Edge computing-based intelligent socket electric arc fault detection method, device and equipment and storage medium

Abstract

The application provides an edge computing-based intelligent socket electric arc fault detection method and device, equipment and a storage medium, relates to the electric arc detection technical field, and obtains detection data of the intelligent socket through a sampling channel. After a zero-crossing alignment window is determined and screened, an effective determination window is generated, and an operation signal of each effective determination window is constructed. A load segment marker is determined, a stable conduction position set is determined, local port reference data is established in the stable conduction position set, contact-related disturbance markers are calculated, and are organized into a contact constraint operation summary. A continuous aggregation additional coefficient is selected according to a temperature background marker, and a contact form quantity is generated in combination with the number of occurrences and the longest continuous segment length. A normal allowable boundary is constructed based on the contact form quantity. The normal allowable boundary is used to determine a fault arc risk level, a local disposal action of each effective determination window is generated by reading an action configuration table, and fault arc detection is completed.

Description

Technical Field

[0001] This invention relates to the field of arc detection technology, and more particularly to a method, apparatus, equipment and storage medium for detecting arc faults in smart sockets based on edge computing. Background Technology

[0002] With the widespread adoption of smart home devices, commercial safe electricity use, and remote energy management equipment, smart sockets have evolved from simple on / off controllers into edge power terminals with capabilities such as power metering, overload protection, over-temperature protection, timer control, and local safety warnings. In practical use, smart sockets are the connection nodes between plugs, sockets, conductive springs, relay contacts, and external loads. Their bodies not only perform power supply control functions but can also become locations where fault arcs occur. Especially after a plug is not fully inserted, socket springs experience long-term fatigue, plug metal contacts oxidize, relay contacts wear, or after prolonged high-power load operation or frequent plugging and unplugging, the internal conductive structure of the socket or the contact interface between the plug and socket is prone to phenomena such as decreased contact pressure, increased contact resistance, localized temperature rise, and intermittent unstable conductivity, leading to early contact-related fault arcs. Existing fault arc detection technologies typically collect current, voltage, or high-frequency components and then use threshold judgment, spectrum analysis, waveform similarity calculation, or classification models to identify arc anomalies. These methods have certain applicability in circuit breakers, distribution boxes, or dedicated arc protection devices, but their direct application to smart sockets still presents significant engineering challenges. The types of loads connected to smart sockets change frequently. Devices such as hair dryers, vacuum cleaners, small motors, switching power adapters, rice cookers, and LED drivers generate startup surges, speed fluctuations, switching spikes, periodic heating variations, or non-stationary current waveforms during normal operation. These normal load changes can easily be confused with early fault arcing in terms of port voltage and load current. Meanwhile, before posing a significant hazard, contact-related fault arcing in smart sockets often manifests as a slight dip in port voltage during stable conduction, minor fluctuations in load current, and abnormal internal temperature background. Judging solely based on a single instantaneous waveform or a uniform threshold makes it difficult to distinguish between normal changes caused by load switching and abnormal disturbances caused by deterioration of the socket's contact interface. Furthermore, smart sockets are low-cost, long-term online devices requiring rapid response. Therefore, a lightweight judgment process utilizing existing port voltage, load current, and internal temperature sampling is more suitable for practical products, rather than relying on complex external testing equipment or large-scale waveform analysis. Therefore, the main shortcoming of the existing technology is that it usually treats smart sockets as ordinary electrical signal acquisition entry points, lacks the constraint expression of the socket contact interface as a fault source, and also lacks a mechanism to use load segments, temperature background and contact disturbance location together to generate normal boundaries, which leads to problems such as false alarms, missed alarms or inaccurate handling timing in complex load scenarios. Summary of the Invention

[0003] The purpose of this invention is to provide a method, apparatus, device, and storage medium for detecting arc faults in smart sockets based on edge computing, so as to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, this application proposes a method for detecting arc faults in smart sockets based on edge computing, comprising:

[0005] The detection data of the smart socket is obtained through the sampling channel. The local edge processor determines the zero-crossing alignment window based on the detection data. After filtering the zero-crossing alignment window, a valid judgment window is generated. The detection data of the valid judgment window is organized into a detection sequence. The running signal of each valid judgment window is constructed based on the detection sequence.

[0006] Based on the continuous change pattern within the window of the detection sequence, the load segment marker is determined by looking up a table, the set of stable conduction positions is determined based on the load segment marker, local port reference data is established in the set of stable conduction positions, and a temperature background marker is constructed based on the load segment marker and the detection sequence. The contact-related disturbance marker is calculated by combining the local port reference data, the load segment marker, and the temperature background marker, and organized into a contact constraint operation summary.

[0007] The number of times the contact-related disturbance markers appear at the sampling location and the length of the longest continuous segment in the contact constraint operation summary are counted. A continuous aggregation additional coefficient is selected based on the temperature background marker, and a contact morphology quantity is generated by combining the number of occurrences and the length of the longest continuous segment. A normal allowable boundary is constructed based on the contact morphology quantity.

[0008] The processor determines the fault arc risk level based on the normal allowable boundary, reads the action configuration table based on the fault arc risk level, load segment marker, and temperature background marker, and generates local handling actions for each valid judgment window based on the action configuration table to complete the fault arc detection.

[0009] In some embodiments, after organizing the detection data of the effective decision window into a detection sequence, the method further includes:

[0010] Descramble isolated spikes in the sampling link:

[0011] If a sampling point experiences an isolated abrupt change relative to its adjacent sampling points in the same channel and no corresponding disturbance occurs in another electrical channel, then the interpolation between adjacent valid sampling points in the same sampling channel is used to replace it; otherwise, the disturbance is retained.

[0012] In some embodiments, the detection data includes port voltage sample values, load current sample values, and internal temperature sample values, and the detection sequence includes a port voltage sequence, a port current sequence, and an internal temperature sequence.

[0013] In some embodiments, the zero-crossing alignment window uses the zero-crossing position of the port voltage sample value as the window boundary.

[0014] In some embodiments, the normal allowable boundary includes contact morphology quantity, allowable contact morphology boundary, load segment marker, and temperature background marker. The step of determining the fault arc risk level based on the normal allowable boundary specifically includes:

[0015] The measured and protection thresholds are respectively set as contact morphology quantities and normal allowable boundaries, and a scene buffer amount obtained by the joint indexing of the load segment marker and the temperature background marker is added;

[0016] The fault arc risk level is determined based on the comparison results of the measured value, the protection threshold, and the scenario buffer.

[0017] In some embodiments, the action configuration table further includes action confirmation conditions, wherein the action confirmation conditions are in units of the number of valid decision windows;

[0018] When the power outage risk level is reached and the temperature background is marked as a contact heating background, or when multiple valid judgment windows reach the power outage risk level, the power supply cut-off action is directly output.

[0019] When a risk level of alert is reached, the alert action will only be executed after several consecutive valid judgment windows reach the alert level.

[0020] In some embodiments, the normal allowable boundary includes a contact shape quantity, a contact shape boundary, a load segment marker, and a temperature background marker, wherein the contact shape boundary is read from a preset boundary configuration matrix based on the load segment marker and the temperature background marker.

[0021] To achieve the above objectives, this application also proposes an edge computing-based smart socket arc fault detection device, comprising:

[0022] The signal generation module is used to acquire the detection data of the smart socket through the sampling channel. The local edge processor determines the zero-crossing alignment window based on the detection data, filters the zero-crossing alignment window, generates a valid judgment window, organizes the detection data of the valid judgment window into a detection sequence, and constructs the running signal of each valid judgment window based on the detection sequence.

