Intelligent switch cabinet safety monitoring method and system based on data analysis

Abstract

The invention discloses an intelligent switch cabinet safety monitoring method and system based on data analysis, and the method comprises the following steps: collecting multi-mode monitoring signal data in the operation process of an intelligent switch cabinet, and carrying out the preprocessing; executing feature extraction; calculating an information entropy value of the local time window of each monitoring signal channel, and mapping the information entropy value to a unified entropy spectrum space; calculating a cross-modal energy coupling degree; executing continuous time tracking, and generating an energy stability monitoring curve; when the energy stability monitoring curve has a descending trend in a continuous time period, positioning an abnormal monitoring signal channel; entropy feature inversion is carried out, and a monitoring diagnosis result is generated; and sending the abnormal monitoring signal channel, the corresponding component identifier and the monitoring diagnosis result to a monitoring center. According to the method, the multi-modal data analysis and information entropy energy coupling method is utilized to realize accurate hidden fault identification and energy disturbance diagnosis of the intelligent switch cabinet, and the method has the advantages of high accuracy, timely response and high interpretability.

Description

A Data-Based Intelligent Switchgear Safety Monitoring Method and System Technical Field

[0001] This invention relates to the field of power equipment safety monitoring, and in particular to a method and system for intelligent switchgear safety monitoring based on data analysis. Background Technology

[0002] Existing intelligent switchgear safety monitoring technologies mainly rely on single-mode sensor data such as temperature, current, and voltage. They detect anomalies and issue alarms by setting thresholds or based on empirical models. These methods are effective in low-dimensional signal environments, but in complex multi-source monitoring environments, they are difficult to accurately characterize the energy transfer and correlation features between different physical signals. In particular, they cannot reflect the dynamic changes in energy state and multi-modal coupling relationships during operation, making it difficult to detect latent faults in a timely manner.

[0003] As switchgear structures become more complex and operating environments become more varied, traditional methods suffer from problems such as inconsistent data dimensions, difficulty in time alignment, and feature redundancy when dealing with multimodal monitoring signals. They lack systematic energy layer modeling and cross-modal correlation analysis mechanisms. Existing monitoring systems mostly remain at the stage of judging surface feature thresholds, unable to dynamically quantify energy coupling states, or achieve energy stability assessment and inversion diagnosis based on entropy features. This makes it difficult to achieve early identification and accurate location of latent energy disturbances in intelligent switchgear. Summary of the Invention

[0004] One objective of this invention is to propose a data analysis-based method and system for safety monitoring of intelligent switchgear. This invention utilizes multimodal data analysis and information entropy energy coupling method to achieve accurate identification of latent faults and diagnosis of energy disturbances in intelligent switchgear, and has the advantages of high accuracy, timely response and strong interpretability.

[0005] A data analysis-based intelligent switchgear safety monitoring method according to an embodiment of the present invention includes the following steps:

[0006] Collect multimodal monitoring signal data during the operation of the intelligent switchgear, perform preprocessing, and generate standardized multimodal monitoring signal sequences;

[0007] Feature extraction is performed on the standardized multimodal monitoring signal sequence, and the temporal fluctuation characteristics, frequency domain energy characteristics and phase consistency characteristics of each monitoring signal channel are calculated. Time alignment and feature normalization are then performed to form a multimodal monitoring feature matrix.

[0008] Based on the multimodal monitoring feature matrix, the information entropy value is calculated for the local time window of each monitoring signal channel and mapped to a unified entropy spectrum space to generate a multimodal entropy spectrum sequence.

[0009] Calculate the entropy difference sequence between adjacent time points based on the multimodal entropy spectrum sequence, and calculate the cross-modal energy coupling degree to generate the energy coupling matrix;

[0010] Perform continuous-time tracking on the energy coupling matrix, calculate the entropy change rate of each channel pair, and generate energy stability monitoring curves;

[0011] When the energy stability monitoring curve shows a downward trend over a continuous period of time, it is determined that the intelligent switch cabinet is in a state of hidden energy disturbance, the abnormal monitoring signal channel is located, and the corresponding component identifier is generated.

[0012] Entropy feature inversion is performed based on the entropy contribution rate of each monitoring signal channel, and energy contribution inversion and feature decomposition are performed on the abnormal monitoring signal channels to generate monitoring and diagnostic results.

[0013] The abnormal monitoring signal channel, corresponding component identification, and monitoring and diagnostic results are sent to the monitoring center.

[0014] Optionally, the multimodal monitoring signal data includes temperature signal, current signal, voltage signal, partial discharge signal, infrared thermal imaging signal, humidity signal, and gas concentration signal, and the preprocessing includes anomaly removal, time synchronization, noise removal, and amplitude normalization.

[0015] Optionally, the generation of the multimodal monitoring feature matrix specifically includes:

[0016] Feature extraction and initialization are performed on the standardized multimodal monitoring signal sequence, and the time series of each multimodal monitoring signal channel is segmented by a sliding window to construct a time window set;

[0017] For the multimodal monitoring signal data within each time window, the time-series fluctuation characteristics are calculated, including time-series mean characteristics, time-series variance characteristics, and time-series rate of change characteristics;

[0018] Frequency domain energy feature extraction is performed on the standardized multimodal monitoring signal sequence. The fast Fourier transform algorithm is used to perform spectral decomposition on the time series data of each multimodal monitoring signal channel to obtain the frequency domain energy distribution features and extract the main frequency amplitude features, spectral centroid features and frequency band energy ratio features.

[0019] Calculate the phase consistency characteristics for each multimodal monitoring signal channel;

[0020] The time-series mean feature, time-series variance feature, time-series rate of change feature, dominant frequency amplitude feature, spectral centroid feature, frequency band energy ratio feature, and phase consistency feature of each multimodal monitoring signal channel are sequentially concatenated to form the comprehensive feature vector of that multimodal monitoring signal channel. The comprehensive feature vectors of all multimodal monitoring signal channels are then arranged in order to form the multimodal monitoring feature matrix.

[0021] Perform time alignment on the multimodal monitoring feature matrix;

[0022] Feature normalization is performed on the time-aligned multimodal monitoring feature matrix to form a multimodal monitoring feature matrix.

