Intelligent diagnosis method for analyzing line hidden danger based on dynamic electrical fingerprint characteristics

Abstract

The invention discloses an intelligent diagnosis method for analyzing line hidden dangers based on dynamic electrical fingerprint characteristics, and belongs to the technical field of power line state monitoring and fault prediction. According to the method, a dynamic electrical fingerprint is constructed through high-frequency acquisition of line voltage and current instantaneous value signals and extraction of multi-dimensional features such as a time domain, a frequency domain and a time-frequency domain; establishing an electrical fingerprint database and a health model of different load working conditions in a line health state; during on-line monitoring, through three-level analysis of deviation comparison, state classification and trend prediction, graded early warning of line hidden dangers is realized, and abnormal feature items are output for auxiliary diagnosis. According to the invention, subtle changes of electrical parameters can be captured, the problems of low sensitivity, weak anti-interference capability and incapability of early warning in the prior art are solved, early recognition and advanced prediction of latent hidden dangers are realized, and the safety and reliability of a power system are significantly improved.

Description

Technical Field

[0001] This application belongs to the technical field of power line condition monitoring and fault prediction, and particularly relates to an intelligent diagnostic method for line hidden dangers based on dynamic electrical fingerprint feature analysis. Background Technology

[0002] In critical facilities such as ships, aerospace, and data centers, the operational reliability of power lines directly determines the safe and stable operation of the entire system. Over long-term use, power lines are prone to latent problems such as poor contact, minor insulation damage, joint oxidation, and abnormal loads. If these problems are not identified and addressed early, they will gradually develop into serious faults such as short circuits, open circuits, and electrical fires, causing significant losses including equipment damage, system shutdowns, and even personal injury or death.

[0003] Current mainstream line monitoring methods have the following shortcomings: First, traditional monitoring methods focus on macroscopic electrical parameters such as current and voltage RMS values, and can only issue alarms when parameters exceed rated thresholds, failing to capture subtle and dynamic changes in electrical parameters during the development of potential hazards. Second, while periodic inspections can identify some problems by measuring insulation resistance, they suffer from fixed inspection cycles and delayed responses, making it difficult to track the development of potential hazards in real time. Taking loose terminal blocks as an example, before a fault occurs, the contact resistance slowly increases, causing slight distortions in the current waveform, but the RMS current value remains within the normal range. Traditional monitoring methods cannot identify such early anomalies.

[0004] In recent years, diagnostic technologies based on current harmonic monitoring and line temperature monitoring have emerged in the industry. However, these technologies are susceptible to interference from external factors such as load type fluctuations and changes in ambient temperature and humidity, resulting in high false alarm rates. They are unable to meet the high-precision monitoring requirements of critical facilities, and are particularly insensitive to slowly developing latent hazards such as poor contact and insulation aging, failing to provide early warnings. "Electrical fingerprints" are a unique and repeatable set of electrical characteristics exhibited by lines or electrical equipment under specific operating conditions, and their changes are directly related to the line's condition. How to scientifically extract electrical fingerprint feature sets, establish a mapping relationship between features and line conditions, and achieve early warning of potential hazards through long-term trend analysis has become a core technical challenge that urgently needs to be overcome in the field of power line condition monitoring. Summary of the Invention

[0005] The purpose of this application is to provide an intelligent diagnostic method for line hazards based on dynamic electrical fingerprint feature analysis. By constructing a dynamic electrical fingerprint of the line and monitoring the trend of feature changes, it can achieve accurate identification and early warning of hazards in their early stages, thereby solving the technical problems of low sensitivity, weak anti-interference ability, and inability to achieve early warning in existing line monitoring technologies.

