Electric leakage detection system for electrical equipment of transformer substation
By integrating multi-sensors and deep learning analysis technology in the substation to identify and locate leakage phenomena, the problem of low leakage detection efficiency in the existing technology is solved, and high-precision and fast-responsive leakage detection is achieved, ensuring the safe operation of the substation.
Patent Information
- Application Number
- CN202410843196.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-27
- Publication Date
- 2024-11-15
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The prior art is inefficient in leakage detection in substations, and cannot quickly and accurately identify and locate leakage phenomena, especially when dealing with multi-dimensional and large amounts of data, there are problems such as delay and inability to effectively identify complex electrical abnormalities.
Using multi-sensor data acquisition module, multi-level leakage analysis module, adaptive calibration module and intelligent decision-making control module, combined with current, voltage, temperature and humidity and electromagnetic interference sensors, the leakage phenomenon is identified through deep learning and multi-scale time series analysis, and through The wireless communication module sends real-time early warnings to the operation and maintenance personnel.
It significantly improves the accuracy and response speed of leakage detection, reduces safety accidents and economic losses caused by leakage, and provides strong technical support to ensure the safe operation of the substation.
Smart Images

Figure CN118962273A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of power system monitoring technology, and in particular to a leakage current detection system for electrical equipment in substations. Background Technology
[0002] With the continuous expansion of power systems and technological advancements, the safety and stable operation of substations are receiving increasing attention. In traditional substation management, leakage detection typically relies on regular manual inspections and basic automated monitoring systems. These methods suffer from inefficiency and slow response when processing large-scale power grid data. Especially against the backdrop of increasingly complex power systems, timely detection and accurate location of leakage phenomena become particularly difficult. Furthermore, existing technologies often lack sufficient accuracy in real-time data processing and anomaly identification, making it impossible to achieve rapid response to potential safety risks.
[0003] Faced with these challenges, existing technical solutions are insufficient to meet the need for rapid and accurate identification and location of leakage phenomena. Existing systems often cannot provide comprehensive data analysis, especially when dealing with multi-dimensional and large amounts of data. In addition, traditional methods have significant limitations in real-time data analysis and processing, such as high latency and inability to effectively identify complex electrical anomalies.
[0004] Therefore, it is particularly urgent to develop a system that can integrate advanced data analysis technology, real-time monitoring and intelligent decision support to ensure the safe operation of substations and effectively improve the accuracy and response speed of leakage current detection. Summary of the Invention
[0005] To achieve the above objectives, the present invention provides a leakage current detection system for electrical equipment in substations.
[0006] The substation electrical equipment leakage current detection system includes a multi-sensor data acquisition module, a multi-level leakage current analysis module, an adaptive calibration module, an intelligent decision control module, and a wireless communication module; among which:
[0007] Multi-sensor data acquisition module: Employs current sensors, voltage sensors, temperature and humidity sensors, and electromagnetic interference sensors to monitor multi-dimensional data of electrical equipment in real time, including current, voltage parameters, ambient temperature, humidity, and electromagnetic interference data;
[0008] Multi-level leakage current analysis module: Receives multi-dimensional data from the multi-sensor data acquisition module, and performs hierarchical analysis on the data using traditional methods based on statistical features and anomaly detection algorithms based on deep learning. This analysis is achieved through multi-scale time series analysis, anomaly pattern recognition, and feature extraction to identify potential leakage current phenomena. The analysis results are then sent to the adaptive calibration module.
[0009] Adaptive calibration module: After the multi-level leakage current analysis module detects a potential leakage current signal, the adaptive calibration module automatically triggers the calibration program. By sending low-frequency test signals and high-frequency pulse signals, it confirms the location and degree of leakage current. The calibration results are used to guide the intelligent decision control module to adjust the system settings.
[0010] Intelligent decision control module: Receives the results from the multi-level leakage current analysis module and the adaptive calibration module, performs comprehensive evaluation and decision-making, dynamically adjusts the inspection robot's patrol path and inspection frequency based on the leakage current analysis results and calibration information, and sends the decision results to the wireless communication module;
[0011] Wireless communication module: When leakage or other abnormal electrical conditions are detected, it sends real-time alerts to maintenance personnel via integrated wireless communication technology, and provides specific equipment location and abnormality details.
[0012] Furthermore, the multi-sensor data acquisition module includes a current sensing unit, a voltage sensing unit, a temperature and humidity sensing unit, an electromagnetic interference sensing unit, and a wireless transmission unit; wherein:
[0013] Current sensing unit: includes a Hall effect current sensor for real-time monitoring of current parameters of electrical equipment. The Hall effect current sensor measures current by detecting changes in magnetic field and converts it into a corresponding voltage signal. The output signal of the Hall effect current sensor is converted into a digital signal by an analog-to-digital converter.
[0014] Voltage sensing unit: includes a high-precision voltage divider and a precision amplifier, used to monitor the voltage parameters of electrical equipment in real time. The voltage divider divides the high voltage signal to a safe range, and then the precision amplifier amplifies it before sampling. The output signal of the voltage sensor is converted into a digital signal by an analog-to-digital converter.
[0015] Temperature and humidity sensing unit: includes an integrated temperature sensor and a capacitive humidity sensor, used to monitor the temperature and humidity parameters of the environment around electrical equipment in real time. The temperature sensor uses the characteristic of the resistance of semiconductor materials changing with temperature to measure the temperature, and the humidity sensor measures the ambient humidity by the change of capacitance. The collected analog signal is converted into a digital signal by an analog-to-digital converter.
[0016] Electromagnetic interference sensing unit: includes a wideband electromagnetic interference sensor, used to monitor electromagnetic interference around electrical equipment in real time. The wideband electromagnetic interference sensor receives electromagnetic wave signals through an antenna, extracts interference signals using a tuning circuit and amplifier, and converts them into digital signals through an analog-to-digital converter.