[0023] Optionally, the time-series mean feature is the average value of all monitored signal values ​​within the time window, the time-series variance feature is the average deviation of the monitored signal from the mean within the time window, and the time-series rate of change feature is the rate of change obtained by dividing the difference in amplitude of the monitored signal between the start and end sampling points of the time window by the window length.

[0024] Optionally, the generation of the multimodal entropy spectrum sequence specifically includes:

[0025] Entropy spectrum mapping is performed based on the multimodal monitoring feature matrix. The data of each multimodal monitoring signal channel in the multimodal monitoring feature matrix is ​​used as an independent input channel, and a time series structure is constructed according to the sampling time index.

[0026] The time series of each multimodal monitoring signal channel is segmented according to a fixed time window to form multiple local time windows;

[0027] Within each local time window, the occurrence ratio of multimodal monitoring feature values ​​is statistically analyzed, the distribution of each feature value is calculated, and the corresponding feature probability distribution is obtained.

[0028] The information entropy value is calculated based on the characteristic probability distribution of the local time window, and the local information entropy sequence of each multimodal monitoring signal channel in different time windows is obtained.

[0029] The local information entropy sequences of all multimodal monitoring signal channels are standardized, and the information entropy values ​​between different channels are numerically normalized.

[0030] The normalized information entropy sequence is smoothed by interpolation in the time dimension to form a continuous and smooth information entropy time series distribution.

[0031] The smooth information entropy temporal distribution of each multimodal monitoring signal channel is mapped to a unified entropy spectrum space. Within the unified entropy spectrum space, the information entropy values ​​of different channels are time-aligned and energy scales are unified to establish a cross-channel energy distribution correspondence.

[0032] Based on the temporal alignment results in the unified entropy spectrum space, the entropy spectrum information of all multimodal monitoring signal channels is aggregated to generate a multimodal entropy spectrum sequence.

[0033] Optionally, the generation of the energy coupling matrix specifically includes:

[0034] Energy coupling analysis is performed based on the multimodal entropy spectrum sequence, using the entropy spectrum values ​​of each multimodal monitoring signal channel at each sampling time in the multimodal entropy spectrum sequence as input data;

[0035] In the multimodal entropy spectrum sequence, select any two adjacent sampling times and calculate the change in information entropy of each multimodal monitoring signal channel between adjacent sampling times;

[0036] Arrange the entropy changes of all multimodal monitoring signal channels in order of channel number to form an entropy difference sequence matrix;

[0037] Based on the entropy difference sequence matrix, calculate the cross-modal energy coupling degree between any two multimodal monitoring signal channels;

[0038] The cross-modal energy coupling degree between all multimodal monitoring signal channels is matrixed according to the channel combination method to establish an energy coupling matrix.

[0039] Optionally, the generation of the energy stability monitoring curve specifically includes:

[0040] Within a preset continuous sampling time period, the energy coupling matrix is ​​calculated based on the entropy difference sequence of all multimodal monitoring signal channels within that time period.

[0041] Using time periods as a sliding window, the energy coupling matrix is ​​dynamically updated according to the time series order to construct an energy coupling matrix sequence;

[0042] In the energy coupling matrix sequence, for any pair of multimodal monitoring signal channels, the energy coupling degree of the channel pair under all sliding time periods is extracted to form a time series of channel pair energy coupling degree.

[0043] Perform continuous time difference processing on the channel pair energy coupling time series, calculate the change in energy coupling between adjacent sliding time intervals, and obtain the channel pair energy change series;

[0044] Perform sliding time window smoothing and averaging operations on the energy change sequence of the channel pairs, calculate the average value of the energy change amplitude of the channel pairs, obtain the entropy change rate of the channel pairs, and sum up the entropy change rates of all channel pairs within the same sliding time period to obtain the average entropy change rate of that time period.

[0045] The average entropy change rate of each time period is arranged in the order of the sliding time period to generate a time series of average entropy change rate.

[0046] Plot an energy stability monitoring curve with the center time of the sliding time interval as the horizontal axis and the corresponding average entropy change rate as the vertical axis.

[0047] Optionally, the generation of the corresponding component identifier specifically includes:

[0048] Trend analysis is performed on the energy stability monitoring curve. The average slope of the sliding sampling time period of the energy stability monitoring curve is calculated within a continuous sampling time period. When the average slope is lower than the preset drop threshold for several consecutive sampling time periods, the intelligent switch cabinet is determined to be in a state of hidden energy disturbance.

[0049] After determining the state of latent energy disturbance, the energy anomaly impact index of each multimodal monitoring signal channel is calculated based on the rate of change of energy coupling degree of each multimodal monitoring signal channel pair in the energy coupling matrix sequence.

[0050] All multimodal monitoring signal channels were sorted in descending order based on the energy anomaly impact index, and the multimodal monitoring signal channel with the highest energy anomaly impact index was selected as the anomaly monitoring signal channel.

[0051] Based on the channel number of the abnormal monitoring signal channel in the energy coupling matrix and its energy coupling relationship with other channel pairs, the location of the physical component corresponding to the abnormal monitoring signal channel is determined, and a component identifier corresponding to the physical component is generated.

[0052] Optionally, the generation of the monitoring and diagnostic results specifically includes:

[0053] Based on the channel number of the abnormal monitoring signal channel in the multimodal monitoring feature matrix, the standardized multimodal monitoring signal sequence of the abnormal monitoring signal channel in the continuous sampling time period is extracted, and the information entropy value sequence corresponding to each sampling time period is calculated.

[0054] Within the same sampling time period, the overall information entropy distribution is calculated based on the information entropy value sequence of all multimodal monitoring signal channels, the entropy contribution rate of each multimodal monitoring signal channel in the overall information entropy distribution is determined, and an entropy contribution rate vector is constructed.

[0055] Based on the entropy contribution rate of the anomaly monitoring signal channel, entropy feature inversion is performed from the entropy contribution rate vector space to recover the energy distribution characteristics of the anomaly monitoring signal channel in the energy coupling structure, and the energy contribution inversion result of the anomaly monitoring signal channel is generated.

