Switch cabinet arc light detection method and system based on multispectral fusion

Abstract

The invention discloses a switch cabinet electric arc light detection method and system based on multispectral fusion, and relates to the technical field of electric power system intelligent detection, and the method comprises the following steps: S1, building a cross-spectrum unified time reference, and carrying out the synchronous comparison of an infrared light signal, a visible light signal and an ultraviolet light signal, extracting a temperature transition trigger feature and a strong light disturbance spectrum feature, and generating a time synchronization deviation vector; and S2, constructing an anti-fact spectrum playback model based on the time synchronization deviation vector, generating a shadow spectrum sequence, positioning an infrared response threshold lifting reason through energy difference analysis, calculating a spectrum overlapping risk interval and duration, and outputting a threshold disturbance distribution diagram. According to the method, infrared anomalies are identified through cross-spectrum time alignment, anti-fact playback and differential analysis, channel recovery is guided in combination with springback prediction and time sequence correction, and a spectrum broken chain window is constructed through dislocation sampling and phase regulation and control, so that the accuracy and the anti-interference capability of arc identification are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent detection technology for power systems, specifically to a method and system for detecting arc light in switchgear based on multispectral fusion. Background Technology

[0002] Arc light detection in switchgear based on multispectral fusion is an intelligent detection method that uses multispectral information, including visible light, infrared light, and ultraviolet light, to accurately identify and provide real-time early warning of arc discharge phenomena inside switchgear. This method involves deploying an intelligent sensing system within the switchgear. This system consists of a multispectral sensing unit, a signal synchronization unit, and a fusion computing unit, used to synchronously acquire radiation characteristic signals across multiple wavelengths. The system utilizes a time synchronization algorithm and a spectral feature fusion model to jointly analyze the transient intense light, thermal radiation, and electromagnetic characteristics generated by the arc light, thereby accurately distinguishing normal operating light signals from abnormal arc discharge events in complex electromagnetic environments and under strong noise backgrounds. Through the multispectral information fusion processing of the intelligent sensing system, the detection's anti-interference capability and response speed can be significantly improved, enabling early identification and intelligent early warning of arc light faults, providing reliable protection for the safe operation of switchgear and power systems.

[0003] The existing technology has the following shortcomings: In existing technologies, arc flash detection methods for switchgear based on multispectral fusion typically rely on the response of the infrared channel to temperature radiation characteristics to identify arc discharge events. However, under transient conditions of strong light interference, the response threshold of the infrared channel can be instantaneously raised by the system's adaptive algorithm due to rapid changes in ambient temperature, causing the detection device to lose sensitivity to subtle radiation differences for a short period. When high-intensity external light sources (such as sunlight reflection, welding arc light, or metal flash) cause drastic fluctuations in background thermal radiation, the energy distribution of the infrared band will overlap with the arc flash signal spectrum, causing the actual arc discharge characteristics to be misjudged as background thermal noise in the fusion model. At this time, the fusion model exhibits a backward learning convergence phenomenon in the high-frequency signal processing stage, that is, the algorithm identifies abnormally high energy mutations as negligible background changes, causing the detection results to shift to the low-sensitivity region, forming a dynamic detection blind zone. This problem prevents the system from triggering effective early warnings at the critical moment when the arc flash actually occurs, seriously affecting the real-time identification accuracy of arc faults and the operational safety of the switchgear.

[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for detecting arc light in switchgear based on multispectral fusion, so as to solve the problems in the background art mentioned above.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for detecting arc light in switchgear based on multispectral fusion, comprising the following steps: S1. Establish a unified time reference across the spectrum, synchronously compare infrared, visible and ultraviolet light signals, extract temperature transition triggering features and strong light disturbance spectral features, and generate a time synchronization deviation vector. S2, construct a counterfactual spectral playback model based on the time synchronization deviation vector, generate a shadow spectral sequence, locate the cause of the infrared response threshold rise through energy difference analysis, calculate the spectral overlap risk range and duration, and output a threshold perturbation distribution map; S3, construct a rebound prediction model based on the threshold perturbation distribution map, input the historical response curve and the current change slope, estimate the infrared threshold regression trajectory, and generate time series correction instructions; S4, according to the time series correction command, performs time-frequency synchronization rearrangement on the multispectral channels, reconstructs the sampling time grid, corrects the response deviation, outputs the fusion weight matrix, and calculates the dynamic rebound state; S5, based on the fusion weight matrix and dynamic rebound state, initiates micro-period misalignment gating and phase conjugate image stabilization control, constructs a spectral topology chain break window, blocks the overlapping paths of spectral energy, and achieves stable identification of arc light signals.

[0007] Preferably, step S1 includes: The infrared light receiving unit, visible light imaging unit and ultraviolet light detection unit are respectively deployed in different positions inside the switch cabinet. The viewing angle is calibrated according to the preset geometric calibration protocol, and microsecond-level synchronous acquisition is achieved through high-precision temperature-controlled crystal oscillator and distributed synchronization mechanism. Time series data frameworks for three types of spectral signals are constructed and mapped to the main time axis. Based on the radiative response formed by temperature transition events in the infrared channel, key response features of the three types of spectral signals are extracted and time correspondence is compared. The time offset vector is calculated based on the response time difference, intensity change and duration of the spectral signal to describe the dynamic response deviation of each spectral channel under typical physical disturbances. A dynamic synchronization deviation vector is generated based on the time offset vector, and it is written back into the spectral acquisition structure for time-domain registration and fusion processing sequence optimization. It is continuously updated in subsequent detection to achieve dynamic correction of response delay.

