A fault analysis system and method for power equipment operation state perception

Abstract

The application discloses a kind of power equipment operating state perception-oriented fault analysis system and method, it is related to power equipment fault analysis field.The application includes: the target monitoring point of power equipment is deployed first type sensor and second type sensor, the first type sensor is used to collect electrical quantity parameter;Extract the sensor and the physical parameter monitored by it existing physical correlation, construct influence factor node graph;Through the first type sensor and second type sensor of adjacent or shared data convergence node of physical location are deployed, and construct influence factor node graph based on physical correlation, realize the deep fusion analysis of electrical quantity and non-electrical quantity parameter, simultaneously innovatively introduce sensor itself state parameter monitoring mechanism, when sensor occurs physical disconnection or failure, it can isolate or reduce the weight of its node in time, effectively avoid the fault misjudgment caused by sensor itself exception, greatly improve system anti-interference ability and diagnostic reliability.

Description

Technical Field

[0001] This invention relates to the field of power equipment fault analysis, specifically to a fault analysis system and method for power equipment operating status perception. Background Technology

[0002] During the operation of a power system, the safety and reliability of power equipment directly affect the stable operation of the entire power grid. With the continuous expansion of the power grid and the improvement of its intelligence level, higher demands are placed on the real-time monitoring and fault diagnosis capabilities of power equipment operating status. Traditionally, power equipment monitoring mainly relies on single-type sensors, such as collecting only electrical parameters like current and voltage, or only non-electrical parameters like temperature and vibration. In recent years, some monitoring systems have begun to attempt to integrate multi-source sensor information in order to improve the accuracy and response speed of fault identification. However, existing systems are still relatively rudimentary in terms of sensor deployment, data fusion, and anomaly identification, making it difficult to achieve comprehensive perception and accurate analysis of complex fault scenarios.

[0003] In existing technologies, some fault analysis methods deploy multiple sensors on power equipment to collect electrical and non-electrical parameters, and determine whether the equipment has malfunctioned based on preset thresholds or simple logical rules. For example, when the measured value of a sensor exceeds its normal range, the system triggers an alarm. In addition, some methods attempt to construct sensor networks, using graph structures to describe the relationships between sensors, and weighting scores based on the degree of anomaly at each node to generate a comprehensive anomaly score. These methods have improved the sensitivity and robustness of fault detection to some extent, and are gradually becoming a research hotspot in power equipment condition monitoring.

[0004] However, the aforementioned existing technologies still have significant shortcomings. Existing methods generally ignore potential physical disconnection or failure events that may occur with the sensors themselves, such as power supply anomalies, communication interruptions, or excessive internal temperatures. When a sensor fails, its collected abnormal data may be misjudged as a fault in the equipment itself, leading to false alarms or misjudgments. When constructing correlations between sensors, existing methods often rely solely on fixed topologies or preset weights, failing to fully consider the dynamic response relationships between different physical parameters in actual operation. For example, the physical coupling characteristics between current and temperature, and voltage and vibration, drift with changes in operating conditions, making it difficult for fixed weights to accurately reflect real-time correlation strength. Existing methods have limited ability to distinguish fault types (such as equipment-related faults versus external electrical disturbances), easily leading to deviations in maintenance strategies. Therefore, a power equipment fault analysis scheme that can simultaneously consider sensor state perception, dynamic weighted modeling, and fault type identification is needed. Summary of the Invention

[0005] Based on this, the purpose of the present invention is to provide a fault analysis system and method for power equipment operation status perception, so as to solve the technical problems in the prior art that are prone to fault misjudgment due to ignoring the failure state of the sensor itself, cannot adaptively adjust the analysis model according to the dynamic correlation between electrical and non-electrical parameters, and have difficulty in accurately distinguishing equipment failures from external electrical disturbances.

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

[0007] The first aspect of this application involves deploying a first type of sensor and a second type of sensor at the target monitoring point of power equipment. The first type of sensor is used to collect electrical quantity parameters, and the second type of sensor is used to collect non-electrical quantity parameters. The first type of sensor and the second type of sensor are physically adjacent or share the same data aggregation node. Sensors with physical correlation and their monitored physical parameters are extracted to construct an influence factor node diagram. Nodes in the influence factor node diagram correspond to different physical parameters, and the edges between nodes correspond to physical proximity or electrical coupling relationships. Each edge is assigned an initial influence weight. The self-state parameters include power supply voltage, communication signal strength, and internal temperature. When the self-state parameter of any sensor exceeds a preset threshold, it is determined that the sensor has experienced a physical disconnection or physical failure event. The node corresponding to the sensor is temporarily isolated in the influence factor node diagram, or the weight of its associated edge is reduced to zero. For unisolated nodes, the time series of their electrical quantity parameters and non-electrical quantity parameters are extracted, and the deviation between the actual response ratio of each pair of nodes with edge connections and the historical stable baseline is calculated. The real-time weight of the corresponding edge is dynamically adjusted according to the deviation. The greater the deviation, the lower the real-time weight; the smaller the deviation, the higher the real-time weight. The calculation formula for dynamic adjustment is as follows:

