A substation analog load test fault early warning method based on big data analysis

Abstract

The application discloses a substation simulation load test fault early warning method based on big data analysis and belongs to the technical field of power system automation test and fault diagnosis. The application establishes standard transient response parameter benchmarks under different load levels by constructing a test loop characteristic benchmark library; collects bus voltage, loop current and contact temperature transient data after load step, calculates voltage recovery hysteresis, current overshoot amplitude and contact abnormal temperature rise rate and other multi-dimensional mismatch characteristics; identifies insulation hazards and poor contact risks based on a multi-dimensional evidence fusion rule library, executes graded early warning and generates a fault snapshot; compares the fault snapshot with a historical fault case mapping model, and outputs fault point positioning and maintenance strategies. The application realizes sensitive capture and automatic fault tracing of equipment deterioration precursors in transient response, shortens fault troubleshooting time, and improves substation commissioning efficiency and test safety.

Description

Technical Field

[0001] This invention relates to the field of substation electrical testing technology, and in particular to a fault early warning method for substation simulated load testing based on big data analysis. Background Technology

[0002] As the core node for the collection and distribution of electrical energy in a power system, the operational reliability of a substation's primary equipment and secondary control circuits directly affects the overall safety and stability of the power grid. Before a new substation is put into operation, or after an existing substation's main equipment (such as transformers, busbars, and switchgear) has undergone major overhauls, simulated load testing (i.e., on-load testing) is typically conducted in engineering projects to verify whether key parameters such as voltage, current, impedance, and temperature rise meet design requirements, and to verify the correctness of the action direction, polarity, and logic of protection and control devices. This test simulates changes in electrical quantities under actual operating conditions by injecting step loads into the primary circuit in stages.

[0003] Currently, conventional methods for safety monitoring of simulated load tests mainly rely on two approaches: First, test personnel manually read data from the control panel or monitoring backend to conduct intermittent, random parameter comparisons; second, fixed static safety thresholds (such as overcurrent and overvoltage tripping settings) are preset in the protection or control devices as passive cut-off measures when a fault occurs.

[0004] However, the aforementioned existing technical solutions have significant shortcomings, specifically: First, there is a lack of monitoring capabilities for dynamic response processes. At the moment of step load application, equipment experiences intense electromagnetic and thermal transients. Some early latent defects, such as a slight increase in joint contact resistance, mild localized insulation dampness or aging, may not exceed electrical parameters limits under steady-state operating conditions, but can lead to significantly prolonged transient recovery times or unexpected short-term thermal overshoots. Traditional static threshold methods only focus on steady-state final values ​​and limit boundaries, completely failing to capture abnormal changes in these transient response characteristics, resulting in inherently delayed fault warnings. When such hidden dangers accumulate and evolve into explicit faults such as insulation breakdown or equipment damage, on-site testing often can only be passively interrupted by tripping, failing to effectively guarantee the safety of the testing process and the integrity of the equipment.

[0005] Secondly, post-fault troubleshooting is inefficient and heavily reliant on manual experience. Simulated load test circuits encompass transformers, busbars, circuit breakers, voltage / current transformers, and multiple protection and control devices, all with tight electromagnetic coupling. Once protection actions or parameter anomalies are triggered during testing, the fault typically manifests as multiple electrical quantities, such as voltage, current, and phase angle, simultaneously exceeding limits. A complex nonlinear mapping relationship exists between the fault source and these characteristic quantities. Currently, on-site troubleshooting relies entirely on technicians manually interpreting and comparing chaotic fault waveforms and protection action timing alarm logs. There is a lack of an automated decision-making mechanism that can match and retrieve real-time measured anomalies with existing historical fault cases. This directly leads to difficulties in locating the root cause of the fault, a lengthy troubleshooting process, and consequently, a significantly prolonged power outage testing window, substantially reducing the commissioning efficiency of newly built or overhauled substations. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention aims to provide a fault early warning method for substation simulated load testing based on big data analysis. This method can achieve early automatic warning of anomalies before parameter mismatch and improve the safety, intelligence, and commissioning efficiency of the simulated load testing process.

[0007] The technical solution provided by this invention, as one aspect of this invention, is to overcome the above-mentioned defects of the prior art and achieve the above-mentioned objectives. This invention provides the following technical solution: a substation simulated load test fault early warning method based on big data analysis, comprising: Step S1: Obtain historical load step commands, load test levels, and transient following characteristic data of the target test circuit under historical normal conditions in the substation, and construct a test circuit characteristic benchmark library based on the transient following characteristic data of the target test circuit under different load test levels. Step S2: During the simulated load test of the target test circuit based on the current load step command, the instantaneous bus voltage, instantaneous circuit current and real-time temperature of key contacts are collected simultaneously after the current load step command is triggered, forming a transient monitoring feature set corresponding to the current load step command; Step S3: Based on the transient monitoring feature set and the test circuit characteristic benchmark library, calculate the bus voltage recovery hysteresis, circuit current overshoot amplitude and contact abnormal temperature rise rate under the current load step command, and form a working condition mismatch feature set characterizing the current circuit operating status. Step S4: Based on the operating condition mismatch feature set, identify the insulation hazards or poor contact risks in the target test circuit. When the abnormal temperature rise rate of the contact or the overshoot amplitude of the circuit current exceeds the preset safety tolerance threshold, automatically intercept the subsequent load step command and trigger the circuit disconnection warning. When the circuit disconnection warning is triggered, generate a fault snapshot record based on the operating condition mismatch feature set at the trigger time. Step S5: Compare the fault snapshot record with the pre-built fault cause mapping model, and output the location coordinates of the suspected fault point of the target test circuit and the maintenance and handling guidance strategy based on the comparison results.

[0008] Furthermore, methods for constructing a test loop characteristic benchmark library include: Acquire all historical load step commands executed by the target test circuit under historical normal conditions, and obtain the load test level corresponding to each historical load step command; wherein, the historical load step command includes the command trigger time, the command target load value, the command loading slope, and the command duration; acquire the transient waveform data recorded by the target test circuit during each historical load step command execution, and extract the corresponding transient follow-up characteristic data, including voltage recovery time, current settling time, current peak overshoot, voltage drop depth, and contact temperature rise slope; Each historical load step command is associated with the corresponding load test level and transient following characteristic data to form a transient characteristic record. All transient characteristic records are grouped according to the load test level to form multiple characteristic record groups. For each characteristic record group corresponding to the load test level, statistical analysis is performed on all transient following characteristic data, and the characteristic mean and characteristic standard deviation of each transient following characteristic data are calculated. The corresponding transient characteristic benchmark is formed by integration. The transient characteristic benchmarks of all load test levels are summarized to establish a test loop characteristic benchmark library.

[0009] Furthermore, methods for forming transient monitoring feature sets include: When the current load step command is triggered, the corresponding command trigger time is obtained as the current trigger time. Continuous high-frequency sampling and continuous sampling are performed starting from the current trigger time to obtain the instantaneous bus voltage sequence, instantaneous loop current sequence, and real-time temperature sequence of each key contact. Each instantaneous bus voltage in the instantaneous bus voltage sequence includes the sampling time and the instantaneous voltage value; each instantaneous loop current in the instantaneous loop current sequence includes the sampling time and the instantaneous current value; and each real-time temperature of each key contact in each real-time temperature sequence includes the sampling time and the temperature value. The target load value corresponding to the current load step command is taken as the current target load value, and the rated capacity of the target test circuit is obtained. Based on the current target load value and the rated capacity of the circuit, the load test level corresponding to the current load step command is calculated and marked as the current test level. The instantaneous bus voltage sequence, instantaneous circuit current sequence, real-time temperature sequence of each key contact, current trigger time, current test level and current target load value are integrated to form a transient monitoring feature set.

[0010] Furthermore, methods for forming a set of operating condition mismatch features include: Based on the transient monitoring feature set and the test circuit characteristic benchmark library, the voltage recovery feature set, current overshoot feature set, and contact temperature rise feature set are calculated sequentially, and cross-dimensional abnormal mode correlation analysis is performed to identify collaborative abnormal modes. The voltage recovery feature set, current overshoot feature set, contact temperature rise feature set, collaborative abnormal modes, and current test level are integrated to form the operating condition mismatch feature set. The calculation method for the voltage recovery feature group is as follows: Transient feature point detection is performed on the instantaneous bus voltage sequence to identify the lowest voltage drop point and the steady-state voltage recovery point; based on the lowest voltage drop point and the steady-state voltage recovery point, the current voltage recovery time is calculated, and combined with the characteristic mean of the voltage recovery time in the transient characteristic benchmark, the bus voltage recovery hysteresis is calculated; the bus voltage recovery hysteresis is then standardized and converted to obtain the voltage recovery deviation. For the instantaneous bus voltage sequence, multi-stage recovery curve decomposition analysis and voltage recovery curve smoothness analysis are performed to extract the morphological features and voltage recovery fluctuation of the voltage recovery process. Based on the morphological features of the voltage recovery process, the monotonicity of the recovery slope is checked on the instantaneous bus voltage sequence to obtain the slope anomaly type. The bus voltage recovery hysteresis, voltage recovery deviation, slope anomaly type and voltage recovery fluctuation are integrated to form a voltage recovery feature group.

[0011] Furthermore, the calculation method for the current overshoot characteristic set includes: Envelope peak detection is performed on the instantaneous loop current sequence to identify the current overshoot peak point; the steady-state target current value is calculated, and the current overshoot is calculated based on the absolute value of the steady-state target current value and the instantaneous current value corresponding to the current overshoot peak point; the loop current overshoot amplitude is calculated based on the characteristic mean of the current peak overshoot in the transient characteristic reference; and the current overshoot deviation rate is calculated based on the loop current overshoot amplitude and the steady-state target current value. The instantaneous loop current sequence is subjected to current overshoot waveform morphology analysis to extract the overshoot duration; the instantaneous loop current sequence is subjected to spectrum analysis to extract abnormal harmonic features, and the type of harmonic indication defect is determined based on the abnormal harmonic features; the instantaneous loop current sequence is subjected to current waveform oscillation characteristic analysis to obtain oscillation anomaly indicators; the loop current overshoot amplitude, current overshoot deviation rate, overshoot duration, harmonic indication defect type and oscillation anomaly indicators are integrated to form a current overshoot feature group.

[0012] Furthermore, the calculation method for the contact temperature rise characteristic set includes: Linear fitting is performed on the real-time temperature sequence of each key junction to calculate the current temperature rise slope of each key junction. Combined with the characteristic mean and characteristic standard deviation of the junction temperature rise slope in the transient characteristic benchmark, the temperature rise slope deviation and temperature rise deviation multiple of each key junction are calculated. Based on the temperature rise deviation multiple of each key junction, the temperature rise abnormal junction is identified from all key junctions. The temperature rise abnormal junction with the largest temperature rise slope deviation is marked as the highest temperature rise junction. The temperature rise slope deviation corresponding to the highest temperature rise junction is used as the junction abnormal temperature rise rate, and the junction position code of the highest temperature rise junction is recorded. For each real-time temperature sequence, the absolute value of the junction temperature is checked sequentially to obtain the corresponding maximum temperature exceedance. For each junction with abnormal temperature rise, the junction heat diffusion mode analysis is performed sequentially to obtain the heat source characteristics corresponding to each junction with abnormal temperature rise. For the real-time temperature sequence of the junction with the highest temperature rise, the nonlinear feature extraction of the temperature curve is performed to obtain the corresponding temperature rise acceleration ratio. The junction abnormal temperature rise rate, maximum temperature exceedance, heat source characteristics of all junctions with abnormal temperature rise, junction position encoding of the junction with the highest temperature rise, and temperature rise acceleration ratio are integrated to form a junction temperature rise feature group.