[0006] To achieve the above-mentioned objectives, the technical solution adopted in this application is as follows: In a first aspect, embodiments of this application provide an intelligent diagnostic method for potential circuit hazards based on dynamic electrical fingerprint feature analysis, comprising the following steps: T1: High-frequency acquisition of instantaneous voltage and current signals of the line under monitoring, and preprocessing of the acquired raw signals; T2: Extract time-domain features, frequency-domain features, time-frequency-domain features, power features, and impedance features from the preprocessed signal to construct a dynamic electrical fingerprint feature vector characterizing the instantaneous operating state of the line; T3: Under the condition of line health, simulate different typical load conditions, collect multiple sets of health fingerprint samples, establish the benchmark fingerprint corresponding to each condition, and form an electrical fingerprint library; at the same time, use the health fingerprint samples to train the health status model. T4: During the online monitoring phase, the dynamic electrical fingerprint feature vector of the line is extracted in real time, and the deviation is compared with the benchmark fingerprint under the same working conditions. The health status model is used to classify the status, and the time series of the deviation value is used to predict the trend. T5: Based on the deviation comparison results, state classification results, and trend prediction results, trigger different levels of early warning and output abnormal feature items and contribution.

[0007] As an optional implementation of this embodiment, in step T1, the sampling frequency of high-frequency acquisition is not less than 10kHz; the preprocessing includes digital filtering and denoising processing, the digital filtering adopts finite impulse response filtering, and the denoising processing adopts wavelet threshold denoising.

[0008] As an optional implementation of this embodiment, in step T2, the time-domain features include the skewness, kurtosis, amplitude standard deviation, and zero-crossing rate of the voltage / current waveform; the frequency-domain features include the amplitude and phase of the fundamental and odd harmonics, the total harmonic distortion rate, and the spectral energy distribution of a specific frequency band; the time-frequency domain features are the energy entropy of each sub-band after the current signal is transformed by wavelet packet transformation; the power features are the fluctuation characteristics of instantaneous active power and reactive power; and the impedance features are the equivalent impedance spectrum characteristics of the line during the load stabilization period.

[0009] As an optional implementation of this embodiment, in step T3, typical load conditions include no load, light load, rated load, and 30% overload; K-Means clustering analysis is performed on multiple sets of health fingerprint samples under the same condition, and the cluster centroid is taken as the reference fingerprint; the health status model is trained using a one-class support vector machine (One-ClassSVM) algorithm.

[0010] As an optional implementation of this embodiment, in step T4, the deviation comparison is achieved by calculating the Mahalanobis distance or cosine similarity between the real-time fingerprint feature vector and the reference fingerprint feature vector; the trend prediction uses the ARIMA algorithm or LSTM algorithm to fit the time series of deviation values ​​and predict future values.

[0011] As an optional implementation of this embodiment, in step T5, the warning levels include attention warning, abnormal alarm, and critical warning; the trigger condition for attention warning is that the real-time deviation value exceeds the first threshold L1 and the health status model determines it to be normal; the trigger condition for abnormal alarm is that the real-time deviation value exceeds the second threshold L2 or the health status model determines it to be abnormal; the trigger condition for critical warning is that the trend prediction result shows that the deviation value will exceed the second threshold L2 in the future Δt time; the first threshold L1 and the second threshold L2 are obtained by machine learning algorithm based on historical data training, and L1 <L2。

[0012] As an optional implementation of this embodiment, in step T5, the abnormal feature items and contribution degree are obtained by calculating the contribution weight of each feature to the deviation value. The contribution weight is calculated using a feature importance evaluation algorithm (such as random forest feature importance).

[0013] Secondly, embodiments of this application also provide an intelligent diagnostic system for circuit hazards based on dynamic electrical fingerprint feature analysis, including modules for implementing the method provided in the first aspect above.

[0014] Thirdly, embodiments of this application also provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method provided in the first aspect.

[0015] Fourthly, embodiments of this application also provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method provided in the first aspect above.

[0016] Compared with the prior art, the intelligent diagnostic method for circuit hazards based on dynamic electrical fingerprint feature analysis provided in this application has the following beneficial effects: (1) Extremely high sensitivity and early warning: Through high-frequency sampling and multi-dimensional feature extraction, it can capture subtle electrical waveform distortions that cannot be identified by traditional effective value monitoring, and achieve accurate identification of hidden dangers in the "budding stage". Compared with existing technologies, the warning time is advanced.

[0017] (2) Strong anti-interference capability: The electrical fingerprint database is constructed by adopting the principle of "comparison under the same working conditions" to effectively avoid the interference of load changes on the monitoring results.