[0008] ;

[0009] in, For a moment node With nodes Real-time weights of the edges between them As the initial influence weight, The preset sensitivity coefficient, For a moment node With nodes The degree of deviation between the edges; the comprehensive anomaly score of the influence factor node diagram is calculated using a weighted summation method, and the calculation formula is as follows:

[0010] ;

[0011] in, For a moment The overall abnormality score, The total number of nodes. For nodes There exists a set of neighboring nodes connected by an edge. node At any moment Normalized outliers, when When the value exceeds a preset first threshold, the power equipment is determined to have an abnormal operating status; further distinction is made between fault types: if the real-time weights of both electrical quantity nodes and non-electrical quantity nodes are greater than a preset second threshold and the overall abnormal score is... If the value exceeds the limit, it is determined to be a fault in the equipment itself; if only the real-time weight of the edge related to the electrical quantity node drops below the preset third threshold, the overall abnormal score is not considered. If the value exceeds the limit, it is determined to be an external electrical disturbance.

[0012] The second aspect of this application: a first type of sensor and a second type of sensor are deployed at target monitoring points of power equipment, wherein the first type of sensor collects electrical quantity parameters and the second type of sensor collects non-electrical quantity parameters; the first type of sensor and the second type of sensor are physically adjacent to each other or share the same data aggregation node;

[0013] The node graph construction module is used to extract sensors with physical relationships and the physical parameters they monitor, and construct an influence factor node graph. The nodes in the influence factor node graph correspond to different physical parameters, and the edges between the nodes correspond to physical proximity or electrical coupling relationships. An initial influence weight is assigned to each edge.

[0014] The sensor self-test module is used to monitor the status parameters of each sensor in real time.

[0015] The node isolation module is used to temporarily isolate the node corresponding to any sensor in the influence factor node graph or reduce the weight of its associated edge to zero when the state parameter of any sensor exceeds a preset threshold.

[0016] The deviation calculation module is used to extract the time series of electrical and non-electrical parameters of unisolated nodes, and calculate the degree of deviation between the actual response ratio and the historical stable baseline for each pair of nodes with edge connections. ;

[0017] The dynamic weighting module is used to dynamically adjust the real-time weights of the corresponding edges according to the degree of deviation, using the following calculation formula:

[0018] ;

[0019] The comprehensive scoring module is used to calculate the comprehensive anomaly score of the impact factor node diagram using a weighted summation method. The calculation formula is:

[0020] ;

[0021] The fault classification module is used when When the value exceeds a preset threshold, it distinguishes between equipment malfunction and external electrical disturbance.

[0022] In summary, the present invention has the following main beneficial effects:

[0023] This invention achieves deep fusion analysis of electrical and non-electrical parameters by deploying first-type and second-type sensors with physically adjacent or shared data aggregation nodes and constructing an influence factor node graph based on physical correlation. It also innovatively introduces a sensor self-state parameter monitoring mechanism, which can promptly isolate or reduce the weight of nodes when sensors experience physical disconnection or failure, effectively avoiding misjudgments caused by sensor anomalies and significantly improving the system's anti-interference capability and diagnostic reliability. The invention calculates the deviation between the actual response ratio of adjacent nodes and the historical stable baseline, dynamically adjusting the real-time weights of edges—lower weights for larger deviations and higher weights for smaller deviations—and using a weighted summation method to obtain a comprehensive anomaly score. This allows for sensitive capture of subtle shifts in the correlation between different physical parameters, overcoming the shortcomings of traditional fixed-weight models that cannot adapt to dynamic changes in operating conditions. Furthermore, based on the difference in real-time weights between electrical and non-electrical nodes, this invention can accurately distinguish between equipment-related faults and external electrical disturbances, providing maintenance personnel with more targeted fault judgment criteria, thereby significantly improving the adaptability, robustness, and engineering practicality of power equipment fault analysis. Attached Figure Description

[0024] Figure 1 This is the control flowchart of the present invention. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0026] The embodiments of the present invention will now be described.

[0027] A fault analysis system and method for power equipment operation status perception, such as... Figure 1 As shown, it includes:

[0028] Example 1

[0029] This embodiment provides a fault analysis method for power equipment operation status perception. The following section combines... Figure 1 The flowchart shown provides a detailed explanation of the method.

[0030] I. Sensor Deployment and Establishment of Physical Correlation

[0031] First, deploy both Type I and Type II sensors at the target monitoring points of the power equipment. Type I sensors are used to collect electrical parameters, such as current transformers (CT, unit: ampere, A), voltage transformers (PT, unit: volt, V), or partial discharge sensors (unit: picocoulomb, pC). Type II sensors are used to collect non-electrical parameters, such as thermocouples or infrared temperature sensors (unit: degree Celsius, ℃), piezoelectric accelerometers (vibration amplitude, unit: meters per second squared, m / s²), or acoustic emission sensors (acoustic emission signal intensity, unit: decibel, dB).