[0023] The runtime summary construction module is used to determine the load segment marker by looking up a table based on the continuous change pattern within the window of the detection sequence, determine the set of stable conduction positions based on the load segment marker, establish local port reference data in the set of stable conduction positions, construct temperature background markers based on the load segment markers and the detection sequence, calculate contact-related disturbance markers by combining the local port reference data, load segment markers and temperature background markers, and organize them into a contact constraint runtime summary.

[0024] The allowable boundary generation module is used to count the number of times the contact-related disturbance markers appear at the sampling location and the length of the longest continuous segment in the contact constraint operation summary, select the continuous aggregation additional coefficient according to the temperature background marker, and generate the contact morphology quantity by combining the number of occurrences and the length of the longest continuous segment, and construct the normal allowable boundary based on the contact morphology quantity.

[0025] The arc detection module is used to determine the fault arc risk level based on the normal allowable boundary. The processor reads the action configuration table based on the fault arc risk level, load segment marker and temperature background marker, and generates local handling actions for each valid judgment window based on the action configuration table to complete the fault arc detection.

[0026] To achieve the above objectives, this application also proposes an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-described method.

[0027] To achieve the above objectives, this application also proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0028] Compared with the prior art, the beneficial effects of the present invention are:

[0029] This application first utilizes the port voltage sampling channel, load current sampling channel, and internal temperature sampling channel to form a zero-crossing alignment window in the local edge processor according to the zero-crossing position of the port voltage. The port voltage sequence, load current sequence, and internal temperature sequence within the same window are then organized into an operating signal, ensuring that subsequent judgments are all under the same AC cycle reference. Subsequently, the processor extracts load segment markers, temperature background markers, and contact-related disturbance markers from the operating signal. It then links the local port voltage dip, local load current disturbance, and internal temperature background in the stable conduction position to generate a contact constraint operating summary. This allows the contact state of the plug, socket, conductive spring, or internal switch contact to be included in the fault arc judgment process in a calculable summary form.

[0030] Building upon this, the present invention further converts the contact disturbance marker sequence into contact morphology quantities, and reads the corresponding permissible contact morphology boundaries from the boundary configuration matrix in the firmware based on the load segment markers and temperature background markers, forming the normal permissible boundaries for the current window. Thus, the same disturbance location distribution will correspond to stricter boundaries under stable conduction and contact heating backgrounds, while corresponding to boundaries more suitable for normal load variations under speed regulation segments or normal temperature backgrounds. Finally, the processor determines the fault arc risk level based on the relationship between the contact morphology quantities and the normal permissible boundaries, and, in conjunction with the load segment and temperature background, reads the action configuration table to output actions such as maintaining power supply, local notification, local alarm, or power cut-off.

[0031] This application transforms the physical cause of contact interface degradation in smart sockets into a contact constraint operation summary, and further transforms this summary into normal allowable boundaries that vary with load segments and temperature background. This allows fault arc detection to move beyond simple waveform thresholds or ordinary classification judgments, and instead complete the continuous processing of "contact disturbance identification, scene boundary generation, risk level determination, and local action execution" within a local edge processor. In this way, smart sockets can more effectively distinguish between normal disturbances under complex loads and pre-arc precursors of contact-related faults, making them particularly suitable for deployment in smart sockets, smart power strips, and safe power terminals to improve the timeliness and accuracy of local fault arc detection and handling. Attached Figure Description

[0032] Figure 1 Flowchart of the edge computing-based smart socket arc fault detection method provided in this application embodiment;

[0033] Figure 2 A diagram of an edge computing-based smart socket arc fault detection device provided in an embodiment of this application;

[0034] Figure 3 The zero-crossing window contact perturbation fingerprint provided for the embodiments of this application. Detailed Implementation

[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] refer to Figure 1 As shown, one embodiment of this application proposes a smart socket arc fault detection method based on edge computing, including:

[0037] S1: The local edge processor of the smart socket generates an operating signal and acquires the detection data of the smart socket through the sampling channel. Based on the detection data, the local edge processor determines a zero-crossing alignment window, filters the zero-crossing alignment window, generates a valid judgment window, organizes the detection data of the valid judgment window into a detection sequence, and constructs an operating signal for each valid judgment window based on the detection sequence, specifically including:

[0038] This step, as the first step of the entire method, is used to generate the first step from the smart socket itself. The running signal of a zero-crossing aligned window The operating signal consists of a port voltage sequence, a load current sequence, and an internal temperature sequence, and serves as the input for subsequently generating a contact constraint operating summary. The smart socket internally incorporates port voltage sampling channels, load current sampling channels, and internal temperature sampling channels. The port voltage sampling channel can be composed of an output voltage divider circuit, a protection circuit, and an analog-to-digital converter interface of the local edge processor, used to acquire port voltage sample values. The load current sampling channel can be composed of a sampling resistor, a current transformer, or a Hall current sensor, along with corresponding signal conditioning circuitry, used to acquire load current sample values. The internal temperature sampling channel can be composed of a thermistor, digital temperature sensor, or integrated temperature detection chip located near the conductive spring of the socket, near the relay contacts, or near the power devices, used to acquire internal temperature sample values. The port voltage sample values ​​are used to form window boundaries and preserve the port voltage fluctuation pattern; the load current sample values ​​are used to preserve the load conduction change pattern; and the internal temperature sample values ​​are used to preserve the background of heating changes in the socket's contact parts or conductive structures. The three types of sampled values ​​can be analog-to-digital converted sample counts, or they can be converted into physical quantities such as volts, amperes, and degrees Celsius after front-end calibration. In the same implementation, the sampled values, local reference, and voltage threshold in the port voltage channel use the same scale; the sampled values, local reference, and current threshold in the load current channel use the same scale; and the temperature readings and temperature background threshold in the temperature channel use the same temperature scale. The three types of sampled values ​​are organized in the local edge processor according to the same window number, so that subsequent steps can process voltage, current, and temperature changes under the same AC cycle reference, and avoid the ambiguity of threshold comparison caused by mixing units from different channels.

[0039] The local edge processor continuously reads port voltage samples and uses the zero-crossing position of the port voltage as the window boundary. The zero-crossing position is confirmed by the sign change of adjacent sampling points and the continuous change trend near the zero-crossing. For example, if the port voltage sampling sequence enters the zero-crossing range from the negative side between adjacent sampling points, and the processor confirms that this is a valid zero-crossing position when subsequent sampling points leave this range and enter the positive side. The basis for dividing the window based on the zero-crossing position comes from the periodic characteristics of the AC signal, which is a conventional mathematical processing of phase alignment for periodic signals. This application uses this phase alignment method for data processing before the fault arc detection of the smart socket, so that subsequent comparisons of port voltage micro-drops, load current micro-disturbances, and internal temperature changes can be made at the same AC phase position. The zero-crossing range, the minimum effective zero-crossing interval, and the number of window sampling points are also considered. The parameters are given by the smart socket firmware parameter table and are matched with the power grid frequency, processor sampling frequency, and analog-to-digital conversion cache length. If the adjacent zero-crossing interval significantly deviates from the allowable range of the power grid cycle, or if abnormalities such as open circuits, saturation, or continuous missing sampling occur in the port voltage sampling, the processor will not treat this window as a valid fault arc determination window, but will mark it as an invalid sampling window and wait for the next valid zero-crossing window. In determining the... After zero-crossing alignment windows, the processor organizes the port voltage sample values ​​within the window into... to Organize the load current sampling values ​​into to If the actual number of sampling points within the window exceeds the preset limit... The processor extracts corresponding sampling points according to their positions within the window; if the actual number of sampling points within the window is less than... However, if the minimum number of valid samples configured in the firmware is still reached, the processor interpolates between adjacent valid sample points on the same channel; if the number of valid samples is less than the minimum, the window will not output high-risk results. The update frequency of internal temperature sample values ​​is usually lower than that of port voltage and load current. The processor maps temperature readings near the window to the same window; when there is a temperature reading within the window, the reading is expanded to each sampling position within the window; when there are two temperature readings before and after the window, they are expanded within the window according to their reading time. This process ensures that the internal temperature channel has a consistent window position number with the port voltage channel and load current channel, facilitating the subsequent formation of a contact constraint operation summary from the same operating signal.