[0018] (3) No need for precise modeling and strong versatility: It only relies on the health data of the line itself to establish a benchmark, without the need to obtain specific parameters of the line (such as conductor material and length) and precise load model. It is applicable to AC lines of different voltage levels from 220V to 10kV, with low deployment cost and strong universality.

[0019] (4) High interpretability: When the warning is issued, the abnormal feature items and contribution are output simultaneously, providing clear diagnostic directions for operation and maintenance personnel and shortening the troubleshooting time.

[0020] (5) Support predictive maintenance: The technology upgrade from "condition monitoring" to "trend prediction" is realized through trend prediction algorithm, providing core technical support for predictive maintenance of power system and reducing unplanned outage rate. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This illustration shows a schematic diagram of the overall process of an intelligent diagnostic method for circuit hazards based on dynamic electrical fingerprint feature analysis, provided in an embodiment of this application.

[0023] Figure 2 A schematic diagram of dynamic electrical fingerprint feature extraction provided in an embodiment of this application is shown.

[0024] Figure 3 The diagram illustrates the principle of establishing an electrical fingerprint database and conducting online monitoring and comparison, as provided in an embodiment of this application. Detailed Implementation

[0025] To make the technical problems, technical solutions, and beneficial effects to be solved by this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and are not intended to limit the scope of this application.

[0026] It should be noted that when a component is referred to as being "fixed to" or "set on" another component, it can be directly on or indirectly on that other component. When a component is referred to as being "connected to" another component, it can be directly connected to or indirectly connected to that other component.

[0027] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0028] This invention belongs to the field of power line condition monitoring and fault prediction technology, and relates to an intelligent diagnostic method for potential line hazards based on dynamic electrical fingerprint feature analysis. By analyzing subtle changes in the electrical parameters of the line, this intelligent diagnostic method can achieve early identification of potential hazards and is applicable to power line monitoring scenarios of critical facilities such as ships, aerospace, and data centers.

[0029] To illustrate the technical solution described in this application, specific embodiments are provided below.

[0030] Please see Figures 1 to 3 As shown, this application provides an intelligent diagnostic method for circuit hazards based on dynamic electrical fingerprint feature analysis. Specifically, Figure 1 The overall flowchart of the method is shown, covering five core steps: signal acquisition, feature extraction, baseline establishment, online monitoring, and early warning diagnosis. Figure 2 A schematic diagram of dynamic electrical fingerprint feature extraction is shown, demonstrating the extraction process and data flow of features in the time domain, frequency domain, and time-frequency domain. Figure 3 The diagram illustrates the principle of establishing an electrical fingerprint database and online monitoring and comparison, explaining the generation method of the reference fingerprint under different load conditions and the comparison logic between the real-time fingerprint and the reference fingerprint.

[0031] Specifically, the intelligent diagnostic method for potential circuit hazards based on dynamic electrical fingerprint feature analysis includes the following steps: T1: High-frequency acquisition of instantaneous voltage and current signals of the line under monitoring, and preprocessing of the acquired raw signals.

[0032] Step T1 involves high-frequency acquisition and preprocessing of multi-dimensional electrical signals. For example, a high-frequency data acquisition device is deployed at the incoming end of the line to be monitored, synchronously acquiring the instantaneous voltage value sequence u(t) and the instantaneous current value sequence i(t) of the line at a sampling frequency of not less than 10kHz (far higher than the power frequency of 50Hz). The acquired raw signals are then subjected to digital filtering (such as finite impulse response filtering) and denoising processing (such as wavelet thresholding) to eliminate electromagnetic interference and noise signals introduced during high-frequency sampling, ensuring the accuracy of subsequent feature extraction.

[0033] T2: Extract time-domain features, frequency-domain features, time-frequency-domain features, power features, and impedance features from the preprocessed signal to construct a dynamic electrical fingerprint feature vector that characterizes the instantaneous operating state of the line.

