Method and device for troubleshooting tripping circuit of relay protection device

Abstract

The invention provides a relay protection device tripping circuit troubleshooting method and device, and the method comprises the following steps: deploying a sensor at a key measurement point of a tripping circuit, and collecting the voltage, current, contact resistance, insulation resistance and environment temperature and humidity parameters of the circuit in real time; feature extraction is carried out on the collected parameters, static feature parameters, dynamic feature parameters and environment correlation parameters are extracted, and tripping loop health state feature vectors are constructed; the static characteristic parameters comprise a loop voltage deviation rate, a current change rate, a contact resistance and an insulation resistance reduction rate; the dynamic characteristic parameters comprise trip coil action current characteristics, action delay time and current waveform characteristics. The method for troubleshooting the tripping circuit of the relay protection device aims to solve the problem that real-time online monitoring and hidden fault active early warning capabilities are lacked in the running state of the tripping circuit.

Description

Technical Field

[0001] This invention belongs to the field of power system relay protection technology, specifically relating to a method and device for troubleshooting trip circuit faults in relay protection devices. Background Technology

[0002] The tripping circuit of a relay protection device is the execution channel for protection actions in a power system, and its reliability directly affects the safe and stable operation of the power grid. Traditional fault diagnosis of tripping circuits mainly employs methods such as periodic power outage testing, post-fault analysis, and logical relationship judgment. With the development of smart grid technology, some technical solutions have introduced deep learning algorithms to achieve intelligent fault location after faults by establishing fault location matrices and training neural network models; other technical solutions have developed fault diagnosis methods for alarm channels during the testing phase based on matrix operations. These technologies aim to improve fault diagnosis efficiency and reduce reliance on human experience, but they are mainly applied to fault analysis after protection device operation failures or fault detection during equipment testing.

[0003] However, it is impossible to continuously monitor the electrical parameters and status characteristics of the trip circuit during normal operation, and can only perform passive fault diagnosis after the protection device fails to operate or malfunctions; there is a lack of early warning means for latent faults such as increased contact resistance at the connection point, decreased coil insulation, and loose terminals, which leads to the gradual deterioration of these latent faults until they cause the protection device to fail to operate; it fails to integrate static parameters, dynamic characteristics and environmental factors for multi-dimensional health status assessment, resulting in insufficient fault diagnosis accuracy and location precision; and it lacks an adaptive early warning mechanism based on parameter trend analysis, making it impossible to dynamically adjust the diagnostic strategy according to the equipment deterioration pattern. Summary of the Invention

[0004] The present invention aims to at least partially solve one of the technical problems in the related art.

[0005] Therefore, the first objective of this invention is to provide a method for troubleshooting tripping circuit faults in relay protection devices.

[0006] The purpose of this invention is to solve the problem of lacking real-time online monitoring and proactive early warning capabilities for latent faults in the operation of tripping circuits, and to propose a method for troubleshooting tripping circuit faults in relay protection devices.

[0007] The second objective of this invention is to provide a fault diagnosis device for the tripping circuit of a relay protection device.

[0008] The third objective of this invention is to provide a computer device.

[0009] A fourth objective of this invention is to provide a non-transitory computer-readable storage medium.

[0010] To achieve the above objectives, a first aspect of the present invention provides a method for troubleshooting faults in the tripping circuit of a relay protection device, comprising: S1. Deploy sensors at key measurement points in the trip circuit to collect circuit voltage, circuit current, contact resistance, insulation resistance and ambient temperature and humidity parameters in real time; S2. Extract features from the collected parameters, including static feature parameters, dynamic feature parameters, and environmental correlation parameters, and construct a health status feature vector for the trip circuit; S3. Based on the health status feature vector, the health index of the trip circuit is calculated using a weighted fusion method, and the status is graded and evaluated according to the health index threshold. S4. Perform time series analysis on the health status feature vector, and achieve proactive early warning of latent faults through parameter trend discrimination and threshold comparison; S5. Perform fuzzy matching between the current feature vector and the preset fault mode feature library to identify the fault type; combine the tripping circuit topology and the location of abnormal parameter measurement points to locate the fault point and output early warning information through parameter tracing method. S6. Feed back the actual fault handling results to the fault mode feature library, and adaptively adjust the feature weights and diagnostic threshold task decoder base layer weights to perform cross-task constraints, so that the feature mapping of different tasks remains consistent.

[0011] In one embodiment of the present invention, S1 includes: Under normal operating conditions of the trip circuit, the sampling frequency is 1Hz; when a protection device start signal or a sudden change in circuit parameters is detected, it automatically switches to high-speed sampling mode, increasing the sampling frequency to 10kHz, and continuously collecting data until 500ms after the protection action is completed; the key measuring points include the DC power supply terminal, the protection device output terminal, both ends of the trip coil, the circuit breaker operating circuit terminal, and the intermediate connection terminal of the circuit.

[0012] In one embodiment of the present invention, S2 includes: The static characteristic parameters include loop voltage deviation rate, current change rate, contact resistance, and insulation resistance reduction rate. The dynamic characteristic parameters include the trip coil operating current characteristics, operating delay time, and current waveform characteristics; The environmental parameters are characterized by a comprehensive temperature and humidity index. ,in For temperature, For relative humidity, and These are weighting coefficients determined based on historical data statistics.

[0013] In one embodiment of the present invention, S3 further includes: The method for calculating the health index of the tripping circuit is as follows:

[0014]

[0015]

[0016] For the first The deviation value of each static characteristic parameter. For the corresponding weights, For the first The deviation value of each dynamic characteristic parameter. For the corresponding weights; The status grading assessment divides the health index into five levels: healthy state, good state, attention state, abnormal state, and fault state.

[0017] In one embodiment of the present invention, S4 includes: The proactive early warning of latent faults is achieved by establishing a multi-level early warning mechanism; The time series analysis uses the sliding window method with a window length of 30 days, and the analysis results are updated daily.