[0032] In terms of physical deployment, the first type of sensor and the second type of sensor are required to be physically adjacent, for example, with a distance not exceeding 0.5 meters (m), or both should share the same data aggregation node (i.e., signal aggregation through the same data acquisition terminal). The purpose of this is to ensure a clear physical coupling relationship between the physical parameters collected by the two types of sensors, which facilitates the subsequent construction of the influence factor node diagram.

[0033] Sensors with physical relationships and their monitored physical parameters are extracted to construct an influence factor node graph. The influence factor node graph is a graph structure model where nodes represent different physical parameters (e.g., node 1 is current I, node 2 is temperature T, node 3 is voltage V, and node 4 is vibration B). Edges between nodes represent physical adjacency (e.g., two sensors mounted on the same device housing, 0.3 m apart) or electrical coupling (e.g., coupling between current and temperature due to the Joule heating effect). This graph describes the topology of sensor relationships and forms the basis for subsequent dynamic weighting and comprehensive scoring calculations. An initial influence weight is assigned to each edge. The initial influence weight refers to the fixed weight value pre-assigned to each edge in the influence factor node graph based on the spatial distance between sensors, the overlap of signal transmission paths, and the mechanical coupling of physical mounting brackets under normal equipment conditions. This value is dimensionless and is used to characterize the correlation strength between two sensors in the initial state.

[0034] Initial influence weight Determined by the following formula:

[0035] ;

[0036] in, For nodes With nodes The spatial distance between corresponding sensors is expressed in meters. For example, a current sensor and a temperature sensor are mounted on the same busbar, and the center-to-center distance between them is... m;

[0037] The signal transmission path overlap is defined as follows, with a value range of [value range missing]. Dimensionless. This value is determined by analyzing the geometric overlap of the sensor signal transmission paths: if the signal cables of the two sensors are completely co-located, then... If completely separated, then When there is partial overlap, take the median value, for example... .

[0038] The mechanical coupling degree of the physical mounting bracket, with a value range of [value missing]. Dimensionless. This value is determined by the structural rigidity of the mounting bracket and the connection method: if the two sensors are rigidly fixed to the same metal bracket, then If they are installed on different flexible supports and there is no direct mechanical connection, then For example, the vibration sensor and the voltage sensor are fastened to the same housing by the same bolt, which can be taken as... .

[0039] , , The preset weighting coefficients satisfy... For example, it can be set , , .

[0040] II. Sensor Self-Test and Node Isolation Mechanism

[0041] The system collects operational data from each sensor in real time and monitors the individual sensor's status parameters. These status parameters include:

[0042] Supply voltage, in volts (V), normal range is, for example, 3-3.6V;

[0043] Communication signal strength, measured in decibels (dBmW). The normal range is, for example, -70dBm to -30dBm;

[0044] Internal temperature, in degrees Celsius (°C), normal range for example -20°C to 85°C;

[0045] When the state parameters of any sensor exceed a preset threshold, the sensor is determined to have experienced a physical disconnection or physical failure event. A physical disconnection or physical failure event refers to a fault state in which the sensor is unable to normally collect or transmit measurement data due to reasons such as abnormal power supply, communication interruption, or excessive internal temperature. For example, if the power supply voltage is lower than 2.8V or higher than 3.8V, or the communication signal strength is lower than −80dBm, or the internal temperature is 90℃, a failure determination will be triggered.

[0046] Once a fault is detected, the node corresponding to that sensor is temporarily isolated in the influence factor node graph (i.e., it does not participate in subsequent calculations), or the weight of its associated edges is reduced to zero. This avoids false abnormal signals caused by sensor malfunctions interfering with the overall fault assessment.

[0047] III. Deviation Calculation and Real-time Weight Dynamic Adjustment

[0048] For nodes that are not isolated, extract the time series of their electrical and non-electrical parameters. Assume a sampling period of 0.1 seconds, and analyze the data series from the most recent 10 seconds. Calculate the deviation between the actual response ratio and the historical stable baseline for each pair of connected nodes. , The larger the value, the greater the difference between the physical relationship between the two nodes and the normal state.

[0049] degree of deviation Calculated using the following formula:

[0050] ;

[0051] in, For nodes With nodes The actual response ratio is the ratio of the measured physical quantities of two physically related sensor nodes (such as adjacent locations or electrical coupling) during the operation of power equipment, occurring at the same moment or within a very short time window. This ratio is used to quantify the real-time coupling strength or response characteristics between two different physical parameters. Under normal operating conditions, there is usually a relatively stable proportional relationship or conversion relationship between two physically related parameters. For example, there are theoretical or empirical correspondences between Joule heat generated by current and temperature rise, voltage and mechanical vibration, and partial discharge pulse and acoustic emission signal. Next, node The measured value (or its power function) and the node The ratio between the measured values ​​(or their powers) is calculated using a formula that is defined separately for each combination of different physical parameter types represented by the nodes.