[0008] Preferably, step S2 includes: Historical response data of infrared, visible and ultraviolet light signals are retrieved based on the time synchronization deviation vector, and time-domain unified correction is performed to construct a shadow spectral sequence covering the abnormal window. The shadow spectral sequence is compared frame by frame with the currently acquired multispectral response sequence on the synchronous time axis to identify the energy surge region of the infrared channel and extract its boundary, intensity and duration. By combining the variation trajectories of visible light and ultraviolet light signals, the starting position, ending position, and duration of the multispectral overlap risk zone are calculated, and the period of infrared channel response rise is calibrated. By integrating the time positioning information and energy change curves of each channel, a threshold perturbation distribution map on a unified time axis is constructed, and the boundaries of the interference intensity level and the response offset paths of the spectral channels are clearly defined.

[0009] Preferably, the shadow spectral sequence is composed of historical sampling frames. The ambient temperature, light intensity and electromagnetic interference status corresponding to the selected sampling frames are consistent with the current observation conditions. During the construction process, the complete anomaly trigger window is covered and all spectral channel data are retained.

[0010] Preferably, step S3 includes: Complete response data of infrared light signals under temperature disturbance and strong light interference in multiple detection cycles are retrieved, and the intensity change characteristics of the infrared channel before, during and after the interference are analyzed to form a representative response rebound trajectory model. By combining the infrared anomaly segments identified in the threshold perturbation distribution map, the response rise rate, duration and offset amplitude in the current period are extracted, and a similarity comparison is performed with the historical trajectory model to calculate the trajectory shape and recovery time of the infrared response during the regression process. Infrared threshold rebound prediction curves are constructed on a unified time axis, time reconfiguration intervals are determined based on rebound trend change points, and time correction offsets are allocated to multispectral channels to correct the temporal structure. Based on the infrared rebound prediction curve and the time offset allocation results, a time series correction instruction is generated to clarify the sampling delay compensation amount, response window offset amplitude, and fusion timing priority of each spectral signal in the recovery stage.

[0011] Preferably, step S4 includes: Based on the sampling offset parameters in the time series correction instruction, the infrared light signal, visible light signal and ultraviolet light signal are time-mapped and adjusted, and time is rearranged frame by frame to complete the unified sampling reference alignment. Based on the rearranged sampling time grid structure, a global scan analysis is performed to identify fusion errors and perform interpolation correction and frequency fine-tuning on the error segments to form a multi-channel spectral response matrix in a unified format. Based on the energy distribution characteristics of each channel in the spectral response matrix, the signal contribution ratio of each sampling window is calculated, a fusion weight matrix covering the entire time axis is generated, and the weight inflection point of the infrared channel rebound stage is marked. By combining the weight change trend of the infrared channel with the rebound prediction trajectory, the return state of the infrared channel is determined, and a multi-channel collaborative weight adjustment framework is constructed based on the fusion weight matrix to achieve dynamic equilibrium in the fusion recognition process.

[0012] Preferably, the weight change trend of the infrared light signal in the fusion weight matrix is ​​used to determine its response rebound state. When the weight changes from a continuous decrease to a stable increase and remains stable, it is determined that the infrared light signal has completed the rebound and triggers the collaborative weight redistribution operation of the multispectral channels.

[0013] Preferably, step S5 includes: Based on the rebound state of the infrared channel in the fusion weight matrix, a time-series scan is performed to identify weight mutation points and perform micro-periodic sampling beat misalignment operation, so that the spectral signals are staggered in the high-risk window to break up the energy peak synchronization. By combining the time sampling grid and the trend of fusion weight changes, phase conjugate modulation is performed on the spectral signal in the strong interference region to filter out peak overlap points in reverse and limit the propagation path of energy between channels. A spectral topology chain break window is constructed by integrating the energy transfer characteristics between spectral channels, defining the overlap risk zone and setting an energy upper limit. When the channel signal exceeds the upper limit, the chain break window automatically closes the transmission path. By combining the activation behavior of the chain break window with the fusion weight state, the stability of signal recognition is evaluated. The subsequent sampling strategy and energy management process are optimized through a feedback mechanism to ensure the stable recognition capability of the arc light signal.

[0014] Preferably, in the micro-periodic sampling beat misalignment operation, the offset of the sampling trigger time adopts a periodic incremental strategy, and a unified time calibration is performed after each round of offset, so that the infrared light signal, visible light signal and ultraviolet light signal remain relatively interleaved in multiple sampling cycles and ensure that the overall synchronization structure remains unchanged.

[0015] A switchgear arc light detection system based on multispectral fusion includes a time-scale synchronization module, an interference identification module, a rebound prediction module, a response correction module, and a steady-state control module. The time synchronization module establishes a unified time reference across the spectrum, performs synchronous comparison of infrared, visible and ultraviolet light signals, extracts temperature transition triggering features and strong light disturbance spectral features, and generates a time synchronization deviation vector. The interference identification module constructs a counterfactual spectral playback model based on the time synchronization deviation vector, generates a shadow spectral sequence, locates the cause of the infrared response threshold rise through energy difference analysis, calculates the spectral overlap risk range and duration, and outputs a threshold perturbation distribution map. The rebound prediction module constructs a rebound prediction model based on the threshold perturbation distribution map, inputs the historical response curve and the current change slope, estimates the infrared threshold regression trajectory, and generates time series correction instructions. The response correction module performs time-frequency synchronization rearrangement of the multispectral channels according to the time-series correction instructions, reconstructs the sampling time grid, corrects the response deviation, outputs the fusion weight matrix, and calculates the dynamic rebound state. The steady-state control module, based on the fusion weight matrix and dynamic rebound state, initiates micro-period misalignment gating and phase conjugate image stabilization control, constructs a spectral topology chain break window, blocks the overlapping paths of spectral energy, and achieves stable identification of arc light signals.