[0040] Temperature expansion employs one-dimensional linear interpolation from numerical analysis. The original linear interpolation is used to estimate the value at a point within an interval based on two known points. This application uses it to map low-speed temperature readings to the corresponding sampling positions within a zero-crossing aligned window, specifically:

[0041] ;

[0042] in, Indicates the first The first window The internal temperature value corresponding to each sampling location is obtained from the internal temperature sampling channel and mapped through a window. and These represent two temperature values ​​located before and after the current sampling position, which are actually read by the internal temperature sampling channel. and They represent and The reading time is recorded by the sampling timer of the local edge processor; Indicates the first The corresponding time of each sampling position within the current window is determined by the window start point and the sampling position. This formula performs interval estimation within the internal temperature channel; the right side is an expansion of the ratio of known temperature values ​​to the temperature differences of similar values, and the calculation result still represents the internal temperature value. , and All use the same temperature unit. , and All use the same unit of time, and the time difference ratio is a dimensionless proportion. This interpolation is only used when... and lie in and Used between; if If backward temperature readings are missing or the time interval between forward and backward temperature readings exceeds the firmware-allowed temperature hold time, the processor will not use the fractional calculation. Instead, it will use the most recent valid temperature reading for expansion, or mark the temperature background of the current window as an uncertain temperature data state and use a conservative configuration for subsequent boundary reads. equal The result is ;like equal The result is In one implementation, a zero-crossing alignment window needs to be organized. The sampling position, the first sampling position after the start of the window. Temperature values ​​were read from each timing position. The temperature value was read at the end of the window. , No. If the sampling location is located in the middle of two temperature readings, then the interpolation value is obtained. When there is only one temperature reading near the window. hour, to All readings are taken from this value. This expansion method allows the temperature channel to participate in subsequent summary generation according to its window position, while maintaining the meaning of temperature as a slowly changing channel.

[0043] For isolated spikes in the sampling link, the processor processes them according to the continuity of adjacent sampling points in the same channel. If a sampling point experiences an isolated abrupt change relative to its preceding and following adjacent sampling points in the same channel, and this abrupt change does not continue to occur at adjacent positions, nor does it have a corresponding disturbance in another electrical channel, then it is replaced by interpolation between adjacent valid sampling points in the same channel. If the disturbance occurs consecutively at multiple sampling positions, or if the port voltage and load current change simultaneously at similar positions, then the disturbance is retained for subsequent processing. For example, if a sampling point in the load current sequence suddenly jumps, while the preceding and following sampling points remain stable, and the port voltage sequence does not change accordingly, it can be treated as a sampling link spike. If the load current exhibits fragmented spikes at multiple consecutive sampling positions, and the port voltage experiences a slight dip at similar positions, then this change is retained in the running signal. After completing the above alignment and processing, the local edge processor generates the first... The running signals for each window:

[0044] ;

[0045] in, Indicates the first The running signal for a zero-crossing aligned window is generated by the local edge processor; Indicates the first The first window The port voltage sampled values ​​at each sampling location are obtained from the port voltage sampling channel; Indicates the first The first window The load current sampled values ​​at each sampling location are from the load current sampling channel; Indicates the first The first window The internal temperature values ​​at each sampling location are derived from the internal temperature sampling channel and mapped according to the window position; This indicates the number of sampling locations retained in each window, which is pre-configured by the local edge processor; This indicates the sampling position number within the window. The initial source of this formula is matrix permutation representation in linear algebra. This application arranges the sampling sequences within the three same windows into a fixed input format by channel. In this application, it is used as a data structure with channel identifiers. Cross-channel addition and subtraction operations are not required for the voltage, current, and temperature rows; each channel is subsequently compared separately according to its own unit or sampling count scale. For example, the port voltage channels are organized within this window. to The load current path is sorted out. to The internal temperature channel is obtained through the aforementioned mapping. to The processor arranges the three sequences according to the above formula. The output of this step is the [number]. The running signal of a zero-crossing aligned window The next step is to Generate a runtime summary of contact constraints for the input.

[0046] S2: Based on the continuous change pattern within the window of the detection sequence, determine the load segment marker by looking up a table, determine the set of stable conduction positions based on the load segment marker, establish local port reference data in the set of stable conduction positions, and construct a temperature background marker based on the load segment marker and the detection sequence. Combine the local port reference data, load segment marker, and temperature background marker to calculate the contact-related disturbance marker, and organize it into a contact constraint operation summary, specifically including:

[0047] This step follows the sequence formed by S1. The running signal of a zero-crossing aligned window The contact constraint operation summary is generated by the local edge processor of the smart socket. In S1 The port voltage sequence has been set. to Load current sequence to and internal temperature sequence to Placed within the same AC cycle window, this step directly extracts the constraint information required for contact-type fault arcs from this window. The actual formation process of contact-type fault arcs in smart sockets typically exhibits distinct scenario characteristics: when the contact pressure between the plug and socket decreases, the conductive spring fatigues, the contact surface oxidizes, or the internal switch contacts wear, the power supply circuit will experience localized conduction instability under relatively stable load conditions. This manifests as a short-term dip in port voltage, minor disturbances in load current at similar locations, and may be accompanied by a heating background in the socket's internal contact areas; while load startup, shutdown, speed adjustment, and gear switching generate significant and continuous electrical changes. This step will then... Convert to This allows the next step of generating normal allowable boundaries to separate the handling of "normal changes caused by load switching" and "contact disturbances under stable power supply conditions".