[0013] Furthermore, methods for identifying insulation defects or risks of poor contact include: Multiple rules for determining insulation defects and poor contact are pre-built. Each rule for determining insulation defects has a corresponding defect type label, defect severity level, and rule priority. Each rule for determining poor contact has a corresponding risk type label, risk severity level, and rule priority. The working condition mismatch feature set is matched sequentially with each insulation hazard judgment rule and contact failure judgment rule to obtain the successfully matched insulation hazard judgment rule and contact failure judgment rule, and marked as insulation success rule and contact success rule respectively; the number of insulation success rule and contact success rule is counted to obtain the cumulative number of insulation evidence and the cumulative number of contact evidence. Based on the severity level and priority of the hazard corresponding to each insulation success rule, and the risk severity level and priority of the contact success rule, the main insulation rule and the main contact rule are determined respectively; the hazard type label and hazard severity level corresponding to the main insulation rule are used as the insulation hazard identification result, and the risk type label and risk severity level corresponding to the main contact rule are used as the contact failure identification result; The results of insulation hazard identification, the cumulative number of insulation evidence, the results of poor contact identification, and the cumulative number of contact evidence are integrated to form the circuit risk identification results.

[0014] Furthermore, methods for pre-constructing fault cause mapping models include: Historical fault case data is acquired, which includes multiple historical fault records. Each historical fault record contains historical fault condition characteristics, historical fault cause tags, historical fault location, historical maintenance measures, and historical repair effect evaluations. Data cleaning and quality screening are performed on the historical fault case data. For all retained historical fault records, min-max normalization is performed on the historical fault condition characteristics. Based on the historical fault cause tags, all retained historical fault records are grouped into multiple fault cause category groups. For each fault cause category group, calculate the feature center value and feature distribution interval of each historical fault condition feature parameter. For each fault cause category group, perform feature discrimination evaluation to obtain the feature discrimination capability matrix. For each fault cause category group, construct the corresponding condition feature matching template based on the feature discrimination capability matrix. Construct a fault cause hierarchical classification tree and summarize all condition feature matching templates and fault cause hierarchical classification trees to form a fault cause mapping model.

[0015] Furthermore, methods for comparing fault snapshot records with fault cause mapping models include: The fault snapshot record includes the operating condition mismatch feature set, circuit risk identification results, risk evolution trend, number of consecutive high-risk events, and topology state diagram at the time of the fault. From the fault snapshot record, all parameters in the voltage recovery feature group, current overshoot feature group, and contact temperature rise feature group are extracted and marked as operating condition mismatch feature parameters. Based on the circuit risk identification results, hierarchical screening and comparison are performed on all operating condition feature matching templates to obtain a candidate matching template set. For each operating condition feature matching template in the candidate matching template set, a priority comparison of key indicator features is performed to obtain a pre-screened template set. For each working condition feature matching template in the pre-screened template set, the full feature interval fall-in degree is calculated to obtain the corresponding comprehensive matching score, and the working condition feature matching template with the highest comprehensive matching score is selected as the best matching template. If the feature fall-in degree of the best matching template is greater than or equal to the preset minimum matching confidence threshold, the comparison is determined to be successful; if the feature fall-in degree of the best matching template is less than the minimum matching confidence threshold, the comparison is determined to be unsuccessful, and a similar case reference set is generated.

[0016] Furthermore, methods for outputting the coordinates of suspected fault locations and maintenance guidance strategies include: If the comparison is successful, the set of typical fault location and the set of typical maintenance measures are obtained from the best matching template, and the maintenance urgency label is determined based on the circuit risk identification results; based on the set of typical fault location, it is determined whether to perform multi-evidence fault point screening, and the location coordinates and location credibility of suspected fault points are determined; based on the location coordinates of suspected fault points, a list of recommended maintenance tools and safety protection requirements are obtained; maintenance risk warnings are generated and integrated with the set of typical maintenance measures, maintenance urgency labels, list of recommended maintenance tools and safety protection requirements to form a maintenance handling guidance strategy; If the comparison fails, the location code of the highest temperature rise contact point is used as the location coordinates of the primary suspected fault point; based on the similar case reference set, an auxiliary fault point location set is established and integrated with the location coordinates of the primary suspected fault point to form the location coordinates of the suspected fault point, and the location reliability is determined; based on the circuit risk identification results, the dominant fault type label is determined, and the matching standard troubleshooting steps are obtained; historical maintenance measures for each historical fault record are extracted from the similar case reference set as reference maintenance measures; the standard troubleshooting steps and reference maintenance measures are integrated to form a maintenance handling guidance strategy.

[0017] Implementing the embodiments of the present invention has the following beneficial effects: First, this invention establishes a benchmark library of test circuit characteristics, creating standard transient response characteristic parameters for the target test circuit under different load levels. This effectively overcomes the shortcomings of traditional methods that rely solely on static thresholds and cannot detect dynamic response anomalies. Based on this, multi-stage recovery curve decomposition analysis is used to extract the morphological features of voltage recovery. Fast Fourier Transform is used to extract current anomaly harmonic components, and a harmonic anomaly pattern rule library is used to identify potential equipment defect types. Simultaneously, contact heat diffusion pattern analysis is combined to distinguish between localized and regional heat source anomalies, and the risk of positive feedback deterioration is identified based on the temperature rise acceleration ratio. This achieves sensitive capture of the precursory information of equipment degradation hidden in the transient response waveform.

[0018] Secondly, this invention constructs a multi-dimensional evidence fusion rule base for insulation degradation and contact degradation. Based on multi-dimensional feature combination patterns and rule priorities, it performs multi-evidence fusion adjudication, effectively improving the accuracy and reliability of risk identification. Furthermore, through voltage-current-temperature three-dimensional coordinated anomaly detection, it identifies multi-dimensional coupled fault modes such as all-dimensional coordinated anomalies, electrical coordinated anomalies, electrothermal coordinated anomalies, and pressure-thermal coordinated anomalies. Simultaneously, by leveraging risk evolution trend assessment and continuous high-risk frequency statistics, it achieves real-time tracking and monitoring of the dynamic evolution of risks. A progressive early warning level determination mechanism is designed, implementing graded early warning responses of Level 1, Level 2, and Level 3 based on the severity of the risk, effectively avoiding the traditional single-response mode of immediate shutdown without early warning.

[0019] Finally, this invention constructs a fault cause mapping model that includes feature distribution intervals, feature center values, key indicator feature sets, and a fault cause hierarchical classification tree. This enables coarse-to-fine hierarchical screening and comparison, as well as priority comparison of key features, significantly improving comparison efficiency and accuracy. Furthermore, by utilizing topological path reasoning for localization and fault propagation link analysis, it can accurately screen faults in multi-candidate fault point scenarios based on current paths and upstream / downstream relationships, identifying the path of fault propagation from the source to other parts of the circuit. The final output is a fault warning and handling report containing the location coordinates of suspected fault points, location confidence, fault propagation link description, and priority-ordered maintenance measures. This effectively shortens fault investigation time and improves substation commissioning efficiency and overall safety during testing. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, obtaining other drawings based on these drawings without creative effort still falls within the scope of the present invention. Figure 1 This is a flowchart of the substation simulated load test fault early warning method based on big data analysis according to Embodiment 1 of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings.

[0022] Example 1 Please see Figure 1 The image illustrates a substation simulated load test fault early warning method based on big data analysis as described in this embodiment. In this embodiment, the method includes: Step S1: Obtain historical load step commands, load test levels, and transient following characteristic data of the target test circuit under historical normal conditions in the substation, and construct a test circuit characteristic benchmark library based on the transient following characteristic data of the target test circuit under different load test levels. Step S2: During the simulated load test of the target test circuit based on the current load step command, the instantaneous bus voltage, instantaneous circuit current and real-time temperature of key contacts are collected simultaneously after the current load step command is triggered, forming a transient monitoring feature set corresponding to the current load step command; Step S3: Based on the transient monitoring feature set and the test circuit characteristic benchmark library, calculate the bus voltage recovery hysteresis, circuit current overshoot amplitude and contact abnormal temperature rise rate under the current load step command, and form a working condition mismatch feature set characterizing the current circuit operating status. Step S4: Based on the operating condition mismatch feature set, identify the insulation hazards or poor contact risks in the target test circuit. When the abnormal temperature rise rate of the contact or the overshoot amplitude of the circuit current exceeds the preset safety tolerance threshold, automatically intercept the subsequent load step command and trigger the circuit disconnection warning. When the circuit disconnection warning is triggered, generate a fault snapshot record based on the operating condition mismatch feature set at the trigger time. Step S5: Compare the fault snapshot record with the pre-built fault cause mapping model, and output the location coordinates of the suspected fault point of the target test circuit and the maintenance and handling guidance strategy based on the comparison results.

[0023] The following will illustrate each step in this embodiment with specific examples.

[0024] Specifically, step S1: Obtain historical load step commands, load test levels, and transient following characteristic data of the target test circuit under historical normal conditions in the substation, and construct a test circuit characteristic benchmark library based on the transient following characteristic data of the target test circuit under different load test levels.

[0025] The methods for obtaining historical load step commands and load test levels include: Retrieve all historical load step commands executed under historical normal conditions from the test command archive database of the substation integrated automation system. The target test circuit refers to the primary or secondary circuit of the substation currently undergoing simulated load testing, such as the main transformer circuit, bus circuit, outgoing line circuit, and station service transformer circuit. Historical normal condition refers to the normal operating state of the target test circuit during historical simulated load testing, without any fault warnings, protection actions, or parameter over-limits. The historical load step command includes the command trigger time, the command target load value, the command loading slope, and the command duration. The command trigger time is the moment when the current step load begins to load; the command target load value is the target power value for this step load loading; the command loading slope is the loading rate at which the load jumps from zero to the command target load value; and the command duration is the length of time the current step load is maintained at the command target load value.

[0026] The load test level corresponding to each historical load step command is retrieved from the test command archive database. The load test level characterizes the ratio of the step load to the circuit's rated capacity, including light load, half load, heavy load, and full load levels. The circuit's rated capacity refers to the maximum power that the target test circuit can carry under normal operating conditions, determined by the rated parameters of the main equipment in the circuit and stored in the substation equipment ledger. The light load level corresponds to a step load of less than 25% of the circuit's rated capacity; the half load level corresponds to a step load of 25% to 50% of the circuit's rated capacity; the heavy load level corresponds to a step load of 50% to 80% of the circuit's rated capacity; and the full load level corresponds to a step load of 80% to 100% of the circuit's rated capacity. The load test level is automatically classified and stored by the test command archive database based on the ratio of the target load value of the command to the circuit's rated capacity.

[0027] The methods for obtaining transient tracking characteristic data include: The transient waveform data recorded during each historical load step command execution of the target test circuit is obtained from the historical waveform storage library of the substation fault recording device. The transient waveform data includes bus voltage transient waveforms, loop current transient waveforms, and contact temperature transient curves. The bus voltage transient waveform is a high-frequency sampling sequence of the bus voltage at the instant of step load loading, with a sampling frequency of not less than 10kHz. The loop current transient waveform is a high-frequency sampling sequence of the loop current at the instant of step load loading. The contact temperature transient curve is time-series sampling data of the critical contact temperatures before and after step load loading, with a sampling period of not more than 1 second. Critical contacts refer to electrical connection points in the target test circuit that are prone to poor contact, such as switch contacts, bus lap joints, and cable terminations. The specific location of critical contacts is defined by the contact distribution list in the substation equipment ledger.

[0028] Based on transient waveform data, transient following characteristic data corresponding to each historical load step command is extracted. This transient following characteristic data includes voltage recovery time, current settling time, peak current overshoot, voltage sag depth, and contact temperature rise slope. Specifically, the rated bus voltage of the target test circuit is obtained from the substation equipment ledger. The rated bus voltage refers to the nominal voltage value of the bus connected to the target test circuit under normal operating conditions. The steady-state target current value is calculated based on the commanded target load value and the rated bus voltage. Specifically, if the target test circuit is an AC circuit, the steady-state target current value is calculated using the apparent power formula; if the target test circuit is a DC circuit, the steady-state target current value is calculated using the division relationship between power and voltage. The voltage recovery time is the time required for the bus voltage to recover from the lowest point of the step load to 95% of the rated bus voltage. This is obtained by detecting the time difference between the lowest point of the bus voltage sag and the first time the rated bus voltage reaches 95% in the transient waveform of the bus voltage. The current settling time is the time required for the loop current to rise to 90% of the steady-state target current value from the moment the command is triggered. The peak current overshoot is the difference between the maximum peak value in the transient waveform of the loop current and the steady-state target current value. The voltage sag depth is the difference between the lowest value in the transient waveform of the bus voltage and the rated voltage of the bus. The contact temperature rise slope is the average rate of temperature rise of the contact within a preset observation window after a step load is applied. It is obtained by calculating the difference between the end temperature and the start temperature of the observation and dividing it by the observation window duration; the observation window duration is preset by those skilled in the art based on the thermal inertia characteristics of the contact.