[0056] The energy contribution inversion results are subjected to eigenvalue decomposition to extract the main characteristic components characterizing the energy disturbance properties, forming an energy disturbance feature set for the anomaly monitoring signal channel;

[0057] Monitoring and diagnostic results are generated based on the energy disturbance feature set.

[0058] According to an embodiment of the present invention, an intelligent switchgear safety monitoring system based on data analysis includes:

[0059] The data acquisition and preprocessing module is used to acquire multimodal monitoring signal data during the operation of the intelligent switchgear and generate standardized monitoring signal sequences;

[0060] The feature extraction module is used to extract the temporal fluctuation features, frequency domain energy features, and phase consistency features of each monitoring signal channel to form a multimodal monitoring feature matrix;

[0061] The entropy spectrum construction module is used to calculate the information entropy value of the monitoring signal channel and map it to a unified entropy spectrum space to generate a multimodal entropy spectrum sequence.

[0062] The energy coupling analysis module is used to calculate the cross-modal energy coupling degree and generate the energy coupling matrix.

[0063] The energy stability monitoring module is used to calculate the entropy change rate of the channel pair and generate energy stability monitoring curves.

[0064] Anomaly identification module is used to locate the anomaly monitoring signal channel and generate component identification when the energy stability monitoring curve shows a downward trend;

[0065] The energy inversion and diagnosis module is used to perform energy inversion and eigenvalue decomposition based on the entropy contribution rate and generate monitoring and diagnosis results.

[0066] The monitoring center sending module is used to send abnormal monitoring signal channels, corresponding component identifiers, and monitoring diagnostic results to the monitoring center.

[0067] The beneficial effects of this invention are:

[0068] This invention constructs a multi-layered energy monitoring and diagnostic framework for intelligent switchgear by introducing a technical system that combines information entropy theory with multimodal data analysis. It achieves end-to-end safety monitoring from the signal layer, energy layer to the semantic layer. Compared with traditional monitoring methods based on single-modal features or fixed thresholds, this invention achieves data standardization and time alignment in the multi-source signal acquisition and processing stage, ensuring synchronous analysis of multiple signals such as temperature, current, voltage, partial discharge, infrared, humidity, and gas concentration under a unified time reference. This lays the foundation for subsequent energy modeling and correlation calculation. By constructing a multimodal monitoring feature matrix and mapping it to a unified entropy spectrum space, it establishes a co-scale representation of energy across different monitoring channels, realizing a unified expression of cross-modal energy distribution characteristics.

[0069] At the energy correlation analysis level, the energy coupling matrix proposed in this invention quantitatively characterizes the energy interaction intensity and dynamic relationship between various monitoring signals by calculating the cross-modal energy coupling degree between channels. Then, based on the energy coupling matrix, an energy stability monitoring curve is generated, realizing the time-series visualization tracking of the system's energy state. When the monitoring curve shows a downward trend in a continuous time period, it can automatically identify that the system is in a state of latent energy disturbance. Combined with the energy anomaly impact index, it can accurately locate abnormal channels and physical components, improving the accuracy of latent fault identification and location.

[0070] Furthermore, this invention introduces an entropy feature inversion mechanism and energy disturbance feature set modeling in the diagnostic process, decomposes the complex energy disturbance process into three interpretable indicators: intensity, stability, and volatility, and constructs a hierarchical risk diagnosis model based on the comprehensive disturbance index. This model can automatically generate semantic monitoring and diagnostic results based on changes in energy disturbance characteristics, and realize intelligent judgment of fault severity and early warning. Attached Figure Description

[0071] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0072] Figure 1 is a flowchart of a data analysis-based intelligent switchgear safety monitoring method proposed in this invention;

[0073] Figure 2 is a schematic diagram of energy inversion based on entropy contribution rate in a smart switchgear safety monitoring method based on data analysis proposed in this invention.

[0074] Figure 3 is a schematic diagram of the generation of the energy stability monitoring curve of the intelligent switchgear safety monitoring method based on data analysis proposed in this invention. Detailed Implementation

[0075] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0076] Referring to Figures 1-3, a data analysis-based intelligent switchgear safety monitoring method includes the following steps:

[0077] Collect multimodal monitoring signal data during the operation of the intelligent switchgear, perform preprocessing, and generate standardized multimodal monitoring signal sequences;

[0078] Feature extraction is performed on the standardized multimodal monitoring signal sequence, and the temporal fluctuation characteristics, frequency domain energy characteristics and phase consistency characteristics of each monitoring signal channel are calculated. Time alignment and feature normalization are then performed to form a multimodal monitoring feature matrix.

[0079] Based on the multimodal monitoring feature matrix, the information entropy value is calculated for the local time window of each monitoring signal channel and mapped to a unified entropy spectrum space to generate a multimodal entropy spectrum sequence.

[0080] Calculate the entropy difference sequence between adjacent time points based on the multimodal entropy spectrum sequence, and calculate the cross-modal energy coupling degree to generate the energy coupling matrix;

[0081] Perform continuous-time tracking on the energy coupling matrix, calculate the entropy change rate of each channel pair, and generate energy stability monitoring curves;

[0082] When the energy stability monitoring curve shows a downward trend over a continuous period of time, it is determined that the intelligent switch cabinet is in a state of hidden energy disturbance, the abnormal monitoring signal channel is located, and the corresponding component identifier is generated.

[0083] Entropy feature inversion is performed based on the entropy contribution rate of each monitoring signal channel, and energy contribution inversion and feature decomposition are performed on the abnormal monitoring signal channels to generate monitoring and diagnostic results.

[0084] The abnormal monitoring signal channel, corresponding component identification, and monitoring and diagnostic results are sent to the monitoring center.

[0085] In this embodiment, the multimodal monitoring signal data includes temperature signal, current signal, voltage signal, partial discharge signal, infrared thermal imaging signal, humidity signal, and gas concentration signal. The preprocessing includes anomaly removal, time synchronization, noise removal, and amplitude normalization.