[0034] Step T2 involves the construction and extraction of a dynamic electrical fingerprint feature set. For example, multi-dimensional and multi-scale feature extraction is performed on the preprocessed voltage and current sequences to construct a dynamic electrical fingerprint feature set F that comprehensively characterizes the current operating state of the line. Specifically, this includes: Time-domain characteristics: encompassing statistical features of voltage / current waveforms such as skewness, kurtosis, amplitude standard deviation, and zero-crossing rate, reflecting the distribution pattern and fluctuation characteristics of the signal in the time dimension; Frequency domain characteristics: Perform Fast Fourier Transform (FFT) on the current signal to extract the amplitude and phase parameters of the fundamental, third, fifth and other odd harmonics, calculate the total harmonic distortion (THD), and analyze the spectral energy distribution of a specific frequency band of 0-5kHz to capture the frequency domain characteristics of the signal. Time-frequency domain features: Wavelet packet transform is performed on the current signal, and the energy entropy of each sub-band is extracted after decomposition to a preset number of layers. This effectively captures the transient non-stationary features of the signal and is suitable for identifying subtle signal changes in the early stages of potential hazards. Power characteristics: Calculate the fluctuation amplitude and frequency of instantaneous active and reactive power to reflect the power transmission stability of the line; Impedance characteristics: During the load stabilization period, the equivalent impedance spectrum of the line is calculated by the ratio of the voltage and current waveforms, and the characteristics of impedance amplitude and phase variation with frequency are extracted to characterize the impedance characteristics of the line.

[0035] The above features are combined according to preset rules (such as time domain-frequency domain-time-frequency domain-power-impedance order) to form a high-dimensional feature vector F=[f1, f2, ..., fn], which is the "dynamic electrical fingerprint" of the line at the current moment.

[0036] The definition of fn is: a "specific electrical feature value" in a single dimension. Specifically, fn is the nth feature data point in the high-dimensional feature vector F, representing the quantization result of a specific dimension of electrical feature extracted from the voltage / current signal after line preprocessing (e.g., RMS voltage, RMS current, or reactive power). The meaning of n is: "n" corresponds to the "feature number," arranged sequentially from 1 to n (e.g., f1 is the first feature, f2 is the second feature, ..., fn is the last feature), and the total number of n equals the total number of extracted features.

[0037] T3: Under the condition of line health, simulate different typical load conditions, collect multiple sets of health fingerprint samples, establish the benchmark fingerprint corresponding to each condition, and form an electrical fingerprint library; at the same time, use the health fingerprint samples to train the health status model.

[0038] Step T3 involves establishing the electrical fingerprint database and health model. For example, when the line is confirmed to be in a healthy state after testing, different typical load conditions are simulated (including but not limited to no-load, light load, rated load, 30% overload, etc.), and steps T1-T2 are repeated to collect multiple sets of dynamic electrical fingerprint samples under healthy conditions. K-Means clustering algorithm is used to perform cluster analysis on multiple sets of fingerprint feature vectors under the same load condition, and the cluster centroid is used as the baseline fingerprint E_base for that condition. The baseline fingerprints of all load conditions are integrated to construct the electrical fingerprint database.

[0039] Meanwhile, using fingerprint feature vectors in a healthy state as training samples, a one-class support vector machine (SVM) algorithm is used to train a health model. This model learns the distribution pattern of health features and defines the normal range boundary of the health status feature vector as a criterion for judging the health status of the line.

[0040] T4: In the online monitoring phase, the dynamic electrical fingerprint feature vector of the line is extracted in real time, and the deviation is compared with the benchmark fingerprint under the same working conditions. The health status model is used to classify the status, and the time series of deviation values ​​is used to predict the trend.

[0041] Step T4 involves online monitoring and trend analysis. For example, after entering the online monitoring stage, the high-frequency acquisition device collects the line's electrical signals in real time, and extracts the real-time dynamic electrical fingerprint feature vector E_real according to step T2. The specific analysis process is as follows: Operating condition matching: The load condition of the current line is determined by real-time collected load power data, and the corresponding base fingerprint E_base is called from the electrical fingerprint database; First-level analysis (bias comparison): Calculate the Mahalanobis distance or cosine similarity between the real-time fingerprint E_real and the baseline fingerprint E_base. Mahalanobis distance fully considers the correlation between features and, compared to Euclidean distance, better reflects the essence of feature differences, thus improving the accuracy of bias judgment. Secondary analysis (state classification): Input the real-time fingerprint E_real into the trained One-Class SVM health model to determine whether it falls within the normal range of health features, thus achieving a preliminary classification of the line status; Level 3 analysis (trend prediction): Continuously record the deviation values ​​(Mahavir distance or cosine similarity) obtained from each analysis to form a time series {D1, D2, ..., Dt}; use time series prediction algorithms such as ARIMA (Autoregressive Integral Moving Average Model) or LSTM (Long Short-Term Memory Neural Network) to fit the series, predict the changing trend of the deviation values ​​within a preset time period in the future, and identify the development trend of potential risks in advance.