[0018] In one embodiment of the present invention, S5 includes: The fuzzy matching uses a similarity calculation method:

[0019] in For the first One measured feature parameter value, The first in the fault mode library One standard feature parameter value, Select similarity as the feature weight. The most significant failure mode is used as the diagnostic result; The parameter tracing and localization method includes: Establish a tripping circuit topology diagram and mark the complete path from the DC power supply to the circuit breaker tripping coil; Based on the location of the abnormal parameter measurement points, mark the abnormal area on the topology diagram. When multiple measurement points are abnormal, analyze the transmission path of the parameter in the loop and determine the common upstream node as the fault point. For parallel redundant loops, the fault path is isolated by comparing the parameter differences between healthy and abnormal loops using the differential location method; for series loops, the segment-by-segment elimination method is used to determine the fault segment based on the continuity of parameters of adjacent measuring points and output the fault location description. The differential positioning method calculates... Locate the additional resistor in the faulty section.

[0020] In one embodiment of the present invention, S6 includes: Establish a database of fault cases, regularly analyze the diagnostic accuracy rate, and when the accuracy rate is lower than the preset threshold, perform feature analysis on the misjudged cases to identify the key feature parameters that led to the misjudgment. The feature weight coefficients are automatically adjusted using gradient descent. ,in Let be the loss function, representing the deviation between the diagnostic result and the actual fault. This is the learning rate.

[0021] To achieve the above objectives, a second aspect of the present invention provides a fault diagnosis device for a relay protection device tripping circuit, comprising: The sensor deployment and parameter acquisition module deploys sensors at key measurement points in the trip circuit to collect circuit voltage, circuit current, contact resistance, insulation resistance, and ambient temperature and humidity parameters in real time. The feature extraction and vector construction module extracts features from the collected parameters, including static feature parameters, dynamic feature parameters, and environmental correlation parameters, and constructs a health status feature vector for the trip circuit. The health index calculation and status assessment module calculates the health index of the trip circuit based on the health status feature vector using a weighted fusion method, and performs status classification assessment according to the health index threshold. The time series analysis and fault identification module performs time series analysis on the health status feature vector, and realizes proactive early warning of latent faults through parameter trend discrimination and threshold comparison. The fuzzy matching and early warning output module identifies the fault type by performing fuzzy matching between the current feature vector and the preset fault mode feature library; it also locates the fault point and outputs early warning information by combining the topology of the tripping circuit and the location of abnormal parameter measurement points through parameter tracing methods. The result feedback and adaptive adjustment module feeds back the actual fault handling results to the fault mode feature library, and adaptively adjusts the feature weights and the decoder base layer weights of the diagnostic threshold task to perform cross-task constraints, so that the feature mapping of different tasks remains consistent.

[0022] The present invention provides a method and apparatus for troubleshooting tripping circuits of relay protection devices. Compared with the prior art, the technical solution of this application has the following beneficial technical effects: This invention achieves 24 / 7 continuous monitoring of the trip circuit's operating status by deploying sensors at key measurement points in the trip circuit to collect multi-dimensional parameters in real time, fundamentally changing the limitations of existing passive fault diagnosis technologies. By extracting static, dynamic, and environmentally relevant parameters and constructing a health status feature vector, it achieves deep integration of electrical parameters, operational characteristics, and environmental factors, solving the accuracy problem caused by the single-dimensional diagnosis of existing technologies. By employing a weighted fusion method to calculate a health index and perform status grading assessment, the health status of the trip circuit is quantified into a traceable scoring indicator, providing a scientific basis for preventative maintenance. By performing time-series analysis and parameter trend discrimination on the health status feature vector, proactive early warning of latent faults such as increased contact resistance at connection points, decreased coil insulation, and loose terminals is achieved. This enables early warning signals to be issued in the early stages of fault deterioration, effectively preventing protection device malfunctions. Through fuzzy matching of fault mode feature libraries and parameter tracing methods, accurate fault location from the circuit level to the component level is achieved. By feeding back actual fault handling results and adaptively adjusting feature weights and diagnostic thresholds, a self-learning mechanism based on equipment deterioration patterns is established. This mechanism can continuously optimize diagnostic strategies without the need for extensive training with historical data, significantly improving adaptability to new equipment and new faults.

[0023] To achieve the above objectives, a third aspect of this application provides a computer device, including a processor and a memory; wherein the processor reads executable program code stored in the memory to run a program corresponding to the executable program code, so as to implement a method for troubleshooting a tripping circuit of a relay protection device as described in the first aspect embodiment.

[0024] To achieve the above objectives, the fourth aspect of this application provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a method for troubleshooting a tripping circuit of a relay protection device as described in the first aspect embodiment.

[0025] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0026] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart of a method for troubleshooting a tripping circuit of a relay protection device according to an embodiment of the present invention; Figure 2 This is a structural diagram of a fault diagnosis device for a relay protection device tripping circuit according to an embodiment of the present invention; Figure 3 It is a computer device according to an embodiment of the present invention. Detailed Implementation

[0027] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0028] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0029] The following describes, with reference to the accompanying drawings, a method and apparatus for troubleshooting trip circuit faults in a relay protection device according to an embodiment of the present invention.

[0030] Figure 1 This is a flowchart of a method for troubleshooting a tripping circuit of a relay protection device according to an embodiment of the present invention, as shown below. Figure 1 As shown, it includes: S1. Deploy sensors at key measurement points in the trip circuit to collect circuit voltage, circuit current, contact resistance, insulation resistance and ambient temperature and humidity parameters in real time; S2. Extract features from the collected parameters, including static feature parameters, dynamic feature parameters, and environmental correlation parameters, and construct a health status feature vector for the trip circuit; S3. Based on the health status feature vector, the health index of the trip circuit is calculated using a weighted fusion method, and the status is graded and evaluated according to the health index threshold. S4. Perform time series analysis on the health status feature vector, and achieve proactive early warning of latent faults through parameter trend discrimination and threshold comparison; S5. Perform fuzzy matching between the current feature vector and the preset fault mode feature library to identify the fault type; combine the tripping circuit topology and the location of abnormal parameter measurement points to locate the fault point and output early warning information through parameter tracing method. S6. Feed back the actual fault handling results to the fault mode feature library, and adaptively adjust the feature weights and diagnostic threshold task decoder base layer weights to perform cross-task constraints, so that the feature mapping of different tasks remains consistent.