[0086] In this embodiment, the generation of the multimodal monitoring feature matrix specifically includes:

[0087] Feature extraction and initialization are performed on the standardized multimodal monitoring signal sequence, and the time series of each multimodal monitoring signal channel is segmented by a sliding window to construct a time window set;

[0088] For the multimodal monitoring signal data within each time window, the time-series fluctuation characteristics are calculated, including time-series mean characteristics, time-series variance characteristics, and time-series rate of change characteristics;

[0089] Frequency domain energy feature extraction is performed on the standardized multimodal monitoring signal sequence. The fast Fourier transform algorithm is used to perform spectral decomposition on the time series data of each multimodal monitoring signal channel to obtain frequency domain energy distribution features. The dominant frequency amplitude feature, spectral centroid feature, and frequency band energy ratio feature are extracted. The dominant frequency amplitude feature represents the amplitude of the dominant frequency of the energy distribution, the spectral centroid feature represents the location of the frequency domain energy center, and the frequency band energy ratio feature represents the ratio of high-frequency energy to low-frequency energy.

[0090] A phase consistency feature is calculated for each multimodal monitoring signal channel, and the phase consistency feature is used to characterize the phase synchronization degree of the multimodal monitoring signal channel;

[0091] The generation of the phase consistency feature specifically includes: extracting the instantaneous phase sequence of each multimodal monitoring signal channel within each time window; vectorizing the instantaneous phase sequence within each time window, converting each instantaneous phase into a direction vector while maintaining a unit length; performing vector synthesis on all direction vectors within the same time window to obtain the phase synthesis amplitude corresponding to that time window, which is used to characterize the concentration of phase directions; normalizing the phase synthesis amplitude based on the total number of phase vectors within that time window to obtain the phase consistency ratio of that time window; and arranging the phase consistency ratios of all time windows in chronological order to form the phase consistency feature.

[0092] The time-series mean feature, time-series variance feature, time-series rate of change feature, dominant frequency amplitude feature, spectral centroid feature, frequency band energy ratio feature, and phase consistency feature of each multimodal monitoring signal channel are sequentially concatenated to form the comprehensive feature vector of that multimodal monitoring signal channel. The comprehensive feature vectors of all multimodal monitoring signal channels are then arranged in order to form the multimodal monitoring feature matrix.

[0093] Time alignment is performed on the multimodal monitoring feature matrix to align the features of all multimodal monitoring signal channels under a unified time reference;

[0094] Feature normalization is performed on the time-aligned multimodal monitoring feature matrix to form a multimodal monitoring feature matrix.

[0095] In this embodiment, the time-series mean feature is the average value of all monitored signal values ​​within the time window, the time-series variance feature is the average deviation of the monitored signal from the mean within the time window, and the time-series rate of change feature is the rate of change obtained by dividing the difference in amplitude of the monitored signal between the start and end sampling points of the time window by the window length.

[0096] In this embodiment, the generation of the multimodal entropy spectrum sequence specifically includes:

[0097] Entropy spectrum mapping is performed based on the multimodal monitoring feature matrix. The data of each multimodal monitoring signal channel in the multimodal monitoring feature matrix is ​​used as an independent input channel, and a time series structure is constructed according to the sampling time index.

[0098] The time series of each multimodal monitoring signal channel is segmented according to a fixed time window to form multiple local time windows, each of which contains a continuous set of monitoring feature points;

[0099] Within each local time window, the occurrence ratio of multimodal monitoring feature values ​​is statistically analyzed, the distribution of each feature value is calculated, and the corresponding feature probability distribution is obtained, which is used to characterize the distribution state of monitoring signal energy within that local time window.

[0100] The information entropy value is calculated based on the characteristic probability distribution of the local time window, and the local information entropy sequence of each multimodal monitoring signal channel in different time windows is obtained, which is used to describe the uncertainty characteristics of the energy distribution of the channel.

[0101] The local information entropy sequences of all multimodal monitoring signal channels are standardized, and the information entropy values ​​between different channels are numerically normalized.

[0102] The normalized information entropy sequence is smoothed by interpolation in the time dimension to eliminate the influence of local mutations and sampling errors, forming a continuous and smooth information entropy time series distribution.

[0103] The smooth information entropy temporal distribution of each multimodal monitoring signal channel is mapped to a unified entropy spectrum space. Within the unified entropy spectrum space, the information entropy values ​​of different channels are time-aligned and energy scales are unified to establish a cross-channel energy distribution correspondence.

[0104] The unified entropy spectrum space is a spatial representation domain used to uniformly characterize the energy distribution characteristics of different signal channels by normalizing and aligning the information entropy values ​​of each multimodal monitoring signal channel to the same energy scale.

[0105] Based on the temporal alignment results in the unified entropy spectrum space, the entropy spectrum information of all multimodal monitoring signal channels is aggregated to generate a multimodal entropy spectrum sequence, which is used to describe the overall energy distribution evolution characteristics of the intelligent switchgear during operation.

[0106] In this embodiment, the generation of the energy coupling matrix specifically includes:

[0107] Energy coupling analysis is performed based on the multimodal entropy spectrum sequence. The entropy spectrum values ​​of each multimodal monitoring signal channel in the multimodal entropy spectrum sequence at each sampling time are used as input data to calculate the energy correlation between different channels. The entropy spectrum value refers to the information entropy value calculated based on the characteristic distribution of the multimodal monitoring signal within the time window.

[0108] In the multimodal entropy spectrum sequence, any two adjacent sampling times are selected, and the change in information entropy of each multimodal monitoring signal channel between adjacent sampling times is calculated. The change in information entropy is equal to the entropy spectrum value of the next sampling time minus the entropy spectrum value of the previous sampling time, which is used to reflect the change in energy distribution of the channel within adjacent time windows.

[0109] The entropy changes of all multimodal monitoring signal channels are arranged in order of channel number to form an entropy difference sequence matrix, which is used to describe the energy change trend of each channel in the time dimension.