[0029] The method for constructing a test loop characteristic benchmark library includes: associating each historical load step command with the corresponding load test level and transient following characteristic data to form transient characteristic records; grouping all transient characteristic records according to the load test level to form light load characteristic record group, half load characteristic record group, heavy load characteristic record group, and full load characteristic record group; for each characteristic record group corresponding to each load test level, performing statistical analysis on each transient following characteristic data in all transient characteristic records, and calculating the characteristic mean and characteristic standard deviation of each transient following characteristic data; integrating the characteristic mean and characteristic standard deviation of each transient following characteristic data corresponding to each load test level to form the transient characteristic benchmark for the corresponding load test level; and summarizing the transient characteristic benchmarks for all load test levels to establish a test loop characteristic benchmark library.

[0030] Specifically, step S2: During the simulated load test of the target test circuit based on the current load step command, the instantaneous bus voltage, instantaneous circuit current and real-time temperature of key contacts after the current load step command is triggered are collected simultaneously to form a transient monitoring feature set corresponding to the current load step command.

[0031] The methods for acquiring instantaneous bus voltage include: A high-precision voltage transformer and a high-speed data acquisition unit are deployed on the bus side of the target test circuit. When the current load step command is triggered, the command trigger time of the current load step command is acquired and marked as the current trigger time. Taking the current trigger time as the acquisition start point, the bus voltage signal output by the high-precision voltage transformer is continuously sampled at high frequency by the high-speed data acquisition unit to obtain multiple instantaneous bus voltages. The sampling frequency corresponding to the continuous high-frequency sampling is consistent with the sampling frequency of the fault recording device in step S1. All instantaneous bus voltages between the acquisition start point and the preset transient observation cutoff time constitute an instantaneous bus voltage sequence. The transient observation cutoff time is the time corresponding to the preset transient observation duration after the acquisition start point. The transient observation duration is preset by those skilled in the art based on the typical transient response cycle of the target test circuit. Each instantaneous bus voltage in the instantaneous bus voltage sequence includes the sampling time and the instantaneous voltage value.

[0032] The target load value corresponding to the current load step command is taken as the current target load value. Based on the current target load value and the rated capacity of the target test loop, the load test level corresponding to the current load step command is calculated and marked as the current test level. Specifically, the ratio of the current target load value to the rated capacity of the loop is calculated to obtain the load ratio; the corresponding load test level is determined according to the level range into which the load ratio falls; wherein, the rule for dividing the level range is the same as the rule for dividing the load test level in step S1.

[0033] The method for acquiring instantaneous loop current includes: deploying a high-precision current transformer and a high-speed data acquisition unit on the main current path of the target test loop; taking the current triggering time as the acquisition starting point, continuously sampling the loop current signal output by the high-precision current transformer through the high-speed data acquisition unit to obtain multiple instantaneous loop currents; constructing an instantaneous loop current sequence from the acquisition starting point to the transient observation cutoff time; each instantaneous loop current in the instantaneous loop current sequence includes the sampling time and the instantaneous current value.

[0034] The method for acquiring real-time temperatures of key contacts includes: deploying infrared or fiber optic temperature sensors at each key contact location in the target test circuit; taking the current triggering time as the acquisition starting point, continuously sampling the surface temperature of the corresponding key contact through each temperature sensor to obtain multiple real-time temperatures of key contacts; wherein, the sampling period corresponding to the continuous sampling is consistent with the sampling period of the transient temperature curve of the contact in step S1; the real-time temperatures of all key contacts between the acquisition starting point and the transient observation cutoff time constitute the real-time temperature sequence of each key contact; each real-time temperature of each key contact in each real-time temperature sequence includes the sampling time and temperature value.

[0035] The method for forming a transient monitoring feature set includes: integrating the instantaneous bus voltage sequence, instantaneous loop current sequence, real-time temperature sequence of each key contact, current trigger time, current test level, and current target load value to form a transient monitoring feature set; wherein, the transient monitoring feature set is used to completely record the raw data of the transient response of the target test loop and the test condition parameters after the current load step command is triggered.

[0036] Specifically, step S3: Based on the transient monitoring feature set and the test circuit characteristic benchmark library, calculate the bus voltage recovery hysteresis, circuit current overshoot amplitude and contact abnormal temperature rise rate under the current load step command, and form a working condition mismatch feature set characterizing the current circuit operating status.

[0037] The methods for calculating the bus voltage recovery hysteresis include: From the transient monitoring feature set, the instantaneous bus voltage sequence and the current test level are obtained. Transient feature point detection is performed on the instantaneous bus voltage sequence to identify the lowest voltage drop point and the voltage recovery steady-state point. Specifically, all instantaneous bus voltages in the instantaneous bus voltage sequence are traversed, and the instantaneous bus voltage with the smallest corresponding instantaneous voltage value is identified and marked as the lowest voltage drop point. The sampling time and instantaneous voltage value corresponding to the lowest voltage drop point are marked as the lowest voltage drop time and the lowest voltage drop value, respectively. The rated bus voltage of the target test circuit is obtained from the substation equipment ledger. 95% of the rated bus voltage is calculated to obtain the steady-state recovery judgment threshold. From the instantaneous bus voltages after the lowest voltage drop point, the first instantaneous bus voltage with a voltage instantaneous value greater than or equal to the steady-state recovery judgment threshold is retrieved in chronological order and marked as the voltage recovery steady-state point. The sampling time and instantaneous voltage value corresponding to the voltage recovery steady-state point are marked as the steady-state recovery time and the steady-state recovery voltage value, respectively.

[0038] The difference between the steady-state recovery time and the lowest voltage drop time is calculated to obtain the current voltage recovery time. The transient characteristic benchmark corresponding to the current test level is obtained from the test loop characteristic benchmark library, and the average characteristic value of the voltage recovery time is obtained from the transient characteristic benchmark to obtain the benchmark voltage recovery time. The difference between the current voltage recovery time and the benchmark voltage recovery time is calculated to obtain the bus voltage recovery hysteresis. The bus voltage recovery hysteresis characterizes the delay in the transient recovery speed of the bus voltage under the current load step command relative to the historical normal state. A positive bus voltage recovery hysteresis value indicates that the current voltage recovery speed is slower than the historical normal level, which may indicate abnormal loop impedance or changes in load characteristics.

[0039] The bus voltage recovery hysteresis is standardized and converted to obtain the voltage recovery deviation. Specifically, the standard deviation of the voltage recovery time is obtained from the transient characteristic reference as the reference voltage recovery standard deviation, and the ratio of the bus voltage recovery hysteresis to the reference voltage recovery standard deviation is calculated to obtain the voltage recovery deviation. The voltage recovery deviation is used to eliminate the influence of differences in the dimensions of voltage recovery time under different load test levels, making the hysteresis levels comparable. The larger the absolute value of the voltage recovery deviation, the more significantly the current voltage recovery characteristic deviates from the historical normal range.

[0040] Multi-stage recovery curve decomposition analysis is performed on the instantaneous bus voltage sequence to extract the morphological features of the voltage recovery process. Specifically, the voltage recovery segment is constructed based on all instantaneous bus voltages between the lowest voltage drop point and the steady-state voltage recovery point. The voltage recovery segment is then divided into a first recovery interval, a middle recovery interval, and a last recovery interval in chronological order. For each recovery interval, the least squares method is used to perform linear fitting on all instantaneous bus voltages to obtain the corresponding fitted straight line. The slope values ​​of each fitted straight line are calculated sequentially to obtain the slope of the first, middle, and last recovery segments.

[0041] Based on the morphological characteristics of the voltage recovery process, a monotonicity check of the recovery slope is performed on the instantaneous bus voltage sequence to obtain the slope anomaly type. Specifically, under normal circumstances, the voltage recovery process exhibits an exponential decay characteristic of being fast at first and slow at the end, i.e., the recovery slope in the first stage should be greater than the recovery slope in the middle stage, and the recovery slope in the middle stage should be greater than the recovery slope in the last stage. The recovery slopes in the first, middle, and last stages are compared separately. If the recovery slope in the first stage is less than or equal to the recovery slope in the middle stage, the slope anomaly type is the middle stage sluggish type, indicating that the voltage recovery process is delayed in the middle stage, which may be due to dynamic changes in the loop impedance during the recovery process. If the recovery slope in the middle stage is less than or equal to the recovery slope in the last stage, the slope anomaly type is the last stage rebound type, indicating that the voltage recovery process experiences abnormal fluctuations in the later stage, which may be due to unstable load characteristics or abnormal response of the voltage regulator. If the recovery slope in the first stage is greater than the recovery slope in the middle stage and the recovery slope in the middle stage is greater than the recovery slope in the last stage, the slope anomaly type is the normal type throughout.

[0042] For the voltage recovery segment, a smoothness analysis of the voltage recovery curve is performed to obtain the voltage recovery fluctuation. Specifically, the standard deviation of the difference between the instantaneous values ​​of adjacent instantaneous bus voltages in the voltage recovery segment is calculated to obtain the voltage recovery fluctuation. The voltage recovery fluctuation is used to characterize the smoothness of the voltage recovery process. The larger the voltage recovery fluctuation, the more high-frequency disturbance components are present in the recovery process, which may be due to small arc discharges caused by poor contact or loose parts in the circuit.

[0043] The bus voltage recovery hysteresis, voltage recovery deviation, slope anomaly type, and voltage recovery fluctuation are integrated to form a voltage recovery characteristic group.

[0044] The methods for calculating the overshoot value of the loop current include: Instantaneous loop current sequences are obtained from the transient monitoring feature set. Envelope peak detection is performed on the instantaneous loop current sequences to identify current overshoot peak points. Specifically, all instantaneous loop currents in the instantaneous loop current sequence are traversed, and the absolute value of the instantaneous current value corresponding to each instantaneous loop current is taken to obtain the corresponding absolute current value. The instantaneous loop current with the largest corresponding absolute current value is identified as the current overshoot peak point. The sampling time and current absolute value corresponding to the current overshoot peak point are marked as peak time and peak current value, respectively. The difference between the peak current value and the steady-state target current value is calculated and its absolute value is taken to obtain the current overshoot. The characteristic mean of the current peak overshoot is obtained from the transient characteristic reference and marked as the reference current overshoot. The difference between the current current overshoot and the reference current overshoot is calculated to obtain the loop current overshoot amplitude. The loop current overshoot amplitude is used to characterize the degree of deviation of the transient impact intensity of the loop current under the current load step command from the historical normal state. A positive loop current overshoot amplitude indicates that the current impact intensity exceeds the historical normal level. Calculate the ratio of the loop current overshoot amplitude to the steady-state target current value, and multiply it by 100% to obtain the current overshoot deviation rate. The current overshoot deviation rate is used to convert the absolute overshoot amplitude into a percentage form relative to the steady-state target current, which facilitates a unified comparison with the safety tolerance threshold under different load levels.

[0045] An overshoot waveform morphology analysis is performed on the instantaneous loop current sequence to extract the overshoot duration characteristics. Specifically, the product of the steady-state target current value and a preset overshoot judgment factor is calculated to obtain the overshoot judgment line; the overshoot judgment factor is preset by those skilled in the art based on the transient current withstand capability of the loop equipment; the moment when the absolute value of the current first crosses the overshoot judgment line is identified from the instantaneous loop current sequence and marked as the overshoot start time; the sampling moment when the absolute value of the current first crosses the overshoot judgment line from the instantaneous loop current after the overshoot start time is identified and marked as the overshoot end time; the difference between the overshoot end time and the overshoot start time is calculated to obtain the overshoot duration. The overshoot duration characterizes the length of the current overshoot state on the time axis; a longer overshoot duration indicates a weaker ability of the loop to suppress inrush current.

