Ring main unit fault processing method and device, storage medium and program product

Abstract

The invention discloses a ring main unit fault processing method and device, a storage medium and a program product, and belongs to the technical field of power distribution automation. The method comprises the following steps: deploying a current sensor, a voiceprint sensor, a vibration sensor and an arc light sensor in a ring main unit, and synchronously collecting an electrical signal, a sound signal, a vibration signal and an optical signal when a fault occurs; carrying out preprocessing and feature extraction on the collected multi-modal data to obtain multi-dimensional feature parameters such as current amplitude, spectrum features, vibration energy and light intensity features; and calculating fault confidence based on a weighted fusion algorithm, and outputting fault type and position information when the confidence exceeds a judgment threshold. By fusing electrical quantity and non-electrical quantity information, real fault and non-fault disturbance can be distinguished, atypical faults such as high-resistance grounding and arc faults can be accurately identified, the false alarm rate is reduced, the reliability and the intelligent level of a ring main unit protection system are improved, and the method is suitable for intelligent operation and maintenance scenes of an urban power distribution network.

Description

Technical Field

[0001] This application relates to the field of power distribution automation technology, and in particular to a method, device, storage medium and program product for handling ring main unit faults. Background Technology

[0002] As a critical node in urban medium-voltage power distribution networks, the stable operation of ring main units is essential for ensuring power supply to residents and businesses. Therefore, the ability to quickly and accurately identify and isolate electrical faults within ring main units or their connected lines is one of the core technological directions for improving the reliability and automation level of the entire power distribution system.

[0003] Currently, some technologies for fault diagnosis of ring main units rely on the monitoring of electrical parameters. For example, protective relays or intelligent terminals are used to monitor electrical quantities such as current and voltage in real time. When the monitored current value instantaneously exceeds the preset threshold, the system determines that a short circuit or ground fault has occurred in the line and immediately activates the protection mechanism, controlling the switching equipment to perform a tripping action to disconnect the faulty line.

[0004] However, in actual operating power distribution networks, there are numerous non-faulty transient disturbances, such as the starting and stopping of large motors, capacitor bank switching operations, or transient overvoltages and inrush currents caused by factors such as remote line faults and lightning strikes. These electrical disturbances exhibit highly similar waveform characteristics to those of real faults. When relying solely on electrical quantity information for judgment, these normal transient disturbances are easily misjudged as real faults, leading to unnecessary malfunctions in protection devices. Such erroneous tripping not only causes unexpected power outages for downstream users but also increases the workload of maintenance personnel in troubleshooting and restoring power, directly reducing the accuracy of fault diagnosis and the overall operating efficiency of the power distribution system. Summary of the Invention

[0005] This application provides a method, device, storage medium, and program product for troubleshooting ring main units, which can improve the efficiency and accuracy of fault diagnosis for ring main units.

[0006] In a first aspect, this application provides a ring main unit fault handling method, applied to a first ring main unit device. The method includes: the first ring main unit acquiring corresponding electrical parameter data and at least two types of non-electrical sensor data through electrical and non-electrical sensors; the first ring main unit determining whether the electrical parameter data contains a preset fault electrical characteristic; if the electrical parameter data contains a preset fault electrical characteristic, the first ring main unit cross-compares the fault electrical characteristic with the at least two types of non-electrical sensor data to determine whether a definitive fault identifier is generated; after generating a definitive fault identifier, the first ring main unit controls an internal first switch to perform a tripping action, disconnecting the first ring main unit from the downstream faulty line; and the first ring main unit sending an isolation command message to the second ring main unit of the downstream faulty line.

[0007] This embodiment introduces a cross-comparison step with at least two non-electrical sensor data after detecting fault characteristics in electrical parameters. This expands the triggering condition for protection action from a single electrical quantity criterion to a multi-physical domain collaborative verification mechanism. This mechanism requires the system not to immediately trip when an electrical anomaly is detected. Instead, it logically verifies the electrical characteristic with physical phenomena captured by at least two non-electrical sensors, such as acoustic signatures, vibrations, and arc flashes. Only when the multi-modal data form a self-consistent chain of evidence at the physical mechanism level is a definitive fault identifier generated and protection initiated. Real faults (such as arc short circuits) simultaneously generate multiple observable physical phenomena, including electrical anomalies, characteristic acoustic and vibration signals, and arc flash radiation. Non-faulty transient disturbances (such as capacitor switching and motor starting), while potentially exhibiting similar electrical waveform characteristics, lack corresponding non-electrical physical accompanying phenomena. Through multi-physical domain cross-verification, the judgment basis delves from the surface characteristics of electrical phenomena to the essential level of their physical causes, reducing the probability of protection maloperation caused by transient interference and improving the accuracy of fault identification and the reliability of protection actions.

[0008] In conjunction with some embodiments of the first aspect, in some embodiments, the non-electrical sensor data includes acoustic signature sensor data, vibration sensor data, and arc light sensor data. The first ring network box cross-compares the electrical characteristics of the fault with at least two types of non-electrical sensor data to determine whether a definitive fault identifier is generated. Specifically, this includes: performing joint time-domain and frequency-domain analysis on the acoustic signature sensor data and vibration sensor data respectively to generate their corresponding three-dimensional spectrum maps; performing spectral component analysis on the arc light sensor data to generate a spectral intensity distribution map; and performing a slice-style scan of the three-dimensional spectrum map along the time axis, calculating the time slice where the identified instantaneous energy peak is located. The spectral energy concentration of time slices is determined. When the spectral energy concentration is lower than a preset diffusion threshold, the time slice is determined to have full-spectrum diffusion characteristics, and the quantized value of the instantaneous energy peak and diffusion characteristics are recorded together as the first set of non-electrical characteristic parameters. In the distribution map of spectral intensity changing with time, the peak intensity and its first derivative of the ultraviolet spectral band are identified and quantified to obtain the light intensity jump rate as the second set of non-electrical characteristic parameters. The quantized values ​​of the first set of non-electrical characteristic parameters and the second set of non-electrical characteristic parameters are matched with the preset fault physical model template. When the matching degree calculation result exceeds the preset collaborative confidence threshold, a conclusive fault identifier is generated.

[0009] This embodiment generates a three-dimensional spectrum map by performing time-domain and frequency-domain joint analysis on acoustic and vibration sensor data, and performs slice scanning along the time axis to extract the spectral energy concentration at the instantaneous energy peak. At the same time, it performs spectral component analysis on arc light data to quantify the peak intensity and its first derivative in the ultraviolet band, transforming the raw signals of non-electrical sensors into quantitative characteristic parameters with fault physical diagnosis significance. The core of this transformation lies in not only capturing the amplitude anomalies of the signal at a given time point, but also exploring the dispersion of frequency domain energy distribution and the dynamic changes in spectral components. When an arc short circuit occurs, the discharge process excites a wide-band acoustic and vibration energy release (manifested as a full-spectrum dispersion feature with low spectral energy concentration) and a strong surge in ultraviolet light (manifested as a high light intensity jump rate). These characteristic parameters directly map the physical process of real faults, while the energy of non-fault interference (such as mechanical impact, switching operation) is mostly concentrated in a specific frequency band and lacks ultraviolet light response, exhibiting a completely different pattern in spectral concentration and light intensity jump rate. By calculating the matching degree between the quantified values ​​of the first set of parameters (instantaneous energy peak and dispersion characteristics) and the second set of parameters (light intensity jump rate) and a preset fault physics model template, the system achieves a qualitative leap from "whether there is an anomaly" to a quantitative "degree of conformity to the fault physics pattern". Through the dual constraints of pattern matching with multi-parameter combinations and collaborative confidence thresholds, the discrimination logic has a stronger ability to distinguish the differences between interference signals and real faults at the physical mechanism level, improving the accuracy and anti-interference robustness of the generation of definitive fault identifiers.

[0010] In conjunction with some embodiments of the first aspect, in some embodiments, the fault physical model template is established through the following preliminary steps: acquiring a first set of multimodal data under preset fault conditions and a second set of multimodal data under non-fault conditions, wherein the multimodal data includes at least electrical features, acoustic data, vibration data, and arc light data; performing analysis and identification operations on the first and second sets of multimodal data to extract their respective non-electrical feature parameter sets; performing statistical analysis on the non-electrical feature parameter sets of the first and second sets of multimodal data to calculate and determine the classification thresholds that can maximize the distinction between the two datasets for instantaneous energy peak, spectral energy concentration, ultraviolet spectral peak intensity, diffusion characteristics, and light intensity jump rate; and integrating multiple classification thresholds to construct a fault physical model template.

[0011] This embodiment collects multimodal data under preset fault and non-fault conditions, and performs statistical analysis on the extracted non-electrical feature parameter set to calculate the classification threshold that maximizes the distinction between the two datasets, achieving data-driven adaptive generation of the discrimination criteria. The core of this process lies in the fact that the classification threshold is based on the statistical distribution characteristics of real operating data in dimensions such as instantaneous energy peak, spectral energy concentration, and ultraviolet spectral peak intensity. By quantifying the separability of fault and non-fault conditions in each parameter space, a judgment boundary that maximally distinguishes the two types of conditions is established, and these thresholds are integrated into a fault physical model template. By incorporating actual operating condition data into the template construction process, the classification threshold is tightly coupled with the physical characteristics of the equipment and its operating environment, reducing the risk of misjudgment and missed judgment due to improper parameter settings, and improving the scenario adaptability and discrimination accuracy of the fault identification model.