[0052] As a historical stable baseline, the actual response ratio is collected at no less than 1000 sampling times within a continuous period during which the power equipment is operating normally and has no fault records. Take the arithmetic mean of the data set as... This value is dimensionless and serves as a reference benchmark for determining whether the current response has drifted.

[0053] It is a small positive integer used to prevent the denominator from being zero, and its range is [value range missing]. The specific value depends on The magnitude is determined when: Time to take ,when Time to take .

[0054] Actual response ratio Based on the combination of node types, one of the following formulas shall be used:

[0055] 1. When node For current nodes, nodes This is a temperature node.

[0056] In this case, the actual response ratio is defined as the ratio of the measured temperature value to the square of the current value, used to characterize the temperature rise produced by a unit current square under the Joule heating effect. Its physical meaning is that for a conductor, its heat generation is directly proportional to the square of the current, and the temperature rise has an approximately linear relationship with the heat generation. Therefore, the ratio of temperature to the square of the current can reflect the heat dissipation conditions or load state of the conductor. The calculation formula is as follows:

[0057] ;

[0058] in, The temperature measurement value is the Celsius temperature value directly read by a thermocouple or thermistor sensor, and the unit is Celsius (°C).

[0059] The measured current value is obtained by averaging the squares of the instantaneous current values ​​over one power frequency cycle and then taking the square root. The unit is ampere squared. );

[0060] For the preset small positive numbers, take This is used to avoid the denominator being zero when the current is zero.

[0061] This ratio is not a direct division of the two original measurements, but a response relationship quantity constructed based on physical laws, hence it is called the response ratio.

[0062] 2. When node For voltage nodes, nodes It is a vibration node.

[0063] In this case, the actual response ratio is defined as the ratio of the vibration amplitude to the measured voltage value, used to characterize the intensity of the mechanical vibration response excited by a unit voltage. Its physical meaning is that electrical equipment (such as transformer cores and reactor windings) generates electromagnetic force under the action of an alternating electric field, which in turn induces mechanical vibration. The vibration amplitude is directly proportional to the square of the voltage amplitude; here, the first power ratio is used as a simplified linear model. The calculation formula is as follows:

[0064] ;

[0065] in, The vibration amplitude is the peak value of the vibration signal collected by the accelerometer in the time domain, and the unit is meters per second squared (m / s²).

[0066] This is a voltage measurement, in volts (V).

[0067] Take 0.01V.

[0068] 3. When the node For partial discharge nodes, nodes This is a sound emission node.

[0069] In this case, the actual response ratio is defined as the ratio of the acoustic emission signal intensity to the partial discharge pulse count rate, used to characterize the acoustic emission energy intensity generated per unit partial discharge event. Its physical meaning lies in the fact that when a partial discharge occurs, both electrical pulses and ultrasonic signals are generated simultaneously, and there is a positive correlation between their intensities. This ratio reflects the discharge type and the attenuation characteristics of the discharge source's propagation path. The calculation formula is as follows:

[0070] ;

[0071] in, The intensity of the acoustic emission signal is the amplitude or root mean square value of the signal envelope output by the acoustic emission sensor, and the unit is decibel (dB).

[0072] The partial discharge pulse count rate is the number of partial discharge pulses exceeding a preset amplitude threshold detected per unit time, measured in pulses per second. ).

[0073] The actual response ratio is a dynamically coupled characteristic quantity with different construction forms depending on the combination type of node physical parameters. It truly reflects the actual response relationship between two different sensors driven by physical laws, rather than a simple ratio of measured values. This actual response ratio is compared with a historical stable baseline obtained statistically during normal equipment operation. By comparing, the degree of deviation can be calculated. This allows for a quantitative assessment of whether the correlation between the two parameters has shifted significantly.

[0074] Obtaining the degree of deviation Then, the real-time weights of the corresponding edges are dynamically adjusted based on the degree of deviation. The following calculation formula is used in this embodiment:

[0075] ;

[0076] in, For a moment node With nodes The real-time weight of the edges between them refers to the weight at time t. Based on the degree of deviation at the current moment Initial influence weights The edge weights, obtained after dynamic adjustment, are dimensionless. The greater the deviation, the lower the real-time weight; the smaller the deviation, the higher the real-time weight. This dynamically reflects the reliability of the association between two nodes at the current moment.

[0077] The preset sensitivity coefficient is a preset positive number, dimensionless, used to control the sensitivity of the real-time weight to the degree of deviation. The larger the value, the faster the real-time weight decreases as the degree of deviation increases.