[0046] Spectral analysis is performed on the instantaneous loop current sequence to extract abnormal harmonic features, and the type of harmonic indication defect is determined based on these features. Specifically, a fast Fourier transform is performed on the instantaneous loop current sequence to obtain the current spectrum distribution. A preset harmonic detection threshold is established, which is pre-set by a person skilled in the art based on the harmonic background level during normal loop operation. The rated power frequency of the target test loop is obtained from the substation equipment register. In the current spectrum distribution, the frequency component with the largest amplitude within a preset tolerance range, centered on the rated power frequency, is selected as the fundamental frequency, and the corresponding frequency value is taken as the fundamental frequency, and the corresponding amplitude is taken as the fundamental amplitude. Frequency components whose frequency values ​​fall within the preset tolerance range of the fundamental frequency are removed from the current spectrum distribution, and frequency components with amplitudes exceeding the harmonic detection threshold are identified from the remaining frequency components to obtain significant harmonic components. The system identifies the harmonic component with the largest amplitude from the significant harmonic components and obtains its corresponding frequency and amplitude, which are labeled as the dominant harmonic frequency and dominant harmonic amplitude, respectively. The number of significant harmonic components is counted to obtain the number of abnormal harmonics. The frequency values ​​corresponding to all significant harmonic components are counted, and the difference between the largest and smallest frequency values ​​is calculated to obtain the abnormal harmonic distribution bandwidth. If the dominant harmonic frequency is an odd multiple of the fundamental frequency and the dominant harmonic amplitude exceeds a preset proportion of the fundamental amplitude, the harmonic indication defect type is determined to be core magnetic saturation. If the ratio of the dominant harmonic frequency to the fundamental frequency is not an integer, the harmonic indication defect type is determined to be partial discharge or arcing. If the number of abnormal harmonics exceeds a preset quantity threshold and the abnormal harmonic distribution bandwidth is greater than a preset bandwidth threshold, the abnormal harmonic components are determined to be distributed within a wide frequency band, and the harmonic indication defect type is determined to be insulation degradation. If none of the above conditions are met, the harmonic indication defect type is determined to be no abnormality. The preset proportion, quantity threshold, and bandwidth threshold are all preset by those skilled in the art based on the typical characteristics of harmonic properties and the required identification sensitivity under each abnormal state.

[0047] An oscillation characteristic analysis is performed on the instantaneous loop current sequence to obtain an oscillation anomaly flag. Specifically, for all instantaneous loop currents after the current overshoot peak point in the instantaneous loop current sequence, the instantaneous loop currents with corresponding absolute values ​​of local maxima are detected and marked as local maxima points. The number of local maxima points is counted to obtain the number of local maxima. If the number of local maxima is greater than or equal to 2, the number of local maxima is reduced by one to obtain the number of oscillation periods. If the number of local maxima is less than 2, the number of oscillation periods is set to 0. The average of the differences in the absolute values ​​of the currents corresponding to all adjacent local maxima points is calculated to obtain the oscillation attenuation amplitude. If the number of oscillation periods is greater than a preset upper limit for the number of normal oscillations or the oscillation attenuation amplitude is less than a preset lower limit for normal attenuation, the oscillation anomaly flag is marked as present. If the number of oscillation periods is less than or equal to the upper limit for the number of normal oscillations and the oscillation attenuation amplitude is greater than or equal to the lower limit for normal attenuation, the oscillation anomaly flag is marked as absent. The oscillation anomaly flag is used to indicate whether the loop damping characteristics have deteriorated. The upper limit for the number of normal oscillations and the lower limit for normal attenuation are both preset by those skilled in the art based on the damping oscillation characteristics of the target test loop under normal conditions.

[0048] The overshoot amplitude of the loop current, the overshoot deviation rate, the overshoot duration, the type of harmonic indication defect, and the oscillation anomaly indicator are integrated to form a current overshoot characteristic group.

[0049] The methods for calculating the abnormal temperature rise rate of the contact include: The real-time temperature sequence of each key node is obtained from the transient monitoring feature set; for the real-time temperature sequence of each key node, the least squares method is used to perform linear fitting on the real-time temperature of all key nodes to obtain the temperature fitting line corresponding to each key node; the slope value of each temperature fitting line is calculated in turn to obtain the current temperature rise slope of each key node.

[0050] The mean and standard deviation of the temperature rise slope of the contact are obtained from the transient characteristic reference and labeled as the reference temperature rise slope and reference temperature rise standard deviation, respectively. For each critical contact, the difference between the current temperature rise slope and the reference temperature rise slope is calculated to obtain the temperature rise slope deviation. The ratio of the temperature rise slope deviation to the reference temperature rise standard deviation is calculated to obtain the temperature rise deviation multiple. If the temperature rise deviation multiple is greater than a preset temperature rise anomaly judgment threshold, the corresponding critical contact is marked as a temperature rise anomaly contact. If the temperature rise deviation multiple is less than or equal to the temperature rise anomaly judgment threshold, the corresponding critical contact is not marked. The temperature rise anomaly judgment threshold is preset by those skilled in the art based on the thermal stability requirements of the critical contact.

[0051] For all abnormal temperature rise contacts, the maximum value of the temperature rise slope deviation is obtained and used as the abnormal temperature rise rate of the contact. The abnormal temperature rise rate characterizes the degree of temperature rise acceleration of the critical contact with the most severe temperature rise in the target test circuit relative to its historical normal state. The abnormal temperature rise contact with the largest temperature rise slope deviation is marked as the highest temperature rise contact, and its location code is recorded. The location code is obtained from the contact distribution list in the substation equipment ledger.

[0052] For each real-time temperature sequence, the absolute value of the contact temperature is checked sequentially to obtain the corresponding maximum temperature exceedance. Specifically, from the real-time temperature sequence of each critical contact, the temperature value corresponding to the latest real-time temperature of the critical contact at the corresponding sampling time is obtained and used as the transient end temperature; from the substation equipment ledger, the allowable temperature limit of the equipment materials corresponding to each critical contact is obtained; the difference between the transient end temperature and the allowable temperature limit is calculated to obtain the temperature margin; if the temperature margin is negative, the corresponding critical contact is marked as a temperature exceedance contact; for all temperature exceedance contacts, the maximum absolute value of the corresponding temperature margin is obtained and used as the maximum temperature exceedance.

[0053] For each abnormal temperature rise contact, a contact heat diffusion mode analysis is performed sequentially to obtain the corresponding heat source characteristics. Specifically, the spatial topology relationship of each key contact is obtained from the substation equipment ledger. The spatial topology relationship is used to describe the adjacent connection relationship of each key contact in the circuit physical structure. For each abnormal temperature rise contact, the key contacts adjacent to it are obtained from the spatial topology relationship to form a neighborhood contact set. The difference between the current temperature rise slope of the abnormal temperature rise contact and the current temperature rise slope of each key contact in the corresponding neighborhood contact set is calculated to obtain multiple temperature rise slope differences. If all temperature rise slope differences are greater than the product of the baseline temperature rise standard deviation and a preset multiple, the heat source characteristics of the corresponding temperature rise anomaly junction are marked as local heat source type, indicating that the corresponding temperature rise anomaly junction is an independent heating point. If the number of critical junctions in the neighboring junction set that meet the temperature rise condition is greater than or equal to a preset similar number threshold, the heat source characteristics of the corresponding temperature rise anomaly junction are marked as regional heat source type, indicating that there is a regional thermal anomaly with multiple critical junctions generating heat together. Specifically, the temperature rise condition is that the absolute value of the temperature rise slope difference is less than or equal to the product of the baseline temperature rise standard deviation and a preset similar multiple. If none of the above conditions are met, the heat source characteristics of the corresponding temperature rise anomaly junction are marked as transitional diffusion type, indicating that the thermal anomaly of the corresponding temperature rise anomaly junction is in a transitional state from local heating to neighboring diffusion. All relevant thresholds are preset by those skilled in the art according to the accuracy requirements of thermal diffusion analysis.

[0054] For the real-time temperature sequence of the highest temperature rise contact point, nonlinear feature extraction of the temperature curve is performed to obtain the corresponding temperature rise acceleration ratio. Specifically, for the real-time temperature sequence of the highest temperature rise contact point, the temperature sequence is divided into a first half and a second half; linear fitting and slope calculation are performed on the first half and the second half respectively to obtain the temperature rise slope of the first half and the temperature rise slope of the second half; the ratio of the temperature rise slope of the second half to the temperature rise slope of the first half is calculated to obtain the temperature rise acceleration ratio. The temperature rise acceleration ratio is used to characterize whether the temperature rise process is uniform or accelerated; a temperature rise acceleration ratio greater than 1 indicates that the temperature rise process exhibits accelerated characteristics, which may indicate that the contact resistance further increases with increasing temperature, posing a risk of positive feedback deterioration.

[0055] The abnormal temperature rise rate of the contact, the maximum temperature exceedance, the heat source characteristics of all abnormal temperature rise contacts, the contact location code of the highest temperature rise contact, and the temperature rise acceleration ratio are integrated to form a contact temperature rise feature group.

[0056] The methods for forming a set of operating condition mismatch features characterizing the current operating state of the circuit include: Based on the voltage recovery characteristic group, current overshoot characteristic group, and contact temperature rise characteristic group, cross-dimensional anomaly pattern correlation analysis is performed to identify collaborative anomaly patterns. Specifically, it is determined whether the voltage recovery deviation, current overshoot deviation rate, and contact abnormal temperature rise rate simultaneously exceed their respective preset collaborative anomaly thresholds. If all three exceed the corresponding collaborative anomaly thresholds, the collaborative anomaly pattern is determined to be a full-dimensional collaborative anomaly, indicating significant deviations in all three dimensions: voltage, current, and temperature. If only the voltage recovery deviation and current overshoot deviation rate exceed the corresponding collaborative anomaly thresholds, the collaborative anomaly pattern is determined to be an electrical collaborative anomaly. If only the current overshoot deviation rate and contact abnormal temperature rise rate exceed the corresponding collaborative anomaly thresholds, the collaborative anomaly pattern is determined to be an electrothermal collaborative anomaly. If only the voltage recovery deviation and contact abnormal temperature rise rate exceed the corresponding collaborative anomaly thresholds, the collaborative anomaly pattern is determined to be a pressure-thermal collaborative anomaly. If only one parameter exceeds the corresponding collaborative anomaly threshold, the collaborative anomaly pattern is determined to be a single-dimensional anomaly. If all three are less than or equal to the corresponding collaborative anomaly thresholds, the collaborative anomaly pattern is determined to be without collaborative anomaly. The thresholds for each collaborative anomaly are pre-set by those skilled in the art based on the strictness of the multi-dimensional collaborative anomaly determination.

[0057] The voltage recovery feature set, current overshoot feature set, contact temperature rise feature set, collaborative abnormal mode, and current test level are integrated to form the operating condition mismatch feature set. Among them, the operating condition mismatch feature set is used to comprehensively characterize the degree of deviation and specific abnormal characteristics of the target test circuit under the current load step command relative to the historical normal state in three dimensions: voltage response, current response, and thermal response.

[0058] Specifically, step S4: Based on the operating condition mismatch feature set, identify the insulation hazards or poor contact risks in the target test circuit. When the abnormal temperature rise rate of the contact or the overshoot amplitude of the circuit current exceeds the preset safety tolerance threshold, automatically intercept the subsequent load step command and trigger the circuit disconnection warning. When the circuit disconnection warning is triggered, generate a fault snapshot record based on the operating condition mismatch feature set at the trigger time.

[0059] The method for identifying insulation hazards in the target test circuit includes: pre-constructing a multi-dimensional evidence fusion rule base for insulation degradation. This rule base contains multiple rules for determining insulation hazards, along with corresponding hazard type labels, hazard severity levels, and rule priorities. Each insulation hazard determination rule is pre-defined by a person skilled in the art based on a combination of multi-dimensional characteristics of insulation faults in substation equipment. Different rules have different priorities, which are pre-set by a person skilled in the art based on the rate of hazard evolution and the urgency of on-site emergency response.