[0048] The local edge processor first reads Load current sequence in to The load segment marker is determined based on the continuous change pattern of the current level within the window. The smart socket firmware includes a pre-set segment judgment configuration table, which at least includes a continuous rising length threshold, a continuous falling length threshold, a wide-range periodic change threshold, and a stable fluctuation range threshold. These thresholds are tuned during product debugging through prototype bench testing and written into the firmware parameter area, and are read along with the parameter table version number during runtime. During tuning, current sequences of typical loads are collected under stable conduction, startup, shutdown, upshift, downshift, speed regulation, or periodic operation conditions. The continuous unidirectional change length, window current span, and stable conduction fluctuation amplitude at adjacent sampling positions are statistically analyzed. The continuous rising length threshold and continuous falling length threshold are taken as the lower limit of the continuous change length in the normal startup or upshift, normal shutdown or downshift window, respectively, and are higher than the occasional continuous change length generated by stable conduction noise. The wide-range periodic change threshold is taken as the lower limit of the current span in the speed regulation or periodic operation window, and is higher than the current span in the stable conduction window. The stable fluctuation range threshold is taken as the upper limit of the deviation of the stable conduction window current from the local reference. Taking a zero-crossing alignment window with 8 to 32 sampling positions as an example, the continuous rising length threshold and the continuous falling length threshold can be taken from 3 to 8 sampling positions, the wide-amplitude periodic change threshold can be taken from 15% to 40% of the local current reference value, and the stable fluctuation range threshold can be taken from 2% to 8% of the local current reference value; when expressed in terms of the number of continuous windows, the threshold for the number of continuous rising or continuous falling windows can be taken from 2 to 5 effective windows; among which, the threshold involving the current amplitude is related to... Using the same current sampling scale, the threshold involving continuous length is represented by the number of sampling positions or the number of consecutive windows. The processor determines the threshold by looking up a table based on the direction, duration, and amplitude of current change at adjacent sampling positions. When the current level varies slightly around the same conduction state within the window, and the amplitude of the variation is within the threshold of the stable fluctuation range, The stable conduction state is taken; when the current level rises across multiple consecutive sampling positions and the continuous length reaches the continuous rise length threshold, The system determines whether the current level is in the starting or increasing range state; when the current level decreases across multiple consecutive sampling positions, and the consecutive length reaches the threshold for the consecutive decrease length, Select the off or downshift state; when the current level exhibits a periodic wide-range change corresponding to speed control, and the change amplitude reaches the wide-range periodic change threshold, The processor selects between speed adjustment and periodic adjustment states. If multiple segment conditions are met simultaneously within the same window, the processor executes according to the priority preset in the firmware, prioritizing the identification of load switching segments such as start-up, shutdown, increasing / decreasing speed, or speed adjustment. Only when the switching segment conditions are not met will a stable conduction state be output, to avoid misinterpreting normal load switching as a contact disturbance background. For example, when a hair dryer switches from low to high speed... It will rise at multiple consecutive positions, and the processor will adjust the rise length based on a threshold value. Marked as up mode; after the hair dryer is running stably, The processor will adjust the frequency based on a stable fluctuation range threshold, with the frequency fluctuating around a certain conduction level. It is marked as a stable conduction state. Subsequently, the processor... and to Determine the set of stable conduction positions This set includes sampling positions where the current within the window is near the same conduction level and does not fall into the start-up, turn-off, upshift, downshift, or speed regulation transition zone. When When the number of sampling positions in the window is lower than the minimum number of stable positions preset in the firmware, the processor treats the window as a load switching background and prevents the corresponding position from entering the stable conduction discrimination in subsequent contact-related disturbance marking.

[0049] processor in Internal local port voltage reference and local load current reference . Depend on to The Middle The adjacent stable segments near each sampling location are obtained and used to represent the local reference of the port voltage near that location; Depend on to The Middle The adjacent stable conducting segments near the sampling location are obtained and used to represent the local reference of the load current near that location. In actual implementation, the processor prioritizes the sampling of the adjacent stable conducting segments near the sampling location. Several sampling locations were selected before and after each sampling position. Adjacent sampling points within the range form a reference value; when the first... When a sampling position is close to the window boundary, the processor selects the nearest stable sampling point on the same side to form a reference value; when there are insufficient stable sampling points near the current position, the processor uses the nearest stable sampling point within the same window. The nearest valid stable segment at each location forms a reference value; when the number of stable sampling points in the entire window is less than the minimum number of stable positions, the processor assigns the contact-related disturbance markers of that window to the load segment state constraints, and does not generate contact disturbance markers under stable conduction positions within that window. The port voltage reference comes only from the port voltage channel, and the load current reference comes only from the load current channel, ensuring that the deviation judgment for each channel retains its original engineering meaning. For loads such as mobile phone chargers, hair dryers, and rice cookers, the stable operating phase... It can reflect the load conduction level within this window. It can reflect the local level of port voltage within the window, and the two together are used to determine the synchronization characteristics of whether a contact-type disturbance occurs at a certain sampling location.

[0050] The processor also reads Internal temperature sequence to Combined with load fragment tags Forming temperature background markers The smart socket firmware includes a pre-configured temperature background table, which at least includes the normal temperature range, contact heat-concealment interval, load heat-matching interval, and criteria for determining the direction of temperature change in adjacent windows. The direction of temperature change in adjacent windows is determined by several valid window temperature values ​​in the processor's local rolling cache; if the number of valid temperature windows in the cache is insufficient, the temperature sensor is offline, or the temperature reading exceeds the sensor's range, the processor will... This is recorded as an uncertain background for temperature data, and a conservative boundary, no lower than that of a normal temperature background, or a warning-level handling is adopted when subsequently reading the boundary configuration matrix. The processor determines the background based on the internal temperature sequence, load current sequence, and... Determine by looking up the table When the current window is in a stable conducting state, and the internal temperature sequence is within the contact heating concern range, or when the internal temperature in adjacent windows shows a continuous upward trend without a corresponding increase in load current, Take the background state of contact heating; when the current window is in the start-up, speed increase, speed adjustment, or high-power stable segment, and the internal temperature change and the overall load current level fall within the load heating matching range, Under the load heating background condition; when the internal temperature is within the normal temperature range. Take a normal temperature background state. For smart socket scenarios, with the same port voltage sag and load current disturbance, if accompanied by a background of contact heating, the correlation with plug loosening, spring pressure reduction, or contact deterioration is stronger; if it is under a normal temperature rise background of high-power load operation, the next step can generate corresponding boundaries according to the load operation segment. Temperature information is thus entered in the form of background constraints. .

[0051] Contact related disturbance markers The calculation originates from the local deviation threshold discrimination in signal detection. Classical threshold discrimination typically compares the current sampled value with a local reference value; this application builds upon this by extrapolating to intelligent socket contact-type fault arcs, extending the single-channel deviation to a joint condition of "stable conduction position, local voltage sag, local current disturbance, temperature background selection threshold, and load segment selection threshold." Specifically, the derivation is as follows: the port voltage sag is determined by the local port voltage reference... With current port voltage The difference is reflected in the load current disturbance, which is determined by the current load current. With local load current reference The differences are reflected in the fact that contact-type disturbances require both to simultaneously meet the corresponding conditions near the same sampling location, and that location must be within the set of stable conduction locations. In the middle. Therefore, the first... Contact-related disturbance markers at each sampling location:

[0052] ;

[0053] in, Indicates the first The first window The contact-related perturbation markers at each sampling location are calculated by the local edge processor; Indicates the sampling position number within the window, as shown in S1. The sampling location numbers are consistent; Represents the set of stable conduction positions, by to The continuously changing shape within the window is obtained; express The Middle Port voltage sample values ​​at each sampling location; Indicates the first The local port voltage reference at each sampling location is determined by... to The adjacent stable segments are obtained; Indicates temperature background marker The selected port voltage sag detection threshold is read from the threshold configuration table in the smart socket firmware; express The Middle Load current sampled values ​​at each sampling location; Indicates the first The local load current reference at each sampling location is determined by... to The adjacent stable conducting segments are obtained; Indicates that it is marked by the load fragment. The selected load current disturbance discrimination threshold is read from the threshold configuration table in the smart socket firmware. and , Use the same port voltage scale. and , The same load current scale is used. If a certain temperature background or load segment does not have a corresponding item in the threshold configuration table, the processor does not use a temporary threshold or an unfounded threshold, but instead uses a conservative default threshold preset in the firmware and records the missing item in the parameter table. In this formula, the port voltage is compared with the port voltage reference and the port voltage threshold, and the load current is compared with the load current reference and the load current threshold. The temperature background and load segment are selected through threshold selection. The calculation relationship corresponds to the engineering characteristic of "voltage sag and current disturbance synchronous phenomenon in stable power supply position" when the smart socket is unstable.

[0054] In one implementation, within a zero-crossing alignment window The processor according to to Judgment No. To the Each sampling location belongs to ,at the same time Take a stable conduction state. Take the background state of contact heating. Firmware configuration table according to Read the port voltage sag threshold ,according to Read the load current disturbance discrimination threshold If the first Local port voltage reference at each sampling location Port voltage sampling value Then it satisfies Meanwhile, the local load current reference Load current sampling value Then it satisfies And the first Each sampling location belongs to ,therefore Pick If the first A current change was observed at one sampling location, but it was within the initial transition phase and had not yet entered the transition phase. ,but Pick The values ​​in this example correspond to the sample count values ​​after analog-to-digital conversion. The threshold and the corresponding sampling channel use the same counting scale, which is written into the firmware configuration table during the product debugging phase.