[0110] Based on the entropy difference sequence matrix, calculate the cross-modal energy coupling degree between any two multimodal monitoring signal channels;

[0111] The generation of the cross-modal energy coupling degree specifically includes: based on the entropy difference sequence matrix, selecting any two multimodal monitoring signal channels sequentially, and extracting the information entropy difference sequence of each channel at all sampling times; performing difference calculation on the two information entropy difference sequences at the same sampling time to obtain the channel pair time-series difference sequence; performing absolute value and normalization on the channel pair time-series difference sequence to generate a standardized channel difference sequence; performing amplitude averaging calculation on all sampling points of the standardized channel difference sequence in the time dimension to obtain the channel pair average entropy difference value, which is used to characterize the average energy difference intensity of the channel pair in the entire sampling period; further performing amplitude averaging calculation on the information entropy change of each multimodal monitoring signal channel between consecutive sampling times to obtain the channel average entropy change amplitude, which is used to characterize the energy fluctuation intensity of a single channel; and performing a ratio calculation between the channel pair average entropy difference value and the channel average entropy change amplitude to obtain the cross-modal energy coupling degree of the channel pair, which is used to quantify the degree of energy coupling correlation between different monitoring signal channels.

[0112] The cross-modal energy coupling degree between all multimodal monitoring signal channels is matrixed according to the channel combination method to establish an energy coupling matrix. The rows and columns of the energy coupling matrix correspond to the multimodal monitoring signal channel numbers, and each element of the matrix is ​​used to represent the energy coupling strength between the corresponding channel pairs, which is used to characterize the energy interaction relationship between each multimodal monitoring signal channel of the intelligent switch cabinet.

[0113] In this embodiment, the generation of the energy stability monitoring curve specifically includes:

[0114] Within a preset continuous sampling time period, an energy coupling matrix is ​​calculated based on the entropy difference sequence of all multimodal monitoring signal channels within that time period. This matrix is ​​used to characterize the energy correlation strength of each multimodal monitoring signal channel within that time period.

[0115] Using time periods as sliding windows, the energy coupling matrix is ​​dynamically updated according to the time sequence to construct an energy coupling matrix sequence, which is used to record the evolution of the energy coupling state of the intelligent switchgear under different sliding time periods;

[0116] In the energy coupling matrix sequence, for any pair of multimodal monitoring signal channels, the energy coupling degree of the channel pair is extracted under all sliding time periods to form a channel pair energy coupling degree time series, which is used to characterize the continuous change of channel pair energy interaction intensity over time.

[0117] Perform continuous time difference processing on the channel pair energy coupling time series, calculate the change in energy coupling between adjacent sliding time intervals, and obtain the channel pair energy change series;

[0118] Perform sliding time window smoothing and averaging operations on the energy change sequence of the channel pairs, calculate the average value of the energy change amplitude of the channel pairs, obtain the entropy change rate of the channel pairs, and sum up the entropy change rates of all channel pairs within the same sliding time period to obtain the average entropy change rate of that time period.

[0119] The entropy change rate of the channel pair is the average rate at which the energy coupling degree between a pair of multimodal monitoring signal channels changes over time during a continuous sampling period. It is used to quantify the speed and stability of the change in energy transfer intensity between the channel pairs.

[0120] The average entropy change rate of each time period is arranged in the order of sliding time periods to generate an average entropy change rate time series. The average entropy change rate time series is used to reflect the continuous change trend of energy transfer stability of the intelligent switch cabinet system over time.

[0121] An energy stability monitoring curve is plotted with the center moment of the sliding time interval as the horizontal axis and the corresponding average entropy change rate as the vertical axis. The energy stability monitoring curve is used to reflect the stable trend and dynamic fluctuation of energy transfer during the operation of the intelligent switch cabinet system.

[0122] In this embodiment, the generation of the corresponding component identifier specifically includes:

[0123] Trend analysis is performed on the energy stability monitoring curve. The average slope of the sliding sampling time period of the energy stability monitoring curve is calculated within a continuous sampling time period. When the average slope is lower than the preset drop threshold for several consecutive sampling time periods, the intelligent switch cabinet is determined to be in a state of hidden energy disturbance.

[0124] After determining the state of latent energy disturbance, the energy anomaly impact index of each multimodal monitoring signal channel is calculated based on the rate of change of energy coupling degree of each multimodal monitoring signal channel pair in the energy coupling matrix sequence. Specifically, for each multimodal monitoring signal channel in the energy coupling matrix sequence, the average amplitude of the rate of change of energy coupling degree between it and all other channels is calculated in each sampling time period to characterize the instantaneous energy disturbance degree of the channel in that sampling time period. Then, the average amplitude over all sampling time periods is accumulated over time to obtain the energy anomaly impact index of the channel, which is used to describe the cumulative intensity of energy disturbance and overall instability of the channel in continuous sampling time periods.

[0125] All multimodal monitoring signal channels were sorted in descending order based on the energy anomaly impact index, and the multimodal monitoring signal channel with the highest energy anomaly impact index was selected as the anomaly monitoring signal channel.

[0126] Based on the channel number of the abnormal monitoring signal channel in the energy coupling matrix and its energy coupling relationship with other channel pairs, the location of the physical component corresponding to the abnormal monitoring signal channel is determined, and a component identifier corresponding to the physical component is generated. Specifically, this includes: based on the channel number of the abnormal monitoring signal channel, retrieving the row data corresponding to the channel number in the energy coupling matrix, extracting the energy coupling degree between the abnormal monitoring signal channel and all other multimodal monitoring signal channels, forming an energy coupling vector for the abnormal monitoring signal channel; sorting the energy coupling degrees in the energy coupling vector in descending order, selecting several channel numbers with high energy coupling degree values, and constructing an abnormal channel association set; querying the corresponding physical component number set according to the index relationship between each channel number in the abnormal channel association set and the physical component mapping table; determining the target physical component number corresponding to the abnormal monitoring signal channel in the physical component number set according to the one-to-one mapping relationship between each channel number and the physical component number; generating a unique component identifier according to the index position of the target physical component number in the component identifier index table, wherein the component identifier is used to mark the location of the physical component corresponding to the abnormal monitoring signal channel in the intelligent switch cabinet.

[0127] In this embodiment, the generation of the monitoring and diagnostic results specifically includes:

[0128] Based on the channel number of the abnormal monitoring signal channel in the multimodal monitoring feature matrix, the standardized multimodal monitoring signal sequence of the abnormal monitoring signal channel in the continuous sampling time period is extracted, and the information entropy value sequence corresponding to each sampling time period is calculated.