[0031] The following describes in detail, with reference to the accompanying drawings, a method for troubleshooting the tripping circuit of a relay protection device according to an embodiment of the present invention.

[0032] The present invention provides a method for troubleshooting trip circuit faults in a relay protection device, comprising the following steps: S10. Deploy sensors at key measurement points in the trip circuit to collect circuit voltage, circuit current, contact resistance, insulation resistance and ambient temperature and humidity parameters in real time. S20. Extract features from the collected parameters, extract static feature parameters, dynamic feature parameters, and environmental correlation parameters, and construct a health status feature vector of the trip circuit; the static feature parameters include circuit voltage deviation rate, current change rate, contact resistance, and insulation resistance reduction rate; the dynamic feature parameters include trip coil operating current characteristics, operating delay time, and current waveform characteristics. S30. Based on the health status feature vector, a weighted fusion method is used to calculate the health index of the trip circuit, and a status classification assessment is performed according to the health index threshold. S40. Perform time series analysis on the health status feature vector, and achieve proactive early warning of latent faults through parameter trend discrimination and threshold comparison; S50. Perform fuzzy matching between the current feature vector and the preset fault mode feature library to identify the fault type; S60. By combining the topology of the tripping circuit and the location of abnormal parameter measurement points, the fault point is located through parameter tracing methods. S70. Feed back the actual fault handling results to the fault mode feature library and adaptively adjust the feature weights and diagnostic thresholds.

[0033] The real-time acquisition in S10 adopts a time-division sampling strategy: under normal operating conditions of the trip circuit, the sampling frequency is 1Hz; when a protection device start signal or a sudden change in circuit parameters is detected, it automatically switches to high-speed sampling mode, increasing the sampling frequency to 10kHz, and continues to acquire data until 500ms after the protection action is completed. The key measurement points include the DC power supply terminal, the protection device output terminal, both ends of the trip coil, the circuit breaker operating circuit terminal, and the intermediate connection terminal of the circuit. During the data acquisition process of the trip circuit, an adaptive time-division sampling strategy is adopted to balance data integrity and storage efficiency. Under normal operating conditions, the sensor acquires parameters at each measurement point at a frequency of 1Hz, and the purpose of sampling at this time is to monitor the gradual change trend of the parameters. When the system detects any condition triggered by the protection device start signal, the circuit voltage drop exceeding 10%, the circuit current sudden change exceeding 50% of the reference value, or the circuit breaker action feedback signal, it automatically switches to high-speed sampling mode.

[0034] In high-speed sampling mode, the sampling frequency is increased to 10kHz, and the acquisition time window starts from the trigger moment and lasts for 500ms after the protection action is completed. This time setting ensures complete recording of the dynamic characteristics of the trip coil from energization to release.

[0035] The deployment of key measuring points covers the entire electrical path of the trip circuit: voltage and current sensors are installed at the DC power supply end to monitor the power supply quality; the output end of the protection device monitors the issuance status of protection action signals; voltage and current sensors are deployed at both ends of the trip coil to capture the coil action characteristics; the circuit breaker operating circuit end monitors the execution of the trip command; and the intermediate connection ends of the circuit monitor important intermediate relay contacts, terminal blocks, and other locations to form a complete parameter acquisition chain.

[0036] The method for extracting static feature parameters in S20 is as follows: Circuit voltage deviation rate:

[0037] Inspection current change rate: ,in The average current over the first 30 consecutive days during the initial operation of the equipment; Contact resistance: The contact resistance is calculated using the voltage drop method by setting voltage measurement points at both ends of the contact. Insulation resistance reduction rate: ; The dynamic characteristic parameters are extracted during the protection operation, including the peak value, rise time, integral area, and decay time constant of the trip coil current waveform, as well as the action delay time from contact closure to opening; the environmental correlation parameters are characterized by a comprehensive temperature and humidity index. ,in For temperature, For relative humidity, and These are weighting coefficients determined based on historical data statistics. Static characteristic parameters are extracted under normal inspection mode. The loop voltage deviation rate reflects power supply stability, and the calculation formula is:

[0038] in, The actual measured voltage at present. The rated operating voltage of the system (usually 110V or 220V DC).

[0039] The rate of change of the monitored current is used to identify changes in loop impedance.

[0040] Reference current The average value of the first 30 days of equipment operation was taken to eliminate unstable factors in the initial operation.

[0041] Contact resistance is measured using the voltage drop method, with a voltage measurement point set before and after the contact.

[0042] This method avoids disconnecting the measurement loop and enables online monitoring.

[0043] The rate of decrease in insulation resistance characterizes the degree of insulation degradation.

[0044] The reference insulation resistance is taken as the initial measurement value when the equipment is put into operation.

[0045] The extraction of dynamic characteristic parameters is performed during the actual tripping process of the protection device, by capturing the current waveform of the trip coil through high-speed sampling. The extracted parameters include: peak current reflecting the coil excitation capability, rise time reflecting the circuit response speed, integral area representing input energy, and decay time constant reflecting energy release characteristics. Simultaneously, the action delay from the issuance of the closing command to the actual opening of the contact is recorded; this parameter directly affects the protection action speed.

[0046] Environmental parameters quantify environmental impact through a comprehensive temperature and humidity index:

[0047] in, The ambient temperature (°C) is the ambient temperature. Relative humidity (%), weighting factor and Based on historical data fitting, humidity typically has a significant impact on insulation performance. The value is set too high.

[0048] The method for calculating the health index of the tripping circuit in S30 is as follows:

[0049] in,

[0050] For the first The deviation value of each static characteristic parameter. The corresponding weights are used; for example, the weight of a decrease in insulation resistance is higher than that of a small voltage fluctuation.