[0055] Contact-related perturbation marker sequence to After generation, the processor retains its phase position distribution within the AC cycle window. Contact-type fault arcs in smart sockets often exhibit characteristics of being short, aperiodic, and locally occurring; the position distribution in the marked sequence can reflect which phase positions the disturbance occurs at within the window. If If the disturbances occur in clusters at several adjacent sampling locations, it indicates the presence of localized concentrated disturbances within that window; if The discrete appearance of multiple phase positions within the window indicates that the window exhibits non-periodic contact perturbation characteristics. The processor then associates this marker sequence with... , This information is retained so that the next step can process the location of contact disturbances, load fragment background, and temperature background when generating normal allowable boundaries.

[0056] After completing the above processing, the local edge processor organizes the contact-related perturbation marker sequence, load fragment marker, and temperature background marker into the first... Summary of contact constraint operation for each window:

[0057] ;

[0058] in, Indicates the first The contact constraint operation summary for each window is generated by the local edge processor based on the operation signal output by S1. generate; to Indicates the first The contact-related disturbance markers at each sampling position within a window are derived from a combined judgment of port voltage sag, load current disturbance, stable conduction position, and threshold selection. Indicates the first The load fragment markers for each window are determined by... to The overall shape changes within the window; Indicates the first The temperature background marker for each window is... to and to The operational context represented is jointly determined; This represents the number of sampling locations retained in each window of S1. This formula originates from linear algebra and ordered tuple representations in data structures. This application organizes the contact disturbance location, load operation segment, and temperature background into a single summary. The logical relationship between the two formulas is as follows: the first formula is used to generate the data point by point. Then the second formula will complete the process. to and , Organization This transforms the local sampling location determination into a window-level contact constraint runtime summary.

[0059] Taking a smart plug connected to a hair dryer as an example, when the hair dryer is at a stable setting, Marked as a stable conducting state; if the contact pressure between the socket spring and the plug metal plate decreases, the port voltage experiences brief dips at certain sampling locations, the load current exhibits minor disturbances at similar locations, and the internal temperature is under contact heating background, the processor will [process] at these sampling locations. Marked as and will The background is marked as contact heating. If the hair dryer is in the speed switching phase, the load current changes overall across multiple consecutive sampling locations. The status of this load segment will be recorded, and the next step can be based on it. Generate the normal allowable boundaries corresponding to this fragment. Therefore, The running signal of S1 Converted into a window-level summary for arc detection of contact-type faults in smart sockets, the next step is to... Generate normal allowed boundaries for the input.

[0060] S3: Statistically count the number of times contact-related disturbance markers appear at the sampling location and the length of the longest continuous segment in the contact constraint operation summary. Select a continuous aggregation additional coefficient based on the temperature background marker, and generate a contact morphology quantity by combining the number of occurrences and the length of the longest continuous segment. Construct a normal allowable boundary based on the contact morphology quantity, specifically including:

[0061] S3 inherits the output of S2. Summary of contact constraint operation for each window The normal allowable boundaries are generated by the local edge processor of the smart socket. S2 has organized the contact-related perturbation marker sequence, load fragment marker, and temperature background marker within the same zero-crossing alignment window into... Therefore, this step directly uses All information in: to Used to describe the location where contact disturbances occur within the AC cycle window. Used to describe whether the current window belongs to a load segment such as stable conduction, start-up, shutdown, speed regulation, or periodic adjustment. This describes whether the current window is under a normal temperature background, a load-heating background, or a contact-heating background. The loads connected to smart sockets vary greatly in actual operation, and the same disturbance marker sequence has different meanings under different load segments. Furthermore, contact-related fault arc precursors caused by loose plugs, fatigue of socket springs, or deterioration of internal contacts often manifest as short disturbances at local phase positions in stable conduction segments, and exhibit a stronger risk connotation under a contact-heating background. Based on this engineering characteristic, this step will... Convert to the normal allowed boundary corresponding to the current window This allows the next step to directly determine whether the current disturbance pattern exceeds the normal operating range of the window.

[0062] The processor first reads binary perturbation marker sequence to The algorithm extracts the quantitative and continuous clustering features of contact disturbances from the sequence. The algorithm is derived from binary state counting and run-length statistics in discrete signal analysis: binary state counting is used to count the number of times a state appears in a sequence, and run-length statistics are used to describe the longest consecutive segment of that state. This application extrapolates these two classical discrete statistics to the zero-crossing alignment window of the smart socket to characterize the scale and local clustering of contact-related disturbances within an AC cycle reference, and further introduces features marked by temperature background. The chosen continuous aggregation coefficient allows the contact heating background to exert a stronger constraint on continuous disturbances. This is because when the contact surface of a smart socket deteriorates, the disturbances are often not uniformly distributed across the entire window, but rather concentrated near certain conduction phases; when this concentrated disturbance is accompanied by the contact heating background, its distinguishability compared to ordinary load noise is higher. Therefore, the processor generates the first... Contact shape quantity of each window :

[0063] ;

[0064] in, Indicates the first The contact pattern of each window is determined by the local edge processor based on the output of S2. to Calculated; Indicates the first The first window The contact-related disturbance markers at each sampling location are derived from the output of S2. ; This indicates the number of sampling locations retained in each window, consistent with the meaning in S1 and S2; Indicates the sampling position number within the window; Represents a set consisting of consecutive sampling locations, used to represent Continuous take A fragment; This indicates the number of sampling locations contained in the continuous segment; Indicates temperature background marker The selected continuous aggregation additional coefficient is read from the configuration table in the smart socket firmware. Sourced from S2 output .like to There is no value in The sampling position, then the length of the longest continuous perturbation segment is calculated according to... deal with, Correspondingly To avoid missing calculation scope due to empty sets; This is a dimensionless, non-negative coefficient, provided by the firmware configuration table. By default, a smaller additional coefficient corresponding to a normal temperature background is used, or it is directly denoted as an anomaly in the parameter table. The first term of this formula is the number of occurrences of the binary perturbation marker, and the second term is the additional characterization of the longest continuous perturbation segment under the temperature background. Both parts are based on a counting scale formed by the same binary marker sequence. The calculation result still represents a count-type characterization of the contact perturbation morphology within the window, rather than physical quantities such as port voltage, current, or temperature. For example, in the window, if to for The number of times the disturbance occurs is The longest continuous perturbation segment is the first one. To the There are 1 sampling location, and the segment length is 1. ;like Indicating a background of heat generation upon contact, the firmware configuration table provides... ,but ;like Indicating a normal temperature background, the firmware configuration table provides... ,but Therefore, the same disturbance location distributed under different temperature backgrounds results in different contact morphology quantities, adapting to the engineering differences between contact-type fault arcs in smart sockets and ordinary load disturbances.

[0065] The processor then according to and Get the allowed contact shape boundary corresponding to the current window. This boundary comes from the boundary configuration matrix in the firmware. ,matrix The row corresponding to the load fragment tag Column corresponding to temperature background marker The matrix elements represent the boundary quantities of contact forms that are allowed in the corresponding scenario. During the product debugging phase, a smart socket prototype test bench can be configured to collect data under the following conditions: the test bench is connected to typical loads such as mobile phone chargers, switching power adapters, hair dryers, rice cookers, and small motors, and data is collected under stable conduction, start-up, shutdown, speed regulation, and periodic heating conditions. Simultaneously, adjustable contact pressure clamps, aged socket samples, or worn contact sample samples were used to create a contact degradation state, and data were collected under the background of contact heating. The processor adapts to different scenarios. The distribution is configured to cover normal load disturbances and allow contact degradation disturbances to form boundary values ​​for over-limit determination in the next step.