[0042] The definition of Dt is: the "real-time deviation value" at a certain moment. Specifically, Dt is the t-th data point in the time series, representing the quantification result of the deviation between the real-time dynamic electrical fingerprint and the reference fingerprint under the same working conditions during the t-th analysis in the online monitoring stage.

[0043] Basis for calculating the deviation value: The deviation value comes from the "comparison between the real-time fingerprint feature vector (E_real) and the reference fingerprint feature vector under the same working conditions (E_base)", specifically the Mahalanobis distance or cosine similarity.

[0044] The meaning of "t": "t" corresponds to the "time sequence", representing the time node of the t-th deviation analysis (for example, if analyzed once every 10 minutes, t = 1 represents the first analysis; t = 2 represents the second analysis, and so on).

[0045] T5: According to the deviation comparison result, status classification result and trend prediction result, trigger warnings of different levels, and output the abnormal feature items and contribution degrees.

[0046] Step T5 is hidden danger warning and diagnosis. For example, based on the historical operation data and fault cases of the line, train a dynamic threshold model through machine learning algorithms (such as logistic regression, random forest) to determine the thresholds (L1, L2, and L1 < L2) of different warning levels. When any of the following conditions is met, the system triggers the corresponding level of warning and generates a diagnostic report: Attention warning: The real-time deviation value (such as the real-time Mahalanobis distance D) exceeds the threshold L1, but the health model determines that the line is still in a normal state, indicating that the line may have minor abnormalities and requires enhanced monitoring; Abnormal alarm: The real-time deviation value (such as the real-time Mahalanobis distance D) exceeds the threshold L2, or the health model determines that the line is in an abnormal state, indicating that the line has clear potential hidden dangers and requires immediate arrangement for investigation; Critical warning: The trend prediction result shows that the deviation value will exceed the threshold L2 within the future Δt time (such as 72 hours, 1 week, which can be set according to actual needs), realizing the early prediction of hidden dangers and reserving processing time for operation and maintenance personnel.

[0047] The abnormal feature items with the greatest contribution to the deviation (such as "the amplitude of the 5th harmonic increases by 20% compared with the reference value", "the energy entropy of the sub-frequency band of the wavelet packet exceeds the normal range by 15%") need to be clearly marked in the warning information to assist operation and maintenance personnel in quickly locating the type of hidden danger (such as harmonic abnormality corresponding to load equipment failure, impedance abnormality corresponding to poor line contact), and shortening the fault investigation time. The normal feature items and contribution degrees are obtained by calculating the contribution weights of each feature to the deviation value, and the contribution weights are calculated using feature importance evaluation algorithms (such as random forest feature importance).

[0048] In another embodiment of this application, the following specific embodiments are also provided.

[0049] (a) Implementation Scenarios Taking the 220V AC lighting branch in the living area of ​​a certain ship as the monitoring object, the load of this branch is a fluorescent lamp group. Long-term operation is prone to hidden dangers such as ballast performance degradation and loose wiring terminals. This method is needed to achieve early warning.