[0012] In conjunction with some embodiments of the first aspect, in some embodiments, applied to a second ring network box device, the method includes: the second ring network box receiving and parsing an isolation command message, controlling an internal second switch to perform a tripping action, and disconnecting the second ring network box from the upstream faulty line; after performing the tripping action, the second ring network box generating a request for help message containing historical load data of the downstream faulty line before the tripping; the second ring network box broadcasting the request for help message to at least one third ring network box in the tie switch system; the second ring network box receiving and parsing at least one response message, obtaining the remaining capacity data therein as the remaining capacity data to be tested; the second ring network box comparing the remaining capacity data to be tested with the power loss load data in the request for help message according to preset conditions, and filtering the target remaining capacity data; the second ring network box determining the third ring network box corresponding to the target remaining capacity data as the target third ring network box, and sending a takeover confirmation message to the target third ring network box.

[0013] This embodiment enables the second ring network box to immediately generate a request for help message containing historical load data of the downstream faulty line before the trip, and broadcast it to at least one third ring network box in the tie switch system, proactively initiating capacity negotiation. The core of this mechanism is that, while completing fault isolation, the second ring network box carries the actual demand information of its downstream loads and triggers the tie-side ring network box to generate a response message based on its remaining capacity data via broadcast. After receiving multiple response messages, the second ring network box compares and filters the remaining capacity data to be tested with the power-loss load data according to preset conditions, autonomously determines the target third ring network box, and sends a takeover confirmation message, completing the local decision-making for the transfer target. This solution decentralizes the load transfer decision-making logic to the local ring network box, enabling it to quickly initiate the transfer negotiation process after the isolation action is completed, shortening the power restoration waiting time for the power-loss load, and improving the response speed of power restoration after fault isolation and the system's self-healing capability.

[0014] In conjunction with some embodiments of the first aspect, in some embodiments, the step of comparing the remaining capacity data to be measured with the power outage load data in the distress message according to preset conditions and filtering the target remaining capacity data specifically includes: multiplying the power outage load data in the distress message by a preset safety factor to calculate the load demand threshold, wherein the preset safety factor is greater than 1; comparing each received remaining capacity data to be measured with the load demand threshold; storing all remaining capacity data to be measured that are greater than the load demand threshold into a candidate capacity data set; and selecting the data with the largest value from the candidate capacity data set as the target remaining capacity data.

[0015] This embodiment calculates the load demand threshold by multiplying the power outage load data by a preset safety factor greater than 1, and selects the maximum value from the remaining capacity data to be measured that exceeds this threshold as the target remaining capacity data, thus achieving capacity matching screening with margin. This mechanism elevates the judgment criterion from "remaining capacity ≥ power outage load" to "remaining capacity ≥ power outage load × safety factor," forcibly reserving overload margin and prioritizing the largest remaining capacity. Distribution network load fluctuates dynamically with factors such as electricity consumption behavior and ambient temperature, and sensor errors and transmission delays may cause deviations between historical load data and actual load. Schemes that only meet numerical matching may lead to overload of the tie lines after load fluctuation, causing secondary tripping. This scheme embeds an anti-disturbance buffer mechanism into the decision-making process through a safety factor and the "maximum capacity priority" principle, reducing the risk of secondary tripping due to insufficient capacity margin and improving the safety margin and operational stability of the load transfer scheme.

[0016] In conjunction with some embodiments of the first aspect, in some embodiments, the help request message is used to enable the third ring network box to obtain its own rated capacity data and current real-time load data, calculate and generate remaining capacity data, encapsulate it into a response message, and send it to the second ring network box.

[0017] This embodiment enables the third ring network box to autonomously obtain its own rated capacity data and current real-time load data after receiving a request for help, calculate and generate remaining capacity data, and encapsulate it into a response message to feed back to the second ring network box. This realizes the distributed generation and point-to-point interaction of capacity information, avoids the communication overhead of the dispatch center centrally querying and calculating the remaining capacity of each ring network box, and improves the response efficiency of the capacity negotiation process.

[0018] In conjunction with some embodiments of the first aspect, in some embodiments, applied to a target third ring network box, the method includes: after receiving a takeover confirmation message, the target third ring network box detects the voltage value of the interconnection line between the target third ring network box and the second ring network box; when the voltage value of the interconnection line is detected to be zero, the target third ring network box controls its interconnection switch to perform a closing action, and restores power supply to the second ring network box and downstream through the interconnection line.

[0019] This embodiment, upon receiving the takeover confirmation message, first detects the voltage value of the tie line between the target third ring network box and the second ring network box, and only controls the tie switch to close when the detected voltage value is zero. This mechanism inserts a real-time detection step of the tie line voltage status before closing, verifying through physical quantities that there is no potential difference on both sides of the tie line. After fault isolation in the distribution network, if the upstream side fails to complete the power outage due to abnormalities such as switch failure, misoperation, or lost communication commands, or if residual voltage remains on the downstream side due to reverse power supply from distributed power sources, a safety interlock is established between logical judgment and physical state by forcibly requiring a zero voltage value as the physical verification condition for closing. This reduces the risk of unsafe closing due to communication abnormalities or asynchronous equipment states, and improves the safety of load transfer operations.

[0020] Secondly, embodiments of this application provide a ring network box fault handling device, which includes: one or more processors and a memory; the memory is coupled to the one or more processors, and the memory is used to store computer program code, which includes computer instructions, and the one or more processors call the computer instructions to cause the ring network box fault handling device to perform the method described in the first aspect and any possible implementation thereof.

[0021] Thirdly, embodiments of this application provide a computer program product containing instructions that, when the computer program product is run on a ring network box fault handling device, cause the ring network box fault handling device to execute the method described in the first aspect and any possible implementation thereof.

[0022] Fourthly, embodiments of this application provide a computer-readable storage medium including instructions that, when executed on a ring network box fault handling device, cause the ring network box fault handling device to perform the method described in the first aspect and any possible implementation thereof.

[0023] Understandably, the ring main unit fault handling device provided in the second aspect, the computer program product provided in the third aspect, and the computer storage medium provided in the fourth aspect are all used to execute the methods provided in the embodiments of this application. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods, and will not be repeated here.

[0024] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0025] 1. This application achieves multi-physical domain collaborative verification by introducing a cross-comparison mechanism between at least two types of non-electrical sensor data and fault electrical characteristics. Through cross-verification across multiple physical domains, the probability of protection maloperation is reduced, and the accuracy of fault identification is improved.

[0026] 2. This application extracts the spectral energy concentration from acoustic and vibration data, quantifies the light intensity jump rate from arc light data, and transforms the original signal into quantitative feature parameters with fault physical diagnosis significance. It then calculates the matching degree with a preset fault physical model template, enabling the discrimination logic to have a stronger ability to distinguish the differences between interference signals and real faults at the physical mechanism level, thereby improving the accuracy of fault diagnosis.

[0027] 3. This application achieves data-driven generation of discrimination criteria by enabling the fault physical model template to adaptively generate classification thresholds based on actual multimodal data through statistical analysis. By incorporating actual operating condition data into the template construction process, the classification thresholds are tightly coupled with the physical characteristics and operating environment of the equipment, reducing the risk of misjudgment and missed judgment due to improper threshold setting, and improving the adaptability and discrimination accuracy of the fault identification model under different operating scenarios. Attached Figure Description

[0028] Figure 1 This is a flowchart illustrating a ring main unit fault handling method in an embodiment of this application;

[0029] Figure 2 This is a flowchart illustrating a ring main unit fault identification method in an embodiment of this application;

[0030] Figure 3 This is a flowchart illustrating the adaptive fault identification method for ring network boxes in this application embodiment;

[0031] Figure 4This is a schematic diagram of the physical device structure of a ring network box fault handling device in the embodiments of this application. Detailed Implementation

[0032] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification of this application, the singular expressions “a,” “an,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to any or all possible combinations including one or more of the listed items.

[0033] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0034] To facilitate understanding, the application scenarios of the embodiments of this application are described below.

[0035] With the rapid development of urban power distribution network construction, 10kV ring main units play a crucial role in ensuring power supply reliability as key node equipment. Ring main units integrate various electrical equipment such as circuit breakers, load switches, cable joints, and busbars. During long-term operation, they may experience faults such as short circuits, grounding, and insulation breakdown, threatening power grid safety.