[0016] The technical effects and advantages provided by the present invention in the above technical solution are as follows: This invention establishes a unified cross-spectral time reference to ensure strict alignment of infrared, visible, and ultraviolet signals in the time dimension. Combined with counterfactual spectral playback and energy difference analysis, it accurately identifies the causes of infrared threshold rise and spectral overlap risk segments, preventing background noise from masking the arc signal. Furthermore, it utilizes a threshold response rebound prediction mechanism and time series correction commands to controllably guide the dynamic performance recovery process of the infrared channel. Finally, through temporal misalignment sampling and phase conjugate modulation, it constructs a spectral topology chain-breaking window to dynamically block energy overlap propagation paths between multispectral signals, effectively improving the recognition purity and robustness of the multi-channel fusion model under complex interference backgrounds. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0018] Figure 1 This is a flowchart of the method for detecting arc light in switchgear based on multispectral fusion according to the present invention.

[0019] Figure 2 This is a schematic diagram of the module of the switchgear arc light detection system based on multispectral fusion of the present invention. Detailed Implementation

[0020] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0021] This invention provides, for example Figure 1 The method for detecting arc light in switchgear based on multispectral fusion, as shown, includes the following steps: S1. Establish a unified time reference across the spectrum. Through a synchronization mechanism, compare the time of infrared, visible and ultraviolet light signals, extract the triggering features caused by ambient temperature transitions and the spectral features of external strong light source disturbances, and generate a time synchronization deviation vector for dynamic reference. To achieve high-precision identification of electric arc light inside switchgear, the multispectral fusion detection method first constructs a unified cross-spectral time reference standard, aligning infrared, visible, and ultraviolet light signals on the same time axis. Then, features related to the electric arc event are extracted based on synchronous comparison, thereby establishing a precise time synchronization deviation vector. The specific implementation details of this process are as follows: Infrared light receiving units, visible light imaging units, and ultraviolet light detection units are deployed in different locations within the switch cabinet to ensure complementary spectral coverage of the same observation area. Before synchronous acquisition, the receiving angle and installation angle of each spectral channel are calibrated using spatial positioning according to a preset geometric calibration protocol, ensuring that spatial imaging results in different bands correspond to the same target area in physical space. Based on this, clock initialization is initiated for each channel. A reference time signal is generated by introducing a high-precision temperature-controlled crystal oscillator, establishing independent timing control signals for each spectral channel. A distributed synchronization mechanism is then used to synchronously trigger the acquisition of the three types of spectral data at a microsecond-level time granularity. The acquisition period is uniformly set to less than one millisecond to ensure sufficient time resolution to capture the transient characteristics of arc discharge.

[0022] After synchronous acquisition, a time-series data framework was established for each type of spectral signal and aligned to a unified main time axis. To improve the accuracy of the comparison, a physical feature labeling mechanism was introduced, selecting temperature abrupt changes or switching processes in the natural environment as triggering events to mark the enhanced radiation response caused by temperature transitions in the infrared band. In the corresponding visible and ultraviolet light data, image brightness changes and ultraviolet signal intensity surge segments were extracted that occurred simultaneously with the infrared thermal shock changes. By comparing the start time, peak time, and recovery time of these physical changes, the temporal correspondence of key features in the infrared, visible, and ultraviolet spectral channels was constructed. Based on this relationship, the subtle temporal shifts of different spectral signals were further statistically analyzed and measured to establish the relative position difference of each frame of data on the time axis.

[0023] After completing the time alignment and key point marking of the multispectral signals, the spectral perturbations caused by environmental changes were analyzed. In the infrared band, the radiation response peaks induced by rapid temperature transitions were monitored, and the rise slope, peak duration, and fall process response characteristics were identified. In the visible light band, regions of brightness abrupt changes that may be caused by ambient light reflection were analyzed, and their abruptness and high contrast characteristics were determined. In the ultraviolet band, high-intensity narrow pulse signals induced by non-natural electric arcs were extracted, and the intensity of adjacent frames was compared. Based on this, according to the signal mode changes of the three types of spectra at corresponding times, the response deviations between the spectral signals during temperature transitions or external strong light source interference events were calculated, including the difference in response time, the difference in intensity abrupt change amplitude, and the difference in signal duration. These multidimensional deviation data were uniformly mapped into a time offset vector to express the response differences of different spectral signals and their dynamic characteristics over time under the action of typical physical events.

[0024] Using the time offset vector as an input reference, a dynamic synchronization deviation vector is further generated. This vector includes not only the offset at the starting point of time but also the trend of synchronization accuracy changes and the duration of deviation between channels throughout the entire acquisition cycle. Combined with the previously constructed time axis alignment results, this vector is written back into the multispectral acquisition structure and used as a dynamic adjustment basis in subsequent detection processes to guide the optimization of the temporal registration operation and fusion processing sequence of each spectral signal. By continuously updating this synchronization deviation vector, dynamic correction of response delays to sudden events can be achieved, and a quantifiable reference baseline can be provided for interference source identification, abnormal signal localization, and temporal optimization of the fusion model. In this way, even if complex heat source fluctuations or optical interference occur during the operation of power equipment, this synchronization deviation vector can still ensure the temporal uniformity of multispectral data, laying a high-precision temporal foundation for subsequent multidimensional fusion processing.