[0060] Specifically, the rule base for multidimensional evidence fusion of insulation degradation includes the following rules for determining insulation hazards: First rule for determining insulation hazards: If the current overshoot deviation rate is greater than the preset severe overshoot threshold, and the overshoot duration is greater than the preset duration threshold, and the harmonic indication defect type is partial discharge or arc, then the hazard type label is determined to be insulation breakdown precursor hazard, and the hazard severity level is high risk level.

[0061] Second rule for determining insulation hazards: If the current overshoot deviation rate is greater than the severe overshoot threshold, and the coordinated abnormality mode is full-dimensional coordinated abnormality or electrical coordinated abnormality, and the oscillation abnormality mark is present, then the hazard type label is determined to be severe insulation deterioration hazard, and the hazard severity level is high-risk level.

[0062] Third rule for determining insulation hazards: If the current overshoot deviation rate is greater than the preset moderate overshoot threshold and less than or equal to the severe overshoot threshold, and the harmonic indication defect type is insulation degradation, then the hazard type label is determined to be insulation aging hazard, and the hazard severity level is medium risk level.

[0063] Fourth rule for determining insulation hazards: If the current overshoot deviation rate is greater than the moderate overshoot threshold, and the voltage recovery deviation is greater than the preset voltage deviation threshold, and the slope abnormality type is the mid-segment sluggish type, then the hazard type label is determined to be insulation moisture hazard, and the hazard severity level is medium risk level.

[0064] Fifth rule for determining insulation hazards: If the current overshoot deviation rate is greater than the preset slight overshoot threshold, the voltage recovery deviation is greater than the voltage deviation threshold, and the voltage recovery fluctuation is greater than the preset fluctuation threshold, then the hazard type label is determined to be slight insulation degradation hazard, and the hazard severity level is low risk level.

[0065] Sixth rule for determining insulation hazards: If the harmonic indication defect type is core magnetic saturation and the current overshoot deviation rate is greater than the mild overshoot threshold, then the hazard type label is determined to be core insulation deterioration hazard, and the hazard severity level is medium risk level.

[0066] Among them, each overshoot threshold, duration threshold, voltage deviation threshold and fluctuation threshold is preset by those skilled in the art based on the equipment characteristics, safety margin and identification sensitivity requirements of the target test circuit.

[0067] The operating condition mismatch feature set is sequentially matched with each insulation hazard judgment rule in the insulation degradation multi-dimensional evidence fusion rule base to obtain successfully matched insulation hazard judgment rules, which are then marked as insulation success rules. If only one insulation success rule exists, it is directly used as the main insulation rule; if multiple insulation success rules exist, the severity level of the hazards corresponding to all insulation success rules is compared, and the insulation success rule with the highest hazard severity level is selected as the main insulation rule; if multiple insulation success rules have the same hazard severity level, their corresponding rule priorities are further compared, and the insulation success rule with the highest rule priority is selected as the main insulation rule. Hazard type labels and hazard severity levels are obtained from the main insulation rules as the final insulation hazard identification result; the number of all insulation success rules is counted to obtain the cumulative number of insulation evidence; if none of the insulation hazard judgment rules are successfully matched, the insulation hazard identification result is empty, and the cumulative number of insulation evidence is zero.

[0068] The methods for identifying the risk of poor contact in the target test circuit include: A pre-constructed multi-dimensional evidence fusion rule base for contact degradation is established, containing multiple contact failure judgment rules and corresponding risk type labels, risk severity levels, and rule priorities. Each contact failure judgment rule is predefined by those skilled in the art based on the typical performance patterns of temperature rise rate, voltage recovery hysteresis, and abnormal electrothermal coupling when contact failure occurs at each critical contact. The rule priorities corresponding to different contact failure judgment rules are different, and the rule priorities are pre-set by those skilled in the art based on the thermal runaway rate, conductivity degradation degree, and urgency of on-site emergency response for the corresponding contact failure type.

[0069] Specifically, the contact degradation multidimensional evidence fusion rule base includes the following contact degradation judgment rules: First rule for determining poor contact: If the abnormal temperature rise rate of the contact is greater than the preset rapid temperature rise threshold, and the maximum temperature limit is greater than zero, and the temperature rise acceleration ratio is greater than the preset acceleration deterioration threshold, then the risk type label is determined to be serious poor contact risk, and the risk severity level is high risk level.

[0070] The second rule for determining poor contact is: if the abnormal temperature rise rate of the contact is greater than the threshold for rapid temperature rise, and the heat source characteristic of the contact with the highest temperature rise is a local heat source type, and the collaborative abnormality mode is electrothermal collaborative abnormality or all-dimensional collaborative abnormality, then the risk type label is determined to be serious poor contact risk, and the risk severity level is high risk level.

[0071] The third rule for determining poor contact is: if the abnormal temperature rise rate of the contact is greater than the preset significant temperature rise threshold, and the current overshoot deviation rate is greater than the preset slight overshoot threshold, and the temperature rise acceleration ratio is greater than 1, then the risk type label is determined to be moderate poor contact risk, and the risk severity level is medium-risk level.

[0072] Fourth rule for determining poor contact: If the abnormal temperature rise rate of the contact is greater than the significant temperature rise threshold, and the heat source characteristic of the contact with the highest temperature rise is a regional heat source type, then the risk type label is determined to be multi-point contact deterioration risk, and the risk severity level is medium risk level.

[0073] Fifth rule for determining poor contact: If the abnormal temperature rise rate of the contact is greater than the preset slight temperature rise threshold, and the bus voltage recovery hysteresis is greater than the preset hysteresis judgment threshold, and the coordinated abnormal mode is pressure-heat coordinated abnormal, then the risk type label is determined to be potential poor contact risk, and the risk severity level is low risk level.

[0074] Sixth rule for judging poor contact: If the abnormal temperature rise rate of the contact is greater than the slight temperature rise threshold, the voltage recovery deviation is greater than the voltage deviation threshold, and the temperature rise acceleration ratio is greater than 1, then the risk type label is judged as the risk of gradually increasing contact resistance, and the risk severity level is low risk level.

[0075] Among them, the thresholds for rapid temperature rise, significant temperature rise, slight temperature rise, accelerated deterioration, slight overshoot, and hysteresis are all preset by those skilled in the art based on the material characteristics of key contacts, the evolution law of thermal faults, and the requirements for early identification sensitivity.

[0076] The operating condition mismatch feature set is sequentially matched with each contact failure judgment rule in the multi-dimensional evidence fusion rule base for contact degradation to obtain successfully matched contact failure judgment rules, which are then marked as successful contact rules. Based on the successful contact rules, the contact failure identification results and the cumulative number of contact evidence are obtained in the same way as the main insulation rule determination method and the insulation evidence accumulation statistics method; wherein, the contact failure identification results include the risk type label and risk severity level corresponding to the main contact rule.

[0077] The results of insulation hazard identification, the cumulative number of insulation evidence, the results of poor contact identification, and the cumulative number of contact evidence are integrated to form the circuit risk identification results.

[0078] The methods for automatically intercepting subsequent load step commands and triggering circuit disconnection warnings include: Based on the loop risk identification results, a risk evolution trend assessment is performed, and the number of consecutive high-risk cycles is determined. Specifically, a risk history cache is established to store the loop risk identification results corresponding to multiple consecutive load steps during the current simulated load test of the target test loop. After each loop risk identification is completed, the corresponding loop risk identification result is stored in the risk history cache. The loop risk identification results of the most recent preset number of analyses are retrieved from the risk history cache to form a risk history window. The number of analyses is preset by those skilled in the art based on the observation period requirements of the risk evolution trend analysis.

[0079] For each circuit's risk identification results in the risk history window, the corresponding comprehensive risk level is determined. Specifically, if both the insulation hazard identification result and the poor contact identification result in the circuit risk identification results are empty, the corresponding comprehensive risk level is marked as no risk. If only the insulation hazard identification result is not empty, the severity level of the hazard corresponding to the insulation hazard identification result is used as the comprehensive risk level. If only the poor contact identification result is not empty, the severity level of the risk corresponding to the poor contact identification result is used as the comprehensive risk level. If both the insulation hazard identification result and the poor contact identification result are not empty, the higher of the corresponding hazard severity level and the risk severity level is used as the comprehensive risk level. If the hazard severity level and the risk severity level are the same, the severity level corresponding to the larger of the cumulative number of insulation evidence and the cumulative number of contact evidence is used as the comprehensive risk level. The comprehensive risk levels are mapped to progressively decreasing severity values ​​in the order of high risk, medium risk, low risk, and no risk. For example, the severity value corresponding to high risk is 4, the severity value corresponding to medium risk is 3, the severity value corresponding to low risk is 2, and the severity value corresponding to no risk is 1.

[0080] Analyze the changing trends of each severity value in the risk history window: if the last severity value is greater than the first severity value, the risk evolution trend is marked as a worsening trend; if the last severity value is equal to the first severity value and the last severity value is greater than 1, the risk evolution trend is marked as a persistent risk; if the last severity value is less than the first severity value, the risk evolution trend is marked as a mitigating trend; if all severity values ​​in the risk history window are equal to 1, the risk evolution trend is marked as a risk-free trend. Count the number of times the severity value in the risk history window is consecutively greater than or equal to the severity value corresponding to the medium-risk level to obtain the consecutive high-risk count; the consecutive high-risk count is used to assess whether the target test loop is in a persistent high-risk state.

[0081] A pre-built safety tolerance threshold configuration table is provided, which contains the contact temperature rise safety threshold and current overshoot safety threshold corresponding to different load test levels. The safety tolerance threshold configuration table is pre-built by those skilled in the art based on the equipment thermal stability limit, current surge withstand capability and test safety requirements of the target test circuit under different load test levels. According to the current test level, the corresponding contact temperature rise safety threshold and current overshoot safety threshold are obtained from the safety tolerance threshold configuration table.

[0082] A pre-built progressive warning level determination rule set is provided. This rule set contains multiple warning level determination rules, corresponding warning levels, and response actions. It is pre-built by those skilled in the art based on the evolution of anomalies during simulated load testing, the risk of test interruption, and on-site protection response requirements. Warning levels include Level 1, Level 2, and Level 3. The response action for Level 1 is to issue an audible and visual alarm; the response action for Level 2 is to issue an audible and visual alarm and suspend the automatic issuance of subsequent load step commands; the response action for Level 3 is to issue an audible and visual alarm, intercept all subsequent load step commands, and trigger a circuit disconnection operation.

[0083] Specifically, the progressive warning level determination rules set includes the following warning level determination rules: First warning level determination rule: If the abnormal temperature rise rate of the contact is greater than the product of the contact temperature rise safety threshold and the preset first-level ratio and is less than or equal to the product of the contact temperature rise safety threshold and the preset second-level ratio, or if the overshoot amplitude of the circuit current is greater than the product of the current overshoot safety threshold and the preset first-level ratio and is less than or equal to the product of the current overshoot safety threshold and the preset second-level ratio, then a first-level warning is triggered.

[0084] The second warning level determination rule is as follows: If the abnormal temperature rise rate of the contact is greater than the product of the contact temperature rise safety threshold and the preset secondary ratio but less than or equal to the contact temperature rise safety threshold, or the circuit current overshoot amplitude is greater than the product of the current overshoot safety threshold and the preset secondary ratio but less than or equal to the current overshoot safety threshold, or the risk evolution trend is a deteriorating trend and there are medium-risk insulation hazards or poor contact risks in the circuit risk identification results, then a secondary warning is triggered.

[0085] The third warning level determination rule is as follows: if the abnormal temperature rise rate of the contact is greater than the contact temperature rise safety threshold, or the circuit current overshoot amplitude is greater than the current overshoot safety threshold, or the number of consecutive high-risk events is greater than or equal to the preset continuous high-risk threshold, or there are high-risk insulation hazards or poor contact risks in the circuit risk identification results, then a level three warning will be triggered.