[0036] Some related technologies primarily rely on electrical quantities such as current and voltage as criteria, employing traditional relay protection principles like instantaneous overcurrent protection, overcurrent protection, and zero-sequence protection for fault detection. These methods demonstrate high reliability and technological maturity in identifying typical faults such as metallic short circuits. Some advanced devices have introduced auxiliary criteria such as waveform recognition and harmonic analysis, improving protection sensitivity to some extent. However, relying solely on electrical quantity characteristics makes it difficult to effectively distinguish between genuine faults and non-fault disturbances (such as capacitor switching, load surges, and adjacent line operations). These disturbances can generate current surges or voltage dips similar to faults, leading to maloperation or failure to operate by protection devices, resulting in a high false alarm rate. For atypical faults such as high-resistance grounding and arcing faults, electrical quantity characteristics are not obvious, making reliable identification based solely on current and voltage amplitudes difficult, resulting in protection dead zones.

[0037] This invention proposes a fault identification method for ring network boxes based on multimodal data fusion. By synchronously collecting multi-source heterogeneous signals such as electrical characteristics, acoustic patterns, vibration, and arc light, a fault physical model template is established to achieve accurate differentiation between faulty and non-faulty disturbances.

[0038] To facilitate understanding, the method provided in this implementation will be described in detail below, using the above scenario as an example. Please refer to [link / reference]. Figure 1 This is a flowchart illustrating a ring network box fault handling method in an embodiment of this application.

[0039] S101, the first ring network box acquires corresponding electrical parameter data and at least two types of non-electrical sensor data through electrical sensors and non-electrical sensors.

[0040] Electrical sensors refer to measuring devices installed inside ring main units (RMUs) for monitoring electrical parameters, including but not limited to current transformers, voltage transformers, and zero-sequence current sensors. These sensors are used to collect real-time electrical parameters such as current amplitude, voltage amplitude, frequency, phase angle, active power, reactive power, and zero-sequence current. Non-electrical sensors refer to sensing devices that monitor fault symptoms based on non-electrical physical quantities. These include acoustic sensors (which pick up acoustic signals generated by fault discharge), vibration sensors (which detect abnormal equipment vibration), arc light sensors (which capture light radiation during arc discharge), temperature sensors (which monitor localized overheating), and gas sensors (which detect SF6 decomposition products). Electrical parameter data represents a sequence of electrical quantity measurements collected and digitized by electrical sensors, including timestamps, amplitude, and phase information.

[0041] Specifically, the first ring main unit uses an array of electrical sensors deployed inside to perform high-speed synchronous sampling of electrical quantities such as line current and voltage at a preset sampling frequency (e.g., 10kHz), generating an electrical parameter data stream containing amplitude, phase, and harmonic components. Simultaneously, acoustic sensors capture acoustic signals from inside the ring main unit and the busbar area at a sampling rate of no less than 40kHz. Vibration sensors, in the form of accelerometers, are mounted on the surface of key equipment to collect triaxial vibration data. Arc light sensors monitor changes in light intensity in the ultraviolet-visible band in real time through a photodetector array. The output signals of each non-electrical sensor are converted from analog to digital to form a time-aligned non-electrical sensor dataset. A hardware clock synchronization mechanism (such as the IEEE 1588 precision clock protocol) ensures that the alignment accuracy between the electrical parameter data and the non-electrical sensor data on the time axis is better than 1 microsecond, establishing a unified time reference for subsequent cross-comparison. The collected multimodal data is preprocessed (denoised and normalized) and stored in a circular buffer for real-time use by the fault diagnosis algorithm, enabling comprehensive physical quantity monitoring of the ring main unit's operating status.

[0042] S102. The first ring network box determines whether the electrical parameter data contains preset fault electrical characteristics.

[0043] When the ring main unit detects sudden changes, exceeding limits, or abnormal fluctuations in the electrical quantities of the line, the system needs to quickly determine whether the electrical disturbance has fault characteristics in order to decide whether to activate the subsequent multi-modal cross-verification mechanism. Specifically, the first ring main unit extracts the latest electrical parameter data window (such as the sampling data of the last 10 power frequency cycles) from the circular buffer and performs feature extraction on this data window: calculates the effective value, peak value, and rate of change of the current, decomposes the harmonic components through Fast Fourier Transform (FFT), and extracts fault-sensitive quantities such as zero-sequence current and negative-sequence current. The extracted feature values ​​are compared item by item with the criterion thresholds in the preset fault electrical feature library—if the current mutation is greater than 5 times the rated current and the duration exceeds half a cycle, it is determined to meet the "short-circuit current characteristic"; if the zero-sequence current suddenly increases from near zero to more than 10% of the rated current, it is determined to meet the "single-phase grounding characteristic"; if the three-phase voltage imbalance exceeds the set limit, it is determined to meet the "asymmetrical fault characteristic". If any feature criterion is met, the system confirms that the electrical parameter data has a preset fault electrical feature, and records the type identifier of the feature (such as "short circuit feature" or "grounding feature") and its quantification parameters (current amplitude, duration, etc.) as the input basis for the cross-comparison step.

[0044] S103. If the electrical parameter data has preset fault electrical characteristics, the first ring network box will cross-compare the fault electrical characteristics with data from at least two non-electrical sensors to determine whether to generate a definitive fault identifier.

[0045] Specifically, after the first ring cage identifies fault electrical characteristics (such as short-circuit current characteristics), it immediately extracts non-electrical sensor data segments from the data buffer that are strictly time-aligned with the occurrence of the electrical characteristic (the time window range is 50 milliseconds before and after the electrical characteristic is triggered), ensuring that the causal relationship of the physical event is traceable. The system performs time-frequency analysis on the acoustic sensor data to detect whether there are broadband popping sound characteristics (energy concentrated in the 5-20kHz frequency band) corresponding to arc discharge; it performs spectral decomposition on the vibration sensor data to determine whether there are low-frequency high-amplitude vibrations (dominant frequency in the 100-500Hz) corresponding to short-circuit electromagnetic force impact; and it analyzes the spectral composition of the arc light sensor data to verify whether strong pulse radiation in the ultraviolet band (200-400nm) is detected. If the electrical characteristic is "short circuit," the corresponding physical accompanying phenomena should include: instantaneous high-energy broadband sonic booms in acoustic data, electromagnetic force impact pulses in vibration data, and a sudden increase in ultraviolet light intensity detected in arc light data. When at least two non-electrical sensors simultaneously detect characteristic patterns consistent with the physical process of a short circuit, and the consistency error between these features and the electrical characteristics in terms of timestamps is less than 5 milliseconds, the system generates a definitive fault identifier through logical AND operation. Conversely, if the electrical anomaly is caused by capacitor switching, although there is a sudden change in current, the acoustic data is only single-frequency noise from mechanical switch operation, the vibration is low-amplitude mechanical vibration, and the arc light sensor has no response. In this case, the cross-comparison result does not meet the fault physical model, and the system does not generate a definitive fault identifier, thereby suppressing false operation.

[0046] S104. After generating a confirmed fault identifier, the first switch inside the first ring network box controls the tripping action, disconnecting the first ring network box from the downstream faulty line.

[0047] When the system confirms the existence of a fault through multimodal cross-validation, the faulty line must be disconnected as quickly as possible to limit the fault range, protect equipment safety, and maintain power supply to non-faulty areas. Specifically, after receiving a confirmed fault indication, the intelligent control unit of the first ring network box sends a tripping command to the operating mechanism of the first switch (via hardwired connection or fieldbus communication). After tripping, the auxiliary contact state of the first switch flips, feeding back a tripping signal to the control unit, confirming that the electrical connection between the first ring network box and the downstream faulty line has been reliably disconnected. At this time, the short-circuit current at the fault point is interrupted, the fault arc is extinguished, the upstream grid bus voltage is restored, and the non-faulty feeders continue to supply power normally.

[0048] S105, the first ring network box sends an isolation command message to the second ring network box of the downstream faulty line.

[0049] The isolation command message refers to the communication data frame sent by the first ring network box to the downstream adjacent ring network box, used to indicate the fault location and request coordinated isolation. The message content includes fields such as fault type identifier, fault occurrence time, sender equipment ID, and required isolation action, and is encapsulated using industrial communication protocols such as IEC61850 GOOSE (General Object-Oriented Substation Events) or Modbus. The second ring network box refers to the ring network box equipment located downstream of the faulty line, connected in series with the first ring network box. The opening action of its internal switch can cut off the current backflow path from the fault point from the downstream side, achieving double-end blocking of the fault in conjunction with upstream isolation.

[0050] When the upstream side has disconnected the faulty line, if the downstream ring network box does not trip synchronously, the fault point may still be energized due to the reverse power supply formed by the tie line or distributed power source, affecting the isolation effect. Therefore, it is necessary to notify the downstream equipment to perform coordinated tripping through communication. Specifically, after confirming the tripping signal of the first switch, the first ring network box immediately constructs an isolation command message through its communication module (Ethernet interface or Fiber Channel). The message includes key fields such as fault type (e.g., "downstream line short circuit"), fault timestamp (accurate to milliseconds), local device address (used by downstream equipment to identify the source of the command), and isolation action command code (e.g., "immediate tripping"), and calculates the message checksum to ensure data integrity. The message is sent to the communication interface of the second ring network box corresponding to the downstream faulty line via multicast or point-to-point through the distribution automation communication network (e.g., fiber optic industrial Ethernet ring network). After receiving the message, the second ring network box confirms the legality of the command source by parsing the device address in the message header, verifies the integrity of the message, extracts the fault type and action command, and triggers the local protection logic to perform tripping.