[0066] Specifically, for the same combination of load segment markers and temperature background markers, the prototype bench forms a normal window set and a contact anomaly window set, respectively, and calculates the contact morphology quantity of each window according to the aforementioned formula; the boundary element values ​​are based on covering the concentrated area of ​​normal windows and being lower than the concentrated area of ​​contact anomaly windows. Taking a zero-crossing aligned window with 8 to 32 sampling positions as an example, the background value for stable conduction and contact heating can be 3 to 7, the background value for stable conduction and normal temperature can be 5 to 10, and the background value for starting, stopping, or adjusting the speed segment and normal temperature can be 7 to 18; when the window retains 256 sampling positions, the background value for uncertain temperature data or contact heating can be 18 to 35. The above boundaries and contact morphology quantities use the same counting scale.

[0067] In one implementation, the matrix The firmware of the smart socket is written after the bench data is configured during the product debugging phase; during the operation phase, it is written by the local edge processor according to... and from Read the boundary from the matrix. The configuration process corresponds one-to-one with the load segments and temperature background states of the smart socket, ensuring that the boundary values ​​remain the defined configuration items in the firmware. If China lacks and is currently and For the corresponding elements, the processor does not generate sourceless boundaries, but instead reads the default boundary items specified in the firmware and records missing items in the parameter table; for uncertain temperature data backgrounds, the default boundary items can be configured to be no wider than the boundaries corresponding to normal temperature backgrounds, to avoid expanding the normal allowable range due to missing temperature backgrounds. The processor generates the first... Normal allowable boundaries for each window :

[0068] ;

[0069] in, Indicates the first The normal allowable boundaries of each window are determined by the local edge processor based on the output of S2. and the aforementioned generate; This represents the contact shape quantity of the current window, determined by... to and Calculated; This represents the allowed contact pattern boundary under the current load segment and temperature background, configured by the boundary configuration matrix. Read; Representation matrix In and load fragment tags and temperature background marker The corresponding element; Indicates the load segment marker of the S2 output; This represents the temperature background marker output by S2. The formula originates from ordered boundary bag representation and matrix lookup operations in data structures. This application combines the actual contact morphology quantity, scene allowable boundary, load segment, and temperature background into a single boundary bag for direct comparison in the next step. and The logical relationship between the two formulas is: first from... to and Calculate the actual contact shape quantity of the window Then by and from Reading allowed contact shape boundaries Finally , , and The organization is .

[0070] In one specific implementation, the boundary configuration matrix After configuration on the prototype bench, the following scene boundaries are included: the boundary corresponding to stable conduction and a normal temperature background is... The boundary corresponding to stable conduction and contact heating background is The boundary corresponding to the speed regulation segment and the normal temperature background is When the smart socket is connected to a hair dryer and is in a stable setting, the S2 output... To maintain a stable conduction state, if the internal temperature background... Indicates exposure to a heating background, and to for Then, from the aforementioned calculations, we can obtain... , and by Read Processor output This boundary packet indicates that the contact morphology of the current window is higher than the normal allowable boundary under stable conduction and contact heating conditions. If another window has the same... to ,but Indicates a speed adjustment segment. If the background represents a normal temperature, then according to get and by matrix Read Processor output This boundary packet indicates that the same disturbance location distribution falls within a wider acceptable range under speed regulation segments and normal temperature backgrounds. In this way, this step transforms the same contact disturbance marker sequence into different acceptable boundaries under different smart socket usage scenarios.

[0071] The output of this step is the [number]. Normal allowable boundaries for each window . Complete summary of contact constraint operation following S2 output ,in to Transformed into contact morphological quantity , and Used to read the allowed contact shape boundary and with Together they form a boundary packet. The next step is to directly use... As input, the contact pattern quantity of the current window is compared with the normal allowable boundary to determine the fault arc risk level and perform local handling of the smart socket.

[0072] S4: Determine the fault arc risk level based on the normal allowable boundary. The processor reads the action configuration table based on the fault arc risk level, load segment marker, and temperature background marker. Based on the action configuration table, it generates a local handling action for each valid judgment window to complete the fault arc detection, specifically including:

[0073] S4 receives the output from S3. Normal allowable boundaries for each window The fault arc risk level is determined by the local edge processor of the smart socket. and generate local processing actions. S3 output middle, This indicates the contact shape of the current window. This indicates the boundary of the contact morphology that is allowed to occur under the current load segment and temperature conditions. Indicates the load fragment marker, This indicates a temperature background marker. This step uses it directly. The final decision is made on all elements in the dataset: first compare... and The boundary relationship, combined with the and The jointly determined scenario buffer size yields the window-level fault arc risk level. Subsequently, based on the risk level, load segment, and temperature background, the local action configuration table is read to control the smart socket to perform actions such as maintaining power supply, providing a notification, issuing an alarm, or cutting off power supply. Thus, the contact disturbance location, load operating segment, and temperature background formed in the preceding steps are transformed into locally executable safety actions by the smart socket.

[0074] Fault Arc Risk Level The determination originates from the segmented threshold determination in relay protection and embedded safety control. Classical segmented threshold determination typically compares the measured quantity with the protection threshold and classifies the protection level according to the degree of exceedance. This application extends this approach to the arc fault scenario of a smart socket contact type, setting the measured quantity as S3 to obtain the contact form quantity. Set the protection threshold to the normal allowable boundary. And add the load fragment marker. Temperature background marker Scene buffer size obtained from common index The buffer size in this scenario is used to express the difference in risk tolerance under different operating states of the smart socket. For example, a smaller buffer size corresponds to stable conduction and a background of contact heating, while a larger buffer size corresponds to speed regulation and a background of normal temperature. The processor then obtains the following formula: Fault arc risk level for each window:

[0075] ;

[0076] in, Indicates the first The fault arc risk level of each window is output by the local edge processor of the smart socket; This represents the contact shape of the current window, derived from the S3 output. ; This represents the normal allowed boundaries of the current window, derived from the S3 output. ; This indicates the load fragment marker for the current window, derived from S3 output. ; The background marker indicating the current window's temperature, from the S3 output. ; Indicates by and The scene buffer size obtained from the common index is read from the risk configuration table in the smart socket firmware. In this formula, , and All three are at the contact morphology scale formed by S3, and the comparison relationship among them corresponds to the boundary quantity at the same scale; This is a non-negative boundary buffer amount. If an item is missing, a conservative buffer amount from the firmware risk configuration table will be used, or the risk level will be changed to a warning level. Values ​​without a source will not be used. For discrete level labeling, Corresponding to the normal allowable range, Corresponding to warning level risk, Corresponding to the risk level of power outage. The derivation logic of this formula is as follows: First, based on... As the normal boundary of the current scenario, and then... As a boundary of the power outage level, it separates ordinary out-of-bounds and high-risk out-of-bounds situations.

[0077] In one implementation, the smart socket is connected to a hair dryer and is in a stable conduction phase, with S3 outputting... ,in , , Indicates a stable conduction state. This indicates a background of heat generation. The firmware risk configuration table is read for this combination. processor computing ,because Greater than ,get This result corresponds to the risk of power failure under stable conduction and contact heating conditions. If a certain window outputs... And the combination corresponding to ,but The processor gets This result corresponds to a warning level of risk. The calculation example above illustrates that the same type of contact pattern quantity will be mapped to different risk levels under different load segments and temperature backgrounds, adapting to the actual differences when smart sockets are connected to complex loads.