[0086] The first-level ratio and the second-level ratio are both preset by those skilled in the art according to the requirements of the graded early warning strategy. For example, the first-level ratio is 0.6 and the second-level ratio is 0.9. The continuous high-risk threshold is preset by those skilled in the art according to the cumulative effect of failure and the test safety boundary when multiple consecutive load steps occur in a high-risk state.

[0087] The abnormal temperature rise rate of the contact, the overshoot amplitude of the loop current, the risk evolution trend, the number of consecutive high-risk events, and the loop risk identification results are sequentially matched with the various warning level determination rules in the progressive warning level determination rule set. If the third warning level determination rule is successfully matched, the current warning level is determined to be a Level 3 warning. If the third warning level determination rule is not successfully matched but the second warning level determination rule is successfully matched, the current warning level is determined to be a Level 2 warning. If only the first warning level determination rule is successfully matched, the current warning level is determined to be a Level 1 warning. If none of the warning level determination rules are successfully matched, the current warning level is determined to be no warning.

[0088] If the current warning level is Level 3, a command interception signal is sent to the substation simulated load test control system, and a loop disconnection warning signal is sent to the substation integrated automation system. The command interception signal prevents subsequent load step commands from being issued, while the loop disconnection warning signal triggers protection linkage. Upon receiving the command interception signal, the substation simulated load test control system automatically intercepts subsequent load step commands to be executed. Upon receiving the loop disconnection warning signal, the substation integrated automation system initiates the safety disconnection linkage process for the target test loop. Simultaneously, the warning trigger status is marked as triggered, and the current time is recorded as the warning trigger time.

[0089] If the current warning level is Level II, a pause command signal is sent to the substation simulated load test control system to halt the automatic issuance of subsequent load step commands, and an audible and visual alarm is issued simultaneously. If the current warning level is Level I, only an audible and visual alarm is issued.

[0090] The methods for generating fault snapshot records include: If the warning trigger status is not marked as triggered, fault snapshot record generation will not be performed; if the warning trigger status is marked as triggered, fault snapshot record generation will be performed. The circuit topology diagram of the target test circuit is obtained from the substation equipment ledger. This diagram describes the connection relationships and current flow of each device in the target test circuit. The operating status information of the relevant devices in the target test circuit at the time of the warning trigger is obtained from the substation equipment operation monitoring system. This operating status information includes the circuit breaker position status and the disconnector switch position status. The operating status information of each device is superimposed onto the circuit topology diagram to form a fault-time topology diagram. This fault-time topology diagram visually displays the actual operating position and status of each device at the time of the fault. The voltage recovery feature group, current overshoot feature group, contact temperature rise feature group, collaborative abnormal mode, insulation hazard identification results, cumulative number of insulation evidence, poor contact identification results, cumulative number of contact evidence, risk evolution trend, number of consecutive high-risk events, and the topology state diagram at the time of the fault are integrated to form a fault snapshot record. Among them, the fault snapshot record is used to completely preserve the multi-level quantitative mismatch characteristics, multi-evidence risk identification conclusions, risk evolution trend, and circuit topology state at the time of trigger circuit disconnection warning.

[0091] Specifically, step S5: compare the fault snapshot record with the pre-built fault cause mapping model, and output the location coordinates of the suspected fault point of the target test circuit and the maintenance and handling guidance strategy based on the comparison results.

[0092] The methods for pre-constructing fault cause mapping models include: Historical fault case data for the target test circuit and similar circuits are obtained from the substation historical fault archive system. This data includes multiple historical fault records, each containing a historical fault number, time of occurrence, fault condition characteristics, cause label, location, maintenance measures, and repair effectiveness evaluation. The fault condition characteristics are identical to those in the voltage recovery, current overshoot, and contact temperature rise characteristic groups within the condition mismatch characteristic group. The cause label represents the root cause category of the fault, confirmed through manual investigation, including but not limited to main transformer winding insulation breakdown, busbar bushing insulation deterioration, poor switch contact, increased cable head contact resistance, and CT secondary circuit disconnection. The location of the fault point is represented using the substation equipment spatial positioning coding system, including the equipment bay number, equipment type code, and specific contact number. The repair effectiveness evaluation includes complete repair, partial repair, and ineffective repair.

[0093] For historical fault case data, data cleaning and quality screening were performed. Specifically, historical fault records with missing fault condition features, unclear fault cause labels, or ineffective repair evaluations were removed. For all retained historical fault records, min-max normalization was performed on their fault condition features. Based on the fault cause labels, all retained historical fault records were grouped into multiple fault cause category groups. For each fault cause category group, historical fault condition features were extracted from all historical fault records within that group. For each fault condition feature parameter, the lower quartile, median, and upper quartile of the corresponding feature parameter in each fault cause category group were calculated. The median was used as the feature center value, and a feature distribution interval was constructed based on the lower and upper quartiles.

[0094] For each fault cause category group and corresponding operating condition characteristic parameter, a feature discrimination evaluation is performed to obtain a feature discrimination capability matrix. Specifically, for each operating condition characteristic parameter, any two fault cause category groups are selected, and the feature distribution intervals of the two fault cause category groups under the corresponding operating condition characteristic parameter are obtained. The overlap interval length of the two feature distribution intervals is calculated, and then divided by the interval lengths of the two feature distribution intervals respectively. The larger of the two ratios is taken as the feature overlap rate of the corresponding operating condition characteristic parameter between the two fault cause category groups. The feature overlap rate is used to characterize the degree of overlap of the feature distribution intervals of the two fault cause category groups on the operating condition characteristic parameter. The smaller the feature overlap rate, the more significant the difference between the two fault cause category groups on the corresponding operating condition characteristic parameter. The method for calculating the length of the overlapping interval is as follows: the larger value of the lower boundaries of the two feature distribution intervals is taken as the left endpoint of the overlapping interval, and the smaller value of the upper boundaries of the two feature distribution intervals is taken as the right endpoint of the overlapping interval. If the right endpoint of the overlapping interval is greater than the left endpoint, the difference between the right endpoint and the left endpoint is calculated to obtain the length of the overlapping interval. If the right endpoint is less than or equal to the left endpoint, the length of the overlapping interval is set to 0, and the feature overlap rate is also set to 0. If the feature overlap rate is equal to 0, the corresponding operating condition feature parameter is determined to have strong distinguishing ability for differentiating the two types of root causes of faults. If the feature overlap rate is greater than 0 and less than or equal to the preset overlap rate distinction threshold, it is determined to have medium distinguishing ability. If the feature overlap rate is greater than the overlap rate distinction threshold, it is determined to have weak distinguishing ability. The overlap rate distinction threshold is preset by those skilled in the art according to the distinction accuracy requirements of the fault feature parameters. The distinguishing ability of each operating condition feature parameter between each fault cause category group is summarized to form a feature distinction ability matrix.

[0095] For each fault cause category group, a corresponding operating condition feature matching template is constructed. Specifically, for each fault cause category group, operating condition feature parameters with moderate or strong distinguishing ability against all other fault cause category groups are selected from the feature discrimination capability matrix to form the key indicator feature set for that fault cause category group. For each fault cause category group, all historical fault point locations and historical maintenance measures are deduplicated to form the corresponding typical fault point location set and typical maintenance measure set. For each fault cause category group, the number of historical fault records with complete repair evaluation and the total number of historical fault records are counted to obtain the number of complete repairs and the number of fault records. The ratio of the number of complete repairs to the number of fault records is calculated to obtain the historical repair success rate. Based on the historical fault cause labels, key indicator feature sets, typical fault point location sets, typical maintenance measure sets, historical repair success rates, feature distribution intervals, and feature center values ​​of each operating condition feature parameter corresponding to each fault cause category group, the operating condition feature matching template corresponding to each fault cause category group is constructed.

[0096] A hierarchical classification tree for fault causes is constructed, and all operating condition feature matching templates are aggregated with the hierarchical classification tree to form a fault cause mapping model. The hierarchical classification tree specifically comprises three levels: the first level is the fault category, specifically including insulation faults and contact faults; the second level is the fault sub-categories, specifically including breakdown faults, degradation faults, moisture faults, resistance increase faults, and contact loosening faults; breakdown faults, degradation faults, and moisture faults correspond to insulation faults, while resistance increase faults and contact loosening faults correspond to contact faults; the third level is the specific fault cause label.

[0097] The methods for comparing fault snapshot records with fault cause mapping models include: From the fault snapshot records, all parameters in the voltage recovery feature group, current overshoot feature group, and contact temperature rise feature group are extracted and marked as operating condition mismatch feature parameters. Numerical standardization is then performed on the extracted operating condition mismatch feature parameters to obtain a standardized operating condition feature vector. Specifically, a minimum-maximum normalization method is used, employing the minimum and maximum values ​​of each operating condition feature parameter used in the minimum-maximum normalization process applied to historical fault operating condition features when constructing the fault cause mapping model. This is used to normalize and calculate the standardized operating condition feature vector for each operating condition mismatch feature parameter.

[0098] A hierarchical screening and comparison process is performed to obtain a set of candidate matching templates. Specifically, the circuit risk identification results in the fault snapshot records are obtained; if the insulation hazard identification result is not empty and the contact failure identification result is empty, the main fault category is determined to be insulation fault; if the contact failure identification result is not empty and the insulation hazard identification result is empty, the main fault category is determined to be contact fault; if both the insulation hazard identification result and the contact failure identification result are not empty, the accumulated number of insulation evidence and the accumulated number of contact evidence are compared; if the accumulated number of insulation evidence is greater than the accumulated number of contact evidence, insulation fault is taken as the main fault category; if the accumulated number of contact evidence is greater than the accumulated number of insulation evidence, contact fault is taken as the main fault category; if the accumulated number of insulation evidence and the accumulated number of contact evidence are the same, both insulation fault and contact fault are taken as the main fault categories; all operating condition feature matching templates belonging to the main fault category are screened from the fault cause hierarchical classification tree to form a set of candidate matching templates.

[0099] For each working condition feature matching template in the candidate matching template set, a key indicator feature priority comparison is performed to obtain a pre-screened template set. Specifically, the corresponding key indicator feature set is obtained from the working condition feature matching template, and the working condition mismatch feature parameters corresponding to the key indicator feature set are extracted from the standardized working condition feature vector and marked as key feature parameters. Each key feature parameter is compared with the corresponding feature distribution interval, and the number of key feature parameters falling within the corresponding feature distribution interval is counted to obtain the key feature hit count. The ratio of the key feature hit count to the total number of working condition feature parameters in the key indicator feature set is calculated to obtain the key feature hit rate. If the key feature hit rate is less than the preset key feature pre-screening threshold, the corresponding working condition feature matching template is removed. If the key feature hit rate is greater than or equal to the key feature pre-screening threshold, the corresponding working condition feature matching template is retained. The key feature pre-screening threshold is preset by those skilled in the art based on the balance requirements of comparison efficiency and accuracy. All retained working condition feature matching templates are summarized to form a pre-screened template set.

[0100] For each working condition feature matching template in the pre-screened template set, the full feature interval fall-in degree is calculated to obtain the corresponding comprehensive matching score, and the working condition feature matching template with the highest comprehensive matching score is selected as the best matching template. Specifically, each working condition mismatch feature parameter in the standardized working condition feature vector is compared with the corresponding feature distribution interval in the working condition feature matching template; if the working condition mismatch feature parameter falls within the corresponding feature distribution interval, the fall-in mark of the corresponding working condition mismatch feature parameter is set as a hit; the number of all fall-in marks marked as hits and the number of working condition mismatch feature parameters in the standardized working condition feature vector are counted to obtain the number of hit features and the total number of feature parameters; the ratio of the number of hit features to the total number of feature parameters is calculated to obtain the feature fall-in degree. Simultaneously, the feature center values ​​of each working condition feature parameter are obtained from the working condition feature matching template to form the template feature center vector; the Euclidean distance between the standardized working condition feature vector and the template feature center vector is calculated to obtain the center deviation distance; the reciprocal of the center deviation distance is taken to obtain the center proximity; corresponding matching score weights are set for feature fall-in degree and center proximity, respectively; and the feature fall-in degree and center proximity are weighted and summed based on the matching score weights to obtain the comprehensive matching score of the working condition feature matching template. Each matching score weight is pre-set by those skilled in the art based on the degree of influence of fault feature coverage completeness and fault feature typicality on matching accuracy.