[0129] Within the same sampling time period, the overall information entropy distribution is calculated based on the information entropy value sequence of all multimodal monitoring signal channels. The entropy contribution rate of each multimodal monitoring signal channel in the overall information entropy distribution is determined, and an entropy contribution rate vector is constructed to characterize the degree of influence of each multimodal monitoring signal channel on the overall energy state of the intelligent switchgear.

[0130] The entropy contribution rate is calculated by taking the relative proportion of each multimodal monitoring signal channel in the overall information entropy, reflecting the degree of influence of each channel on the energy distribution and dynamic changes of the intelligent switch cabinet;

[0131] Based on the entropy contribution rate of the anomaly monitoring signal channel, entropy feature inversion is performed from the entropy contribution rate vector space to recover the energy distribution characteristics of the anomaly monitoring signal channel in the energy coupling structure, and the energy contribution inversion result of the anomaly monitoring signal channel is generated.

[0132] The generation of the energy contribution inversion result specifically includes: extracting the entropy contribution rate sequence of the abnormal monitoring signal channel within a continuous sampling time period from the entropy contribution rate vector; normalizing the entropy contribution rate sequence to obtain the standardized entropy contribution sequence of the abnormal monitoring signal channel in the overall information entropy distribution; extracting the energy coupling degree sequence between the abnormal monitoring signal channel and all other multimodal monitoring signal channels according to the abnormal monitoring signal channel number in the energy coupling matrix sequence to form an energy coupling vector set for the abnormal monitoring signal channel; matching the standardized entropy contribution sequence with the energy coupling vector set time period by time, calculating the consistency index between the entropy contribution change and the energy coupling change within each sampling time period to form an entropy-energy matching sequence; performing a sliding sampling time period smoothing operation on the entropy-energy matching sequence to obtain the energy distribution characteristics of the abnormal monitoring signal channel in the energy coupling structure; and using the energy distribution characteristics as the energy contribution inversion result of the abnormal monitoring signal channel to characterize the inversion contribution characteristics of the channel to the overall energy transfer of the intelligent switch cabinet within a continuous sampling time period.

[0133] The generation of the entropy-energy matching sequence specifically includes: performing time alignment processing on the standardized entropy contribution rate sequence and the energy coupling vector set; calculating the change in the standardized entropy contribution rate relative to the previous sampling time period within each sampling time period. If the current entropy contribution rate is higher than the previous sampling time period, the change direction is recorded as upward; if it is lower than the previous sampling time period, it is recorded as downward. At the same time, the difference between the corresponding energy coupling vector and the energy coupling vector of the previous sampling time period is calculated, and its change direction is determined. If the overall change is upward, it is recorded as upward; if the overall change is downward, it is recorded as downward. The change direction of the entropy contribution rate is compared with the change direction of the energy coupling. If the directions are consistent, the consistent direction state is recorded as 1 in that sampling time period; if the directions are opposite, it is recorded as 0. The number of channel pairs with a consistent direction state of 1 in each sampling time period is counted, and the proportion of the channel pairs is calculated to obtain the consistency index value of that sampling time period. The consistency index values ​​of all sampling time periods are arranged in chronological order to form an entropy-energy matching sequence, which is used to characterize the dynamic coordination relationship between the change in standardized entropy contribution rate and the change in energy coupling in the time dimension of the multimodal monitoring signal.

[0134] The energy contribution inversion results are subjected to eigenvalue decomposition to extract the main characteristic components characterizing the energy disturbance properties, forming an energy disturbance feature set for the anomaly monitoring signal channel;

[0135] The generation of the energy disturbance feature set specifically includes: extracting the values ​​of the energy contribution inversion results over all sampling time periods to construct an energy distribution time series; calculating the energy distribution change amplitude between adjacent sampling time periods to form an energy disturbance change sequence; performing amplitude analysis on the energy disturbance change sequence to extract the dominant change component with the largest change amplitude as the main characteristic component of the energy disturbance; calculating the average change amplitude of this main characteristic component over all sampling time periods to obtain the energy disturbance intensity value; then statistically analyzing the proportion of the main characteristic component whose change direction remains consistent over all sampling time periods to obtain the energy disturbance stability value; calculating the proportion of the change amplitude in the energy disturbance change sequence that exceeds the average level to obtain the energy disturbance volatility value; and combining the energy disturbance intensity value, energy disturbance stability value, and energy disturbance volatility value to form the energy disturbance feature set of the anomaly monitoring signal channel, which is used to characterize the amplitude characteristics, duration characteristics, and volatility characteristics of the energy disturbance.

[0136] The monitoring and diagnostic results are generated based on the energy disturbance feature set. Specifically, for each abnormal monitoring signal channel, the energy disturbance intensity value, energy disturbance stability value, and energy disturbance volatility value are extracted from its energy disturbance feature set. The energy disturbance intensity value is compared with a preset energy disturbance judgment threshold. If the energy disturbance intensity value is greater than the threshold, the channel is determined to be in an abnormal energy disturbance state; if it is less than the threshold, it is determined to be in a normal state. Subsequently, for channels in an abnormal energy disturbance state, their disturbance comprehensive index is calculated. The disturbance comprehensive index is determined by the weighted average of the energy disturbance intensity value and the energy disturbance volatility value, and is used to reflect the overall significance of the energy disturbance. The risk level is divided according to the size range of the disturbance comprehensive index. When the disturbance comprehensive index is in the low range, it is determined to be a slight risk; when it is in the medium range, it is determined to be a medium risk; and when it is in the high range, it is determined to be a serious risk. The monitoring and diagnostic results are generated, including the abnormal monitoring signal channel number, energy disturbance intensity value, disturbance comprehensive index, and risk level, which are used to reflect the energy disturbance characteristics and risk level of the intelligent switch cabinet system in the current operating state.

[0137] A data analysis-based intelligent switchgear safety monitoring system includes:

[0138] The data acquisition and preprocessing module is used to acquire multimodal monitoring signal data during the operation of the intelligent switchgear and generate standardized monitoring signal sequences;

[0139] The feature extraction module is used to extract the temporal fluctuation features, frequency domain energy features, and phase consistency features of each monitoring signal channel to form a multimodal monitoring feature matrix;

[0140] The entropy spectrum construction module is used to calculate the information entropy value of the monitoring signal channel and map it to a unified entropy spectrum space to generate a multimodal entropy spectrum sequence.