[0078] IV. Comprehensive Anomaly Score Calculation and Fault Judgment

[0079] The comprehensive anomaly score of the impact factor node plot is calculated using a weighted summation method. , refers to the time The dimensionless score is calculated by weighted summation, taking into account the normalized outliers of all nodes in the influencing factor node graph and the real-time weights of their associated edges. This score is used to comprehensively evaluate whether the power equipment has any abnormal operating status at the current moment; the larger the value, the higher the degree of abnormality. The calculation formula is:

[0080] ;

[0081] in, For a moment The comprehensive anomaly score is dimensionless.

[0082] The total number of nodes refers to the total number of nodes that are not isolated in the influencing factor node graph at the current moment;

[0083] For nodes The set of neighboring nodes that have edges connected to each other in the influence factor node graph refers to the set of nodes that are connected to each other by edges. The set of all other nodes connected by edges;

[0084] node At any moment The normalized outlier refers to the value at time t. ,node The corresponding sensor measurement value The anomaly index obtained after normalization is dimensionless.

[0085] Normalized outliers Calculated using the following formula:

[0086] ;

[0087] in: For nodes The corresponding sensor at time The measured values ​​are in units that vary depending on the type of sensor (e.g., current is measured in A, temperature in °C).

[0088] For nodes The corresponding normal operating range, The lower threshold is... The upper limit threshold refers to the normal operating range of the node. The corresponding sensor's measurement value is within the allowable normal fluctuation range under normal operating conditions of the power equipment. This range is determined by the equipment's factory parameters, historical operating data, or relevant industry standards.

[0089] Calculate the comprehensive anomaly score Later, when When the value exceeds a preset first threshold, the power equipment is determined to have an abnormal operating status. The first threshold is a preset critical value used to determine whether the power equipment has an abnormal operating status, and it is dimensionless.

[0090] V. Fault Type Differentiation

[0091] When the overall anomaly score exceeds the standard, further differentiation of the fault type is required:

[0092] Equipment failure: If the real-time weights of both electrical and non-electrical quantity nodes are greater than the preset second threshold and the overall anomaly score is... If the value exceeds the limit, it is determined to be a fault in the equipment itself. The second threshold is a preset critical value used to determine whether the real-time weights of electrical and non-electrical quantity nodes are sufficiently high, and it is dimensionless. When the real-time weights of both types of nodes are greater than this threshold, it is used as one of the conditions for determining a fault in the equipment itself.

[0093] Equipment malfunction refers to a fault state in which the internal structure or function of power equipment malfunctions or is damaged, causing its operating parameters to deviate from the normal range. The root causes of this type of fault lie in internal factors such as physical defects in the equipment itself, material aging, insulation deterioration, mechanical wear, or heat accumulation. Typical equipment malfunctions include, but are not limited to: inter-turn short circuits in transformer windings, multiple grounding points in the iron core, aggravated partial discharge, insulation breakdown, wear of motor bearings, overheating and oxidation of cable joints, and burnt-out of switch contacts.

[0094] A significant characteristic of equipment failure is that the fault source is located inside the equipment. Therefore, it not only causes abnormalities in electrical parameters (such as current, voltage, and partial discharge), but also causes corresponding changes in non-electrical parameters (such as increased temperature, intensified vibration, and enhanced acoustic emission) through energy transfer or physical coupling mechanisms.

[0095] External electrical disturbances: If the real-time weight of the edge related only to the electrical quantity node drops below the preset third threshold, and the overall anomaly score is also affected... If the value exceeds the limit, it is determined to be an external electrical disturbance. The third threshold is a dimensionless threshold used to determine whether the real-time weight of the edges related to the electrical quantity node has dropped to a sufficiently low preset threshold. When only the real-time weight of the edges related to the electrical quantity node is less than this threshold, it is used as one of the conditions for determining an external electrical disturbance.

[0096] External electrical disturbances refer to transient or steady-state abnormal events in the power grid caused by external factors, not by faults in the equipment itself. The root causes of these disturbances include external factors such as changes in the power grid structure, sudden load changes, lightning strikes, switching operations, and the propagation of faults from nearby equipment. Typical external electrical disturbances include, but are not limited to: system-side voltage sags / surges, frequency fluctuations, harmonic injection, three-phase imbalance, current surges caused by external short-circuit faults, and lightning overvoltages.

[0097] A significant characteristic of external electrical disturbances is that the disturbance source is located outside the equipment and usually affects electrical parameters (such as voltage and current) first and foremost, while non-electrical parameters (such as temperature, vibration, and acoustic emission) may not show significant changes in the initial stage of the disturbance or throughout the entire duration of the disturbance due to the existence of thermal inertia and mechanical inertia.

[0098] VI. Formula for Adjusting Alternative Dynamic Weights

[0099] As an alternative to the above dynamic adjustment formula, the real-time weights can be calculated using an exponential decay method:

[0100] ;

[0101] in, The preset attenuation factor has a value range of [value range missing]. , is a preset positive number in the exponential decay form of the dynamic weight calculation formula. It is dimensionless and is used to control the rate at which the real-time weight decays exponentially as the degree of deviation increases. The larger the value, the faster the decay.