[0051] S106. The second ring network box receives and parses the isolation command message, controls the internal second switch to perform a tripping action, and cuts off the connection between the second ring network box and the upstream faulty line.

[0052] When the upstream ring main unit detects a fault and trips, the downstream ring main unit must quickly respond to the isolation command, disconnecting the faulty line from the downstream side to form a double-ended isolation with the upstream, preventing the fault point from remaining energized due to interconnected power supply or distributed power backflow. Specifically, after receiving the isolation command message through the Ethernet physical layer, the communication module of the second ring main unit first performs a CRC check to confirm that the message has not been damaged during transmission. Then, it parses the message frame structure according to the IEC61850 or Modbus protocol specifications, extracting key information such as the fault type field (confirming a downstream line fault), the action command code (confirming a trip command), and the sender device ID (verified as the upstream first ring main unit). After confirming the legality of the command, the control logic unit immediately sends an operating current to the trip coil of the second switch, or issues a trip command through the electronic trip unit, driving the operating mechanism of the second switch to perform the trip action. After the trip is completed, the second ring main unit confirms the trip is in place through the status feedback of the switch auxiliary contacts, and can selectively send a confirmation message back to the first ring main unit, forming a closed-loop communication.

[0053] S107. After the second ring network box performs the tripping action, it generates a help message containing historical load data of the downstream faulty line before the tripping.

[0054] Historical load data refers to the real-time load power or current data of downstream lines recorded by the second ring network box before the tripping action. This includes quantitative indicators such as average active power, reactive power, and effective value of three-phase current within a time window before the tripping (e.g., the most recent 15 minutes), used to characterize the actual power demand in the power-loss area. The request for assistance message indicates a communication data frame initiated by the second ring network box after completing fault isolation, requesting power transfer support from the tie-line switching system. The message content includes fields such as power-loss load data, the device address, power-loss time, and request for power transfer flag. It is sent via broadcast or multicast to trigger responses from candidate power transfer sources, enabling the third ring network box to obtain its rated capacity data and current real-time load data, calculate and generate remaining capacity data, encapsulate it into a response message, and send it to the second ring network box.

[0055] Specifically, after confirming the trip signal of the second switch, the second ring network box immediately extracts historical load data prior to the trip from the local data recording module: it reads the active power sampling sequence within 15 minutes before the trip, calculates its time average as the baseline value of the power outage load (e.g., average 2.5MW), and records the peak power within this time window (e.g., peak 2.8MW) as the upper limit reference for load fluctuation. The system encapsulates the power outage load data (baseline 2.5MW, peak 2.8MW), the unique identifier of this device (e.g., device ID: RMU-02), the time of power outage (UTC timestamp), and the transfer request flag into a request for help message according to a predetermined protocol format. The power outage load data in the message is quantized in power units (kW) to ensure that the receiver can directly use it for capacity matching calculations. The generated request for help message is sent via multicast to the predefined tie switch system address group through the distribution automation communication network, triggering all potential transfer sources (third ring network box) to receive and process the request.

[0056] S108. The second ring network box broadcasts a distress message to at least one third ring network box in the tie switch system.

[0057] The tie switch system refers to the collection of tie lines and their control switches in a distribution network used to connect different power supply zones or feeders from different power sources. Under normal operation, the tie switches are in the open state to maintain independent operation of each zone; in case of a fault, load transfer can be achieved by closing the switches. The third ring network box represents a ring network box device capable of providing transfer power to the second ring network box. Its internal tie switches are connected to an upstream power source different from that of the second ring network box, and it has remaining capacity to accommodate transferred loads.

[0058] When the second-ring network box requires power transfer support, it can broadcast a request for assistance to the communication system, simultaneously triggering multiple candidate power transfer sources to perform capacity self-checks and responses. This enables parallel discovery of power transfer resources and shortens the power transfer decision time. Specifically, the second-ring network box sends an encapsulated request message through its communication module to a predefined communication system multicast address (e.g., 224.0.1.100) or broadcast address (e.g., subnet broadcast of 255.255.255.255). The message is transmitted via the distribution automation communication network (fiber optic Ethernet or wireless private network) and is synchronously received by the communication interfaces of all online third-ring network boxes in the communication switch system. The broadcast method ensures that even if the second-ring network box is unaware of which third-ring network boxes have power transfer capabilities, all potential power transfer sources can receive the request and decide whether to respond based on their own status, thus realizing an automatic discovery mechanism for power transfer resources. The time-to-live (TTL) of the message at the network layer is set to a reasonable value (e.g., 5 hops) to limit the broadcast range to the local distribution automation network and avoid cross-regional diffusion that could cause network congestion.

[0059] S109. The target third ring network box sends a response message to the second ring network box.

[0060] The response message refers to the response message generated by the third ring network box after receiving the request for help, based on its own operating status. This message includes remaining capacity data and a device identifier, and is used to inform the second ring network box whether the device has the capacity to transfer power and the available capacity. The remaining capacity data represents the additional load power that the third ring network box can currently accept. It is calculated by subtracting the current real-time load from the rated capacity, quantifying the upper limit of the load that the device can handle without overload.

[0061] Specifically, the communication module of the third-ring network enclosure receives and parses the distress message, extracting the power loss load data (e.g., 2.5MW) and the requesting device ID (RMU-02). The control unit immediately reads its own rated capacity parameters (e.g., rated capacity 5MW, stored in the device configuration file) and current real-time load data (e.g., current load 1.8MW, measured in real-time by a current sensor) from the local data acquisition system, and calculates the remaining capacity through subtraction: 5MW - 1.8MW = 3.2MW. If the remaining capacity is greater than zero (indicating potential for power transfer), the third-ring network enclosure encapsulates the remaining capacity data (3.2MW), its own device ID (RMU-03), and the current timestamp into a response message, and sends it via unicast to the address of the second-ring network enclosure identified in the distress message (point-to-point communication to avoid network storms caused by broadcast replies). If the remaining capacity is less than or equal to zero or the device is under maintenance, no response message is sent, and the device automatically withdraws from the power transfer candidate list.

[0062] S110, the second ring network box receives and parses at least one response message, and obtains the remaining capacity data in it as the remaining capacity data to be tested.

[0063] When multiple third-ring network enclosures respond to the supply transfer request, the second-ring network enclosure needs to aggregate all response information, extract key capacity data, and provide a complete candidate set for subsequent capacity matching and screening. Specifically, after sending a request for help, the second-ring network enclosure starts a timer (with a timeout set to 500 milliseconds to balance response speed and network latency), during which it continuously listens for response messages received by the communication interface. Upon receiving each response message, the communication module immediately performs CRC check and protocol parsing: verifying message integrity, confirming the message type as "supply transfer response," extracting the remaining capacity field value (e.g., 3.2MW), the sender device ID (e.g., RMU-03), timestamp, and other information, and storing this data as a record in the local candidate list. After the timer expires, the system stops receiving new responses and summarizes all records in the candidate list: Assuming three response messages are received, from RMU-03 (3.2MW remaining), RMU-04 (1.5MW remaining), and RMU-05 (4.0MW remaining), the set of remaining capacity data to be tested is {3.2MW, 1.5MW, 4.0MW}, and the corresponding set of device IDs is {RMU-03, RMU-04, RMU-05}.

[0064] S111, the second ring network box compares the remaining capacity data to be tested with the power failure load data in the distress message according to preset conditions, and filters the target remaining capacity data.

[0065] Among them, the preset conditions refer to the comparison rules used to determine whether the remaining capacity data to be tested meets the transfer requirements. These can be logical constraints such as capacity threshold calculation methods (e.g., multiplying by a safety factor), comparison operators (greater than, greater than or equal to), and filtering priorities (e.g., maximum value priority), without any limitations.

[0066] Specifically, the second ring network enclosure first extracts the power loss load data (e.g., a baseline value of 2.5MW) from the distress call message, multiplies it by a preset safety factor (e.g., 1.2) to calculate the load demand threshold: 2.5MW × 1.2 = 3.0MW. This threshold serves as the entry threshold for candidate resources. The system iterates through the set of remaining capacity data to be tested {3.2MW, 1.5MW, 4.0MW}, performing a greater than operation on each data point: 3.2MW > 3.0MW (satisfied), 1.5MW > 3.0MW (not satisfied), 4.0MW > 3.0MW (satisfied). The data points {3.2MW, 4.0MW} that meet the conditions are stored in the candidate capacity data set, while the RMU-04 corresponding to 1.5MW is removed. In the candidate set, a maximum value search is performed: 3.2MW and 4.0MW are compared, and 4.0MW is determined to be the maximum value. It is marked as the target remaining capacity data, and the corresponding device ID is recorded as RMU-05. If the candidate set is empty (i.e. no third ring cage meets the threshold requirement), the system triggers a transfer failure alarm and can choose to reduce the security factor for re-screening or report to the dispatch center for manual intervention.