[0025] S2, construct a counterfactual spectral playback model based on the time synchronization deviation vector, generate the corresponding shadow spectral sequence, perform energy difference analysis on the shadow spectral sequence and the current spectral signal to locate the cause of the rise in the infrared band response threshold, calculate the start and end positions and duration of the overlap risk interval between the infrared light signal and other spectral signals, and output the threshold perturbation distribution map; Based on the establishment of a unified time base and the generation of a multispectral time synchronization deviation vector, the expected signal response is reconstructed to identify the root cause of abnormal threshold changes in the infrared band. A control spectral sequence is generated through counterfactual spectral playback, and energy differential analysis is performed in conjunction with current signal data to track infrared response drift and accurately define risk zones. The specific steps are as follows: Based on the generated time synchronization deviation vector, the original response data of infrared, visible, and ultraviolet light signals are retrieved within several sampling periods before and after the key triggering event. The time domain of these three types of spectral signals is then uniformly corrected according to the time alignment results to ensure consistency of the observation data at each moment in the time dimension. On this basis, a counterfactual spectral playback structure is constructed. Historical sampling frames without infrared anomalous rise within the reference time window are used as the basic template. Normal spectral response sequences under temperature background, illumination intensity, and electromagnetic interference conditions closest to the current observation conditions are selected to form the standard response sequence under the theoretically expected conditions, i.e., the shadow spectral sequence. During the construction of this sequence, it is necessary to ensure that it covers the entire anomalous triggering window defined by the synchronization deviation vector, while retaining the complete frame structure of all infrared, visible, and ultraviolet spectral data to construct a reference spectral trajectory comparable to the current observation sequence.

[0026] The currently acquired multispectral response sequence and the constructed shadow spectral sequence are compared frame-by-frame on a fully synchronized time axis. The changes in energy response intensity in the infrared band are analyzed to identify abnormal energy surge regions in the infrared channel. The temporal boundaries, maximum intensity, rise rate, and duration of these abnormal regions are then precisely extracted. After identifying time periods with significant energy changes, the response states in the visible and ultraviolet bands are combined to analyze whether there are synergistic changes caused by external strong light interference or heat source reflection. By cross-validating the energy fluctuation characteristics of the three types of spectral signals within the same time period, occasional fluctuations caused solely by natural heating, arc signals, or random noise can be effectively eliminated. This allows for the accurate localization of response anomalies with a high probability of infrared threshold drift as the affected area of ​​infrared band response threshold elevation.

[0027] Based on the identified anomalous regions, the energy change trend of the infrared channel in the current spectral signal is further analyzed, and this trend is used as the main basis for the infrared threshold rise. By continuously tracking the response evolution of this infrared channel, the time points of the anomaly onset, peak appearance, and signal recovery are extracted. Based on this, the start time, end time, and total duration of the threshold rise phenomenon are determined. Combining the change trajectories of the corresponding time periods in the visible and ultraviolet signals, it is determined whether there is a multispectral overlap phenomenon, and whether the infrared response masks or suppresses the arc characteristics in other spectral channels during this overlap phase. Further statistical analysis is conducted on the degree of overlap of the energy values ​​of the three types of spectral signals in the time dimension, the difference in signal abrupt change intensity, and the relative delay distribution within the overlap region. This allows for the calculation of the potential spectral overlap risk range boundary, clearly defining the start, end, and duration of the spectral overlap risk interval.

[0028] Based on the dual results of abnormal infrared response rise and spectral overlap risk, a threshold perturbation distribution map is constructed by integrating the temporal positioning information and energy intensity variation curves of each channel. This distribution map uses a unified time axis as the horizontal axis, marking the energy rise region of the infrared light signal with a color gradient, and overlaying the corresponding energy overlap curves of the visible and ultraviolet light signals. This visually presents the interference paths, risk regions, and response drift processes between multispectral channels. The map clearly defines the boundaries of interference levels at different intensity levels, further refining them into an analytical basis for subsequent response correction, threshold adjustment, and channel rearrangement. This distribution map not only quantifies the potential identification risks caused by infrared threshold fluctuations within the current detection cycle but also provides clear input for subsequent rebound response prediction and dynamic correction mechanisms, thereby enabling comprehensive tracking and compensation control of hidden misjudgment regions in arc light detection.

[0029] S3, based on the threshold perturbation distribution map, constructs a threshold response rebound prediction model, takes the historical spectral response curve and the current change slope as input, estimates the regression trajectory of the infrared band response threshold, and generates time series correction instructions for channel correction; After obtaining the threshold perturbation distribution map, a rebound prediction mechanism for the threshold response is established to restore the detection sensitivity of the infrared channel. This mechanism, by modeling the historical spectral response behavior and combining it with the current trend of the perturbation slope, determines whether the response threshold of the infrared channel can return to the normal detection level after the perturbation ends, and generates time-series correction instructions for controlling the adjustment of the multispectral sampling channels. The specific steps are as follows: Complete response data sequences of infrared light signals that have experienced similar temperature disturbances and strong light interference events across multiple historical detection periods are retrieved, and a comprehensive analysis of the signal intensity changes in the infrared channels within these response sequences is performed. Each set of historical data must include a baseline segment before interference, a disturbance segment during the interference, and a recovery segment after the interference subsides, ensuring that the dynamic change process of the threshold rebound can be extracted from a complete time dimension. During the analysis, the initial value, peak value, rate of change, and time length required for the infrared channel response intensity to return to the baseline level in each segment are clearly defined, and these parameters are used as typical response feature templates for unified summarization. In this way, multiple representative response rebound trajectory models are formed, establishing an empirical foundation for subsequent prediction calculations.