[0051]

[0052] For the first The deviation value of each dynamic characteristic parameter. For the corresponding weights; By performing linear regression analysis on each characteristic parameter, the slope of the parameter change trend is calculated. The larger the absolute value of the slope, the lower the health index. The status grading assessment divides the health index into five levels: In good health; In good condition; To monitor the situation, a warning signal is issued; An alarm signal is issued if the condition is abnormal. In a fault state, an emergency alarm signal is issued; the weight The health index is dynamically adjusted based on equipment type and service life. New equipment focuses on static parameters, while older equipment focuses on trend parameters. The trend health index is calculated by establishing time series models for each characteristic parameter, using least squares linear regression, and calculating the slope of the parameter change trend. A larger absolute value of the slope indicates faster deterioration, and the health index decreases accordingly. A five-level assessment system is established for status grading: Healthy status indicates normal equipment operation; Good status indicates slight deviations that do not affect operation; Caution status issues a yellow warning, suggesting attention; Abnormal status issues an orange alarm, requiring maintenance; Fault status issues a red emergency alarm, requiring immediate action.

[0053] The weighting coefficients are dynamically adjusted based on the equipment's operational phase: newly commissioned equipment (within 3 years) is set... Focusing on the current state; setting up older equipment that has been in operation for more than 5 years. Strengthen trend monitoring.

[0054] The latent fault proactive early warning in S40 is achieved by establishing a multi-level early warning mechanism: For contact resistance, the normal range is set to less than 100mΩ. When the contact resistance is detected to be increasing for 7 consecutive days and exceeding 200mΩ, a first-level warning is issued. When the contact resistance exceeds 500mΩ or the daily increase exceeds 100mΩ, a second-level alarm is issued. For abnormal inspection current, the normal fluctuation range is set to ±10% of the benchmark value. When the fluctuation exceeds ±20% or shows a continuous upward or downward trend for 14 consecutive days, a first-level warning is issued. When the sudden change exceeds ±50%, a second-level alarm is issued. For the decrease in insulation resistance, the normal value range is set to be greater than 10MΩ. When the insulation resistance continuously decreases to below 5MΩ, a first-level warning is issued, and when the insulation resistance is below 2MΩ, a second-level alarm is issued. For deterioration of trip coil operating characteristics, an early warning is issued when the peak trip coil operating current decreases by more than 10% compared to the reference value or the operating delay increases by more than 5ms. The time series analysis uses a sliding window method with a window length of 30 days, and the analysis results are updated daily. Proactive early warning for latent faults is achieved through a differentiated multi-level early warning mechanism. For contact resistance, a dynamic threshold is set: the normal range is less than 100mΩ. When the system detects that the contact resistance shows a monotonically increasing trend for 7 consecutive days and the value exceeds 200mΩ, it is judged as gradual deterioration, and a level one early warning is issued, prompting the scheduling of routine maintenance. When the contact resistance exceeds 500mΩ or the daily increase exceeds 100mΩ, it is judged as an acute fault risk, and a level two alarm is issued, requiring handling within 24 hours.

[0055] The normal fluctuation range for abnormal inspection current is set at ±10% of the baseline value, taking into account the influence of environmental factors such as temperature. A Level 1 warning is issued when the current fluctuation exceeds ±20% or when a continuous unidirectional change trend is detected for 14 consecutive days using the sliding window method; a Level 2 alarm is issued when the current sudden change exceeds ±50%.

[0056] The insulation resistance reduction warning is designed for slow-changing processes such as moisture and pollution: the normal value is greater than 10MΩ. When the insulation resistance continues to drop below 5MΩ, a first-level warning is issued; when it drops below 2MΩ, a second-level alarm is issued, at which point it is approaching the safety threshold.

[0057] The warning of deterioration of trip coil operating characteristics is achieved by comparing with historical benchmark values: a drop in peak operating current of more than 10% indicates a decay in coil excitation capacity, and an increase in operating delay of more than 5ms may cause the protection time to exceed the standard. A warning is issued when any of the conditions are met.

[0058] The time series analysis uses a 30-day sliding window to calculate the regression trend daily, ensuring both statistical validity and sufficient sensitivity.

[0059] The fault mode feature library in S50 includes: Fault modes of contact oxidation or poor contact are characterized by increased contact resistance, decreased inspection current, and increased action delay. The characteristic combination of the trip coil insulation degradation fault mode is reduced insulation resistance, increased inspection current, and increased coil temperature. The terminal loosening fault mode is characterized by large fluctuations in contact resistance and intermittent drops in circuit voltage. The cable strand breakage or wire breakage fault mode is characterized by an increase in total circuit resistance, a decrease in inspection current, and local features without contact points. DC power supply fault mode, characterized by a combination of factors including an overall drop in circuit voltage and simultaneous abnormalities in all circuits; Multi-point concurrent failure mode, characterized by multiple characteristic parameters being abnormal at the same time, and significant differences in parameters between parallel redundant circuits; The fuzzy matching uses a similarity calculation method:

[0060] in For the first One measured feature parameter value, The first in the fault mode library One standard feature parameter value, Select similarity as the feature weight. The most significant fault mode is used as the diagnostic result, and the fault mode feature library is built based on feature combinations of typical fault types. The feature combination of contact oxidation or poor contact is manifested as: a significant increase in contact resistance (usually exceeding 200mΩ), which causes a decrease in inspection current, and a slower contact closing speed during operation, resulting in an increased operation delay.

[0061] The characteristic combination of insulation degradation in the trip coil is: decreased insulation resistance (below 5MΩ), increased leakage current leading to an abnormally high inspection current, and increased coil power consumption causing a rise in temperature. If this fault is not addressed promptly, it may develop into an inter-turn short circuit.

[0062] The characteristics of a loose terminal fault are intermittent: the contact resistance fluctuates greatly, the resistance value jumps when there is vibration or temperature change, and the circuit voltage drops intermittently, but the duration is short.

[0063] Cable strand breakage or wire breakage faults manifest as: increased total circuit resistance and correspondingly decreased inspection current. However, since the breakage location is inside the cable, it is impossible to detect local features through the connection point measurement point. It is necessary to judge in combination with the resistance distribution of the cable section.