[0067] S112. The second ring network box identifies the third ring network box corresponding to the target remaining capacity data as the target third ring network box and sends a takeover confirmation message to the target third ring network box.

[0068] Specifically, after determining the target remaining capacity data as 4.0MW and the corresponding device ID as RMU-05, the second ring network box marks RMU-05 as the target third ring network box. The system immediately constructs a takeover confirmation message: filling in the transfer confirmation flag (indicating "request to execute transfer"), power outage load data (2.5MW, used for secondary verification of the target device's capacity), suggested closing time (current time + reserved detection time, such as current time + 2 seconds), and the device ID (RMU-02, used by the target device to identify the requester), and calculating the message checksum. The message is sent unicast to the communication address of RMU-05 via the distribution automation communication network (point-to-point transmission to avoid malfunctions of other devices). After receiving the takeover confirmation message, RMU-05 parses and confirms that it has been selected as the transfer source, and begins to execute the tie line voltage detection and closing preparation process. Meanwhile, the second ring network box can selectively send a "cancel standby" message to the unselected third ring network box (RMU-03) to notify it to exit the transfer candidate status and release communication resources.

[0069] S113. After receiving the takeover confirmation message, the target third ring network box detects the voltage value of the connection line between the target third ring network box and the second ring network box.

[0070] Specifically, after parsing the takeover confirmation message and confirming that it has been selected as the power transfer source, the target third ring network box immediately sends a voltage detection command to the tie switch control module. The control module reads the output signals of the voltage transformers installed on both sides of the tie switch to obtain the voltage amplitude U1 at the upper end of the switch (connected to the busbar side of this ring network box) and the voltage amplitude U2 at the lower end of the switch (connected to the tie line to the second ring network box). The system calculates the absolute value of the voltage difference between the two sides: |U1-U2|, and compares it with the preset zero voltage threshold (e.g., 50V, considering measurement error). If |U1-U2|<50V, the tie line voltage is determined to be zero, indicating that the potentials at both ends are equal (usually, the second ring network box side has lost power to zero voltage due to upstream tripping, while the busbar on the third ring network box side is energized, and the induced voltage at the lower end after passing through the line impedance is also close to zero). If a non-zero voltage value is detected (e.g., U2 still maintains a high voltage, possibly due to incomplete power disconnection on the second ring network box side or the presence of distributed power supply backflow), the system will prohibit closing the circuit breaker and trigger an alarm, waiting for the voltage to drop to zero or for manual handling.

[0071] S114. When the voltage value of the connecting line is detected to be zero, the target third ring network box controls its connecting switch to perform a closing action, and restores power supply to the second ring network box and downstream through the connecting line.

[0072] When safety conditions are met, the target third-ring network box extends its power supply to the power-loss area through the closing tie switch, completing a closed loop from fault isolation to power restoration and realizing the self-healing function of the distribution network. Specifically, after the voltage detection module confirms that |U1-U2|<50V (voltage value is zero), the control logic unit of the target third-ring network box immediately sends an operating current to the closing coil of the tie switch, or issues a closing command through the electronic controller. At the moment of closing, the voltage of the third-ring network box busbar (e.g., 10kV) is conducted to the second-ring network box busbar through the tie line, and the busbar voltage of the second-ring network box rapidly rises from zero to the rated value of 10kV. The downstream feeder switch of the second-ring network box (if it did not trip during fault isolation) automatically resumes power supply to the downstream load as the busbar becomes energized, and the voltage and frequency of the power-loss users return to normal, realizing power restoration. The target third-ring network box monitors the current transformer of the tie switch to confirm that the line current rises to a value matching the transferred load after the switch is closed (e.g., the current rises from 0 to approximately 250A, corresponding to a 2.5MW load), thus verifying successful power transfer. Simultaneously, the system can report the completed power transfer status to the distribution automation master station to update the network topology and power flow distribution.

[0073] The above embodiments, through the deployment, data acquisition, and feature extraction of a multimodal sensor network, established a fault physical model template based on statistical analysis, providing a basic model framework for ring main unit fault identification. However, in practical applications, the structures of different models of ring main units vary significantly, and the multimodal characteristics of the same fault type may deviate on different devices, making a single template difficult to adapt to all scenarios. Therefore, this application further proposes an adaptive optimization scheme to improve the method's versatility and robustness.

[0074] The following provides a more detailed description of the process of the method provided in this implementation. Please refer to [link / reference]. Figure 2 This is a flowchart illustrating a cage fault identification method in an embodiment of this application.

[0075] S201. Perform joint time-domain and frequency-domain analysis on the acoustic sensor data and vibration sensor data respectively to generate their respective three-dimensional spectrum maps. Perform spectral component analysis on the arc light sensor data to generate a spectral intensity distribution map.

[0076] When electrical criteria initially identify an anomaly, the system needs to perform refined analysis on non-electrical signals within the same time window, transforming the original time-domain waveform into a time-frequency domain feature spectrum with fault diagnosis significance, providing standardized input for subsequent physical pattern recognition. Specifically, the first ring network box extracts acoustic sensor data windows (e.g., 200ms of data before and after the fault trigger, 100ms each, sampling rate 40kHz, totaling 8000 sampling points) and vibration sensor data windows (e.g., triaxial acceleration data, sampling rate 10kHz) that are strictly aligned with the electrical fault feature time from the data buffer. A Short-Time Fourier Transform (STFT) is performed on the acoustic data: the 200ms signal is divided into multiple overlapping time windows (e.g., window length 25ms, overlap rate 75%), and an FFT is performed on the signal within each window to obtain the spectrum (frequency resolution 40Hz). The spectra of each window are arranged in chronological order to construct a three-dimensional matrix of time-frequency-amplitude, where the time axis resolution is approximately 6ms, the frequency axis covers 0-20kHz, and the amplitude axis uses a logarithmic scale (dB), generating a three-dimensional acoustic spectrum. The same time-frequency analysis process is performed on the vibration data, using continuous wavelet transform (CWT) or STFT to generate a three-dimensional vibration spectrum covering 0-5kHz (capturing low-frequency mechanical vibration and mid-frequency electromagnetic vibration). Spectral decomposition is performed on the arc light sensor data: if a multi-channel photodetector is used, the light intensity time series of each band (UV 200-280nm, 280-400nm; visible light 400-500nm, 500-600nm, 600-700nm) are directly read; if a single-channel broadband detector is used, the components of each band are acquired by a multi-channel ADC after spectral dispersion using a filter bank or grating. The light intensity data of each band are aligned along the time axis to construct a two-dimensional matrix of band-time-light intensity, generating a spectral intensity distribution map. The band axis is discretized into 5-10 characteristic bands, the time axis resolution is consistent with the electrical data (e.g., 1ms), and the light intensity axis uses normalized amplitude or absolute irradiance (μW / cm²).

[0077] S202. Perform a slice scan of the three-dimensional spectrum along the time axis, and calculate the spectral energy concentration of the time slice where the identified instantaneous energy peak is located.

[0078] When a sudden energy change occurs in the acoustic signature or vibration signal during its time evolution, the system needs to accurately locate the moment of the change and extract the spectral distribution characteristics at that moment to distinguish between different physical processes (such as the broadband burst of arc discharge and the narrow-frequency resonance of mechanical impact). Specifically, the first ring cage performs a time-axis traversal of the acoustic signature's three-dimensional spectrum: starting from the initial time t=0, with a step size of 5ms, it sequentially extracts two-dimensional frequency-amplitude slices at each time point t=0ms, 5ms, 10ms...200ms. For each slice, it calculates the instantaneous total energy E(t)=Σ[A(f,t)²], where A(f,t) is the amplitude of frequency f at time t, and sums it up across all frequency bins (e.g., 500 frequency points). Identify local maxima in the instantaneous energy sequence E(t): If E(t_peak) > E(t_peak-1) and E(t_peak) > E(t_peak+1), then t_peak is the moment of the instantaneous energy peak, and its quantized value E_peak is recorded (e.g., a normalized energy value of 0.85 indicates that it accounts for 85% of the total energy in the entire time window). For the identified peak moment t_peak, extract the time slice spectrum vector S_peak = [A(f1, t_peak), A(f2, t_peak), ..., A(f500, t_peak)]. Calculate the energy normalization for this spectrum vector: Pi = A(fi, t_peak)² / Σ[A(fj, t_peak)²], so that the sum of the energy proportions of each frequency bin is 1. Substituting into the Shannon entropy formula, the spectral energy concentration C = -Σ(Pi·log2(Pi)) is calculated. The entropy value ranges from 0 to log2(500) ≈ 8.97 bits: if the energy is concentrated at a single frequency point, C is close to 0 (high concentration); if the energy is evenly distributed across the entire frequency band, C is close to 8.97 (low concentration, i.e., high dispersion). The same slice scanning and concentration calculation process is performed on the three-dimensional vibration spectrum.

[0079] S203. When the spectral energy concentration is lower than the preset diffusion threshold, the time slice is determined to have full-spectrum diffusion characteristics, and the quantization value and diffusion characteristics of the instantaneous energy peak are recorded together as the first set of non-electrical characteristic parameters.