[0078] get Then, the processor according to , and Read Action Configuration Table , generate the first Local processing actions for each window This action generation originates from state lookup table control in embedded control. Classic lookup table control reads the corresponding control action based on a discrete state index; this application expands the lookup table index to include fault arc risk level, load segment, and temperature background, enabling the smart socket to execute the corresponding action based on the current operating scenario even under the same risk level. Specifically:

[0079] ;

[0080] in, Indicates the first The local processing actions corresponding to each window are output by the local edge processor of the smart socket to the indicator light driver circuit, buzzer driver circuit, relay driver circuit or electronic switch driver circuit. This represents the action configuration table in the smart socket firmware, which is obtained from the product security policy configuration. The table entries include maintaining power supply, local notification, local alarm, and cutting off power supply. This indicates the fault arc risk level obtained from the aforementioned formula; Indicates the load segment marker of the S3 output; This represents the temperature background marker output by S3. The initial source of this formula is discrete state lookup table control in embedded systems. This application uses the fault arc risk level, load segment, and temperature background as a joint index, allowing the risk determination result to be directly mapped to the local action of the smart socket. The logical relationship between the two formulas is: first from... In , , and calculate Then by , and Read Action Configuration Table Output .

[0081] Action Configuration Table Configuration can be made in the smart plug firmware according to hardware capabilities and security policies. For smart plugs equipped with indicator lights and buzzers, It can be configured to flash the indicator light and emit a short beep; for smart sockets with relays, It can be configured to disconnect the relay and maintain the fault indication status; for smart sockets using electronic switches, It can be configured to disable the electronic switch drive signal and illuminate the fault indicator. To reduce false alarms caused by a single abnormal sampling window, the action configuration table can also include action confirmation conditions, such as... The alert level risk requirement stipulates that the alert should only be maintained after several consecutive valid windows reach the alert level, and a higher confirmation count should be set for windows in the speed regulation segment and with a normal temperature background; for and If the system encounters a heat-generating background or if multiple consecutive valid windows reach a power-off risk level, a power-off action will be directly output. The number of confirmations mentioned above is given by the firmware security policy table and is expressed in units of valid windows, without changing the aforementioned... The calculation results only affect The final execution timing. Taking the aforementioned example of a hair dryer being stably powered on and in contact with a heated background, the processor obtains... Then, read Output To cut off power and trigger a local alarm, in a relay-type smart socket, the edge processor outputs a disconnect control signal to the relay drive circuit, disconnecting the main power supply circuit while the fault indicator light remains illuminated. Taking the aforementioned example with a speed-adjustable segment and a normal temperature background, the processor obtains... Then, read Output To maintain power supply. Taking the warning-level risk example, the processor receives... Then, read Output This is a local prompt; users can check the plug insertion status or change the socket position.

[0082] During prototype debugging, an adjustable contact pressure clamp can be used to simulate the decrease in contact pressure between the plug and the socket. Stable conduction, speed regulation, and start-up segments can be formed using loads such as hair dryers, mobile phone chargers, and small motors, and the corresponding window values ​​can be recorded. , and For example, under stable conduction and contact heating conditions, when multiple consecutive windows output similar values... When the boundary packet is received, the processor gets And execute power cut-off; under the speed regulation segment and normal temperature background, when the window output At that time, the processor receives And maintain power supply. This commissioning method can verify the response differences of the risk configuration table and action configuration table in different load segments, and make the final action consistent with the target of arc detection for contact-type faults in smart sockets.

[0083] The output of this step is the [number]. Fault arc risk level of each window and local handling actions . From the output of S3 middle , , and Segmentation determination, From , and Action Configuration Table The processor can also check the current status, overload protection status, and manual reset status of the relay or electronic switch before outputting the power cut-off action; if the execution end feedback is not disconnected or communication is abnormal, it will maintain the fault prompt and record the execution failure mark to prevent subsequent windows from treating the unconfirmed power cut-off status as a completed action. Through this step, the smart socket converts the operating signals, contact constraint operating summaries, and normal allowable boundaries formed in the previous steps into local fault arc risk judgment and local handling results.

[0084] refer to Figure 2 As shown, another embodiment of this application proposes a smart socket arc fault detection device based on edge computing, comprising:

[0085] The signal generation module is used to acquire the detection data of the smart socket through the sampling channel. The local edge processor determines the zero-crossing alignment window based on the detection data, filters the zero-crossing alignment window, generates a valid judgment window, organizes the detection data of the valid judgment window into a detection sequence, and constructs the running signal of each valid judgment window based on the detection sequence.

[0086] The runtime summary construction module is used to determine the load segment marker by looking up a table based on the continuous change pattern within the window of the detection sequence, determine the set of stable conduction positions based on the load segment marker, establish local port reference data in the set of stable conduction positions, construct temperature background markers based on the load segment markers and the detection sequence, calculate contact-related disturbance markers by combining the local port reference data, load segment markers and temperature background markers, and organize them into a contact constraint runtime summary.

[0087] The allowable boundary generation module is used to count the number of times the contact-related disturbance markers appear at the sampling location and the length of the longest continuous segment in the contact constraint operation summary, select the continuous aggregation additional coefficient according to the temperature background marker, and generate the contact morphology quantity by combining the number of occurrences and the length of the longest continuous segment, and construct the normal allowable boundary based on the contact morphology quantity.

[0088] The arc detection module is used to determine the fault arc risk level based on the normal allowable boundary. The processor reads the action configuration table based on the fault arc risk level, load segment marker and temperature background marker, and generates local handling actions for each valid judgment window based on the action configuration table to complete the fault arc detection.

[0089] The following specific examples illustrate this application:

[0090] To demonstrate that S1 to S4 can form a recalculated detection chain in the local edge processor, this embodiment selects a set of traceable arc test waveform records for recalculation. This record contains port voltage and current sample values ​​from the same sampling sequence. The source record file is PTFE_P01_D01_S1.tab, corresponding to the test conditions of PTFE insulation material, 0.1 bar air pressure, 1 mm electrode spacing, 20.8°C ambient temperature, and 9.2% humidity. This waveform record is used to verify the processing chain of "port voltage sag, load current disturbance, contact-related disturbance marker, contact morphology quantity, normal boundary, and local action" in this application. It should be noted that this source waveform record only contains voltage and current channels and does not include readings from the internal temperature sensor of the smart socket; therefore, this embodiment does not reconstruct the internal temperature sequence but performs conservative boundary recalculation according to the aforementioned branch with uncertain temperature data background, ensuring consistency between the embodiment and the source data range.

[0091] Referring to Table 1, this embodiment selects the first waveform from the source waveform. To the Each sampling point is used as a zero-crossing alignment processing window, and the window retains... Each sampling location. Within this window, the processor establishes local references for both port voltage and load current, and generates a sampling rate using port voltage sag thresholds and load current perturbation thresholds consistent with the sampling scale. In this embodiment, the port voltage sag threshold... Configured as Voltage sampling count, load current disturbance threshold Configured as Each current sampling unit, temperature background is processed according to the uncertain background of temperature data, and a continuous aggregation additional coefficient is applied. Configured as After recalculation, there are a total of [number] windows in this window. Each sampling location satisfies the contact-related disturbance labeling condition, and the longest continuous disturbance segment contains Each sampling location, with contact morphology quantity as follows: Under this temperature condition, the normal allowable boundary given by the local boundary configuration matrix is: This value falls within a conservative boundary configuration for temperature data uncertainty under a window of 256 sampling locations, and is within the aforementioned recommended range of 18 to 35. The scenario buffer given in the risk configuration table is... ,therefore The processor obtains the fault arc risk level. The local processing action is output from the action configuration table. To cut off power and trigger a local alarm.