[0050] (II) Implementation Steps 1. Signal acquisition device deployment: A high-frequency data acquisition device is installed at the lower end of the circuit breaker of the lighting branch, with a sampling rate of 20kHz. The instantaneous voltage value sequence u(t) and the instantaneous current value sequence i(t) are acquired simultaneously. The acquired data is transmitted to the monitoring terminal via industrial Ethernet. 2. Dynamic electrical fingerprint extraction: Select one power frequency cycle (20ms) as the data window, and extract 35-dimensional features from the preprocessed signal, including: skewness and kurtosis of voltage / current waveforms (time domain features), amplitude and phase of fundamental to 21st harmonic, total harmonic distortion rate (frequency domain features), energy entropy of each sub-band after wavelet packet 4-level decomposition (time and frequency domain features), instantaneous active power fluctuation amplitude (power features), and equivalent impedance spectrum when the load is stable (impedance features), forming a 35-dimensional dynamic electrical fingerprint feature vector F; 3. Establishment of Electrical Fingerprint Database and Health Model: After the new line is laid and confirmed to be in a healthy state through insulation and continuity tests, three typical load conditions are simulated: 500 sets of health fingerprint samples were collected under each working condition. The K-Means algorithm was used to cluster the samples under each working condition, and the cluster centroids were used as the baseline fingerprints for the corresponding working conditions to construct an electrical fingerprint database. At the same time, a One-Class SVM health model was trained with 1500 sets of health samples, with a confidence level of 99.5%. Operating condition 1: All fluorescent lights are off (no load); Operating Condition 2: Turn on one-third of the fluorescent lamps (light load); Operating Condition 3: Turn on all fluorescent lights (full load).

[0051] 4. Online Monitoring and Trend Analysis: One year after the system was put into operation, online monitoring data showed that under "full load" conditions, the Mahalanobis distance between the real-time fingerprint and the baseline fingerprint increased slowly from the initial 1.2 to 2.8 (thresholds L1=2.5, L2=4.0), and the health model still determined that the line was in a normal state. The LSTM algorithm was used to predict the trend of the Mahalanobis distance time series. The results showed that at the current rate of increase, the Mahalanobis distance would reach 4.1 in 15 days, exceeding the threshold L2.

[0052] 5. Hazard Warning and Diagnosis: The system triggers an "Attention Warning," and the warning report shows that "the 3rd harmonic amplitude has the largest contribution, increasing by 18% compared to the baseline value." Based on the warning information, maintenance personnel investigated and found that the internal capacitor of a large fluorescent lamp ballast was aging, which led to an increase in the 3rd harmonic current. After replacing the aging ballast, the real-time Mahalanobis distance dropped back to 1.3, the warning was lifted, and the branch tripping fault caused by excessive harmonics was successfully avoided.

[0053] (III) Implementation Results This implementation verified the effectiveness of the methods described in the above embodiments: it enabled early identification of potential ballast aging hazards, with an early warning time 15 days earlier than traditional monitoring methods; the false alarm rate was 0, and the troubleshooting time was shortened from 4 hours in traditional methods to 1.5 hours, significantly improving the operation and maintenance efficiency and safety of ship power systems.

[0054] In summary, the intelligent diagnostic method for circuit hazards based on dynamic electrical fingerprint feature analysis provided in this application has the following beneficial effects: (1) Extremely high sensitivity and early detection: Through high-frequency sampling of 10kHz and above and multi-dimensional feature extraction, it can capture subtle electrical waveform distortions that cannot be identified by traditional effective value monitoring, and achieve accurate identification of the "budding stage" of hidden dangers. Compared with existing technologies, the warning time is advanced by 1-5 fault development cycles.

[0055] (2) Strong anti-interference capability: The electrical fingerprint database is constructed by adopting the principle of "comparison under the same working conditions" to effectively avoid the interference of load changes on the monitoring results; at the same time, the application of Mahalanobis distance eliminates the coupling interference between features and reduces the false alarm rate to less than 1%.

[0056] (3) No need for precise modeling and strong versatility: It only relies on the health data of the line itself to establish a benchmark, without the need to obtain specific parameters of the line (such as conductor material and length) and precise load model. It is applicable to AC lines of different voltage levels from 220V to 10kV, with low deployment cost and strong universality.

[0057] (4) High interpretability: When the warning is issued, the abnormal feature items and contribution are output simultaneously, providing clear diagnostic directions for operation and maintenance personnel and shortening the troubleshooting time by more than 60%.

[0058] (5) Support predictive maintenance: The technology upgrade from "condition monitoring" to "trend prediction" is realized through trend prediction algorithm, providing core technical support for predictive maintenance of power system and reducing unplanned outage rate.