[0102] VII. Data estimation steps when a node fails

[0103] When node After a sensor is isolated due to its own state parameters exceeding limits, the isolated node refers to a node in the influence factor node graph that is temporarily excluded from the subsequent comprehensive anomaly score calculation process because the corresponding sensor has experienced a physical disconnection or failure event. Isolation is not a physical removal of the sensor or disconnection of its electrical connection, but a logical operation at the data processing level. That is, the measurement data of this node no longer participates in the calculation of the comprehensive anomaly score, and the edges between this node and its neighboring nodes no longer contribute to the weighted summation. To avoid interrupting the monitoring of this physical parameter, the influence factor node graph is used to... Given all neighboring nodes connected by an edge, the node values ​​are estimated using the following weighted regression formula. Expected measurement value :

[0104] ;

[0105] in, For nodes The expected measurement value refers to the value at the node. After a sensor is isolated due to failure, the node is estimated using weighted regression based on measurements from its neighboring nodes. The predicted value. The unit of this value is relative to the node. The units of the original measurements are consistent.

[0106] For nodes There exists a set of neighboring nodes connected by an edge;

[0107] For nodes With nodes Initial influence weights between them;

[0108] For neighboring nodes At any moment The actual measured value, the unit depends on the node. Type;

[0109] Nodes fitted based on historical data Measurement to node The mapping function for measured values ​​refers to the function obtained by fitting historical data and used to map values ​​from neighboring nodes. Measured values Mapped to node The functional relationship between the measured values.

[0110] Example 2

[0111] This embodiment provides a fault analysis system for power equipment operation status perception, which is used to implement the method of Embodiment 1.

[0112] The system includes the following components:

[0113] 1. Type I sensors and Type II sensors

[0114] Deployed at target monitoring points of power equipment, the first type of sensor collects electrical parameters (such as current, voltage, and partial discharge), while the second type of sensor collects non-electrical parameters (such as temperature, vibration, and acoustic emission). The first and second types of sensors are physically adjacent (distance ≤ 0.5m) or share the same data aggregation node.

[0115] 2. Node Graph Construction Module

[0116] This module extracts physically related sensors and the physical parameters they monitor, constructing an influence factor node graph. Nodes in the graph correspond to different physical parameters, and edges between nodes correspond to physical adjacency or electrical coupling. Each edge is assigned an initial influence weight. The calculation method is the same as in Example 1.

[0117] 3. Sensor self-test module

[0118] This module is used to monitor the status parameters of each sensor in real time, including power supply voltage (V), communication signal strength (dBm), and internal temperature (°C), and updates the status once per second.

[0119] 4. Node isolation module

[0120] When the state parameters of any sensor exceed a preset threshold (e.g., power supply voltage <2.8V or >3.8V), the module temporarily isolates the node corresponding to that sensor in the influence factor node graph (marks it as unavailable), or reduces the weight of its associated edge to zero.

[0121] 5. Deviation Calculation Module

[0122] For nodes that are not isolated, this module extracts the time series of their electrical and non-electrical parameters and calculates the actual response ratio between each pair of nodes with edge connections. Compared with historical stable baseline degree of deviation between The specific calculation formula is the same as in Example 1.

[0123] 6. Dynamic weighting module

[0124] This module dynamically adjusts the real-time weights of the corresponding edges based on the degree of deviation, using the following calculation formula:

[0125] ;

[0126] The definitions, units, and sources of each parameter are completely consistent with those in Example 1.

[0127] 7. Comprehensive Scoring Module

[0128] This module uses a weighted summation method to calculate the comprehensive anomaly score of the influence factor node diagram. :

[0129] ;

[0130] The definitions, units, and sources of each parameter are completely consistent with those in Example 1.

[0131] The comprehensive scoring module also includes an adaptive threshold update unit. This unit dynamically updates the judgment threshold based on the comprehensive anomaly score at historical normal operation times. The adaptive threshold update unit refers to updating the threshold at each historical time. The judgment threshold is dynamically calculated based on the statistical characteristics of comprehensive anomaly scores during historical normal operation, and is dimensionless. This threshold replaces the fixed first threshold, adapting to fluctuations in the normal operating status of the equipment and improving the sensitivity and robustness of anomaly detection. The updated formula is:

[0132] ;

[0133] in, For a moment The threshold for judgment is dimensionless;

[0134] The mean is the average of the comprehensive anomaly scores within the past time window. The mean is the arithmetic average of all comprehensive anomaly scores within the past time window and is dimensionless. The time window is a continuous time interval used to statistically analyze historical data, such as the past 1 hour or the past 24 hours. The length of the window can be preset according to the actual application scenario.

[0135] This is the standard deviation of the comprehensive anomaly scores within the past time window. The standard deviation refers to the standard deviation of all comprehensive anomaly scores within the past time window and is dimensionless.