[0030] By combining the infrared response anomalies identified in the current detection period using the threshold perturbation distribution map, key parameters such as the response intensity increase process, rate of change, and duration of these anomalies are extracted and compared with previously compiled historical templates. By comparing indicators such as the response start slope, maximum response offset, and response duration, it is determined whether the perturbation type experienced by the current infrared channel belongs to a known recoverable type. If so, the similarity position of the current perturbation response in the corresponding historical template is further calculated, and its subsequent possible rebound process is estimated, including the regression path shape, recovery time point, and final response level. This process should also incorporate the changing trends of visible and ultraviolet light signals in the current period to verify whether other spectral channels have completed the absorption or response to strong light interference, thereby confirming the independent validity of the current infrared rebound prediction.

[0031] Based on the comparison and estimation results, a rebound trajectory prediction curve for the infrared response threshold is constructed on a unified time axis. This prediction curve needs to cover the entire expected recovery interval after the end of the current disturbance segment, and mark the time scales of the regression critical point, the stable point, and the transition segment. After the curve is constructed, a phased sampling adjustment plan is formulated based on the rebound trend change points, delineating the data segment range that needs time reconfiguration, and assigning corresponding time correction offsets to each spectral channel within this range. These offsets can guide the data of each channel to move closer to the expected temporal structure during the subsequent spectral signal reconstruction process, avoiding the overall fusion error caused by the drift of a single infrared channel. In addition, it is also necessary to simultaneously evaluate the dynamic change trend of the fusion weight during this stage, and clarify whether the proportion of the infrared channel in the fusion structure needs to be reduced in stages to ensure the dominant identification of other spectral channels during the infrared rebound instability period.

[0032] Based on the infrared response rebound prediction curve and the multispectral time offset allocation results, a complete time-series correction instruction is output. This instruction must cover the time-domain adjustment schemes for all three types of spectral signals in the current recovery stage, including the sampling delay compensation amount, response window offset amplitude, and fusion timing priority for each type of spectral signal. The instruction should clearly indicate the effective time range of the correction operation and the target position of each sampling frame after adjustment, ensuring that in subsequent data fusion and feature extraction processes, each spectral signal can be restored to a time-aligned state and accurately mapped to the predetermined position structure in the multidimensional recognition model. The generation of the time-series correction instruction not only provides accurate time-domain support for the fusion recognition process but also achieves adaptive management of the dynamic coordination mechanism between multispectral channels at the structural level, providing a solid foundation for the continuous and stable recognition of arc light signals.

[0033] S4, according to the time series correction command, performs time frequency locking and synchronization rearrangement operation on the multispectral sampling channel, reconstructs the sampling time grid, corrects the response error caused by spectral delay and bias, outputs the fusion weight matrix, and calculates the dynamic rebound state of the infrared response threshold in combination with the weight change trend. After generating the time-series correction command for channel correction, this command is applied to the multispectral sampling channels to correct for temporal misalignment caused by spectral delay and response bias, and further reconstruct the unified sampling time structure, thereby improving the overall accuracy and stability of spectral fusion recognition. The specific steps are as follows: Based on the sampling offset parameters defined in the time series correction instruction, the original sampling data of infrared, visible, and ultraviolet spectral signals are time-mapped and adjusted. During this adjustment process, the original timestamp of each spectral data frame is compared with the target correction time position frame by frame to calculate the required adjustment time interval, and the signal sampling points are rearranged forward or backward accordingly. To ensure data integrity during the time rearrangement process, each sampling point must retain its original spectral intensity and spatial location information before and after adjustment, avoiding breakpoints or drift during data reconstruction. Special attention is paid to high-response segments in the infrared channel, prioritizing high-precision correction of their corresponding time periods to reduce the impact of delay propagation during abnormal state recovery. Through this round of precise time-frequency locking and rearrangement operations, the three types of spectral signals are realigned onto a unified sampling reference, forming a standardized time sequence structure.

[0034] After time rearrangement, a global scan analysis of the relative temporal relationships of the three types of spectral signals is performed based on the new sampling time grid structure to identify fusion errors caused by data offset. Specifically, by jointly calculating the distribution density, energy overlap, and abrupt change distribution of the reconstructed spectral signals along the time axis, it is determined whether there are still issues such as response redundancy, omissions, or phase drift between sampling points. For detected time error segments, local interpolation and beat fine-tuning methods are used to slightly correct the sampling frequency, ensuring that the time sampling interval of all spectral data tends to be consistent within the error tolerance. Subsequently, the data after sampling time correction is re-aggregated into a unified format multi-channel spectral response matrix, which is then divided into several fixed sampling windows to provide a structured data foundation for subsequent fusion processing.

[0035] Based on the reconstructed temporal sampling grid, the signal contribution ratio of each spectral channel in the current fusion state is calculated according to the energy distribution characteristics of each spectral channel within each window. Specifically, the expressive power of each channel for the overall spectral characteristics is quantified by statistically analyzing the average response intensity, slope of change, and signal fluctuation amplitude of each channel within the same time window. These values ​​are input as weighting factors into the fusion decision logic, ultimately generating a fusion weight matrix covering the entire time axis. This weight matrix uses matrix rows to represent sampling time segments and matrix columns to represent the contribution weight of each spectral channel within that segment, serving as an important basis for evaluating the reliability of fusion recognition. During matrix construction, special attention is paid to the weight change pattern of the infrared channel during the perturbation rebound phase, marking its inflection point from suppression to recovery, and recording this trend in the fusion weight tracking results to provide data support for subsequent threshold state transition judgment.