[0059] In some embodiments of this application, an intelligent diagnostic system for circuit hazards based on dynamic electrical fingerprint feature analysis is also provided, including modules for implementing the methods provided in the above embodiments.

[0060] In some embodiments of this application, an electronic device is also provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method provided in the above embodiments.

[0061] In some embodiments of this application, a computer-readable storage medium is also provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the method provided in the above embodiments.

[0062] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. An intelligent diagnostic method for potential circuit hazards based on dynamic electrical fingerprint feature analysis, characterized in that, Includes the following steps: T1: High-frequency acquisition of instantaneous voltage and current signals of the line under monitoring, and preprocessing of the acquired raw signals; T2: Extract time-domain features, frequency-domain features, time-frequency-domain features, power features, and impedance features from the preprocessed signal to construct a dynamic electrical fingerprint feature vector characterizing the instantaneous operating state of the line; T3: Under the healthy condition of the line, simulate different typical load conditions, collect multiple sets of health fingerprint samples, establish the benchmark fingerprint corresponding to each condition, and form an electrical fingerprint library. At the same time, a health status model is trained using healthy fingerprint samples; T4: During the online monitoring phase, the dynamic electrical fingerprint feature vector of the line is extracted in real time, and the deviation is compared with the benchmark fingerprint under the same working conditions. The health status model is used to classify the status, and the time series of the deviation value is used to predict the trend. T5: Based on the deviation comparison results, state classification results, and trend prediction results, trigger different levels of early warning and output abnormal feature items and contribution.

2. The method according to claim 1, characterized in that, In step T1, the sampling frequency of high-frequency acquisition is not less than 10kHz; preprocessing includes digital filtering and noise reduction.

3. The method according to claim 1, characterized in that, In step T2: Time-domain characteristics include voltage / current waveform skewness, kurtosis, amplitude standard deviation, and zero-crossing rate; Frequency domain characteristics include the amplitude and phase of the fundamental wave and odd harmonics, total harmonic distortion rate, and spectral energy distribution in a specific frequency band; The time-frequency domain characteristic is the energy entropy of each sub-band of the current signal after wavelet packet transform; The power characteristics are the fluctuation characteristics of instantaneous active power and reactive power; The impedance characteristics are the equivalent impedance spectrum characteristics of the line during the load stabilization period.

4. The method according to claim 1, characterized in that, In step T3: Typical load conditions include no-load, light-load, rated load, and 30% overload; K-Means clustering analysis was performed on multiple groups of healthy fingerprint samples under the same working condition, and the centroid of the cluster was taken as the baseline fingerprint under the corresponding working condition. The health status model is trained using a support vector machine algorithm.

5. The method according to claim 1, characterized in that, In step T4, the deviation comparison is achieved by calculating the Mahalanobis distance or cosine similarity between the real-time fingerprint feature vector and the reference fingerprint feature vector; the trend prediction uses the ARIMA algorithm or LSTM algorithm to fit the time series of deviation values ​​and predict future values.

6. The method according to claim 1, characterized in that, In step T5, the warning levels include attention warning, abnormal alarm, and critical warning; Note that the warning is triggered when the real-time deviation value exceeds the first threshold L1 and the health status model determines that it is normal. The trigger condition for an abnormal alarm is that the real-time deviation value exceeds the second threshold L2 or the health status model determines that it is abnormal; The trigger condition for a critical warning is that the trend prediction results show that the deviation value will exceed the second threshold L2 within the next Δt time period; Among them, the first threshold L1 and the second threshold L2 are obtained by training based on historical data using a machine learning algorithm, and L1 <L2。 7. The method according to claim 1, characterized in that, In step T5, the abnormal feature items and their contribution are obtained by calculating the contribution weight of each feature to the deviation value. The contribution weight is calculated using a feature importance evaluation algorithm.

8. An intelligent diagnostic system for potential circuit hazards based on dynamic electrical fingerprint feature analysis, characterized in that, Includes modules for implementing the method as described in any one of claims 1-7.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the method as described in any one of claims 1-7.

Images