[0017] Wireless transmission unit: includes a data collector and a wireless transmitter, used to receive digital signals from each sensing unit and transmit them to the multi-level leakage current analysis module via a wireless communication protocol.
[0018] Furthermore, the multi-level leakage current analysis module includes a data preprocessing unit, a statistical feature extraction unit, a deep learning analysis unit, a multi-scale time series analysis unit, and an abnormal pattern recognition unit; wherein:
[0019] Data preprocessing unit: Receives multidimensional data from the multi-sensor data acquisition module, performs noise reduction, normalization and missing value processing on the multidimensional data, and ensures the integrity and accuracy of the multidimensional data through filtering and interpolation techniques;
[0020] Statistical Feature Extraction Unit: Utilizes traditional methods based on statistical features to extract statistical features from preprocessed data, including mean, variance, skewness, and kurtosis; the extracted statistical features are used to initially identify abnormal patterns in the data, providing an initial feature set for the deep learning analysis unit;
[0021] Deep learning analysis unit: Employs a deep learning model based on attention mechanism to perform in-depth analysis of multidimensional data. By assigning different weights to different parts of the data, it focuses on the characteristics and patterns of leakage current phenomena, thereby improving the accuracy of leakage current detection.
[0022] Multi-scale time series analysis unit: used to decompose data into different time scales and perform time series analysis on the data at each scale. Specifically, it uses sliding window technology and autocorrelation function to detect periodic and trend changes in the data to improve the accuracy of leakage current detection.
[0023] Anomaly pattern recognition unit: Combining the output of a deep learning model with statistical features and attention mechanism, it performs anomaly pattern recognition on the data, and identifies abnormal data points that deviate from the normal range by comparing real-time data with historical normal data patterns, and determines whether there is leakage.
[0024] Furthermore, the deep learning analysis unit includes:
[0025] Data input: Let X = {x1, x2, ..., x} n} represents multidimensional input data, where X is the original input data matrix.
[0026] x i Let i be the i-th feature in the input data;
[0027] Feature transformation: The input data is transformed into a higher-dimensional feature space through a linear transformation layer. The transformation formula is: Z = XW + b, where Z is the feature matrix after linear transformation; W is the feature transformation matrix; and b is the bias vector.
[0028] Calculate attention score: The attention score of each feature is calculated using the softmax function. The score formula is: α = softmax(Z), where α is the attention weight vector of the feature.
[0029] Feature-weighted sum: Calculate the weighted sum to highlight important features. The formula for the weighted sum is: Where A is the weighted feature representation; α i z is the attention weight for the i-th feature; i This represents the i-th feature after transformation;
[0030] Feature and pattern recognition: Finally, a deep neural network layer A is used to process and identify and classify leakage-related features and patterns.
[0031] Furthermore, the multi-scale time series analysis unit includes:
[0032] Data decomposition: The input time series data T = {t1, t2, ..., t...} is decomposed into... m The data is decomposed into multiple subsequences at different time scales. Each subsequence is generated using a sliding window technique, with a window size of w and a sliding step size of s; where T is the original time series data.
[0033] Sliding window technique: This technique uses a sliding window to divide time series data into multiple subsequences, with the formula: S j ={t j ,t j+1 ,…,t j+w-1};j=1,1+s,1+2s,…,m-w+1, where, S j For the j-th sliding window subsequence;
[0034] Autocorrelation function calculation: For each subsequence S j Calculate the autocorrelation function R k The formula is:
[0035] Among them, R k S is the value of the autocorrelation function at lag k; N is the subsequence S. j Length; S j,i For subsequence S j The i-th data point; For subsequence S j The mean; k is the lag time;
[0036] Periodicity and trend detection: By analyzing the autocorrelation function R k The peak and decay patterns are detected to identify periodic and trend changes in the data. A threshold θ is set when R... kWhen the value is greater than θ, a periodic change is determined; and a trend change is identified based on the decay rate and pattern of the autocorrelation function.
[0037] Furthermore, the abnormal pattern recognition unit includes:
[0038] Feature fusion: The output features of the statistical feature extraction unit and the deep learning analysis unit with attention mechanism are fused to form a comprehensive feature vector F. Let the statistical feature vector be S and the deep learning output feature vector be D. Then the comprehensive feature vector F is expressed as: F = [S, D];
[0039] Comparison of real-time and historical data: The feature vectors of real-time data are compared with the feature vectors of historical normal data. The Euclidean distance between the feature vectors of real-time data and historical data is calculated using the following formula:
[0040] Among them, F real-time F represents the feature vector of real-time data. historical d represents the feature vector of historical normal data; d represents the Euclidean distance between real-time data and historical data; n represents the dimension of the feature vector.
[0041] Anomaly detection: Set threshold θ d When the Euclidean distance d exceeds the threshold θ d When this happens, the current data point is determined to be an abnormal data point;
[0042] Leakage current detection: Analyze all detected abnormal data points to determine if leakage current exists. Combine the characteristics output by the multi-level leakage current analysis module to comprehensively judge the source and nature of the abnormal data points. Let the number of abnormal data points be N. abnormal The criteria for determining leakage current are:
[0043] ifN abnormal ≥N threshold Then there is a leakage current phenomenon, where N threshold Threshold for abnormal data points to determine leakage current phenomenon.
[0044] Furthermore, the adaptive calibration module includes a calibration signal generation unit, a signal injection unit, a signal detection unit, a leakage location confirmation unit, and a leakage degree measurement unit; wherein:
[0045] The calibration signal generation unit is used to generate low-frequency test signals and high-frequency pulse signals. The low-frequency test signal is used to detect the location of leakage, and the high-frequency pulse signal is used to measure the degree of leakage. The specific generation formula is as follows:
[0046] V low (t)=A low sin(2πf low t);
[0047] V high (t)=A high sin(2πf high t), where V low (t) represents the low-frequency test signal; A low f is the amplitude of the low-frequency signal. low V represents the frequency of the low-frequency signal. high (t) is a high-frequency pulse signal; A high f is the amplitude of the high-frequency signal; high The frequency of the high-frequency signal;
[0048] Signal injection unit: The generated low-frequency test signal and high-frequency pulse signal are injected into the electrical equipment through the prepared injection device. The specific operation of signal injection is controlled by a microcontroller to ensure the timing and amplitude of signal injection.