[0092] Table 1. Parameter table for zero-crossing alignment processing window

[0093]

[0094] refer to Figure 3 As shown, the port voltage sample value, load current sample value, local reference, contact-related disturbance marker sequence, and local edge decision chain within the same window are displayed on the same sampling position axis. In the upper figure, the red markers correspond to the positions where the port voltage dips relative to the local reference and the load current is disturbed relative to the local reference; the middle figure shows the load current disturbance at the same position; the lower disturbance fingerprint band shows the phase distribution of mkn within the window. The decision chain on the right further provides the number of disturbance markers in Ck, the contact morphology quantity and normal boundary in Bk, and finally Gk and Ak. As can be seen from this embodiment, this method does not directly judge a single current peak value, but first confirms whether the voltage dip and current disturbance occur synchronously within the same window, then converts the number of disturbances and the degree of continuous aggregation into contact morphology quantities, and finally combines the normal boundary and risk configuration table to obtain the local action. The recalculation process corresponds to the technical effect of arc detection for contact-type faults in smart sockets: when voltage / current disturbances exhibit multiple discrete and locally clustered characteristics within the window, even if the temperature channel is temporarily unavailable, the processor can still enter a conservative boundary judgment and output local safety measures, thereby improving the executability of local detection and handling.

[0095] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method.

[0096] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0097] The processor can be a microcontroller within the smart socket, an edge processing chip, an embedded processor with an analog-to-digital converter interface, or a safety control chip connected to the sampling front end. It is configured to read port voltage sampling values, load current sampling values, and internal temperature sampling values, and execute program instructions stored in the storage medium. The storage medium can be on-chip flash memory, external flash memory, read-only memory, or other non-volatile memory. When the stored program instructions are executed by the processor, the processor completes the above steps S1 to S4 and reads configuration items such as N, Δurk, Δisk, λrk, D, ρsk,rk, and P from the parameter table in the storage medium. Therefore, the variable sources, parameter tables, risk levels, and action outputs in the method steps can all be mapped to the device hardware and processor program execution process, avoiding a disconnect between the method scheme and the device, storage medium, and processor implementation.

[0098] It is understood that the content of the above method embodiments is applicable to this device embodiment. The specific functions implemented by this device embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0099] It is understood that the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0100] The above embodiments are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for detecting arc faults in smart sockets based on edge computing, characterized in that, Include: The detection data of the smart socket is obtained through the sampling channel. The local edge processor determines the zero-crossing alignment window based on the detection data. After filtering the zero-crossing alignment window, a valid judgment window is generated. The detection data of the valid judgment window is organized into a detection sequence. The running signal of each valid judgment window is constructed based on the detection sequence. After organizing the detection data of the effective determination window into a detection sequence, the method further includes: Descramble isolated spikes in the sampling link: If a sampling point exhibits an isolated abrupt change relative to its adjacent sampling points in the same channel and no corresponding disturbance appears in another electrical channel, then the interpolation between adjacent valid sampling points in the same sampling channel is used to replace it; otherwise, the disturbance is retained. Based on the continuous change pattern within the window of the detection sequence, the load segment marker is determined by looking up a table, the set of stable conduction positions is determined based on the load segment marker, local port reference data is established in the set of stable conduction positions, and a temperature background marker is constructed based on the load segment marker and the detection sequence. The contact-related disturbance marker is calculated by combining the local port reference data, the load segment marker, and the temperature background marker, and organized into a contact constraint operation summary. The number of times the contact-related disturbance markers appear at the sampling location and the length of the longest continuous segment in the contact constraint operation summary are counted. A continuous aggregation additional coefficient is selected based on the temperature background marker, and a contact morphology quantity is generated by combining the number of occurrences and the length of the longest continuous segment. A normal allowable boundary is constructed based on the contact morphology quantity. The processor determines the fault arc risk level based on the normal allowable boundary, reads the action configuration table based on the fault arc risk level, load segment marker and temperature background marker, and generates local handling actions for each valid judgment window based on the action configuration table to complete the fault arc detection. The normal allowable boundary includes contact morphology quantity, allowable contact morphology boundary, load segment marker, and temperature background marker. The allowable contact morphology boundary is read from a preset boundary configuration matrix based on the load segment marker and the temperature background marker. The determination of the fault arc risk level based on the normal allowable boundary specifically includes: The measured and protection thresholds are respectively set as contact morphology quantities and normal allowable boundaries, and a scene buffer amount obtained by the common index of the load segment marker and the temperature background marker is added; The fault arc risk level is determined based on the comparison results of the measured value, the protection threshold, and the scenario buffer.

2. The edge computing-based smart socket arc fault detection method according to claim 1, characterized in that, The detection data includes port voltage sampling values, load current sampling values, and internal temperature sampling values, and the detection sequence includes port voltage sequence, port current sequence, and internal temperature sequence.

3. The edge computing-based smart socket arc fault detection method according to claim 2, characterized in that, The zero-crossing alignment window uses the zero-crossing position of the port voltage sample value as the window boundary.

4. The edge computing-based smart socket arc fault detection method according to claim 1, characterized in that, The action configuration table also includes action confirmation conditions, which are in units of the number of valid judgment windows; When the power outage risk level is reached and the temperature background is marked as a contact heating background, or when multiple valid judgment windows reach the power outage risk level, the power supply cut-off action is directly output. When a risk level of alert is reached, the alert action will only be executed after several consecutive valid judgment windows reach the alert level.

5. An edge computing-based smart socket arc fault detection device, characterized in that, Include: The signal generation module is used to acquire the detection data of the smart socket through the sampling channel. The local edge processor determines the zero-crossing alignment window based on the detection data, filters the zero-crossing alignment window, generates a valid judgment window, organizes the detection data of the valid judgment window into a detection sequence, and constructs the running signal of each valid judgment window based on the detection sequence. After organizing the detection data of the effective determination window into a detection sequence, the method further includes: Descramble isolated spikes in the sampling link: If a sampling point exhibits an isolated abrupt change relative to its adjacent sampling points in the same channel and no corresponding disturbance appears in another electrical channel, then the interpolation between adjacent valid sampling points in the same sampling channel is used to replace it; otherwise, the disturbance is retained. The runtime summary construction module is used to determine the load segment marker by looking up a table based on the continuous change pattern within the window of the detection sequence, determine the set of stable conduction positions based on the load segment marker, establish local port reference data in the set of stable conduction positions, construct temperature background markers based on the load segment markers and the detection sequence, calculate contact-related disturbance markers by combining the local port reference data, load segment markers and temperature background markers, and organize them into a contact constraint runtime summary. The allowable boundary generation module is used to count the number of times the contact-related disturbance markers appear at the sampling location and the length of the longest continuous segment in the contact constraint operation summary, select the continuous aggregation additional coefficient according to the temperature background marker, and generate the contact morphology quantity by combining the number of occurrences and the length of the longest continuous segment, and construct the normal allowable boundary based on the contact morphology quantity. The arc detection module is used to determine the fault arc risk level according to the normal allowable boundary. The processor reads the action configuration table according to the fault arc risk level, load segment mark and temperature background mark, and generates local handling action for each valid judgment window based on the action configuration table to complete the fault arc detection. The normal allowable boundary includes contact morphology quantity, allowable contact morphology boundary, load segment marker, and temperature background marker. The allowable contact morphology boundary is read from a preset boundary configuration matrix based on the load segment marker and the temperature background marker. The determination of the fault arc risk level based on the normal allowable boundary specifically includes: The measured and protection thresholds are respectively set as contact morphology quantities and normal allowable boundaries, and a scene buffer amount obtained by the common index of the load segment marker and the temperature background marker is added; The fault arc risk level is determined based on the comparison results of the measured value, the protection threshold, and the scenario buffer.

6. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method according to any one of claims 1 to 4.

7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 4.

Images