[0080] The dispersion threshold is a critical value used to determine whether the spectral energy exhibits a uniform distribution across the entire frequency band. This threshold is determined through statistical analysis of the spectral concentration of historical fault and non-fault cases, maximizing the distinction between broadband dispersion (fault characteristic) and narrowband concentration (non-fault characteristic). Numerically, it is typically taken from the mid-to-high range of Shannon entropy (e.g., 6.5 bits, corresponding to energy distribution across a large number of frequency bins). Full-spectrum dispersion indicates that the spectral energy of a time slice exhibits a broadband uniform distribution pattern along the frequency axis; that is, the energy is not concentrated at a few dominant frequencies but is dispersed across most frequency bins. This characteristic is a typical physical marker of impact faults such as arc discharge (the discharge process excites broadband sound waves and vibrations).

[0081] Specifically, the first ring cage compares the spectral energy concentration C_sound (e.g., C_sound=7.2 bits) calculated from the voiceprint data with the preset diffusion threshold (e.g., 6.5 bits). Since C_sound (7.2) > threshold (6.5), the concentration of this time slice is determined to be "low" (high entropy value indicates high dispersion and low concentration), satisfying the full-spectrum diffusion characteristic: energy is distributed in a wide frequency band of 0-20kHz rather than concentrated in a specific frequency. The system marks this slice as having a "diffusion characteristic" flag as True. At the same time, the instantaneous peak energy quantization value E_peak_sound=0.85 (normalized energy) is recorded. The two are combined to form a voiceprint feature parameter pair: (E_peak_sound=0.85, diffusion flag=True, concentration value=7.2). The same judgment is applied to vibration data: if C_vib=6.8>6.5, it is also judged to have diffusion characteristics, and the vibration characteristic parameter pair is recorded: (E_peak_vib=0.72, diffusion flag=True, concentration value=6.8). The characteristic parameter pairs of acoustic and vibration are integrated into the first set of non-electrical characteristic parameters: {acoustic energy 0.85, acoustic diffusion True, acoustic concentration 7.2, vibration energy 0.72, vibration diffusion True, vibration concentration 6.8}. If the concentration of a certain sensor does not exceed the threshold (e.g., the concentration of a certain vibration is only 5.5<6.5), it is judged not to have diffusion characteristics, and the diffusion flag is recorded as False, indicating that the energy is concentrated in a specific frequency band (such as mechanical resonance frequency), and this mode is inconsistent with the physical mechanism of the fault.

[0082] S204. In the distribution diagram of spectral intensity changing with time, identify and quantify the peak intensity and its first derivative in the ultraviolet spectral band, and obtain the light intensity jump rate as the second set of non-electrical characteristic parameters.

[0083] When the arc light sensor captures the optical radiation signal, the system needs to extract the ultraviolet band data with the most significant fault characteristics from the multi-band spectrum and quantify its intensity and dynamic change characteristics to provide a quantitative basis for determining the existence of the arc. Specifically, the first ring network box extracts the light intensity time series I_UV(t) of the ultraviolet band (200-400nm) from the spectral intensity distribution map. This series contains 200ms of sampling data (sampling rate 1kHz, 200 sampling points) before and after the electrical fault characteristics are triggered. Peak detection is performed on the I_UV(t) series: all sampling points are traversed, the maximum value I_peak=max(I_UV(t)) is identified (e.g., I_peak=850μW / cm²), and the time when the peak appears, t_peak, is recorded (e.g., t_peak=52ms, corresponding to 2ms after the start of the electrical short-circuit current). The first derivative of the I_UV(t) sequence is calculated using numerical differentiation: the central difference method is used, dI / dt(ti) = [I_UV(ti+1) - I_UV(ti-1)] / (2Δt), where Δt = 1ms is the sampling interval, resulting in a derivative sequence of dI / dt(t) with a total of 198 points. Maximum value detection is performed on the derivative sequence: the maximum value of dI / dt, max(dI / dt), is identified (e.g., max(dI / dt) = 680 (μW / cm²) / ms). This value is the light intensity jump rate, indicating that the light intensity suddenly increases at a rate of 680 (μW / cm²) / ms at a certain moment. The peak intensity and jump rate are combined to form the second set of non-electrical characteristic parameters: {UV peak intensity 850, light intensity jump rate 680}. If there is no obvious light intensity response in the ultraviolet band (e.g., I_peak < 10 μW / cm², close to the sensor noise floor), then the recorded peak value is close to zero, and the jump rate is also close to zero, indicating that no electric arc occurs.

[0084] S205. Calculate the matching degree between the quantized values ​​of the first group of non-electrical characteristic parameters and the second group of non-electrical characteristic parameters and the preset fault physical model template.

[0085] Specifically, the first ring cage loads the classification thresholds and weight configurations for each parameter from the fault physical model template: acoustic energy threshold range [0.7, 1.0], weight w1=0.15; vibration energy threshold range [0.6, 1.0], weight w2=0.15; diffusion feature requirement is True, weight w3=0.20; ultraviolet peak intensity threshold range [500, +∞) μW / cm², weight w4=0.25; light intensity jump rate threshold range [400, +∞) (μW / cm²) / ms, weight w5=0.25. Matching degree calculation is performed on each of the real-time extracted parameters:

[0086] Voiceprint energy matching degree: The measured value of 0.85 is within the range of [0.7, 1.0]. The normalized matching degree m1 is calculated as (0.85-0.7) / (1.0-0.7)=0.5. If it exceeds the upper limit, then m1=1.0.

[0087] Vibration energy matching degree: The measured value of 0.72 is located in [0.6, 1.0], m2 = (0.72-0.6) / (1.0-0.6) = 0.3;

[0088] Diffusion feature matching degree: The actual measurement is True, consistent with the requirements, m3=1.0;

[0089] UV peak matching degree: measured value 850>500, m4=1.0 (exceeding the threshold is full score);

[0090] Light intensity jump rate matching degree: measured value 680>400, m5=1.0.

[0091] The overall matching degree is calculated using a weighted summation: M = w1·m1 + w2·m2 + w3·m3 + w4·m4 + w5·m5 = 0.15 × 0.5 + 0.15 × 0.3 + 0.20 × 1.0 + 0.25 × 1.0 + 0.25 × 1.0 = 0.075 + 0.045 + 0.20 + 0.25 + 0.25 = 0.82. The matching degree score is M = 0.82 (82%). If a parameter does not meet the threshold (e.g., the UV peak value is only 50 < 500), the corresponding matching degree mi = 0, resulting in a significant decrease in the overall matching degree.

[0092] S206. When the matching degree calculation result exceeds the preset collaborative confidence threshold, a definitive fault identifier is generated.

[0093] When the matching degree between the multi-dimensional non-electrical characteristics and the fault physical model reaches a preset confidence level, the system needs to generate a definitive fault identifier to trigger the protection trip, completing the complete discrimination process from electrical anomaly detection to fault confirmation. Specifically, the first ring network box compares the matching degree calculation result M=0.82 with the preset collaborative confidence threshold (e.g., 0.75). Since M(0.82)>threshold(0.75), the "exceeded" condition is met, and the system determines that the current multi-modal characteristic combination is highly consistent with the fault physical model, confirming that the electrical anomaly is caused by a real physical fault (e.g., busbar short-circuit arc) rather than transient interference. The control logic unit immediately generates a definitive fault identifier (setting the flag FaultConfirmed=True), which serves as the trigger signal for the trip action in step S104. At the same time, it records detailed information such as the fault confirmation time, matching degree score, and quantization values ​​of each parameter to the fault recorder for post-event analysis. If the matching degree does not exceed the threshold (e.g., M=0.65<0.75), it indicates that although an electrical abnormality is detected, the non-electrical evidence is insufficient to support the fault determination (it may be normal operation such as capacitor switching). The system does not generate a conclusive fault indicator, and the protection logic remains in the "monitoring" state without performing a trip to avoid malfunction.

[0094] In some specific instances, the fault physical model template is established through the following preliminary steps:

[0095] a. Acquire the first set of multimodal data under preset fault conditions and the second set of multimodal data under non-fault conditions respectively. The multimodal data includes at least electrical characteristics, acoustic data, vibration data and arc light data.

[0096] The preset fault conditions refer to typical fault scenarios simulated in the laboratory or on decommissioned equipment, including busbar short circuits, cable flashovers, and insulation breakdowns. These must cover samples of varying severity (short-circuit impedance 0.01-0.5Ω) and different locations (incoming / outgoing side). Non-fault conditions represent disturbance scenarios during normal operation, including load switching, voltage sags, and lightning overvoltages—non-fault transient events. These conditions produce signal characteristics similar to faults but do not constitute real faults in essence.