[0049] Signal detection unit: Uses current and voltage sensors to detect the response of injected signals in electrical equipment, and collects response data R(t), including low-frequency and high-frequency responses. The specific detection formula is as follows:
[0050]
[0051] Among them, I response,low (t) represents the low-frequency response current Z. low (t) represents the impedance at low frequencies; I response,high (t) represents the high-frequency response current; Z high (t): Impedance at high frequencies;
[0052] Leakage Location Confirmation Unit: By analyzing low-frequency response data, the leakage location is confirmed using time-domain reflectometry. The specific analysis formula is as follows: Where d is the distance to the leakage location; v is the signal propagation speed; Δt is the time difference between sending and receiving the signal;
[0053] Leakage Current Measurement Unit: By analyzing high-frequency response data, the leakage current is measured using an amplitude attenuation model. The specific measurement formula is as follows: Among them, E leak The degree of leakage current; V in V represents the amplitude of the input signal. out This represents the amplitude of the output signal.
[0054] Furthermore, the intelligent decision-making and control module includes a data receiving unit, a comprehensive evaluation unit, a path optimization unit, a frequency adjustment unit, and a control command generation unit; wherein:
[0055] Data receiving unit: Receives leakage analysis results from the multi-level leakage analysis module and calibration information from the adaptive calibration module, specifically including leakage location, leakage degree, and analysis and calibration results;
[0056] The comprehensive evaluation unit performs a comprehensive evaluation of the received leakage current analysis results and calibration information. The comprehensive evaluation unit calculates the comprehensive risk score using a weighted average method, with the following formula:
[0057] R score =w1·A leak +w2·C leak , where R score For comprehensive risk scoring; w1 is the weight of the leakage current analysis results; A leak w1 represents the leakage risk value in the leakage analysis results; w2 represents the weight of the calibration information; C leak The leakage current calibration value in the calibration information;
[0058] Path optimization unit: Based on the risk score from the comprehensive evaluation unit, the inspection path of the inspection robot is optimized. Specifically, Dijkstra's algorithm is used to calculate the optimal inspection path, and the formula is:
[0059] Among them, P opt The optimal inspection path; d(v i ,v i+1 For the inspection robot at node v i and node v i+1 The distance between them; R score,j The comprehensive risk score for the j-th detection point;
[0060] Control command generation unit: Based on the results of the path optimization unit and the frequency adjustment unit, it generates control commands and sends them to the inspection robot. The control commands include adjustment information for the inspection path and inspection frequency.
[0061] Furthermore, the wireless communication module includes a data receiving unit, an early warning generation unit, a location identification unit, a communication management unit, an early warning notification unit, and an anomaly details display unit; wherein:
[0062] Data receiving unit: used to receive leakage current detection results, calibration information, inspection path and inspection frequency adjustment information from the intelligent decision and control module;
[0063] Early warning generation unit: Based on the received data, it generates detailed early warning information, including the location of the leakage, the degree of leakage, the detection time, detailed information of related equipment, and recommended emergency handling measures;
[0064] Location identification unit: Utilizes the device's geographic location data and power grid topology information to accurately identify the location of the leaking device, ensuring that the warning information includes the specific device location;
[0065] Communication Management Unit: Used to manage the wireless communication network, specifically by sending early warning information to the mobile devices of maintenance personnel through integrated wireless communication technologies, including Wi-Fi, 4G / 5G or a preset wireless network;
[0066] Early warning notification unit: Receives early warning information sent by the communication management unit and transmits the early warning information to operation and maintenance personnel via SMS, email or mobile application notification.
[0067] The beneficial effects of this invention are:
[0068] This invention significantly improves the accuracy and efficiency of leakage current detection by integrating advanced multi-level data analysis technology. By using deep learning analysis and multi-scale time series analysis units, the system can accurately identify abnormal patterns and potential leakage current phenomena from a large amount of complex electrical data. This high-precision analysis capability enables leakage current phenomena to be detected in their early stages, thereby greatly reducing safety accidents and economic losses caused by leakage current.
[0069] This invention, through a dynamic calibration method provided by an adaptive calibration module, can automatically adjust the detection signal based on real-time data, ensuring the accuracy and reliability of the leakage current detection process. By sending low-frequency and high-frequency test signals, the system can not only accurately locate the leakage point but also assess the degree of leakage. This adaptive adjustment capability provides strong technical support for substation operation and maintenance personnel, enabling them to respond to potential risks more quickly.
[0070] This invention, through the introduction of an intelligent decision control module and a wireless communication module, greatly enhances the system's response speed and operational efficiency. Through the collaborative work of these modules, the system can quickly notify maintenance personnel when leakage or other electrical anomalies are detected, while providing detailed anomaly information and handling suggestions. This not only improves the maintenance personnel's ability to handle emergencies but also optimizes the inspection robot's patrol path and inspection frequency, further improving the power system's safety management level and maintenance efficiency. Attached Figure Description
[0071] To more clearly illustrate the technical solutions in this invention or the prior art, 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 for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0072] Figure 1This is a schematic diagram of an electrical equipment leakage current detection system according to an embodiment of the present invention;
[0073] Figure 2 This is a schematic diagram of the multi-level leakage current analysis module according to an embodiment of the present invention. Detailed Implementation
[0074] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.
[0075] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.