[0036] By combining the infrared channel weight change information in the fusion weight matrix, the rebound state after the disturbance ends and the channel enters a stable phase is further calculated. This calculation process comprehensively considers the infrared rebound prediction trajectory generated in the previous stage and the actual trend of the current weight curve, analyzes the consistency between the two, and determines whether the current infrared channel has completed its regression from a high-suppression state to a normal response state. After identifying that the infrared channel has entered a stable state, the deviation between its current response level and the historical baseline state is compared, and a new response threshold correction target is set accordingly, providing a decision reference for the next stage of sampling strategy optimization and signal discrimination mechanism update. Simultaneously, a multi-channel collaborative weight adjustment framework is constructed based on the fusion weight matrix, enabling other spectral channels to automatically adjust their fusion participation according to the recovery rhythm of the infrared channel, thereby achieving a dynamic balance between fusion recognition efficiency and robustness overall. Through the execution of the above entire process, a complete closed loop from time correction to fusion optimization can be achieved, significantly improving the resolution and real-time adaptability of multispectral detection for arc light signals in complex environments.

[0037] S5, based on the fusion weight matrix and dynamic rebound state, triggers the micro-period misalignment gating mechanism and starts the phase conjugate image stabilization control process, establishes a spectral topology chain break window for energy path isolation, dynamically blocks the energy overlap propagation path between different spectral signals, and achieves stable identification of arc light signals. After completing the fusion weight matrix output and obtaining the dynamic rebound state of the infrared channel, the intervention control stage begins. This stage actively guides energy path isolation to prevent interference superposition of spectral signals at critical identification moments. By triggering specific gating mechanisms and phase modulation processes, this stage constructs a spatiotemporal chain break structure for energy flow between spectral channels, thereby restoring the clear identifiability of the arc light signal. The specific steps are as follows: Based on the rebound state change rhythm of the infrared channel as indicated in the fusion weight matrix, a fine-grained temporal scan is performed on the entire multispectral sampling process to identify the overlap trend between weight abrupt changes and adjacent channels, thus locking in high-risk time windows prone to energy aliasing. On this basis, a sampling beat misalignment scheme with micro-intervals is implemented to slightly adjust the sampling trigger times of infrared, visible, and ultraviolet signals within this window. The adjustment method employs a periodic offset strategy, enabling staggered acquisition of the three spectral signals on an extremely short time scale, avoiding response overlap issues in the same instantaneous sampling. This micro-periodic misalignment of the sampling beat effectively disrupts the peak synchronicity of the spectral signals in time, forming a refined temporal misalignment structure, thereby suppressing response distortion caused by the superposition of energy components from different bands during the fusion process. To maintain the consistency of the three spectral signals in the overall time domain, a calibration return period is set in the misalignment operation to ensure that a unified sampling synchronization benchmark is maintained even after long-term operation.

[0038] Based on the micro-periodic misaligned sampling structure, and combined with the temporal sampling grid and fusion weight variation trend constructed in the previous stage, phase conjugate modulation of the spectral signal is performed on the identified regions with strong energy interference. The core of this modulation process lies in the inverse reconstruction of the phase characteristics of the spectral channel signals, enabling the interference energy to cancel each other out in the spectral structure. In practice, phase inversion mapping is performed on the rearranged infrared, visible, and ultraviolet signals respectively, and their temporal evolution direction and spatial response mode are simulated and inversely deduced to identify peak overlap points caused by energy overlap. These points are then filtered out in reverse and synchronously suppressed, limiting the interference energy of different channels to its propagation range within the original channel at the fusion point, preventing it from entering the main identification path of other channels. This process enables the spectral channels to maintain independent response capabilities when encountering strong external disturbances, unaffected by cross-coupling, providing a clear and uncontaminated spectral input foundation for subsequent signal identification.

[0039] After completing timing misalignment and phase modulation, and comprehensively considering the distribution characteristics of energy interaction paths between spectral signals, a spectral topology chain-breaking window is constructed in the multispectral signal processing architecture. This chain-breaking window is a virtual blank transmission segment created by timing cutting and phase shielding, based on the energy conduction paths that spectral channels may form in the multidimensional response space. During construction, firstly, the connection regions with overlapping risks between channels are identified, and the trend direction of energy diffusion from the main channel to secondary channels is analyzed. Subsequently, by setting isolation window boundaries on the time axis and assigning a limited energy upper limit value to the spectral sampling segments within the boundaries, when a channel signal exceeds this upper limit, the window will automatically close its transmission path to other channels. This chain-breaking window has dynamic adjustment capabilities, automatically extending or shrinking its duration and coverage area according to changes in fusion weights, thus flexibly responding to spectral interference events under different environmental disturbances. Through the construction of this structure, energy decoupling on the physical paths between spectral channels is achieved, effectively blocking the conduction interference of non-target signals during the identification period.

[0040] After the energy path isolation mechanism is deployed, the recognition stability under spectral fusion is evaluated in real time by combining the activation frequency and closure time of the chain-breaking window during actual operation. When the spectral chain-breaking window successfully blocks overlapping paths and the fusion weight matrix shows that the infrared channel rebound state has become stable, it indicates that the current recognition channel has high signal purification and response clarity. At this time, by analyzing the real-time capture results of the arc light signal features frame by frame, its independence and continuity in the multi-channel fusion spectrum are confirmed, further improving the confidence level of the recognition decision. At the same time, the triggering behavior of the chain-breaking window is correlated with the arc light detection results to form a closed-loop feedback relationship between interference control and recognition performance, which is used to guide the iterative update of subsequent sampling strategies and the dynamic optimization of energy management mechanisms. Through the implementation of this complete process, not only is the precise isolation of spectral energy channels under strong interference environment achieved, but also the stable, continuous and high-confidence recognition capability of arc light signals in complex backgrounds is finally realized.