[0136] The preset multiplier is a preset positive number, dimensionless, used to control the leniency of the adaptive threshold. The larger the value, the higher the adaptive threshold, and the stricter the judgment. This embodiment uses... This means adopting the classic anomaly detection criterion of the mean plus three times the standard deviation.

[0137] This adaptive threshold can be automatically adjusted according to fluctuations in the normal operating status of the equipment, thereby improving detection sensitivity and robustness.

[0138] 8. Fault Classification Module

[0139] When S(t) is greater than the aforementioned adaptive threshold Th(t), this module further distinguishes the fault type:

[0140] If the real-time weights of both electrical quantity nodes and non-electrical quantity nodes are greater than the second threshold (e.g., 0.5), then the device itself is determined to be faulty.

[0141] If the real-time weight of the edge related to only the electrical quantity node drops to less than the third threshold (e.g., 0.3), it is determined to be an external electrical disturbance.

[0142] Optionally, this system also includes an edge computing node. An edge computing node refers to a field computing device (such as an industrial computer or embedded smart terminal) deployed near the sensor array. It stores local influence factor node graphs and runs core algorithms such as the dynamic weighting module and the comprehensive scoring module. The influence factor node graph is stored in this edge computing node, and the edge computing node only stores node graph data related to its physical area (e.g., a local graph corresponding to a transformer, not the entire substation). The dynamic weighting module and the comprehensive scoring module run within this edge computing node. This edge computing node is deployed in a field cabinet no more than a preset interval (50 meters in this embodiment) from the sensor array, such as in a smart terminal box next to the transformer, to ensure low latency in data acquisition and reliable communication.

[0143] Taking a 10kV distribution transformer as an example, current sensors (phase A, phase B, and phase C) are deployed on its three-phase outgoing lines, temperature sensors are deployed on the winding surfaces, and vibration sensors are deployed on the casing. All of these sensors are connected to the same edge computing node (installed in a cabinet 20 meters away from the transformer). After system startup, the node graph construction module establishes five nodes (IA, IB, IC, T, and B) and the edges between them (e.g., an edge is established between IA and T due to the Joule heating effect). Initial weights. The temperature is calculated based on distance, signal overlap, and mechanical coupling. The sensor self-test module monitors the power supply voltage and internal temperature of each sensor every second. When the power supply voltage of a temperature sensor drops to 2.7V, the node isolation module temporarily isolates that temperature node. Subsequently, the deviation calculation module estimates the temperature value using a weighted regression formula with the remaining normally operating current sensors. The dynamic weighting module calculates the temperature in real time. The comprehensive scoring module calculates... ,when When the adaptive threshold is exceeded, the fault classification module determines whether it is internal transformer overheating (core fault) or external grid voltage sag (external disturbance) based on the distribution of weights on each side. The entire process achieves real-time perception of the operating status of power equipment, adaptive fault identification, and robust fault type differentiation.

[0144] In summary, the fault analysis system and method for power equipment operation status perception provided in this embodiment constructs an influence factor node graph based on physical correlation by deploying electrical quantity and electrical quantity sensors at physically adjacent or shared data aggregation nodes. It combines sensor self-state monitoring and node isolation mechanisms, dynamic weight adjustment based on the deviation between actual response ratios and historical stable baselines, weighted comprehensive anomaly score calculation, and fault type differentiation based on differences in the weights of electrical and non-electrical quantity nodes. This achieves real-time perception, adaptive modeling, and robust fault identification of power equipment operation status, effectively solving the technical problems in existing technologies such as susceptibility to misjudgment due to sensor failure, inability of fixed weights to adapt to dynamic operating conditions, and difficulty in distinguishing between equipment-related faults and external electrical disturbances.