[0101] If the feature inclusion degree of the best matching template is greater than or equal to the preset minimum matching confidence threshold, the comparison is considered successful. The fault cause label, typical fault location set, typical maintenance measure set, and historical repair success rate are obtained from the best matching template. The minimum matching confidence threshold is preset by those skilled in the art based on the reliability requirements for fault cause identification and the requirements for misjudgment control. If the feature inclusion degree of the best matching template is less than the minimum matching confidence threshold, the comparison is considered unsuccessful, and an unknown fault type label is generated. From the historical fault records corresponding to the main fault category, the Euclidean distance between the vector obtained after numerical standardization of the historical fault condition features corresponding to each historical fault record and the standardized operating condition feature vector is calculated sequentially. The three historical fault records with the smallest Euclidean distance are selected to form a similar case reference set. This similar case reference set is used to provide a reference for subsequent fault location and maintenance guidance when the comparison fails.

[0102] The methods for outputting the coordinates of suspected fault points include: If the comparison is successful, a set of typical fault point locations is obtained from the best matching template. If the set of typical fault point locations contains only one historical fault point location, it is directly used as the location coordinates of the suspected fault point, and the location confidence level is marked as high confidence. If the set of typical fault point locations contains multiple historical fault point locations, multi-evidence fault point filtering is performed to determine the location coordinates and location confidence level of the suspected fault point. Specifically, the contact point location code of the highest temperature rise contact is obtained from the fault snapshot record, and the contact point location code is matched with each historical fault point location in the set of typical fault point locations. If there is a historical fault point location with the same contact point location code, the corresponding historical fault point location is used as the location coordinates of the suspected fault point, and the location confidence level is marked as high confidence. If there is no historical fault point location with the same contact point location code, topological path reasoning is performed for location.

[0103] The topology path reasoning and localization process specifically involves: obtaining the topology state diagram at the time of the fault from the fault snapshot record; identifying the location of the highest temperature rise-up contact in the loop topology based on the topology state diagram at the time of the fault; and filtering historical fault point locations in the loop topology that are connected in series with the highest temperature rise-up contact on the same electrical path as the highest temperature rise-up contact from the set of typical fault point locations, forming a candidate set of the same electrical path. Here, the same electrical path refers to the electrical path formed by devices or contacts connected in series along the path of current flow from the power supply side to the load side in the loop topology described by the topology state diagram at the time of the fault. If the candidate set of the same path is not empty, then based on the current flow direction pre-marked in the topology state diagram at the time of the fault, determine the positional relationship between each historical fault point in the candidate set and the highest temperature rise-up point; mark the historical fault point upstream of the current at the highest temperature rise-up point as the upstream candidate position, and mark the historical fault point downstream of the current at the highest temperature rise-up point as the downstream candidate position; if the main fault category is a contact fault, then the upstream candidate position is preferentially selected as the location coordinates of the suspected fault point, and the location confidence is marked as medium confidence; if the main fault category is an insulation fault, then both the upstream and downstream candidate positions are used as the location coordinates of the suspected fault point, and the location confidence is marked as medium confidence. If the candidate set of the same path is empty, then the contact position of the highest temperature rise-up point is directly encoded as the location coordinates of the suspected fault point, and the location confidence is marked as low confidence.

[0104] If the comparison fails, the location of the highest temperature rise contact point is encoded as the coordinates of the main suspected fault point. From the similar case reference set, the historical fault point locations corresponding to each historical fault record are extracted, and auxiliary fault point location sets are formed by merging and deduplication. The coordinates of the main suspected fault point are integrated with the auxiliary fault point location sets to form the coordinates of the suspected fault point, and the location confidence is marked as low confidence.

[0105] Perform fault propagation link analysis to obtain a fault propagation link description. Specifically, from the fault snapshot record, obtain the contact location code and heat source characteristics of each abnormal temperature rise contact; if the number of abnormal temperature rise contacts is greater than or equal to 2, determine whether all abnormal temperature rise contacts are on the same electrical path based on the topology state diagram at the time of the fault; if all abnormal temperature rise contacts are on the same electrical path, sort each abnormal temperature rise contact according to the current flow direction to obtain an ordered sequence of abnormal temperature rise contacts; obtain the abnormal temperature rise contact located at the upstream of the current from the ordered sequence of abnormal temperature rise contacts, and determine whether the heat source characteristics corresponding to the abnormal temperature rise contact located at the upstream of the current are local heat source type; if the heat source characteristics corresponding to the abnormal temperature rise contact located at the upstream of the current are local heat source type, mark the corresponding abnormal temperature rise contact as the fault source point, and integrate the contact location codes of each abnormal temperature rise contact according to the arrangement order in the ordered sequence of abnormal temperature rise contacts to form a fault propagation link description. In the fault propagation link description, the contact with the abnormal temperature rise corresponding to the first contact position code is the fault source point. If the number of abnormal temperature rise contacts is less than 2, or if all abnormal temperature rise contacts are not on the same electrical path, or if the heat source characteristic corresponding to the abnormal temperature rise contact located upstream of the current is not a local heat source type, then no fault propagation link description will be generated, and the fault propagation link description will be set to null.

[0106] The methods for providing maintenance and handling guidance strategies include: If the comparison is successful, a set of typical maintenance measures is obtained from the best matching template, and the maintenance urgency label is determined based on the highest severity level in the circuit risk identification results. High-risk levels correspond to emergency maintenance, medium-risk levels to priority maintenance, and low-risk levels to planned maintenance. Based on the location coordinates of the suspected fault point, the equipment type and specifications of the corresponding equipment are obtained from the substation equipment ledger. Based on the equipment type and specifications, a recommended list of maintenance tools and safety protection requirements are obtained from a pre-built maintenance tool configuration library. This maintenance tool configuration library is pre-built by those skilled in the art based on standard maintenance procedures for various types of substation equipment, common maintenance tool configuration specifications, and on-site safety operation requirements.

[0107] The process generates maintenance risk alerts, including alerts for severe equipment condition deterioration and multi-point fault risks. Specifically, it retrieves the number of consecutive high-risk events and the risk evolution trend from fault snapshot records. If the number of consecutive high-risk events exceeds a preset severe deterioration threshold and the risk evolution trend is deteriorating, a severe equipment condition deterioration alert is generated, prompting maintenance personnel to conduct a comprehensive inspection of the relevant equipment in the target test loop before formal maintenance to eliminate other potential hazards. If the fault propagation link description is not empty, a multi-point fault risk alert is generated, prompting maintenance personnel to check each connection point in the fault propagation link description sequentially. The severe equipment condition deterioration alert and the multi-point fault risk alert are integrated to form a maintenance risk alert; if neither alert is generated, the maintenance risk alert is set to an empty value. The severe deterioration threshold is preset by those skilled in the art based on the empirical relationship between the number of consecutive high-risk events and the accelerated deterioration of equipment condition, as well as the needs of on-site emergency repairs.

[0108] The system integrates typical maintenance measures, maintenance urgency labels, recommended maintenance tool lists, safety protection requirements, and maintenance risk warnings to form a maintenance handling guidance strategy.

[0109] If the comparison fails, the circuit risk identification result is obtained from the fault snapshot record, and the dominant fault type label is extracted from the circuit risk identification result according to the main fault category. Specifically, if the main fault category is an insulation fault, the hazard type label is obtained from the insulation hazard identification result and used as the dominant fault type label; if the main fault category is a contact fault, the risk type label is obtained from the poor contact identification result and used as the dominant fault type label. Standard investigation steps matching the dominant fault type label are obtained from a pre-built general investigation process library. The general investigation process library is pre-built by those skilled in the art based on standardized investigation steps for common fault types in substations, on-site maintenance experience, and accident handling procedures. Historical maintenance measures for each historical fault record are extracted from the similar case reference set as reference maintenance measures; the standard investigation steps and reference maintenance measures are integrated to form a maintenance and handling guidance strategy.

[0110] The system integrates the location coordinates of suspected fault points, location reliability, fault propagation path description, and maintenance and handling guidance strategies to form a fault early warning and handling report.

[0111] This embodiment establishes a standard transient response characteristic parameter benchmark for the target test circuit under different load test levels by constructing a test circuit characteristic benchmark library, effectively overcoming the shortcomings of relying solely on static thresholds and failing to detect dynamic response anomalies. Morphological features of voltage recovery are extracted through multi-stage recovery curve decomposition analysis; abnormal harmonic features of current are extracted through fast Fourier transform; and potential equipment defect types are identified based on a harmonic anomaly pattern rule library. Local heat source type and regional heat source type thermal anomalies are distinguished through contact heat diffusion pattern analysis, and positive feedback deterioration risks are identified through temperature rise acceleration ratio, achieving sensitive capture of equipment degradation precursor information hidden in transient response waveforms. By constructing a multi-dimensional evidence fusion rule library for insulation degradation and contact degradation, multi-evidence fusion adjudication is performed based on multi-dimensional feature combination patterns and rule priorities, effectively improving the accuracy and reliability of risk identification. Multi-dimensional coupled fault modes such as all-dimensional coordinated anomalies, electrical coordinated anomalies, electrothermal coordinated anomalies, and pressure-thermal coordinated anomalies are identified through voltage-current-temperature three-dimensional coordinated anomaly detection. The dynamic evolution process of risks is tracked and monitored through risk evolution trend assessment and continuous high-risk frequency statistics. The designed progressive early warning level determination mechanism implements graded early warning responses of Level 1, Level 2, and Level 3 based on the severity of risk, effectively avoiding the single response mode of shutting down without an early warning. By constructing a fault cause mapping model that includes feature distribution intervals, feature center values, key indicator feature sets, and a fault cause hierarchical classification tree, it achieves coarse-to-fine hierarchical screening and comparison, as well as priority comparison of key features, significantly improving comparison efficiency and accuracy. Through topological path reasoning and fault propagation link analysis, it can accurately screen and identify the path of fault propagation from the source to other parts of the circuit in scenarios with multiple candidate fault points based on current paths and upstream and downstream relationships. Finally, it outputs a fault early warning and handling report that includes the location coordinates of suspected fault points, location confidence, fault propagation link description, and priority-ordered maintenance measures, thereby effectively shortening fault investigation time and improving substation commissioning efficiency and overall safety of the testing process.

[0112] Example 2 This application also provides an electronic device. The electronic device may include one or more processors and one or more memories. The memories store computer-readable code, which, when executed by the one or more processors, can perform the substation simulated load test fault early warning method based on big data analysis as described above.

[0113] The method according to the embodiments of this application can also be implemented using the architecture of the electronic device shown in this application. The electronic device may include a bus, one or more CPUs, ROM, RAM, a communication port connected to a network, input / output, a hard disk, etc. The storage device in the electronic device, such as a ROM or hard disk, may store the substation simulated load test fault early warning method based on big data analysis provided in this application. Furthermore, the electronic device may also include a user interface. Of course, the architecture shown in this application is merely exemplary; when implementing different devices, one or more components of the electronic device shown in this application may be omitted according to actual needs.

[0114] Example 3 Please refer to the accompanying drawings. One embodiment of this application discloses a computer-readable storage medium. The computer-readable storage medium stores computer-readable instructions. When the computer-readable instructions are executed by a processor, they can perform the substation simulated load test fault early warning method based on big data analysis according to an embodiment of this application, as described above. The storage medium includes, but is not limited to, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0115] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 Units that specify functions within one or more boxes.

[0116] The above description is merely a preferred embodiment of the present invention and should not be construed as limiting the scope of the invention. Therefore, any equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.