[0141] The energy coupling analysis module is used to calculate the cross-modal energy coupling degree and generate the energy coupling matrix.

[0142] The energy stability monitoring module is used to calculate the entropy change rate of the channel pair and generate energy stability monitoring curves.

[0143] Anomaly identification module is used to locate the anomaly monitoring signal channel and generate component identification when the energy stability monitoring curve shows a downward trend;

[0144] The energy inversion and diagnosis module is used to perform energy inversion and eigenvalue decomposition based on the entropy contribution rate and generate monitoring and diagnosis results.

[0145] The monitoring center sending module is used to send abnormal monitoring signal channels, corresponding component identifiers, and monitoring diagnostic results to the monitoring center.

[0146] Example 1:

[0147] To verify the feasibility of this invention in practice, it was applied to a safety monitoring scenario of a 10kV intelligent switchgear operating in a high-humidity and high-salt environment in a coastal area. This region has complex climatic conditions, with high air humidity and salt spray concentration. Long-term operation can easily lead to oxidation of internal contact parts, enhanced partial discharge, and abnormal temperature rise in the switchgear. Traditional monitoring systems mainly rely on temperature and current thresholds for alarms, which cannot identify potential risks from energy disturbances, often resulting in delayed alarms, misjudgments, and missed alarms, making it difficult to detect hidden faults in a timely manner. This invention addresses this problem by constructing a multimodal monitoring system. It synchronously collects data from multiple sources, including temperature, current, voltage, partial discharge, infrared thermal imaging, humidity, and gas concentration, forming a multimodal monitoring signal sequence. The system automatically performs preprocessing operations such as anomaly removal, time synchronization, noise removal, and amplitude normalization to ensure the stability and comparability of the data input.

[0148] In this operational scenario, the system extracts and fuses the temporal fluctuations, frequency domain energy, and phase consistency features of each channel based on a multimodal monitoring feature matrix, forming a unified entropy spectrum space to achieve energy scale alignment between different physical signals. Subsequently, by calculating the entropy difference between adjacent time periods and the cross-modal energy coupling degree, an energy coupling matrix is ​​generated to dynamically quantify the energy interaction intensity between different monitoring signals. During continuous operation, the system tracks the changes in the energy coupling matrix in real time and generates an energy stability monitoring curve to reflect the stability and fluctuation trend of energy transfer in the equipment. When the monitoring curve shows a continuous downward trend, the system automatically determines that there is a hidden energy disturbance state and locates the corresponding abnormal monitoring signal channel according to the energy anomaly impact index. Combined with the energy coupling relationship mapping table, the system further identifies the physical component location corresponding to the abnormal channel, generates a unique component identifier, and achieves accurate location of hidden faults.

[0149] Based on this, the system performs entropy feature inversion according to the entropy contribution rate, performs energy contribution inversion and feature decomposition on the abnormal monitoring signal channel, extracts multi-dimensional indicators such as energy disturbance intensity, stability and volatility, constructs an energy disturbance feature set, and can automatically determine the abnormality level by comparing it with a preset threshold and output monitoring and diagnostic results.

[0150] To verify the performance of the present invention, it was compared with traditional security monitoring methods. The comparison results are shown in Table 1.

[0151] Table 1. Performance Comparison of the Invention Method and Traditional Security Monitoring Methods

[0152] Comparison Indicators | Improvement of Traditional Method | Fault Identification Accuracy (%) | 84.6 | 92.8 | +8.2 | Latent Fault Detection Rate (%) | 61.3 | 83.4 | +2 | 2.1 | False Alarm Rate (%) | 11.5 | 3.2 | -8.3 | Fault Location Accuracy (Component Level) | 78.5 | 95.7 | +17.2 | Average Response Latency (seconds) | 4.8 | 2.1 | -2.7 | Data Fusion Stability Index (0-1) | 0.7 | 60.9 | 4 | +0.18 | Abnormal Energy Disturbance Identification Sensitivity (%) | 68.2 | 91.5 | +2 | 3.3 | Number of False Shutdowns During Operation (times / month) | 51 | -4 surface

[0153] As shown in Table 1, the data analysis-based intelligent switchgear safety monitoring method proposed in this invention outperforms the traditional temperature and current threshold method in all key performance indicators. The traditional method relies solely on single-mode features for judgment, resulting in low sensitivity to latent faults and coupled energy disturbances, leading to a latent fault detection rate of only 61.3% and a high false alarm rate of 11.5%. In contrast, this invention utilizes multi-modal signal collaborative analysis and unified entropy spectrum spatial mapping technology to perform energy coupling modeling of multi-source signals such as temperature, current, voltage, partial discharge, and humidity, achieving a unified expression of cross-modal energy features. This improves the fault identification accuracy to 92.8% and the latent fault detection rate to 83.4%.

[0154] In terms of response speed, the traditional threshold method requires multiple levels of judgment to trigger the alarm, with an average response time of 4.8 seconds. However, this invention can trigger the judgment in the early stage of energy stability decline by calculating the energy coupling matrix and entropy change rate curve in real time, shortening the response delay to 2.1 seconds and improving the real-time performance and early warning capability of the system. In terms of fault location accuracy, this invention performs component-level mapping based on the channel energy coupling relationship, avoiding the error of relying on manual inspection in traditional methods, and improving the location accuracy by about 17%.

[0155] Furthermore, in the long-term stability comparison, the present invention achieved a data fusion stability index of 0.94, indicating that the fluctuation of the multimodal monitoring feature fusion process is smaller and the energy feature expression is more stable, thereby reducing the number of false shutdowns, from an average of 5 times per month to 1 time. This improvement is due to the energy disturbance feature set and entropy energy matching mechanism introduced by the system, which enables the system to adaptively identify and dynamically correct potential energy fluctuations.