[0097] Before model training, labeled samples need to be accumulated in advance to provide a data foundation for statistical analysis. When the system is set up for experimental purposes or historical data is reviewed, the representativeness and statistical significance of fault and non-fault samples must be ensured. Specifically, fault experiments are designed on the 10kV simulation platform: busbar short circuits are achieved by applying different impedances (0.01-0.5Ω) using a controllable short-circuit generator, with three phases × 4 impedances × 10 repetitions to obtain 120 sets of samples; cable flashovers are achieved by artificially triggered breakdowns to obtain 30 sets; insulation breakdowns are achieved by triggering pollution cracks to obtain 20 sets, for a total of 170 sets of the first set of data. For each set, three-phase current / voltage waveforms (20kHz, 600ms), acoustic signatures (40kHz), vibration (10kHz), and arc light (6 channels 1kHz) are collected. Non-fault conditions are simulated, including capacitor switching (20 sets), load surges (20 sets), voltage dips (15 sets), lightning overvoltages (10 sets), and proximity operations (10 sets), to obtain 75 sets of the second set of data. The first group of 50 cases and the second group of 30 cases were supplemented with on-site waveform data (through cross-verification via relay protection records), ultimately forming a dataset of 220 groups in the first group and 105 groups in the second group.

[0098] b. Perform analysis and identification operations on the first and second groups of multimodal data to extract their respective non-electrical characteristic parameter sets.

[0099] After data acquisition, the raw signals (tens of thousands of sampling points) need to be converted into structured feature vectors for statistical analysis. When traversing each group of samples, the system performs a unified feature extraction process for each modal data. Specifically, the first group of 220 samples is processed one by one: the acoustic signature data of the i-th sample is processed using STFT (window length 25ms, overlap 75%) to generate a spectrum, the energy peak moment is identified, the normalized value E_peak_sound_i is recorded, and the Shannon entropy C_sound_i of the spectrum slice at that moment is calculated. If C_sound_i > 6.5, the dispersion flag Scatter_sound_i = 1. Vibration data is processed similarly to obtain E_peak_vib_i, C_vib_i, and Scatter_vib_i. For arc light data, the ultraviolet band peak value I_UV_peak_i and the maximum value of the first derivative dI_dt_max_i are extracted. The eight parameters are used to form a feature vector X_i = [E_peak_sound_i, C_sound_i, Scatter_sound_i, E_peak_vib_i, C_vib_i, Scatter_vib_i, I_UV_peak_i, dI_dt_max_i]. This vector is then used to iterate through all samples to form the fault class feature matrix X_fault (220×8). The same process is then performed on the second group of 105 samples to obtain the non-fault class feature matrix Y_normal (105×8). For non-fault samples without arc light, the ultraviolet parameters are recorded as near-zero values ​​(<10 μW / cm²).

[0100] c. Perform statistical analysis on the non-electrical characteristic parameter sets of the first and second sets of multimodal data, and calculate and determine the classification thresholds that can maximize the distinction between the two sets of datasets for instantaneous energy peak, spectral energy concentration, ultraviolet spectral peak intensity, diffusion characteristics, and light intensity jump rate.

[0101] After feature extraction, the optimal decision boundary for each parameter needs to be determined to maximize the separability between the two classes of samples. When performing ROC analysis, the system iterates through candidate thresholds to calculate classification performance metrics. Specifically, for the peak voiceprint energy E_peak_sound: 220 values ​​(mean 0.82, concentrated in the range of 0.70-0.95) are extracted from the fault class and 105 values ​​(mean 0.45, concentrated in the range of 0.20-0.70) from the non-fault class, with an overlap of 0.60-0.75 between the two distributions. The candidate threshold is increased from 0.0 to 1.0 in steps of 0.01, and the TPR, TNR, and Youden index J are calculated for each threshold. The threshold corresponding to the maximum value of J is identified: at 0.68, TPR=0.95, TNR=0.92, and J=0.87 are at their maximum, thus the peak energy threshold is determined to be 0.68. Similarly, the spectral concentration C_sound was analyzed (7.1 bits of high-entropy dispersion for faulty classes and 4.8 bits of low-entropy concentration for non-faulty classes), determining a dispersion threshold of 6.3 bits (the point of maximum J). The ultraviolet peak I_UV_peak reached J=0.98 at 450 μW / cm² (780 for faulty classes and 8 for non-faulty classes), and the light intensity jump rate was J=0.96 at 350 (μW / cm²) / ms (680 for faulty classes and 12 for non-faulty classes). The same analysis was performed on vibration parameters to determine the threshold. Weights were assigned to each parameter according to the Youden index: ultraviolet peak (J=0.98, w=0.25), light intensity jump rate (J=0.96, w=0.25), dispersion characteristics (J=0.87, w=0.20), acoustic energy (J=0.75, w=0.15), and vibration energy (J=0.70, w=0.15).

[0102] d. Finally, integrate multiple classification thresholds to construct a fault physical model template.

[0103] After the threshold is determined, the scattered statistical results need to be organized into an executable decision model to support real-time fault identification. When the protection device is started, the template is loaded to initialize the matching calculation module. Specifically, the template data structure includes six fields: 1. Feature parameter definition: listing eight parameter names and dimensions (e.g., voiceprint energy 0-1, concentration bits, UV peak value μW / cm²); 2. Classification threshold configuration: assigning values ​​to various thresholds (voiceprint energy 0.68, concentration 6.3, UV peak value 450, light intensity jump rate 350, etc.); 3. Discriminant logic definition: direction (if energy / concentration / UV / jump rate are all > threshold, then it is a fault); 4. Weight coefficient configuration: assigning values ​​(UV 0.25, jump rate 0.25, diffusion 0.20, voiceprint energy 0.15, vibration energy 0.15); 5. Collaborative confidence threshold: setting the decision threshold to 0.75 (can be adjusted to 0.85 according to the false alarm rate requirement); 6. Model metadata: recording the training set size (220 faulty cases, 105 non-faulty cases), validation performance (accuracy 95%, recall 96%, F1 score 0.955), and version number. The structure is encoded in JSON format and stored in non-volatile memory, which is loaded and initialized upon startup. During real-time execution, the matching degree mi is obtained by comparing the threshold with the measured value from the template, and the weight wi is read and M=Σ(wi·mi) is executed. The result is then compared with the collaborative confidence threshold to complete the discrimination.

[0104] The above embodiments, through adaptive optimization, address the characteristic differences between different models of ring main units and achieve a universal deployment of the fault physical model. However, during long-term operation, ring main unit equipment will be affected by factors such as aging, environmental changes, and load characteristic evolution, leading to multimodal signal baseline drift, sensor sensitivity attenuation, and potential performance degradation. To ensure the continuous reliability of the protection system, this application further proposes a dynamic model update mechanism to adapt to changes in the operating status throughout the equipment's entire lifecycle.

[0105] The following provides a more detailed description of the process of the method provided in this implementation. Please refer to [link / reference]. Figure 3 This is a flowchart illustrating the adaptive fault identification method for ring network boxes in this application embodiment.

[0106] S301. Multiply the power failure load data in the distress message by a preset safety factor to calculate the load demand threshold. The preset safety factor is greater than 1.

[0107] The preset safety factor represents the capacity margin factor, with a value of 1.1-1.3. It is used to cope with load fluctuations, inrush currents, and measurement errors, ensuring that the transfer lines are not overloaded and that there is a dynamic margin.

[0108] When the first ring network enclosure fails and requires power transfer, the system needs to determine the required capacity for safe transfer, providing a quantitative standard for subsequent capacity selection. Specifically, the second ring network enclosure extracts the power loss load data (e.g., S=950kVA) from the request for assistance message, reads the preset safety factor from the configuration file (K=1.3 for industrial parks, K=1.15 for residential areas, and K=1.2 for commercial areas), and performs the calculation: S_threshold=950×1.3=1235kVA. If the load is three-phase unbalanced, the maximum phase load S_max is multiplied by the factor; if only active power P is available, S=P / cosφ is deduced based on historical power factors. The threshold of 1235kVA is stored for later use, and the calculation basis is recorded in the log for parameter optimization.

[0109] S302. Compare each received remaining capacity data to be tested with the load demand threshold.

[0110] The system needs to evaluate each remaining capacity data point to ensure it meets sufficiency requirements, removing any that do not. Specifically, assume four response messages are received: ENS-003 (1580kVA), ENS-004 (920kVA), ENS-005 (1350kVA), and ENS-006 (1200kVA), with a threshold of 1235kVA. Each message is compared: 1580 > 1235 passes, 920 < 1235 is removed (315kVA shortage), 1350 > 1235 passes, and 1200 < 1235 is removed (35kVA shortage). When removing data, a rejection response is sent to the corresponding ring network box, and the reason for removal is recorded in the log. Data validation is performed: negative values, values ​​exceeding limits, and expired data (timestamp > 5 seconds) are removed. The pass rate is calculated to be 50% (2 / 4). If it is less than 30%, a capacity shortage alarm is triggered.

[0111] S303. Store all remaining capacity data that exceeds the load demand threshold into the candidate capacity data set.