Claims

1. A method for fault early warning in substation simulated load testing based on big data analysis, characterized in that, include: Step S1: Obtain historical load step commands, load test levels, and transient following characteristic data of the target test circuit under historical normal conditions in the substation, and construct a test circuit characteristic benchmark library based on the transient following characteristic data of the target test circuit under different load test levels. Step S2: During the simulated load test of the target test circuit based on the current load step command, the instantaneous bus voltage, instantaneous circuit current and real-time temperature of key contacts are collected simultaneously after the current load step command is triggered, forming a transient monitoring feature set corresponding to the current load step command; Step S3: Based on the transient monitoring feature set and the test circuit characteristic benchmark library, calculate the bus voltage recovery hysteresis, circuit current overshoot amplitude and contact abnormal temperature rise rate under the current load step command, and form a working condition mismatch feature set characterizing the current circuit operating status. Step S4: Based on the operating condition mismatch feature set, identify the insulation hazards or poor contact risks in the target test circuit. When the abnormal temperature rise rate of the contact or the overshoot amplitude of the circuit current exceeds the preset safety tolerance threshold, automatically intercept the subsequent load step command and trigger the circuit disconnection warning. When the circuit disconnection warning is triggered, generate a fault snapshot record based on the operating condition mismatch feature set at the trigger time. Step S5: Compare the fault snapshot record with the pre-built fault cause mapping model, and output the location coordinates of the suspected fault point of the target test circuit and the maintenance and handling guidance strategy based on the comparison results.

2. The method according to claim 1, characterized in that, In step S1, a test loop characteristic benchmark library is constructed, including: Acquire all historical load step commands executed by the target test circuit under historical normal conditions, and obtain the load test level corresponding to each historical load step command; wherein, the historical load step command includes the command trigger time, the command target load value, the command loading slope, and the command duration; acquire the transient waveform data recorded by the target test circuit during each historical load step command execution, and extract the corresponding transient follow-up characteristic data, including voltage recovery time, current settling time, current peak overshoot, voltage drop depth, and contact temperature rise slope; Each historical load step command is associated with the corresponding load test level and transient following characteristic data to form a transient characteristic record. All transient characteristic records are grouped according to the load test level to form multiple characteristic record groups. For each characteristic record group corresponding to the load test level, statistical analysis is performed on all transient following characteristic data, and the characteristic mean and characteristic standard deviation of each transient following characteristic data are calculated. The corresponding transient characteristic benchmark is formed by integration. The transient characteristic benchmarks of all load test levels are summarized to establish a test loop characteristic benchmark library.

3. The method according to claim 2, characterized in that, In step S2, a transient monitoring feature set is formed, including: When the current load step command is triggered, the corresponding command trigger time is obtained as the current trigger time. Continuous high-frequency sampling and continuous sampling are performed starting from the current trigger time to obtain the instantaneous bus voltage sequence, instantaneous loop current sequence, and real-time temperature sequence of each key contact. Each instantaneous bus voltage in the instantaneous bus voltage sequence includes the sampling time and the instantaneous voltage value; each instantaneous loop current in the instantaneous loop current sequence includes the sampling time and the instantaneous current value; and each real-time temperature of each key contact in each real-time temperature sequence includes the sampling time and the temperature value. The target load value corresponding to the current load step command is taken as the current target load value, and the rated capacity of the target test circuit is obtained. Based on the current target load value and the rated capacity of the circuit, the load test level corresponding to the current load step command is calculated and marked as the current test level. The instantaneous bus voltage sequence, instantaneous circuit current sequence, real-time temperature sequence of each key contact, current trigger time, current test level and current target load value are integrated to form a transient monitoring feature set.

4. The method according to claim 3, characterized in that, In step S3, a set of operating condition mismatch features is formed, including: Based on the transient monitoring feature set and the test circuit characteristic benchmark library, the voltage recovery feature set, current overshoot feature set, and contact temperature rise feature set are calculated sequentially, and cross-dimensional abnormal mode correlation analysis is performed to identify collaborative abnormal modes. The voltage recovery feature set, current overshoot feature set, contact temperature rise feature set, collaborative abnormal modes, and current test level are integrated to form the operating condition mismatch feature set. The calculation method for the voltage recovery feature group is as follows: Transient feature point detection is performed on the instantaneous bus voltage sequence to identify the lowest voltage drop point and the steady-state voltage recovery point; based on the lowest voltage drop point and the steady-state voltage recovery point, the current voltage recovery time is calculated, and combined with the characteristic mean of the voltage recovery time in the transient characteristic benchmark, the bus voltage recovery hysteresis is calculated; the bus voltage recovery hysteresis is then standardized and converted to obtain the voltage recovery deviation. For the instantaneous bus voltage sequence, a multi-stage recovery curve decomposition analysis is performed to obtain the initial recovery slope, the middle recovery slope, and the final recovery slope. Voltage recovery curve smoothness analysis is then performed to extract the morphological features and voltage recovery fluctuations of the voltage recovery process. Based on the morphological features of the voltage recovery process, the monotonicity of the recovery slope is checked on the instantaneous bus voltage sequence to obtain the slope anomaly type. The bus voltage recovery hysteresis, voltage recovery deviation, slope anomaly type, and voltage recovery fluctuation are integrated to form a voltage recovery feature group.

5. The method according to claim 4, characterized in that, The calculation of the current overshoot characteristic set includes: Envelope peak detection is performed on the instantaneous loop current sequence to identify the current overshoot peak point; the steady-state target current value is calculated, and the current overshoot is calculated based on the absolute value of the steady-state target current value and the instantaneous current value corresponding to the current overshoot peak point; the loop current overshoot amplitude is calculated based on the characteristic mean of the current peak overshoot in the transient characteristic reference; and the current overshoot deviation rate is calculated based on the loop current overshoot amplitude and the steady-state target current value. The instantaneous loop current sequence is subjected to current overshoot waveform morphology analysis to extract the overshoot duration; the instantaneous loop current sequence is subjected to spectrum analysis to extract abnormal harmonic features, and the type of harmonic indication defect is determined based on the abnormal harmonic features; the instantaneous loop current sequence is subjected to current waveform oscillation characteristic analysis to obtain oscillation anomaly indicators; the loop current overshoot amplitude, current overshoot deviation rate, overshoot duration, harmonic indication defect type and oscillation anomaly indicators are integrated to form a current overshoot feature group.

6. The method according to claim 5, characterized in that, The calculation of the contact temperature rise characteristic set includes: Linear fitting is performed on the real-time temperature sequence of each key junction to calculate the current temperature rise slope of each key junction. Combined with the characteristic mean and characteristic standard deviation of the junction temperature rise slope in the transient characteristic benchmark, the temperature rise slope deviation and temperature rise deviation multiple of each key junction are calculated. Based on the temperature rise deviation multiple of each key junction, the temperature rise abnormal junction is identified from all key junctions. The temperature rise abnormal junction with the largest temperature rise slope deviation is marked as the highest temperature rise junction. The temperature rise slope deviation corresponding to the highest temperature rise junction is used as the junction abnormal temperature rise rate, and the junction position code of the highest temperature rise junction is recorded. For each real-time temperature sequence, the absolute value of the junction temperature is checked sequentially to obtain the corresponding maximum temperature exceedance. For each junction with abnormal temperature rise, the junction heat diffusion mode analysis is performed sequentially to obtain the heat source characteristics corresponding to each junction with abnormal temperature rise. For the real-time temperature sequence of the junction with the highest temperature rise, the nonlinear feature extraction of the temperature curve is performed to obtain the corresponding temperature rise acceleration ratio. The junction abnormal temperature rise rate, maximum temperature exceedance, heat source characteristics of all junctions with abnormal temperature rise, junction position encoding of the junction with the highest temperature rise, and temperature rise acceleration ratio are integrated to form a junction temperature rise feature group.

7. The method according to claim 6, characterized in that, In step S4, identify insulation defects or risks of poor contact, including: Multiple rules for determining insulation defects and poor contact are pre-built. Each rule for determining insulation defects has a corresponding defect type label, defect severity level, and rule priority. Each rule for determining poor contact has a corresponding risk type label, risk severity level, and rule priority. The working condition mismatch feature set is matched sequentially with each insulation hazard judgment rule and contact failure judgment rule to obtain the successfully matched insulation hazard judgment rule and contact failure judgment rule, and marked as insulation success rule and contact success rule respectively; the number of insulation success rule and contact success rule is counted to obtain the cumulative number of insulation evidence and the cumulative number of contact evidence. Based on the severity level and priority of the hazard corresponding to each insulation success rule, and the risk severity level and priority of the contact success rule, the main insulation rule and the main contact rule are determined respectively; the hazard type label and hazard severity level corresponding to the main insulation rule are used as the insulation hazard identification result, and the risk type label and risk severity level corresponding to the main contact rule are used as the contact failure identification result; The results of insulation hazard identification, the cumulative number of insulation evidence, the results of poor contact identification, and the cumulative number of contact evidence are integrated to form the circuit risk identification results.

8. The method according to claim 7, characterized in that, In step S5, a fault cause mapping model is pre-constructed, including: Historical fault case data is acquired, which includes multiple historical fault records. Each historical fault record contains historical fault condition characteristics, historical fault cause tags, historical fault location, historical maintenance measures, and historical repair effect evaluations. Data cleaning and quality screening are performed on the historical fault case data. For all retained historical fault records, min-max normalization is performed on the historical fault condition characteristics. Based on the historical fault cause tags, all retained historical fault records are grouped into multiple fault cause category groups. For each fault cause category group, calculate the feature center value and feature distribution interval of each historical fault condition feature parameter. For each fault cause category group, perform feature discrimination evaluation to obtain the feature discrimination capability matrix. For each fault cause category group, construct the corresponding condition feature matching template based on the feature discrimination capability matrix. Construct a fault cause hierarchical classification tree and summarize all condition feature matching templates and fault cause hierarchical classification trees to form a fault cause mapping model.

9. The method according to claim 8, characterized in that, In step S5, the fault snapshot record is compared with the fault cause mapping model, including: The fault snapshot record includes a set of operating condition mismatch features, circuit risk identification results, risk evolution trends, number of consecutive high-risk events, and a topology diagram at the time of the fault. From the fault snapshot record, all parameters in the voltage recovery feature group, current overshoot feature group, and contact temperature rise feature group are extracted and marked as operating condition mismatch feature parameters. Based on the circuit risk identification results, hierarchical screening and comparison are performed on all operating condition feature matching templates to obtain a candidate matching template set. For each operating condition feature matching template in the candidate matching template set, a priority comparison of key indicator features is performed to obtain a pre-screened template set. For each working condition feature matching template in the pre-screened template set, the full feature interval fall-in degree is calculated to obtain the corresponding comprehensive matching score, and the working condition feature matching template with the highest comprehensive matching score is selected as the best matching template. If the feature fall-in degree of the best matching template is greater than or equal to the preset minimum matching confidence threshold, the comparison is determined to be successful; if the feature fall-in degree of the best matching template is less than the minimum matching confidence threshold, the comparison is determined to be unsuccessful, and a similar case reference set is generated.

10. The method according to claim 9, characterized in that, In step S5, the coordinates of the suspected fault location and the maintenance and handling guidance strategy are output, including: If the comparison is successful, the set of typical fault location and the set of typical maintenance measures are obtained from the best matching template, and the maintenance urgency label is determined based on the circuit risk identification results. If the set of typical fault location contains multiple historical fault location locations, multi-evidence fault point screening is performed to determine the location coordinates and location credibility of suspected fault points. Based on the location coordinates of suspected fault points, a list of recommended maintenance tools and safety protection requirements are obtained. Maintenance risk warnings are generated and integrated with the set of typical maintenance measures, maintenance urgency labels, list of recommended maintenance tools, and safety protection requirements to form a maintenance handling guidance strategy. If the comparison fails, the location code of the highest temperature rise contact point is used as the location coordinates of the primary suspected fault point; based on the similar case reference set, an auxiliary fault point location set is established and integrated with the location coordinates of the primary suspected fault point to form the location coordinates of the suspected fault point, and the location reliability is determined; based on the circuit risk identification results, the dominant fault type label is determined, and the matching standard troubleshooting steps are obtained; historical maintenance measures for each historical fault record are extracted from the similar case reference set as reference maintenance measures; the standard troubleshooting steps and reference maintenance measures are integrated to form a maintenance handling guidance strategy.

Images