[0064] The most obvious characteristics of a DC power supply failure are: the overall circuit voltage drops, and all tripped circuits powered by that power supply simultaneously malfunction. In this case, the power supply system should be checked first.

[0065] The characteristic combination of multi-point concurrent faults is that multiple unrelated characteristic parameters are abnormal at the same time. If there are parallel redundant loops, the comparison parameters show significant differences.

[0066] Fuzzy matching uses weighted similarity calculation:

[0067] in, The current measured number One feature parameter, These are the standard feature values ​​for the corresponding fault modes in the fault database. The feature weights are used to select the fault mode with the highest similarity as the diagnostic result, and the similarity value is output as the diagnostic confidence level.

[0068] The parameter tracing and localization method in S60 includes: First, establish a tripping circuit topology diagram, marking the complete path from the DC power supply to the circuit breaker tripping coil, including all contacts, terminals, cable segments, and sensor measurement point locations; Then, based on the location of the measurement points of the abnormal parameters, the abnormal areas are marked on the topology diagram. When multiple measurement points have abnormal parameters, the transmission path of the parameters in the loop is analyzed, and the common upstream node of the abnormal parameters is determined as the fault point. For parallel redundant circuits, the fault path is isolated by comparing the parameter differences between healthy and abnormal circuits using the differential location method; for series circuits, the fault segment is determined by the segment-by-segment elimination method based on the continuity of parameters at adjacent measuring points. The final output describes the fault location, accurate to the specific device, terminal block, terminal number, or cable section, and also outputs the fault type, characteristic parameter values, health index score, and suggested handling measures; the differential location method calculates... To locate the additional resistor in the faulty section, the parameter tracing and location method first establishes a topology diagram of the tripping circuit. This topology diagram, in single-line form, shows the complete path from the positive and negative DC power supply buses, through the air switch, intermediate relay contacts, protection device output relays, terminal blocks, cables, circuit breaker operating box terminals, and finally to the tripping coil. The installation location and number of all sensor measuring points are marked on the diagram.

[0069] When abnormal parameters occur, mark the abnormal area on the topology diagram. For multiple measurement point anomalies, analyze the electrical connection relationships: if all paths between two measurement points pass through a common node, then that common node is the most likely fault point. For example, if the DC power supply is normal but the protection device output and all subsequent measurement points are abnormal, then the fault lies in the path between the power supply and the protection device.

[0070] For configurations with parallel redundant loops, a differential location method is used by comparing the parameter differences between healthy and abnormal loops:

[0071] The differential resistor is the additional resistor for the faulty section. Combined with the topological relationship, the faulty section can be accurately located.

[0072] For a pure series circuit, a step-by-step elimination method is used: the fault is determined based on the continuity of parameters at adjacent measuring points. If measuring point A is normal while measuring point B is abnormal, the fault is located between A and B; if both measuring points B and C are abnormal but to different degrees, the fault is located between B and C.

[0073] The final output is a detailed fault location report, which includes: the name of the faulty equipment (e.g., the trip circuit of the No. 1 main transformer protection), the specific location (e.g., terminal No. 8 of terminal block No. 2), the fault type (e.g., poor contact), the key characteristic parameter values ​​(e.g., contact resistance 560mΩ), the current health index score, and the recommended handling measures (e.g., tightening the terminal bolts and cleaning the oxide layer).

[0074] The adaptive adjustment mechanism in S70 includes: Establish a fault case database to record complete information for each fault, including fault occurrence time, feature vector, diagnostic results, actual fault location, fault cause, and handling method; Regularly analyze the diagnostic accuracy rate. When the accuracy rate is lower than the preset threshold, perform feature analysis on misjudged cases to identify the key feature parameters that led to the misjudgment. The feature weight coefficients are automatically adjusted using gradient descent. ,in Let be the loss function, representing the deviation between the diagnostic result and the actual fault. The learning rate; For novel faults not included in the fault mode feature library, after confirming the fault type and processing result, the feature mode of the fault is automatically extracted and added to the feature library. For equipment from different manufacturers and of different models, an equipment archive is established to record the rated parameters, baseline parameters, and degradation patterns of the equipment, enabling differentiated parameter threshold configurations. Based on the equipment's operating years, the weighting coefficients of trend analysis and early warning thresholds are dynamically adjusted; the longer the operating years, the higher the trend weight and the more stringent the early warning threshold. An adaptive adjustment mechanism establishes a closed-loop feedback optimization process. The fault case database uses structured storage, with each record containing: a fault occurrence timestamp, time-series data of all characteristic parameters, the fault type and location output by the system diagnostics, the actual fault location and cause found by maintenance personnel, the handling measures taken, and the handling results.

[0075] The system calculates the diagnostic accuracy rate monthly, using the formula: the number of correct diagnoses divided by the total number of diagnoses. When the accuracy rate falls below a preset threshold of 85%, an optimization process is triggered to automatically filter out misdiagnosed cases. Feature comparison analysis is performed on the misdiagnosed cases to identify which feature parameters played a misleading role in the misjudgments.

[0076] Automatic adjustment of feature weights uses the gradient descent algorithm:

[0077] Among them, the loss function The learning rate is defined as the degree of deviation between the diagnostic result and the actual fault. The value is set to 0.01 to ensure adjustment stability. Through iterative optimization, the weight coefficients gradually converge to the optimal value.

[0078] For new faults not found in the fault database, after maintenance personnel confirm the cause of the fault and complete the handling, the system automatically extracts the feature vector at the time the fault occurred, calculates the feature value range of each parameter, and adds it to the feature database as a new fault mode, thereby realizing the automatic expansion of the knowledge base.

[0079] To address the differences between equipment from different manufacturers and models, an equipment archive is established, recording equipment nameplate parameters (rated voltage, rated current, coil resistance, etc.), baseline parameters at commissioning, and degradation curves fitted based on historical data. Parameters are dynamically adjusted according to the equipment's service life: for equipment that has been in operation for more than 10 years, the trend analysis weight is increased from 0.1 to 0.4, and the warning threshold is tightened from 20% deviation from the standard value to 15% deviation.

