds the preset initial sampling analysis threshold, it indicates that the overall stability of sensor acquisition during the detection period is poor, and a poor sampling signal is generated.

[0061] Furthermore, if the initial sampling and analysis value does not exceed the preset initial sampling and analysis threshold, the interval between the production time of sensor i and the current time is obtained and marked as the sensor production value, and the total duration of sensor i in the running state in the historical stage is marked as the sensor running value.

[0062] And obtain various environmental parameters of the environment where sensor i is located. If there are environmental parameters that do not meet the corresponding preset requirements, it is determined that sensor i is in a state of quality damage. Obtain the total duration of sensor i in a state of quality damage in the historical period and mark it as the sensor damage value.

[0063] The sensor detection value is obtained by weighted summation of sensor production value, sensor operation value and sensor damage value. Specifically, the sensor production value, sensor operation value and sensor damage value are each assigned a corresponding preset weight coefficient, and the sensor production value, sensor operation value and sensor damage value are multiplied by the corresponding preset weight coefficient. The sum of the three sets of product results is marked as the sensor detection value.

[0064] It should be noted that the larger the value of the sensor detection, the worse the current quality of sensor i is, and the more difficult it is to ensure continuous and stable operation. The sensor detection value is compared with the corresponding preset sensor detection threshold. If the sensor detection value exceeds the corresponding preset sensor detection threshold, it indicates that the current quality of sensor i is poor and it is difficult to ensure continuous and stable operation. In this case, sensor i is marked as an alarm object.

[0065] The number of alarm objects involved in the multi-dimensional data acquisition module is obtained as the proportion value and marked as the alarm status value. The sensor detection value is calculated by comparing the sensor detection value with the corresponding preset sensor detection threshold. The sensor comprehensive measurement value is calculated by averaging the sensor detection values ​​of all sensors.

[0066] The alarm status value and sensor comprehensive measurement value are compared with the preset alarm status threshold and preset sensor comprehensive measurement threshold respectively. If the alarm status value or sensor comprehensive measurement value exceeds the corresponding preset threshold, it indicates that the stability of sensor acquisition is difficult to guarantee, and a poor acquisition signal is generated. If the alarm status value and sensor comprehensive measurement value do not exceed the corresponding preset threshold, it indicates that the stability of sensor acquisition is guaranteed, and a no acquisition signal is generated.

[0067] The working principle of this invention is as follows: In use, the original signal is output through the multi-dimensional data acquisition module, corrected by the adaptive preprocessing module, and then flows into the multi-modal feature fusion module to generate high-order features. The dynamic health assessment module generates a health index based on the feature vector, and the intelligent decision output module adjusts the data acquisition strategy in reverse according to the assessment results. From the original data acquisition to the final maintenance decision generation, a complete test and analysis link is formed, realizing the quantitative characterization and trend prediction of the performance of the power metering box. While ensuring real-time performance, it improves the accuracy of long-term state prediction and has a high level of intelligence and automation.

[0068] In this invention, the threshold, preset value, or preset range settings are for result comparison and analysis to determine whether the result is good or bad. The magnitude of these values ​​is determined by a combination of large-scale model analysis of sample data and human experience, and can also be appropriately adjusted based on seasonal or common-sense influence conditions. Similarly, the preset weight coefficients and influence factors are assigned specific values ​​based on the magnitude of each parameter's influence on the result, ultimately reflecting the impact on the result. These settings are also determined by a combination of large-scale model analysis of sample data and human experience, and can also be appropriately adjusted based on seasonal or common-sense influence conditions.

[0069] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, enabling those skilled in the art to better understand and utilize it. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A data analysis based performance testing system for electric energy metering box, characterized in that, The system comprises a multi-dimensional data acquisition module, an adaptive preprocessing module, a multi-modal feature fusion module, a dynamic health assessment module, an intelligent decision output module and an operation supervision center; the multi-dimensional data acquisition module synchronously acquires electric parameters, environmental parameters and electromagnetic compatibility parameters through a composite sensor array deployed inside and outside the metering box; After the adaptive preprocessing module receives the original data, it performs preprocessing on the original data, and after the multi-modal feature fusion module receives the preprocessed data, it constructs a time-frequency joint feature set, and transmits the fused feature vector to the dynamic health assessment module through analysis and processing; The dynamic health assessment module loads a pre-trained fuzzy reasoning engine, and after inputting the feature vector, the fuzzy reasoning engine performs state mapping through multi-level fuzzy rules, first determines whether a threshold alarm is triggered, and if not, enters a trend prediction branch, uses a long short-term memory network to predict the future performance degradation trajectory, and performs weighted calculation on the current state value and the predicted degradation speed to generate a dynamic health index, and the health assessment result is transmitted to the intelligent decision output module; the intelligent decision output module automatically matches a maintenance strategy according to the health assessment result and sends it to the operation supervision center.

2. The data analytics based performance testing system for electricity metering box of claim 1, wherein, The sensors involved in the multi-dimensional data acquisition module are designed with non-contact installation, fixed on the surface of the box through magnetic base, and the data sampling frequency is dynamically adjusted according to the parameter type, with 1kHz high-frequency sampling for electric parameters and 10Hz low-frequency sampling for environmental parameters.