[0112] Specifically, initialize the dynamic array and extract the data marked "passed" in S302: insert {ENS-003, 1580kVA} into index 0, insert {ENS-005, 1350kVA} into index 1, and set size = 2. If the set is empty, trigger exception handling: reply with a failure to transfer and report a capacity shortage alarm; if the size = 1, directly determine it as the target and skip S304. Perform preprocessing: sort by capacity in descending order (reduce the extreme value search of S304 from O(n) to O(1)), remove duplicates (only keep the latest data for the same box number), and add a timeliness flag (validity period = current time + 5 seconds). Store the set in shared memory for subsequent access and persist it to the log to record statistical information (maximum capacity 1580kVA, average capacity 1465kVA, margin distribution).

[0113] S304. Select the data with the largest value from the candidate capacity data set and determine it as the target remaining capacity data.

[0114] Specifically, read the candidate set (size n=2) and initialize max_capacity=0. Iterate through the elements: Iteration 1 reads {ENS-003, 1580kVA}, 1580>0, update max=1580, target="ENS-003"; Iteration 2 reads {ENS-005, 1350kVA}, 1350<1580, do not update. Determine the target = {ENS-003, 1580kVA, margin 345kVA (27.9%)}. If there are ties for the maximum value, execute the secondary criteria: physical distance priority (choose the shorter path), device health priority (choose the higher score), timestamp priority (choose the earlier response). If the set is empty, output NULL and trigger the transfer failure process. Write the result to the decision register for subsequent modules to call, and record the alternative solution (ENS-005) and decision time (10ms) to the log.

[0115] The following describes the ring main unit fault handling device in the embodiments of this invention from the perspective of hardware processing. Please refer to [link to relevant documentation]. Figure 4 This is a schematic diagram of the physical device structure of a ring network box fault handling device in the embodiments of this application.

[0116] It should be noted that, Figure 4 The structure of the ring network box fault handling device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0117] like Figure 4 As shown, the ring main unit fault handling device includes a central processing unit (CPU) 401, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 402 or a program loaded from storage section 408 into random access memory (RAM) 403, such as performing the methods described in the above embodiments. The RAM 403 also stores various programs and data required for system operation. The CPU 401, ROM 402, and RAM 403 are interconnected via a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0118] The following components are connected to I / O interface 405: input section 406 including audio input devices, push-button switches, etc.; output section 407 including liquid crystal display (LCD) and audio output devices, indicator lights, etc.; storage section 408 including hard disks, etc.; and communication section 409 including network interface cards such as LAN (Local Area Network) cards, modems, etc. Communication section 409 performs communication processing via a network such as the Internet. Drive 410 is also connected to I / O interface 405 as needed. Removable media 411, such as disks, optical disks, magneto-optical disks, semiconductor memories, etc., are installed on drive 410 as needed so that computer programs read from them can be installed into storage section 408 as needed.

[0119] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing computer programs for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 409, and / or installed from removable medium 411. When the computer program is executed by central processing unit (CPU) 401, it performs the various functions defined in the present invention.

[0120] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0121] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. Each block in a flowchart or block diagram may represent a module, program segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those shown in the drawings.

[0122] Specifically, the ring network box fault handling device in this embodiment includes a processor and a memory. The memory stores a computer program. When the computer program is executed by the processor, it implements the ring network box fault handling method provided in the above embodiment.

[0123] In another aspect, the present invention also provides a computer-readable storage medium, which may be included in the ring main unit fault handling device described in the above embodiments; or it may exist independently and not assembled into the ring main unit fault handling device. The storage medium carries one or more computer programs, which, when executed by a processor of the ring main unit fault handling device, cause the ring main unit fault handling device to implement the ring main unit fault handling method provided in the above embodiments.

[0124] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0125] As used in the above embodiments, depending on the context, the term "when..." can be interpreted as meaning "if...", "after...", "in response to determining...", or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if (the stated condition or event) is interpreted as meaning "if determining...", "in response to determining...", "when (the stated condition or event) is detected", or "in response to detecting (the stated condition or event)".

[0126] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A method for handling ring main unit faults, characterized in that, Applied to the first ring cage equipment, the method includes: The first ring network box acquires corresponding electrical parameter data and at least two types of non-electrical sensor data through electrical sensors and non-electrical sensors; The first ring network box determines whether the electrical parameter data has preset fault electrical characteristics; If the electrical parameter data contains the preset fault electrical characteristics, the first ring network box will cross-compare the fault electrical characteristics with the data from at least two non-electrical sensors to determine whether to generate a definitive fault identifier. After generating the confirmed fault identifier, the first switch inside the first ring network box controls the tripping action to disconnect the first ring network box from the downstream faulty line. The first ring network box sends an isolation command message to the second ring network box of the downstream faulty line.

2. The method according to claim 1, characterized in that, The non-electrical sensor data includes acoustic signature sensor data, vibration sensor data, and arc light sensor data. The first ring network box cross-compares the electrical characteristics of the fault with the at least two types of non-electrical sensor data to determine whether to generate a definitive fault identifier, specifically including: The acoustic sensor data and the vibration sensor data are subjected to joint time-domain and frequency-domain analysis respectively to generate their respective three-dimensional spectrum maps. The arc light sensor data is subjected to spectral component analysis to generate a spectral intensity distribution map. The three-dimensional spectrum is scanned along the time axis in a slice manner, and the spectral energy concentration of the time slice where the identified instantaneous energy peak is located is calculated. When the spectral energy concentration is lower than the preset diffusion threshold, the time slice is determined to have full-spectrum diffusion characteristics, and the quantized value of the instantaneous energy peak and the diffusion characteristics are recorded together as the first set of non-electrical characteristic parameters. In the distribution diagram of the spectral intensity changing with time, the peak intensity and its first derivative of the ultraviolet spectral band are identified and quantified to obtain the light intensity jump rate as the second set of non-electrical characteristic parameters. The quantized values ​​of the first set of non-electrical characteristic parameters and the second set of non-electrical characteristic parameters are matched with the preset fault physical model template to calculate the degree of matching. When the matching degree calculation result exceeds the preset collaborative confidence threshold, the confirmed fault identifier is generated.

3. The method according to claim 2, characterized in that, The fault physical model template is established through the following preliminary steps: The first set of multimodal data under preset fault conditions and the second set of multimodal data under non-fault conditions are acquired respectively. The multimodal data includes at least electrical characteristics, acoustic data, vibration data and arc light data. For the first and second groups of multimodal data, perform analysis and identification operations to extract their respective sets of non-electrical feature parameters; Statistical analysis is performed on the non-electrical characteristic parameter sets of the first group of multimodal data and the second group of multimodal data to calculate and determine the classification thresholds that can maximize the distinction between the two datasets for the instantaneous energy peak, the spectral energy concentration, the ultraviolet spectral peak intensity, the diffusion characteristics, and the light intensity jump rate, respectively. Multiple classification thresholds are integrated to construct the fault physical model template.

4. A method for handling ring main unit faults, characterized in that, Applied to a second ring cage device, the method includes: The second ring network box receives and parses the isolation command message, controls the internal second switch to perform a tripping action, and cuts off the connection between the second ring network box and the upstream faulty line; After the second ring network box performs the tripping action, it generates a help message containing historical load data of the downstream faulty line before the tripping. The second ring network box broadcasts the distress message to at least one third ring network box in the tie switch system; The second ring network box receives and parses at least one of the response messages to obtain the remaining capacity data therein as the remaining capacity data to be tested; The second ring network box compares the remaining capacity data to be tested with the power failure load data in the distress message according to preset conditions, and filters the target remaining capacity data; The second ring network box identifies the third ring network box corresponding to the target remaining capacity data as the target third ring network box and sends a takeover confirmation message to the target third ring network box.

5. The method according to claim 4, characterized in that, The step of comparing the remaining capacity data to be measured with the power outage load data in the distress message according to preset conditions and filtering the target remaining capacity data specifically includes: Multiply the power failure load data in the distress message by a preset safety factor to calculate the load demand threshold, wherein the preset safety factor is greater than 1; Each received remaining capacity data to be tested is compared with the load demand threshold; Store all remaining capacity data that exceeds the load demand threshold into a candidate capacity data set; The data with the largest value is selected from the candidate capacity data set and determined as the target remaining capacity data.

6. The method according to claim 4 or 5, characterized in that, The distress message is used to enable the third ring network box to obtain its own rated capacity data and current real-time load data, calculate and generate remaining capacity data, encapsulate it into a response message, and send it to the second ring network box.

7. A method for handling ring main unit faults, characterized in that, Applied to the target third ring cage, the method includes: After receiving the takeover confirmation message, the target third ring network box detects the voltage value of the communication line between the target third ring network box and the second ring network box; When the voltage value of the connecting line is detected to be zero, the target third ring network box controls its connecting switch to perform a closing action, and restores power supply to the second ring network box and downstream through the connecting line.

8. A ring main unit fault handling device, characterized in that, The ring network box fault handling device includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to cause the ring network box fault handling device to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that, When the instruction is executed on the ring network box fault handling device, the ring network box fault handling device performs the method as described in any one of claims 1-7.

10. A computer program product, characterized in that, When the computer program product is run on the ring network box fault handling device, the ring network box fault handling device performs the method as described in any one of claims 1-7.

Images