[0080] The method also includes a collaborative diagnosis step for multi-point concurrent faults: When two or more measurement points are detected to have abnormal parameters at the same time, first determine whether there is a topological relationship between the abnormal measurement points. If there is a common upstream node, it is determined that the multi-point parameter abnormality is caused by a single point fault, and the fault point is located at the common node. If the abnormal measurement points do not have topological association or are located in different parallel branches, they are determined to be multi-point concurrent faults. Fault mode matching and similarity calculation are performed on each abnormal point in turn to establish a fault point priority queue. The method of isolation and verification is adopted: First, the fault point with the highest similarity is processed. After processing, the parameters are re-collected and the health index is calculated. If other abnormal points return to normal, it is confirmed that the single point fault was misjudged as a multi-point fault. If other abnormal points still exist, it is confirmed as a real multi-point concurrent fault, and the next fault point in the priority queue is processed. In the process of multi-point concurrent fault diagnosis, the comparison data of redundant loops is used to establish a parameter comparison matrix. , where matrix elements Indicates the first The measuring point and the first The correlation coefficients of parameters at individual measurement points are used to classify measurement points with correlation coefficients below a threshold as independent fault points. Collaborative diagnosis of multi-point concurrent faults is designed for complex fault scenarios. When the system detects simultaneous alarms from two or more measurement points, it first performs a topology correlation judgment: tracing the upstream paths of each abnormal measurement point on the topology graph to determine if a common node exists. If a common upstream node exists, it is highly likely that a single-point fault at that node caused abnormal parameters at multiple downstream points, and the fault point is located at the common node.

[0081] If the abnormal measurement points do not have topological associations or are located on different parallel branches, they are determined to be real multi-point concurrent faults. In this case, fault mode matching is performed on each abnormal point, and the similarity between each abnormal point and each mode in the fault database is calculated. A fault point priority queue is established according to the similarity from high to low.

[0082] The confirmation process employs a step-by-step isolation verification method: First, the fault point with the highest similarity (i.e., the highest diagnostic confidence) is processed. After processing, the system is restarted to collect parameters and the health index is recalculated. If other abnormal points subsequently return to normal, it indicates that the initial diagnosis was a misjudgment of a single point of failure causing multiple anomalies; if other abnormal points persist, it is confirmed as a genuine multi-point concurrent fault, and the next fault point is processed according to the priority queue.

[0083] During the diagnostic process, comparative data from redundant loops are used to assist in the judgment. A parameter comparison matrix is ​​established. The matrix dimension is the number of measurement points × the number of measurement points, and the elements are... Indicates the measuring point With measuring points The correlation coefficient of the parameters is obtained by calculating the correlation of the time series of parameter changes. Measurement point pairs with a correlation coefficient higher than 0.8 are identified as having the same source of fault, while measurement point pairs with a correlation coefficient lower than 0.3 are identified as having independent faults. This method effectively distinguishes between cascaded faults and independent concurrent faults.

[0084] The triggering conditions for the high-speed sampling mode include: detecting a circuit voltage drop exceeding 10%, a circuit current mutation exceeding 50% of the reference value, the protection device outputting a trip command signal, or receiving a circuit breaker action feedback signal. High-speed sampling data is used to capture the complete process of the trip coil current waveform. By performing a Fourier transform on the current waveform, frequency domain characteristic parameters, including the main frequency component, harmonic content, and spectral energy distribution, are extracted to identify faults such as inter-turn short circuits and core saturation. After sampling, high-speed and low-speed sampling data are stored separately. High-speed data is retained for 90 days for post-event analysis and fault backtracking. The high-speed sampling mode triggering mechanism has multiple triggering conditions; it starts when any condition is met: detecting a circuit voltage drop exceeding 10% of the rated value indicates a possible large current surge or power supply abnormality; a circuit current mutation exceeding 50% of the reference value indicates a sudden change in circuit impedance or protection action; the protection device outputting a trip command signal indicates an imminent trip; or receiving a circuit breaker action feedback signal indicates that the circuit breaker has started operating.

[0085] High-speed sampling captures the complete dynamic process of the trip coil current waveform, including the excitation rise stage, steady-state holding stage, and energy release decay stage. Fourier transform is performed on the acquired waveform data to convert the time-domain signal into a frequency-domain signal, extracting frequency-domain characteristic parameters: the dominant frequency component reflects the inherent characteristics of the coil; harmonic content can identify nonlinear saturation phenomena in the iron core; and abnormal spectral energy distribution can detect inter-turn short-circuit faults in the coil (the spectrum will show an increase in high-frequency components).

[0086] After sampling, the data is stored in a tiered manner: high-speed sampling data (10kHz sampling rate) is stored for 90 days for post-fault analysis and backtracking; low-speed sampling data (1Hz sampling rate) is stored long-term for trend analysis and equipment health management. This tiered storage strategy effectively controls storage costs while ensuring data integrity.

[0087] Under normal operating conditions, the system collects parameters such as voltage, current, and contact resistance at various measuring points in the trip circuit at a frequency of 1Hz. When triggering conditions such as protection device activation, voltage drop, or current surge are detected, the system immediately switches to a 10kHz high-speed sampling mode to fully capture the dynamic characteristics of the tripping process. The collected data is divided into three categories: static characteristic parameters, dynamic characteristic parameters, and environmental related parameters. A multi-dimensional feature extraction algorithm is used to calculate key indicators such as circuit voltage deviation rate, inspection current change rate, contact resistance, insulation resistance reduction rate, and peak current, rise time, and action delay of the trip coil. Based on the extracted feature parameters, the system uses a multi-dimensional weighted fusion algorithm to calculate the trip circuit health index. This index comprehensively considers three dimensions: static health, dynamic health, and trend health, and dynamically adjusts the weighting coefficients according to the equipment's operating years to achieve a quantitative assessment of the overall circuit status and five-level hierarchical management. In the fault diagnosis stage, the system establishes a feature library covering typical fault modes such as contact oxidation, coil insulation degradation, terminal loosening, cable strand breakage, power supply failure, and multi-point concurrent faults. A fuzzy matching algorithm is used to calculate the similarity between the current abnormal features and each mode in the fault library to identify the fault type. Simultaneously, parameter tracing analysis is performed using the trip circuit topology diagram. For series circuits, a segment-by-segment elimination method is employed, while for parallel redundant circuits, a differential location method is used to accurately pinpoint the fault location. The early warning mechanism sets dynamic thresholds and multi-level early warning rules for different parameters, capturing parameter degradation trends through time series analysis to achieve proactive early warning of latent faults. The system also possesses adaptive learning capabilities, statistically analyzing diagnostic accuracy monthly. When accuracy falls below a threshold, feature weights are automatically optimized, and new fault cases are added to the feature library, continuously improving diagnostic accuracy and adaptability.