[0076] like Figures 1-2 As shown, the substation electrical equipment leakage current detection system includes a multi-sensor data acquisition module, a multi-level leakage current analysis module, an adaptive calibration module, an intelligent decision control module, and a wireless communication module; wherein:
[0077] Multi-sensor data acquisition module: Employs current sensors, voltage sensors, temperature and humidity sensors, and electromagnetic interference sensors to monitor multi-dimensional data of electrical equipment in real time, including current, voltage parameters, ambient temperature, humidity, and electromagnetic interference data, ensuring comprehensive data coverage;
[0078] Multi-level leakage current analysis module: Receives multi-dimensional data from the multi-sensor data acquisition module, and performs hierarchical analysis on the data using traditional methods based on statistical features and anomaly detection algorithms based on deep learning. This analysis is achieved through multi-scale time series analysis, anomaly pattern recognition, and feature extraction to identify potential leakage current phenomena. The analysis results are then sent to the adaptive calibration module.
[0079] Adaptive calibration module: After the multi-level leakage current analysis module detects a potential leakage current signal, the adaptive calibration module automatically triggers the calibration program. By sending low-frequency test signals and high-frequency pulse signals, it confirms the location and degree of leakage current. The calibration results are used to guide the intelligent decision control module to adjust the system settings.
[0080] Intelligent decision control module: Receives the results from the multi-level leakage current analysis module and the adaptive calibration module, performs comprehensive evaluation and decision-making, dynamically adjusts the inspection robot's patrol path and inspection frequency based on the leakage current analysis results and calibration information, and sends the decision results to the wireless communication module;
[0081] Wireless communication module: When leakage or other abnormal electrical conditions are detected, it sends real-time alerts to maintenance personnel via integrated wireless communication technology, and provides specific equipment location and abnormality details.
[0082] The multi-sensor data acquisition module includes a current sensing unit, a voltage sensing unit, a temperature and humidity sensing unit, an electromagnetic interference sensing unit, and a wireless transmission unit; among which:
[0083] Current sensing unit: includes a Hall effect current sensor for real-time monitoring of current parameters of electrical equipment. The Hall effect current sensor measures current by detecting changes in magnetic field and converts it into a corresponding voltage signal. The output signal of the Hall effect current sensor is converted into a digital signal by an analog-to-digital converter (ADC).
[0084] Voltage sensing unit: includes a high-precision voltage divider and a precision amplifier, used to monitor the voltage parameters of electrical equipment in real time. The voltage divider divides the high voltage signal to a safe range, and then the precision amplifier amplifies it before sampling. The output signal of the voltage sensor is converted into a digital signal by an analog-to-digital converter (ADC).
[0085] Temperature and humidity sensing unit: includes an integrated temperature sensor and a capacitive humidity sensor, used to monitor the temperature and humidity parameters of the environment around electrical equipment in real time. The temperature sensor uses the characteristic of the resistance of semiconductor materials changing with temperature to measure the temperature, and the humidity sensor measures the ambient humidity by the change of capacitance. The collected analog signal is converted into a digital signal by an analog-to-digital converter (ADC).
[0086] Electromagnetic interference sensing unit: including a wideband electromagnetic interference (EMI) sensor, used to monitor electromagnetic interference around electrical equipment in real time. The wideband EMI sensor receives electromagnetic wave signals through an antenna, extracts interference signals using a tuning circuit and amplifier, and converts them into digital signals through an analog-to-digital converter.
[0087] The wireless transmission unit, comprising a data collector and a wireless transmitter, receives digital signals from each sensing unit and transmits them to the multi-level leakage current analysis module via wireless communication protocols (such as Wi-Fi or Zigbee). Through the detailed design of the multi-sensor data acquisition module and the specific functions of each unit, multi-dimensional, real-time, and high-precision data monitoring of substation electrical equipment is achieved, providing comprehensive and accurate equipment operating status information. This data provides a solid foundation for subsequent leakage current analysis, adaptive calibration, and intelligent decision-making, significantly improving the detection accuracy and response speed of the leakage current detection system and ensuring the safe operation of the substation.
[0088] The multi-level leakage current analysis module includes a data preprocessing unit, a statistical feature extraction unit, a deep learning analysis unit, a multi-scale time series analysis unit, and an anomaly pattern recognition unit; among which:
[0089] Data preprocessing unit: Receives multidimensional data from the multi-sensor data acquisition module, performs noise reduction, normalization, and missing value processing on the multidimensional data, and uses filtering and interpolation techniques to ensure the integrity and accuracy of the multidimensional data, providing a clear data foundation for subsequent analysis;
[0090] Statistical Feature Extraction Unit: Utilizes traditional methods based on statistical features to extract statistical features from preprocessed data, including mean, variance, skewness, and kurtosis; the extracted statistical features are used to initially identify abnormal patterns in the data, providing an initial feature set for the deep learning analysis unit;
[0091] Deep learning analysis unit: Employs a deep learning model based on attention mechanism to perform in-depth analysis of multidimensional data. By assigning different weights to different parts of the data, it focuses on the characteristics and patterns of leakage current phenomena, thereby improving the accuracy of leakage current detection. The attention mechanism enables the model to accurately identify and highlight abnormal data points in a large dataset.
[0092] Multi-scale time series analysis unit: used to decompose data into different time scales and perform time series analysis on the data at each scale. Specifically, it uses sliding window technology and autocorrelation function to detect periodic and trend changes in the data to improve the accuracy of leakage current detection.
[0093] Anomaly pattern recognition unit: Combining statistical features and the output of a deep learning model based on attention mechanisms, it performs anomaly pattern recognition on the data. By comparing real-time data with historical normal data patterns, it identifies abnormal data points that deviate from the normal range and determines whether leakage current exists. Through the detailed design of the multi-level leakage current analysis module and the specific functions of each unit, the system achieves comprehensive and accurate analysis of multi-dimensional data of substation electrical equipment. Combining statistical features and deep learning algorithms based on attention mechanisms, through multi-scale time series analysis and anomaly pattern recognition, this invention improves the accuracy and response speed of leakage current detection. This method effectively utilizes the attention mechanism in deep learning, enhances the model's performance in complex data environments, and ensures the safe operation of substation equipment.