[0041] This invention establishes a unified cross-spectral time reference to ensure strict alignment of infrared, visible, and ultraviolet signals in the time dimension. Combined with counterfactual spectral playback and energy difference analysis, it accurately identifies the causes of infrared threshold rise and spectral overlap risk segments, preventing background noise from masking the arc signal. Furthermore, it utilizes a threshold response rebound prediction mechanism and time series correction commands to controllably guide the dynamic performance recovery process of the infrared channel. Finally, through temporal misalignment sampling and phase conjugate modulation, it constructs a spectral topology chain-breaking window to dynamically block energy overlap propagation paths between multispectral signals, effectively improving the recognition purity and robustness of the multi-channel fusion model under complex interference backgrounds.

[0042] This invention provides, for example Figure 2 The switchgear arc light detection system based on multispectral fusion shown includes a time-scale synchronization module, an interference identification module, a rebound prediction module, a response correction module, and a steady-state control module. The time synchronization module establishes a unified time reference across the spectrum, performs synchronous comparison of infrared, visible and ultraviolet light signals, extracts temperature transition triggering features and strong light disturbance spectral features, and generates a time synchronization deviation vector. The interference identification module constructs a counterfactual spectral playback model based on the time synchronization deviation vector, generates a shadow spectral sequence, locates the cause of the infrared response threshold rise through energy difference analysis, calculates the spectral overlap risk range and duration, and outputs a threshold perturbation distribution map. The rebound prediction module constructs a rebound prediction model based on the threshold perturbation distribution map, inputs the historical response curve and the current change slope, estimates the infrared threshold regression trajectory, and generates time series correction instructions. The response correction module performs time-frequency synchronization rearrangement of the multispectral channels according to the time-series correction instructions, reconstructs the sampling time grid, corrects the response deviation, outputs the fusion weight matrix, and calculates the dynamic rebound state. The steady-state control module, based on the fusion weight matrix and dynamic rebound state, initiates micro-period misalignment gating and phase conjugate image stabilization control, constructs a spectral topology chain break window, blocks the overlapping paths of spectral energy, and achieves stable identification of arc light signals.

[0043] The switchgear arc light detection method based on multispectral fusion provided in this invention is implemented through the aforementioned switchgear arc light detection system based on multispectral fusion. For details of the specific methods and processes of the switchgear arc light detection system based on multispectral fusion, please refer to the embodiments of the switchgear arc light detection method based on multispectral fusion described above, which will not be repeated here.

[0044] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for detecting arc light in switchgear based on multispectral fusion, characterized in that, Includes the following steps: S1. Establish a cross-spectral unified time reference, synchronously compare infrared, visible and ultraviolet light signals, extract temperature transition triggering features and strong light disturbance spectral features, and generate a time synchronization deviation vector. S2, construct a counterfactual spectral playback model based on the time synchronization deviation vector, generate a shadow spectral sequence, locate the cause of the infrared response threshold rise through energy difference analysis, calculate the spectral overlap risk range and duration, and output a threshold perturbation distribution map; S3, construct a rebound prediction model based on the threshold perturbation distribution map, input the historical response curve and the current change slope, estimate the infrared threshold regression trajectory, and generate time series correction instructions; S4, according to the time series correction command, performs time-frequency synchronization rearrangement on the multispectral channels, reconstructs the sampling time grid, corrects the response deviation, outputs the fusion weight matrix, and calculates the dynamic rebound state; S5, based on the fusion weight matrix and dynamic rebound state, initiates micro-period misalignment gating and phase conjugate image stabilization control, constructs a spectral topology chain break window, and blocks the spectral energy overlap path.

2. The method for detecting arc light in switchgear based on multispectral fusion according to claim 1, characterized in that, Step S1 includes: The infrared light receiving unit, visible light imaging unit and ultraviolet light detection unit are respectively deployed in different positions inside the switch cabinet. The viewing angle is calibrated according to the preset geometric calibration protocol, and microsecond-level synchronous acquisition is achieved through high-precision temperature-controlled crystal oscillator and distributed synchronization mechanism. Time series data frameworks for three types of spectral signals are constructed and mapped to the main time axis. Based on the radiative response formed by temperature transition events in the infrared channel, key response features of the three types of spectral signals are extracted and time correspondence is compared. The time offset vector is calculated based on the response time difference, intensity change and duration of the spectral signal to describe the dynamic response deviation of each spectral channel under typical physical disturbances. A dynamic synchronization deviation vector is generated based on the time offset vector, and it is written back into the spectral acquisition structure for time-domain registration and fusion processing sequence optimization. It is continuously updated in subsequent detection to achieve dynamic correction of response delay.

3. The method for detecting arc light in switchgear based on multispectral fusion according to claim 1, characterized in that, Step S2 includes: Historical response data of infrared, visible and ultraviolet light signals are retrieved based on the time synchronization deviation vector, and time-domain unified correction is performed to construct a shadow spectral sequence covering the abnormal window. The shadow spectral sequence is compared frame by frame with the currently acquired multispectral response sequence on the synchronous time axis to identify the energy surge region of the infrared channel and extract its boundary, intensity and duration. By combining the variation trajectories of visible light and ultraviolet light signals, the starting position, ending position, and duration of the multispectral overlap risk zone are calculated, and the period of infrared channel response rise is calibrated. By integrating the time positioning information and energy change curves of each channel, a threshold perturbation distribution map on a unified time axis is constructed, and the boundaries of the interference intensity level and the response offset paths of the spectral channels are clearly defined.