[0088] To achieve the above embodiments, such as Figure 2As shown, this embodiment also provides a relay protection device trip circuit fault investigation device 10, which includes a sensor deployment and parameter acquisition module 100, a feature extraction and vector construction module 200, a health index calculation and status assessment module 300, a time series analysis and fault identification module 400, a fuzzy matching and early warning output module 500, and a result feedback and adaptive adjustment module 600.

[0089] The sensor deployment and parameter acquisition module 100 deploys sensors at key measurement points in the trip circuit to collect circuit voltage, circuit current, contact resistance, insulation resistance, and ambient temperature and humidity parameters in real time. The feature extraction and vector construction module 200 extracts features from the collected parameters, including static feature parameters, dynamic feature parameters, and environmental correlation parameters, and constructs a health status feature vector for the trip circuit. The health index calculation and status assessment module 300 calculates the health index of the trip circuit based on the health status feature vector using a weighted fusion method, and performs a status classification assessment based on the health index threshold. The time series analysis and fault identification module 400 performs time series analysis on the health status feature vector and achieves proactive early warning of latent faults through parameter trend discrimination and threshold comparison. The fuzzy matching and early warning output module 500 performs fuzzy matching based on the current feature vector and the preset fault mode feature library to identify the fault type; combined with the topology of the tripping circuit and the location of abnormal parameter measurement points, it locates the fault point and outputs early warning information through parameter tracing method. The result feedback and adaptive adjustment module 600 feeds back the actual fault handling results to the fault mode feature library, and adaptively adjusts the feature weights and the decoder base layer weights of the diagnostic threshold task to perform cross-task constraints, so that the feature mapping of different tasks remains consistent.

[0090] Furthermore, the aforementioned sensor deployment and parameter acquisition module 100 is also used for: Under normal operating conditions of the trip circuit, the sampling frequency is 1Hz; when a protection device start signal or a sudden change in circuit parameters is detected, it automatically switches to high-speed sampling mode, increasing the sampling frequency to 10kHz, and continuously collecting data until 500ms after the protection action is completed; the key measuring points include the DC power supply terminal, the protection device output terminal, both ends of the trip coil, the circuit breaker operating circuit terminal, and the intermediate connection terminal of the circuit.

[0091] Furthermore, the aforementioned feature extraction and vector construction module 200 is also used for: The static characteristic parameters include loop voltage deviation rate, current change rate, contact resistance, and insulation resistance reduction rate. The dynamic characteristic parameters include the trip coil operating current characteristics, operating delay time, and current waveform characteristics; The environmental parameters are characterized by a comprehensive temperature and humidity index. ,in For temperature, For relative humidity, and These are weighting coefficients determined based on historical data statistics.

[0092] Furthermore, the aforementioned health index calculation and status assessment module 300 is also used for: The method for calculating the health index of the tripping circuit is as follows:

[0093]

[0094]

[0095] For the first The deviation value of each static characteristic parameter. For the corresponding weights, For the first The deviation value of each dynamic characteristic parameter. For the corresponding weights; The status grading assessment divides the health index into five levels: healthy state, good state, attention state, abnormal state, and fault state.

[0096] Furthermore, the aforementioned time series analysis and fault identification module 400 is also used for: The proactive early warning of latent faults is achieved by establishing a multi-level early warning mechanism; The time series analysis uses the sliding window method with a window length of 30 days, and the analysis results are updated daily.

[0097] Furthermore, the aforementioned fuzzy matching and early warning output module 500 is also used for: The fuzzy matching uses a similarity calculation method:

[0098] in For the first One measured feature parameter value, The first in the fault mode library One standard feature parameter value, Select similarity as the feature weight. The most significant failure mode is used as the diagnostic result; The parameter tracing and localization method includes: Establish a tripping circuit topology diagram and mark the complete path from the DC power supply to the circuit breaker tripping coil; Based on the location of the abnormal parameter measurement points, mark the abnormal area on the topology diagram. When multiple measurement points are abnormal, analyze the transmission path of the parameter in the loop and determine the common upstream node as the fault point. For parallel redundant loops, the fault path is isolated by comparing the parameter differences between healthy and abnormal loops using the differential location method; for series loops, the segment-by-segment elimination method is used to determine the fault segment based on the continuity of parameters of adjacent measuring points and output the fault location description. The differential positioning method calculates... Locate the additional resistor in the faulty section.

[0099] Furthermore, the aforementioned result feedback and adaptive adjustment module 600 is also used for: Establish a database of fault cases, regularly analyze the diagnostic accuracy rate, and when the accuracy rate is lower than the preset threshold, perform feature analysis on the misjudged cases to identify the key feature parameters that led to the misjudgment. The feature weight coefficients are automatically adjusted using gradient descent. ,in Let be the loss function, representing the deviation between the diagnostic result and the actual fault. This is the learning rate.

[0100] This invention discloses a fault diagnosis device for a tripping circuit of a relay protection device, which can realize real-time online monitoring of the operating status of the tripping circuit, effectively identify and warn of hidden faults such as increased contact resistance and insulation deterioration, improve the accuracy of fault diagnosis and positioning, and enhance the reliability and safety of the relay protection system.