[0094] The deep learning analytics unit includes:
[0095] Data input: Let X = {x1, x2, ..., x} n} represents multidimensional input data, where X is the original input data matrix.
[0096] x i Let i be the i-th feature in the input data;
[0097] Feature transformation: The input data is transformed into a higher-dimensional feature space through a linear transformation layer. The transformation formula is: Z = XW + b, where Z is the feature matrix after linear transformation; W is the feature transformation matrix; and b is the bias vector.
[0098] Calculate attention score: The attention score of each feature is calculated using the softmax function. The score formula is: α = softmax(Z), where α is the attention weight vector of the feature.
[0099] Feature-weighted sum: Calculate the weighted sum to highlight important features. The formula for the weighted sum is: Where A is the weighted feature representation used for further analysis; α i z is the attention weight for the i-th feature; i This represents the i-th feature after transformation;
[0100] Feature and pattern recognition: Finally, the deep neural network layer A is used to process and identify and classify leakage-related features and patterns. Through the steps of the above method, the deep learning analysis unit can effectively focus on the features most relevant to leakage phenomena, improving the accuracy and efficiency of leakage detection.
[0101] The multi-scale time series analysis unit includes:
[0102] Data decomposition: The input time series data T = {t1, t2, ..., t...} is decomposed into... mThe data is decomposed into multiple subsequences at different time scales. Each subsequence is generated using a sliding window technique, with a window size of w and a sliding step size of s; where T is the original time series data.
[0103] Sliding window technique: This technique uses a sliding window to divide time series data into multiple subsequences, with the formula: S j ={t j ,t j+1 ,…,t j+w-1};j=1,1+s,1+2s,…,m-w+1, where, S j For the j-th sliding window subsequence;
[0104] Autocorrelation function calculation: For each subsequence S j Calculate the autocorrelation function R k The formula is:
[0105] Among them, R k S is the value of the autocorrelation function at lag k; N is the subsequence S. j Length; S j,i For subsequence S j The i-th data point; For subsequence S j The mean; k is the lag time;
[0106] Periodicity and trend detection: By analyzing the autocorrelation function R k The peak and decay patterns are detected to identify periodic and trend changes in the data. A threshold θ is set when R... k When the value is greater than θ, periodic changes are determined; and trend changes are identified based on the decay rate and pattern of the autocorrelation function. Through the detailed design and algorithm implementation of the multi-scale time series analysis unit, the system can perform multi-scale analysis on the time series data of substation electrical equipment, effectively detecting periodic and trend changes in the data. Combined with the application of sliding window technology and autocorrelation function, the accuracy and reliability of leakage current detection are improved, ensuring the safe operation of the substation.
[0107] The anomaly pattern recognition unit includes:
[0108] Feature fusion: The output features of the statistical feature extraction unit and the deep learning analysis unit with attention mechanism are fused to form a comprehensive feature vector F. Let the statistical feature vector be S and the deep learning output feature vector be D. Then the comprehensive feature vector F is expressed as: F = [S, D];
[0109] Comparison of real-time and historical data: The feature vectors of real-time data are compared with the feature vectors of historical normal data. The Euclidean distance between the feature vectors of real-time data and historical data is calculated using the following formula:
[0110] Among them, F real-time F represents the feature vector of real-time data. historical d represents the feature vector of historical normal data; d represents the Euclidean distance between real-time data and historical data; n represents the dimension of the feature vector.
[0111] Anomaly detection: Set threshold θ d When the Euclidean distance d exceeds the threshold θ d When this happens, the current data point is determined to be an abnormal data point;
[0112] Leakage current detection: Analyze all detected abnormal data points to determine if leakage current exists. Combine the characteristics output by the multi-level leakage current analysis module to comprehensively judge the source and nature of the abnormal data points. Let the number of abnormal data points be N. abnormal The criteria for determining leakage current are:
[0113] ifN abnormal ≥N threshold Then there is a leakage current phenomenon, where N threshold To determine the threshold of abnormal data points for leakage current, the abnormal pattern recognition unit can effectively combine the output of statistical features and attention mechanism deep learning models to perform abnormal pattern recognition on the data. By comparing real-time data with historical normal data patterns, it can accurately identify abnormal data points that deviate from the normal range, thereby determining whether leakage current exists.
[0114] The adaptive calibration module includes a calibration signal generation unit, a signal injection unit, a signal detection unit, a leakage location confirmation unit, and a leakage degree measurement unit; wherein:
[0115] The calibration signal generation unit is used to generate low-frequency test signals and high-frequency pulse signals. The low-frequency test signal is used to detect the location of leakage, and the high-frequency pulse signal is used to measure the degree of leakage. The specific generation formula is as follows:
[0116] V low (t)=A low sin(2πf low t);
[0117] V high (t)=A high sin(2πf high t), where V low (t) represents the low-frequency test signal; A low f represents the amplitude of the low-frequency signal. low V represents the frequency of the low-frequency signal. high (t) is a high-frequency pulse signal; A high f is the amplitude of the high-frequency signal;high The frequency of the high-frequency signal;
[0118] Signal injection unit: The generated low-frequency test signal and high-frequency pulse signal are injected into the electrical equipment through the prepared injection device. The specific operation of signal injection is controlled by a microcontroller (MCU) to ensure the timing and amplitude of signal injection.
[0119] Signal detection unit: Uses current and voltage sensors to detect the response of injected signals in electrical equipment, and collects response data R(t), including low-frequency and high-frequency responses. The specific detection formula is as follows:
[0120]
[0121] Among them, I response,low (t) represents the low-frequency response current Z. low (t) represents the impedance at low frequencies; I response,high (t) represents the high-frequency response current; Z high (t): Impedance at high frequencies;
[0122] Leakage location confirmation unit: By analyzing low-frequency response data, the leakage location is confirmed using time-domain reflectometry (TDR). The specific analysis formula is as follows: Where d is the distance to the leakage location; v is the signal propagation speed; Δt is the time difference between sending and receiving the signal;
[0123] Leakage Current Measurement Unit: By analyzing high-frequency response data, the leakage current is measured using an amplitude attenuation model. The specific measurement formula is as follows: Among them, E leak The degree of leakage current; V in V represents the amplitude of the input signal. out The output signal amplitude is used to generate and inject low-frequency test signals and high-frequency pulse signals. Combined with signal detection and analysis technology, the adaptive calibration module achieves accurate measurement of the leakage location and degree of electrical equipment. The design of this module not only improves the accuracy of leakage detection, but also ensures the reliability of the system in complex power grid environments, providing strong technical support for the safe operation of substation equipment.