[0145] The steps of the methods or algorithms described in this application can be directly embedded in hardware, software units executed by a processor, or a combination of both. Exemplarily, a storage medium can be connected to a processor so that the processor can read information from and write information to the storage medium. Optionally, the storage medium can also be integrated into the processor. Optionally, the processor and the storage medium can also be located in different components within a terminal. These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0146] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and drawings are merely illustrative examples of this application and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Thus, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A fault analysis method for power equipment operation status perception, characterized in that, The method includes: deploying a first type of sensor and a second type of sensor at the target monitoring point of the power equipment, wherein the first type of sensor is used to collect electrical quantity parameters and the second type of sensor is used to collect non-electrical quantity parameters; the first type of sensor and the second type of sensor are physically adjacent to each other or share the same data aggregation node; Extract sensors with physical relationships and the physical parameters they monitor, construct an influence factor node graph, and assign an initial influence weight to each edge; Real-time acquisition of operational data from each sensor and monitoring of the status parameters of each sensor; When the state parameter of any sensor exceeds the preset threshold, the weight of the associated edge of the node corresponding to that sensor in the influence factor node graph is reduced to zero. For unisolated nodes, extract the time series of their electrical and non-electrical parameters, and calculate the deviation between the actual response ratio and the historical stable baseline for each pair of nodes with edge connections. The real-time weight of the corresponding edge is dynamically adjusted based on the degree of deviation. The calculation formula for dynamic adjustment is as follows: ； in, For a moment node With nodes Real-time weights of the edges between them As the initial influence weight, The preset sensitivity coefficient, For a moment node With nodes The degree of deviation between the edges; The comprehensive anomaly score of the impact factor node diagram is calculated using a weighted summation method. The calculation formula is as follows: ； in, For a moment The overall abnormality score, The total number of nodes. For nodes There exists a set of neighboring nodes connected by an edge. node At any moment Normalized outliers, when When the value exceeds a preset first threshold, the power equipment is determined to have an abnormal operating status. If the real-time weights of both electrical quantity nodes and non-electrical quantity nodes are greater than the preset second threshold and the overall anomaly score is... If the value exceeds the limit, it is determined to be a fault in the equipment itself; if only the real-time weight of the edge related to the electrical quantity node drops below the preset third threshold, the overall abnormal score is not considered. If the value exceeds the limit, it is determined to be an external electrical disturbance.

2. The fault analysis method for power equipment operation status perception according to claim 1, characterized in that: The initial influence weight Determined by the following formula: ； in, For nodes With nodes The spatial distance between corresponding sensors, in meters; For signal transmission path overlap, For the mechanical coupling of the physical mounting bracket, , , These are the preset weighting coefficients.

3. The fault analysis method for power equipment operation status perception according to claim 1, characterized in that: The degree of deviation Calculated using the following formula: ； in, For nodes With nodes The ratio of the actual responses between them As a historical stable baseline, It is a positive number.

4. The fault analysis method for power equipment operation status perception according to claim 3, characterized in that: The actual response ratio is calculated using one of the following formulas based on the node type: When node For current nodes, nodes When it is a temperature node: ； in, This is a temperature measurement value. This is the measured current value. Preset positive numbers; When node For voltage nodes, nodes When it is a vibrating node: ； in, The amplitude of the vibration. This is a voltage measurement value; When node For partial discharge nodes, nodes When it is an acoustic emission node: ； in, The intensity of the acoustic emission signal. This represents the partial discharge pulse count rate.

5. The fault analysis method for power equipment operation status perception according to claim 1, characterized in that: The normalized outliers Calculated using the following formula: ； in, For nodes The corresponding sensor at time The measured value, For nodes The corresponding normal operating range, The lower threshold is... This is the upper limit threshold.

6. The fault analysis method for power equipment operation status perception according to claim 1, characterized in that: The calculation formula for dynamically adjusting real-time weights is replaced by an exponential decay form: ； in, This is the preset attenuation factor.

7. The fault analysis method for power equipment operation status perception according to claim 1, characterized in that: It also includes a data estimation step when a node fails: when a node fails After being isolated, the nodes in the impact factor node diagram are used. Given all neighboring nodes connected by an edge, the node values ​​are estimated using the following weighted regression formula. Expected measurement value: ； in, For nodes The expected measurement value, For nodes There exists a set of neighboring nodes connected by an edge. For nodes With nodes Initial influence weights between them For neighboring nodes At any moment The actual measured value, Nodes fitted based on historical data Measurement to node Mapping function for measured values.

8. A fault analysis system for sensing the operating status of power equipment, characterized in that, include: Type I and Type II sensors are deployed at target monitoring points of power equipment; The node graph construction module is used to extract sensors with physical relationships and the physical parameters they monitor, construct an influence factor node graph, and assign an initial influence weight to each edge; The sensor self-test module is used to monitor the status parameters of each sensor in real time. The node isolation module is used to reduce the weight of the associated edge of the node corresponding to any sensor to zero in the influence factor node graph when the state parameter of any sensor exceeds a preset threshold. The deviation calculation module is used to extract the time series of electrical and non-electrical parameters of unisolated nodes and calculate the degree of deviation between the actual response ratio of each pair of nodes with edge connections and the historical stable baseline. A dynamic weighting module is used to dynamically adjust the real-time weight of the corresponding edge according to the degree of deviation; The comprehensive scoring module is used to calculate the comprehensive anomaly score of the impact factor node diagram using a weighted summation method. ; The fault classification module is used when When the value exceeds a preset threshold, it distinguishes between equipment malfunction and external electrical disturbance.

9. A fault analysis system for power equipment operation status perception according to claim 8, characterized in that: The comprehensive scoring module also includes an adaptive threshold update unit, which dynamically updates the preset threshold based on the comprehensive anomaly scores from historical normal operation times. The update formula is as follows: ； in, For a moment The threshold for judgment, This represents the average of the comprehensive anomaly scores over the past time window. The standard deviation of the comprehensive anomaly scores over the past time window. This is a preset multiplier.

Images