[0101] To implement the methods of the above embodiments, the present invention also provides a computer device, such as... Figure 3 As shown, the computer device 600 includes a memory 601 and a processor 602; wherein, the processor 602 reads executable program code stored in the memory 601 to run a program corresponding to the executable program code, so as to implement the various steps of the method described above.

[0102] To implement the above embodiments, this application also proposes a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method described in the foregoing embodiments.

[0103] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0104] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

Claims

1. A method of troubleshooting a trip circuit of a protective relay, the method comprising: include: S1. Deploy sensors at key measurement points in the trip circuit to collect circuit voltage, circuit current, contact resistance, insulation resistance and ambient temperature and humidity parameters in real time; S2. Extract features from the collected parameters, including static feature parameters, dynamic feature parameters, and environmental correlation parameters, and construct a health status feature vector for the trip circuit; S3. Based on the health status feature vector, the health index of the trip circuit is calculated using a weighted fusion method, and the status is graded and evaluated according to the health index threshold. S4. Perform time series analysis on the health status feature vector, and achieve proactive early warning of latent faults through parameter trend discrimination and threshold comparison; S5. Based on the current feature vector and the preset fault mode feature library, perform fuzzy matching to identify the fault type; combine the tripping circuit topology and the location of abnormal parameter measurement points to locate the fault point and output early warning information through parameter tracing method. S6. Feed back the actual fault handling results to the fault mode feature library, and adaptively adjust the feature weights and diagnostic threshold task decoder base layer weights to perform cross-task constraints, so that the feature mapping of different tasks remains consistent.

2. The method of claim 1, wherein, S1 includes: Under normal operating conditions of the trip circuit, the sampling frequency is 1Hz; when a protection device start signal or a sudden change in circuit parameters is detected, it automatically switches to high-speed sampling mode, increasing the sampling frequency to 10kHz, and continuously collecting data until 500ms after the protection action is completed; the key measuring points include the DC power supply terminal, the protection device output terminal, both ends of the trip coil, the circuit breaker operating circuit terminal, and the intermediate connection terminal of the circuit.

3. The method of claim 1, wherein, The S2 includes: The static characteristic parameters include loop voltage deviation rate, current change rate, contact resistance, and insulation resistance reduction rate. The dynamic characteristic parameters include the trip coil operating current characteristics, operating delay time, and current waveform characteristics; The environmental parameters are characterized by a comprehensive temperature and humidity index. ,in For temperature, For relative humidity, and These are weighting coefficients determined based on historical data statistics.

4. The method as described in claim 1, characterized in that, The S3 further includes: The method for calculating the health index of the tripping circuit is as follows: For the first The deviation value of each static characteristic parameter. For the corresponding weights, For the first The deviation value of each dynamic characteristic parameter. For the corresponding weights; The status grading assessment divides the health index into five levels: healthy state, good state, attention state, abnormal state, and fault state.

5. The method as described in claim 1, characterized in that, The S4 includes: The proactive early warning of latent faults is achieved by establishing a multi-level early warning mechanism; The time series analysis uses the sliding window method with a window length of 30 days, and the analysis results are updated daily.

6. The method as described in claim 1, characterized in that, The S5 includes: The fuzzy matching uses a similarity calculation method: in For the first One measured feature parameter value, The first in the fault mode library One standard feature parameter value, Select similarity as the feature weight. The most significant failure mode is used as the diagnostic result; The parameter tracing and localization method includes: Establish a tripping circuit topology diagram and mark the complete path from the DC power supply to the circuit breaker tripping coil; Based on the location of the abnormal parameter measurement points, mark the abnormal area on the topology diagram. When multiple measurement points are abnormal, analyze the transmission path of the parameter in the loop and determine the common upstream node as the fault point. For parallel redundant loops, the fault path is isolated by comparing the parameter differences between healthy and abnormal loops using the differential location method; for series loops, the segment-by-segment elimination method is used to determine the fault segment based on the continuity of parameters of adjacent measuring points and output the fault location description. The differential positioning method calculates... Locate the additional resistor in the faulty section.

7. The method as described in claim 1, characterized in that, The S6 includes: Establish a database of fault cases, regularly analyze the diagnostic accuracy rate, and when the accuracy rate is lower than the preset threshold, perform feature analysis on the misjudged cases to identify the key feature parameters that led to the misjudgment. The feature weight coefficients are automatically adjusted using gradient descent. ,in Let be the loss function, representing the deviation between the diagnostic result and the actual fault. This is the learning rate.

8. A fault diagnosis device for tripping circuits of relay protection devices, characterized in that, include: The sensor deployment and parameter acquisition module deploys sensors at key measurement points in the trip circuit to collect circuit voltage, circuit current, contact resistance, insulation resistance, and ambient temperature and humidity parameters in real time. The feature extraction and vector construction module extracts features from the collected parameters, including static feature parameters, dynamic feature parameters, and environmental correlation parameters, and constructs a health status feature vector for the trip circuit. The health index calculation and status assessment module calculates the health index of the trip circuit based on the health status feature vector using a weighted fusion method, and performs status classification assessment according to the health index threshold. The time series analysis and fault identification module performs time series analysis on the health status feature vector, and realizes proactive early warning of latent faults through parameter trend discrimination and threshold comparison. The fuzzy matching and early warning output module identifies the fault type by performing fuzzy matching between the current feature vector and the preset fault mode feature library; it also locates the fault point and outputs early warning information by combining the topology of the tripping circuit and the location of abnormal parameter measurement points through parameter tracing methods. The result feedback and adaptive adjustment module feeds back the actual fault handling results to the fault mode feature library, and adaptively adjusts the feature weights and the decoder base layer weights of the diagnostic threshold task to perform cross-task constraints, so that the feature mapping of different tasks remains consistent.

9. A computer device, characterized in that, Including processor and memory; The processor reads executable program code stored in the memory to run a program corresponding to the executable program code, so as to implement a method for troubleshooting trip circuits of a relay protection device as described in any one of claims 1-7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements a method for troubleshooting trip circuit faults in a relay protection device as described in any one of claims 1-7.

Images