[0124] The intelligent decision-making and control module includes a data receiving unit, a comprehensive evaluation unit, a path optimization unit, a frequency adjustment unit, and a control command generation unit; wherein:
[0125] Data receiving unit: Receives leakage analysis results from the multi-level leakage analysis module and calibration information from the adaptive calibration module, specifically including leakage location, leakage degree, and analysis and calibration results;
[0126] The comprehensive evaluation unit performs a comprehensive evaluation of the received leakage current analysis results and calibration information. The comprehensive evaluation unit calculates the comprehensive risk score using a weighted average method, with the following formula:
[0127] R score =w1·A leak +w2·C leak , where R score For comprehensive risk scoring; w1 is the weight of the leakage current analysis results; A leak w1 represents the leakage risk value in the leakage analysis results; w2 represents the weight of the calibration information; C leak The leakage current calibration value in the calibration information;
[0128] Path optimization unit: Based on the risk score from the comprehensive evaluation unit, the inspection path of the inspection robot is optimized. Specifically, Dijkstra's algorithm is used to calculate the optimal inspection path, and the formula is:
[0129] Among them, P opt The optimal inspection path; d(v i ,v i+1 For the inspection robot at node v i and node v i+1 The distance between them; R score,j The comprehensive risk score for the j-th detection point;
[0130] Control command generation unit: Based on the results of the path optimization unit and frequency adjustment unit, it generates control commands and sends them to the inspection robot. The control commands include adjustment information for the inspection path and inspection frequency. The intelligent decision-making and control module receives and comprehensively evaluates the results of the multi-level leakage current analysis module and the adaptive calibration module, and effectively optimizes the inspection robot's inspection path and inspection frequency using optimization algorithms and dynamic adjustment mechanisms. This module design not only improves the accuracy and efficiency of leakage current detection, but also ensures the reliability of the system in complex power grid environments, providing important technical support for the safe operation of substation equipment.
[0131] The wireless communication module includes a data receiving unit, an early warning generation unit, a location identification unit, a communication management unit, an early warning notification unit, and an anomaly details display unit; wherein:
[0132] Data receiving unit: Used to receive leakage current detection results, calibration information, inspection path and inspection frequency adjustment information from the intelligent decision and control module, ensuring that all relevant data can be transmitted to the early warning system in real time;
[0133] Early warning generation unit: Based on the received data, it generates detailed early warning information, including the location of the leakage, the degree of leakage, the detection time, detailed information of related equipment, and recommended emergency handling measures;
[0134] Location identification unit: Utilizes the device's geographic location data and power grid topology information to accurately identify the location of the leaking device, ensuring that the warning information includes the specific device location so that maintenance personnel can quickly find the abnormal device;
[0135] Communication Management Unit: Used to manage the wireless communication network. Specifically, through integrated wireless communication technologies, including Wi-Fi, 4G / 5G or a preset wireless network, it sends early warning information to the mobile devices of maintenance personnel. This unit ensures the real-time and reliable transmission of information.
[0136] Early warning notification unit: Receives early warning information sent by the communication management unit and transmits it to maintenance personnel via SMS, email, or mobile application notification. The early warning notification unit ensures that maintenance personnel can receive and respond to early warning information in a timely manner. Through the collaborative work of the above units, the real-time early warning and communication module can quickly generate and send detailed early warning information when leakage or other abnormal electrical conditions are detected, ensuring that maintenance personnel can obtain the specific equipment location and abnormality details in a timely manner, thereby taking appropriate measures quickly.
[0137] This invention is intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this invention should be included within the scope of protection of this invention.
Claims
1. Substation electrical equipment leakage detection system, characterized in that: It includes a multi-sensor data acquisition module, a multi-level leakage analysis module, an adaptive calibration module, an intelligent decision control module and a wireless communication module; among which: Multi-sensor data acquisition module: uses current sensors, voltage sensors, temperature and humidity sensors, and electromagnetic interference sensors to monitor multi-dimensional data of electrical equipment in real time, including current, voltage parameters, ambient temperature, humidity, and electromagnetic interference data; Multi-level leakage analysis module: Receives multi-dimensional data from the multi-sensor data acquisition module, and performs hierarchical analysis on the data through multi-scale time series analysis, abnormal pattern recognition and feature extraction based on traditional methods based on statistical features and anomaly detection algorithms based on deep learning to identify potential leakage phenomena, and sends the analysis results to the adaptive calibration module; Adaptive calibration module: After the multi-level leakage analysis module detects a potential leakage signal, the adaptive calibration module automatically triggers the calibration procedure to confirm the leakage location and leakage degree by sending low-frequency test signals and high-frequency pulse signals. The calibration results are used to guide the intelligent decision control module to adjust the system settings; Intelligent decision control module: receives the results of the multi-level leakage analysis module and the adaptive calibration module, conducts comprehensive evaluation and decision-making, dynamically adjusts the inspection path and inspection frequency of the inspection robot according to the leakage analysis results and calibration information, and sends the decision results to the wireless communication module; Wireless communication module: When leakage or abnormal electrical conditions are detected, the integrated wireless communication module is used to send real-time warnings to operation and maintenance personnel, and provide specific equipment locations and abnormality details.