4. The method for detecting arc light in switchgear based on multispectral fusion according to claim 3, characterized in that, The shadow spectral sequence is composed of historical sampling frames. The ambient temperature, light intensity and electromagnetic interference status corresponding to the selected sampling frames are consistent with the current observation conditions. During the construction process, the complete anomaly trigger window is covered and all spectral channel data are retained.

5. The method for detecting arc light in switchgear based on multispectral fusion according to claim 1, characterized in that, Step S3 includes: Complete response data of infrared light signals under temperature disturbance and strong light interference in multiple detection cycles are retrieved, and the intensity change characteristics of the infrared channel before, during and after the interference are analyzed to form a representative response rebound trajectory model. By combining the infrared anomaly segments identified in the threshold perturbation distribution map, the response rise rate, duration and offset amplitude in the current period are extracted, and a similarity comparison is performed with the historical trajectory model to calculate the trajectory shape and recovery time of the infrared response during the regression process. Infrared threshold rebound prediction curves are constructed on a unified time axis, time reconfiguration intervals are determined based on rebound trend change points, and time correction offsets are allocated to multispectral channels to correct the temporal structure. Based on the infrared rebound prediction curve and the time offset allocation results, a time series correction instruction is generated to clarify the sampling delay compensation amount, response window offset amplitude, and fusion timing priority of each spectral signal in the recovery stage.

6. The method for detecting arc light in switchgear based on multispectral fusion according to claim 1, characterized in that, Step S4 includes: Based on the sampling offset parameters in the time series correction instruction, the infrared light signal, visible light signal and ultraviolet light signal are time-mapped and adjusted, and time is rearranged frame by frame to complete the unified sampling reference alignment. Based on the rearranged sampling time grid structure, a global scan analysis is performed to identify fusion errors and perform interpolation correction and frequency fine-tuning on the error segments to form a multi-channel spectral response matrix in a unified format. Based on the energy distribution characteristics of each channel in the spectral response matrix, the signal contribution ratio of each sampling window is calculated, a fusion weight matrix covering the entire time axis is generated, and the weight inflection point of the infrared channel rebound stage is marked. By combining the weight change trend of the infrared channel with the rebound prediction trajectory, the regression state of the infrared channel is determined, and a multi-channel collaborative weight adjustment framework is constructed based on the fused weight matrix.

7. The method for detecting arc light in switchgear based on multispectral fusion according to claim 6, characterized in that, The trend of weight change of infrared light signal in the fusion weight matrix is ​​used to determine its response rebound state. When the weight changes from a continuous decrease to a stable increase and remains stable, it is determined that the infrared light signal has completed the rebound and triggers the collaborative weight redistribution operation of the multispectral channels.

8. The method for detecting arc light in switchgear based on multispectral fusion according to claim 1, characterized in that, Step S5 includes: Based on the rebound state of the infrared channel in the fusion weight matrix, a time-series scan is performed to identify weight mutation points and perform micro-periodic sampling beat misalignment operation, so that the spectral signals are staggered in the high-risk window to break up the energy peak synchronization. By combining the time sampling grid and the trend of fusion weight changes, phase conjugate modulation is performed on the spectral signal in the strong interference region to filter out peak overlap points in reverse and limit the propagation path of energy between channels. A spectral topology chain break window is constructed by integrating the energy transfer characteristics between spectral channels, defining the overlap risk zone and setting an energy upper limit. When the channel signal exceeds the upper limit, the chain break window automatically closes the transmission path. By combining the activation behavior of the chain break window with the fusion weight state, the signal recognition stability is evaluated. The subsequent sampling strategy and energy management process are optimized through a feedback mechanism to ensure the stable recognition capability of the arc light signal.

9. The method for detecting arc light in switchgear based on multispectral fusion according to claim 8, characterized in that, In the micro-periodic sampling beat misalignment operation, the offset of the sampling trigger time adopts a periodic incremental strategy, and a unified time calibration is performed after each round of offset, so that the infrared light signal, visible light signal and ultraviolet light signal remain relatively interleaved in multiple sampling cycles and the overall synchronization structure remains unchanged.

10. A switchgear arc light detection system based on multispectral fusion, used to implement the switchgear arc light detection method based on multispectral fusion as described in any one of claims 1-9, characterized in that, It includes a time-scale synchronization module, an interference identification module, a rebound prediction module, a response correction module, and a steady-state control module. The time synchronization module establishes a unified time reference across the spectrum, performs synchronous comparison of infrared, visible and ultraviolet light signals, extracts temperature transition triggering features and strong light disturbance spectral features, and generates a time synchronization deviation vector. The interference identification module constructs a counterfactual spectral playback model based on the time synchronization deviation vector, generates a shadow spectral sequence, locates the cause of the infrared response threshold rise through energy difference analysis, calculates the spectral overlap risk range and duration, and outputs a threshold perturbation distribution map. The rebound prediction module constructs a rebound prediction model based on the threshold perturbation distribution map, inputs the historical response curve and the current change slope, estimates the infrared threshold regression trajectory, and generates time series correction instructions. The response correction module performs time-frequency synchronization rearrangement of the multispectral channels according to the time-series correction instructions, reconstructs the sampling time grid, corrects the response deviation, outputs the fusion weight matrix, and calculates the dynamic rebound state. The steady-state control module, based on the fusion weight matrix and dynamic rebound state, initiates micro-period misalignment gating and phase conjugate image stabilization control, constructs a spectral topology chain break window, and blocks the spectral energy overlap path.

Images