3. The data analytics based performance testing system for electric energy metering box of claim 1, wherein, The operation process of the adaptive preprocessing module is as follows: After receiving the original data, first perform time-space alignment to keep the multi-sensor data synchronized in the time dimension; then perform trend analysis on the environmental parameters including temperature and humidity, and when a sudden value is detected, trigger the data validity verification process, automatically mark the suspicious data points, and perform wavelet packet decomposition on the electromagnetic interference data to extract the pulse group features in the specific frequency band, establish an interference attenuation model based on the dielectric constant of the box material, and correct the coupling effect of electromagnetic interference on electric parameter measurement.

4. The data analytics based electric energy metering box performance testing system as claimed in claim 1, wherein, The analysis and processing process of the multi-modal feature fusion module is as follows: Perform short-time Fourier transform on the voltage and current waveforms to extract time-frequency features including harmonic component proportion and waveform distortion rate, and perform envelope demodulation on the vibration signal to identify the mechanical component loosening characteristic frequency, and establish a parameter correlation matrix to analyze the coupling relationship between electric parameters and environmental parameters, and use improved principal component analysis to reduce the dimension of the feature set, and retain the principal components with high variance contribution rate to generate a feature vector.

5. The data analytics based electric energy metering box performance testing system as claimed in claim 1, wherein, The specific operation process of the intelligent decision output module is as follows: When the dynamic health index is at an unqualified level, generate an emergency maintenance process and output the specific fault location and disposal scheme, when the dynamic health index is at a medium level, start the enhanced monitoring mode and adjust the sampling frequency of the key parameters, when the dynamic health index is at a good level, generate a routine inspection plan; And generate a visual test report, including a three-dimensional heat map, a spectrum waterfall chart and a maintenance priority ranking table.

6. The data analytics based electric energy metering box performance testing system as claimed in claim 1, wherein, The operation supervision center communication connection acquisition operation one-by-one judgment module obtains all sensors involved by the multi-dimensional data acquisition module, marks the corresponding sensor as i, and i is a natural number greater than 1; The sensor i is marked as an optimal operation object or a blocked operation object through sensor evaluation analysis, and the marking information of the sensor i is sent to the operation supervision center.

7. The data analytics based performance testing system of electric energy metering box of claim 6, wherein, The specific analysis process of the sensor evaluation analysis is: obtaining the number of acquisition abnormal values exceeding the corresponding preset acquisition abnormal threshold value in a unit time, marking it as an acquisition abnormal frequency value, performing mean value calculation on all acquisition abnormal values in a unit time to obtain an acquisition abnormal value, and marking the maximum acquisition abnormal value in a unit time as an acquisition abnormal amplitude value; The acquisition characteristic value is calculated by weighted summation of the acquisition abnormal frequency value, the acquisition abnormal value and the acquisition abnormal amplitude value, if the acquisition characteristic value exceeds the preset acquisition characteristic threshold value, the sensor i is marked as a blocked operation object; Otherwise, the sensor i is marked as an optimal operation object.

8. The data analytics based performance testing system of electric energy metering box of claim 6, wherein, The acquisition operation one-by-one judgment module is communication connected with the acquisition cooperation analysis module, the acquisition operation one-by-one judgment module sends the marking information of all sensors to the acquisition cooperation analysis module, if the acquisition cooperation analysis module receives a blocked operation object, an acquisition cooperation bad signal is generated; if no blocked operation object is received, acquisition cooperation comprehensive analysis is performed to generate an acquisition cooperation bad signal or an acquisition cooperation no abnormal signal, and the acquisition cooperation bad signal or the acquisition cooperation no abnormal signal is sent to the operation supervision center.

9. The data analytics based electric energy metering box performance testing system as claimed in claim 8, wherein, The specific analysis process of the acquisition cooperation comprehensive analysis is: obtaining the number of fault periods in the detection period and marking it as a fault detection condition value, if the fault detection condition value exceeds the preset fault detection condition threshold value, an acquisition cooperation bad signal is generated; If the fault detection condition value does not exceed the preset fault detection condition threshold value, the acquisition initial analysis value is calculated by weighted summation of the fault detection condition value, the characteristic abnormal frequency value and the characteristic abnormal amplitude value, if the acquisition initial analysis value exceeds the preset acquisition initial analysis threshold value, an acquisition cooperation bad signal is generated.

10. The data analytics based performance testing system of electric energy metering box of claim 9, wherein, If the acquisition initial analysis value does not exceed the preset acquisition initial analysis threshold value, the number of alarm objects involved by the multi-dimensional data acquisition module is obtained, and the number of alarm objects is marked as an alarm occupation condition value, and the sensor detection occupation value of all sensors is calculated to obtain a sensor comprehensive measurement value; if the alarm occupation condition value or the sensor comprehensive measurement value exceeds the corresponding preset threshold value, an acquisition cooperation bad signal is generated; otherwise, an acquisition cooperation no abnormal signal is generated.

Images