2. The substation electrical equipment leakage detection system according to claim 1, characterized in that: The multi-sensor data acquisition module includes a current sensing unit, a voltage sensing unit, a temperature and humidity sensing unit, an electromagnetic interference sensing unit and a wireless transmission unit; wherein: Current sensing unit: including a Hall effect current sensor, which is used to monitor the current parameters of the electrical equipment in real time. The Hall effect current sensor measures the current by detecting the change of the magnetic field and converts it into a corresponding voltage signal. The output signal of the Hall effect current sensor is converted into a digital signal through an analog-to-digital converter; Voltage sensing unit: including high-precision voltage divider and precision amplifier, used to monitor the voltage parameters of electrical equipment in real time. The voltage divider divides the high voltage signal into a safe range, which is then amplified by the precision amplifier for sampling. The output signal of the voltage sensor is converted into a digital signal through an analog-to-digital converter; Temperature and humidity sensing unit: including integrated temperature sensor and capacitive humidity sensor, used to monitor the temperature and humidity parameters of the environment around the electrical equipment in real time. The temperature sensor uses the property that the resistance of semiconductor materials changes with temperature for measurement, and the humidity sensor measures the ambient humidity through the change of capacitance, and converts the collected analog signal into a digital signal through an analog-to-digital converter; Electromagnetic interference sensing unit: including a broadband electromagnetic interference sensor, which is used to monitor the electromagnetic interference around the electrical equipment in real time. The broadband electromagnetic interference sensor receives electromagnetic wave signals through an antenna, extracts interference signals using a tuning circuit and an amplifier, and converts them into digital signals through an analog-to-digital converter; Wireless transmission unit: includes a data collector and a wireless transmitter, used to receive digital signals from each sensor unit and send them to the multi-level leakage analysis module through wireless communication protocol.
3. The substation electrical equipment leakage detection system according to claim 1, characterized in that: The multi-level leakage analysis module includes a data preprocessing unit, a statistical feature extraction unit, a deep learning analysis unit, a multi-scale time series analysis unit and an abnormal pattern recognition unit; wherein: Data preprocessing unit: receives multidimensional data from the multi-sensor data acquisition module, performs denoising, normalization and missing value processing on the multidimensional data, and ensures the integrity and accuracy of the multidimensional data through filtering and interpolation; Statistical feature extraction unit: Statistical features are extracted from the preprocessed data using traditional methods based on statistical features, including mean, variance, skewness, and kurtosis. The extracted statistical features are used to preliminarily identify abnormal patterns in the data and provide an initial feature set for the deep learning analysis unit. Deep learning analysis unit: It uses a deep learning model based on the attention mechanism to conduct in-depth analysis of multi-dimensional data. By assigning different weights to different parts of the data, it focuses on the characteristics and patterns of leakage phenomena, thereby improving the accuracy of leakage detection. Multi-scale time series analysis unit: used to decompose data into different time scales and perform time series analysis on the data at each scale. Specifically, through sliding window technology and autocorrelation function, it detects periodicity and trend changes in the data to improve the accuracy of leakage detection; Abnormal pattern recognition unit: Combines statistical features and the output of the attention mechanism deep learning model to identify abnormal patterns in the data, and by comparing real-time data with historical normal data patterns, identifies abnormal data points that deviate from the normal range and determines whether there is leakage.
4. The substation electrical equipment leakage detection system according to claim 3, characterized in that: The deep learning analysis unit comprises: Data input: Let X = {x1, x2, ..., x n } is the multi-dimensional input data, where X is the original input data matrix, x i is the i-th feature in the input data; Feature transformation: The input data is transformed into a higher-dimensional feature space through a linear transformation layer. The transformation formula is: Z = XW + b, where Z is the feature matrix after linear transformation; W is the feature transformation matrix; and b is the bias vector. Calculate the attention score: Use the softmax function to calculate the attention score of each feature. The scoring formula is: α = softmax(Z), where α is the attention weight vector of the feature; Weighted sum of features: Calculate the weighted sum to highlight important features. The weighted sum formula is: Among them, A is the weighted feature representation; α i is the attention weight of the i-th feature; z i is the i-th feature after transformation; Feature and pattern recognition: The final deep neural network layer processes A to identify and classify leakage-related features and patterns.
5. The substation electrical equipment leakage detection system according to claim 4, characterized in that: The multi-scale time series analysis unit comprises: Data decomposition: Input time series data T = {t1, t2, ..., t m } is decomposed into multiple subsequences of different time scales. Each subsequence is generated by the sliding window technology. Let the window size be w and the sliding step be s; where T is the original time series data; Sliding window technology: Use sliding window technology to divide time series data into multiple subsequences. The formula is: S j ={t j ,t j+1 ,…,t j+w-1 }; j = 1, 1 + s, 1 + 2s, ..., m - w + 1, where S j is the jth sliding window subsequence; Autocorrelation function calculation: For each subsequence S j Calculate the autocorrelation function R k , the formula is: Among them, R k is the value of the autocorrelation function at the time lag k; N is the subsequence S j Length; S j,i is a subsequence S j The i-th data point in ; is a subsequence S j The mean of ; k is the lag time; Periodicity and trend detection: by analyzing the autocorrelation function R k The peak and decay patterns of the data are used to detect periodic and trend changes. The threshold θ is set. When R k When >θ, it is determined that there is a periodic change; and the trend change is identified based on the decay speed and pattern of the autocorrelation function.
6. The substation electrical equipment leakage detection system according to claim 5, characterized in that: The abnormal pattern recognition unit comprises: Feature fusion: The output features of the statistical feature extraction unit and the attention mechanism deep learning analysis unit are fused to form a comprehensive feature vector F. Let the statistical feature vector be S and the deep learning output feature vector be D. The comprehensive feature vector F is expressed as: F = [S, D]; Comparison between real-time data and historical data: Compare the real-time data feature vector with the historical normal data feature vector, and calculate the Euclidean distance between the real-time data feature vector and the historical data feature vector. The formula is: Among them, F real-time is the real-time data feature vector; F historical is the feature vector of historical normal data; d is the Euclidean distance between real-time data and historical data; n is the dimension of the feature vector; Anomaly detection: setting threshold θ d , when the Euclidean distance d exceeds the threshold θ d When , the current data is determined to be an abnormal data point; Leakage phenomenon determination: Analyze all detected abnormal data points to determine whether there is leakage phenomenon. Combine the characteristics of the output of the multi-level leakage analysis module to comprehensively determine the source and nature of the abnormal data points. Suppose the number of abnormal data points is N. abnormal , the conditions for judging leakage are: iF abnormal ≥N threshold , then there is leakage, where N threshold It is the threshold value of abnormal data points for determining leakage phenomenon.
7. The substation electrical equipment leakage detection system according to claim 1, characterized in that: The adaptive calibration module includes a calibration signal generating unit, a signal injecting unit, a signal detecting unit, a leakage position confirming unit and a leakage degree measuring unit; wherein: Calibration signal generation unit: used to generate low-frequency test signals and high-frequency pulse signals. The low-frequency test signal is used to detect the leakage position, and the high-frequency pulse signal is used to measure the leakage degree. The specific generation formula is: V low (t)=A low sin(2πf low t); V high (t) = A high sin(2πf high t), where V low (t) is the low-frequency test signal; A low is the amplitude of the low-frequency signal; f low is the frequency of the low-frequency signal; V high (t) is a high-frequency pulse signal; A high is the amplitude of the high-frequency signal; f high is the frequency of the high-frequency signal; Signal injection unit: The generated low-frequency test signal and high-frequency pulse signal are injected into the electrical equipment through the prepared injection device. The specific operation of signal injection is controlled by the microcontroller to ensure the timing and amplitude of signal injection; Signal detection unit: Use current and voltage sensors to detect the response of the injected signal in the electrical equipment and collect response data R(t), including low-frequency and high-frequency responses. The specific detection formula is: Among them, I response,low (t) is the low frequency response current Z low (t) is the impedance at low frequency; I response,high (t) is the high frequency response current; Z high (t): impedance at high frequency; Leakage location confirmation unit: By analyzing the low-frequency response data, the leakage location is confirmed using time domain reflection technology. The specific analysis formula is: Where d is the distance of the leakage location; v is the signal propagation speed; Δt is the time difference between sending and receiving the signal; Leakage degree measurement unit: By analyzing the high-frequency response data, the leakage degree is measured using the amplitude attenuation model. The specific measurement formula is: Among them, E leak is the leakage degree; V in is the input signal amplitude; V out is the output signal amplitude.
8. The substation electrical equipment leakage detection system according to claim 1, characterized in that: The intelligent decision control module includes a data receiving unit, a comprehensive evaluation unit, a path optimization unit, a frequency adjustment unit and a control instruction generation unit; wherein: Data receiving unit: receiving leakage analysis results from the multi-level leakage analysis module and calibration information from the adaptive calibration module, including leakage location, leakage degree, and analysis and calibration results; Comprehensive evaluation unit: Comprehensively evaluate the received leakage analysis results and calibration information. The comprehensive evaluation unit calculates the comprehensive risk score by weighted average method. The formula is: R score =w1·A leak +w2·C leak , where R score is the comprehensive risk score; w1 is the weight of the leakage analysis result; A leak is the leakage risk value in the leakage analysis result; w2 is the weight of the calibration information; C leak is the leakage calibration value in the calibration information; Path optimization unit: Based on the risk score of the comprehensive evaluation unit, the patrol path of the inspection robot is optimized. Specifically, the Dijkstra algorithm is used to calculate the optimal patrol path. The formula is: Among them, P opt is the optimal patrol path; d(v i ,v i+1 ) is the inspection robot at node v i and node v i+1 The distance between score,j is the comprehensive risk score of the jth detection point; Control instruction generation unit: generates control instructions according to the results of the path optimization unit and the frequency adjustment unit and sends them to the inspection robot. The control instructions include adjustment information of the inspection path and inspection frequency.
9. The substation electrical equipment leakage detection system according to claim 8, characterized in that: The wireless communication module includes a data receiving unit, an early warning generating unit, a location identifying unit, a communication management unit, an early warning notification unit and an abnormality detail display unit; wherein: Data receiving unit: used to receive leakage detection results, calibration information, inspection path and inspection frequency adjustment information from the intelligent decision-making and control module; Warning generation unit: Generates detailed warning information based on the received data, including leakage location, leakage degree, detection time, detailed information of related equipment and recommended emergency treatment measures; Location identification unit: Uses the device's geographic location data and grid topology information to accurately identify the location of the leakage device, ensuring that the warning information contains the specific device location; Communication management unit: used to manage the wireless communication network, specifically to send warning information to the mobile devices of the operation and maintenance personnel through integrated wireless communication technology; Warning notification unit: Receives warning information sent by the communication management unit and transmits it to operation and maintenance personnel via SMS, email or mobile application notification.
10. The substation electrical equipment leakage detection system according to claim 9, characterized in that: The wireless communication technology integrated in the communication management unit includes Wi-Fi, 4G / 5G or a preset wireless network.
Citation Information
Patent Citations
Detection method for fault arc and fault arc circuit-breaker
CN109917241A
Transformer substation electrical equipment comprehensive physical examination system based on cloud side end collaborative perception
CN111784026A
Intelligent substation inspection system based on multiple data sources
CN116865431A
Equipment health analysis method based on multi-source data driving
CN117421582A
Mask time sequence model training method and device, mask time sequence model prediction method and device, equipment and medium
CN117556882A
Cited By
Anti-interference electric leakage protection device and method
CN119134218A
Rapid self-calibration method for field maintenance of water quality sensor
CN119395251A
Remote monitoring system for electrical equipment
CN119651908A
Electric leakage monitoring method and system for smart power grid
CN120370220A
Measurement switch data acquisition and analysis system
CN120891369A