[0156] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A data analysis-based intelligent switchgear safety monitoring method, characterized in that, The process includes the following steps: collecting multimodal monitoring signal data during the operation of the intelligent switchgear, performing preprocessing to generate a standardized multimodal monitoring signal sequence; performing feature extraction on the standardized multimodal monitoring signal sequence, calculating the time-series fluctuation characteristics, frequency domain energy characteristics, and phase consistency characteristics of each monitoring signal channel, and performing time alignment and feature normalization to form a multimodal monitoring feature matrix; Based on the multimodal monitoring feature matrix, the information entropy value is calculated for the local time window of each monitoring signal channel and mapped to a unified entropy spectrum space to generate a multimodal entropy spectrum sequence. Calculate the entropy difference sequence between adjacent time points based on the multimodal entropy spectrum sequence, and calculate the cross-modal energy coupling degree to generate the energy coupling matrix; Perform continuous-time tracking on the energy coupling matrix, calculate the entropy change rate of each channel pair, and generate energy stability monitoring curves; When the energy stability monitoring curve shows a downward trend over a continuous period of time, the intelligent switchgear is determined to be in a state of latent energy disturbance. The abnormal monitoring signal channel is located and the corresponding component identifier is generated. Based on the entropy contribution rate of each monitoring signal channel, entropy feature inversion is performed. Energy contribution inversion and feature decomposition are performed on the abnormal monitoring signal channel to generate monitoring and diagnostic results. The abnormal monitoring signal channel, the corresponding component identifier, and the monitoring and diagnostic results are sent to the monitoring center.

2. The intelligent switchgear safety monitoring method based on data analysis according to claim 1, characterized in that, The multimodal monitoring signal data includes temperature signal, current signal, voltage signal, partial discharge signal, infrared thermal imaging signal, humidity signal, and gas concentration signal. The preprocessing includes anomaly removal, time synchronization, noise removal, and amplitude normalization.

3. The intelligent switchgear safety monitoring method based on data analysis according to claim 1, characterized in that, The generation of the multimodal monitoring feature matrix specifically includes: initializing feature extraction for the standardized multimodal monitoring signal sequence; segmenting the time series of each multimodal monitoring signal channel into sliding windows to construct a time window set; calculating the time-series fluctuation features of the multimodal monitoring signal data within each time window, wherein the time-series fluctuation features include time-series mean features, time-series variance features, and time-series rate of change features; performing frequency domain energy feature extraction on the standardized multimodal monitoring signal sequence, using the Fast Fourier Transform algorithm to perform spectral decomposition on the time-series data of each multimodal monitoring signal channel to obtain frequency domain energy distribution features, and extracting the dominant frequency amplitude features, The system calculates the spectral centroid and band power ratio characteristics; it also calculates the phase consistency characteristics for each multimodal monitoring signal channel; it sequentially concatenates the time-series mean, time-series variance, time-series rate of change, dominant frequency amplitude, spectral centroid, band power ratio, and phase consistency characteristics of each multimodal monitoring signal channel to form a comprehensive feature vector for that channel; it then arranges the comprehensive feature vectors of all multimodal monitoring signal channels in order to form a multimodal monitoring feature matrix; it performs time alignment on the multimodal monitoring feature matrix; and finally, it performs feature normalization on the time-aligned multimodal monitoring feature matrix to form a multimodal monitoring feature matrix.

4. The intelligent switchgear safety monitoring method based on data analysis according to claim 3, characterized in that, The time-series mean feature is the average value of all monitored signal values ​​within the time window; the time-series variance feature is the average deviation of the monitored signal from the mean within the time window; and the time-series rate of change feature is the rate of change obtained by dividing the difference in amplitude of the monitored signal between the start and end sampling points of the time window by the window length.

5. The intelligent switchgear safety monitoring method based on data analysis according to claim 1, characterized in that, The generation of the multimodal entropy spectrum sequence specifically includes: performing entropy spectrum mapping based on the multimodal monitoring feature matrix; using the data of each multimodal monitoring signal channel in the multimodal monitoring feature matrix as an independent input channel; constructing a time series structure according to the sampling time index; segmenting the time series of each multimodal monitoring signal channel according to a fixed time window to form multiple local time windows; statistically analyzing the occurrence ratio of multimodal monitoring feature values ​​within each local time window, calculating the distribution of each feature value, and obtaining the corresponding feature probability distribution; calculating the corresponding information entropy value based on the feature probability distribution of the local time window, thus obtaining the information entropy value of each multimodal monitoring signal channel in different time windows. The local information entropy sequence within the multimodal monitoring signal channels is standardized, and the information entropy values ​​between different channels are numerically normalized. The normalized information entropy sequence is then smoothed by interpolation in the time dimension to form a continuous and smooth information entropy time-series distribution. The smooth information entropy time-series distribution of each multimodal monitoring signal channel is mapped to a unified entropy spectrum space. Within the unified entropy spectrum space, the information entropy values ​​of different channels are time-aligned and energy scales are unified to establish a cross-channel energy distribution correspondence. Based on the time-alignment results in the unified entropy spectrum space, the entropy spectrum information of all multimodal monitoring signal channels is aggregated to generate a multimodal entropy spectrum sequence.

6. The intelligent switchgear safety monitoring method based on data analysis according to claim 1, characterized in that, The generation of the energy coupling matrix specifically includes: performing energy coupling analysis based on the multimodal entropy spectrum sequence, using the entropy spectrum values ​​of each multimodal monitoring signal channel at each sampling time as input data; selecting any two adjacent sampling times in the multimodal entropy spectrum sequence, calculating the information entropy change of each multimodal monitoring signal channel between adjacent sampling times; arranging the information entropy changes of all multimodal monitoring signal channels in channel number order to form an entropy difference sequence matrix; calculating the cross-modal energy coupling degree between any two multimodal monitoring signal channels based on the entropy difference sequence matrix; and matrixing the cross-modal energy coupling degrees between all multimodal monitoring signal channels according to channel combination method to establish the energy coupling matrix.

7. The intelligent switchgear safety monitoring method based on data analysis according to claim 1, characterized in that, The generation of the energy stability monitoring curve specifically includes: calculating the energy coupling matrix based on the entropy difference sequence of all multimodal monitoring signal channels within a preset continuous sampling time period; dynamically updating the energy coupling matrix according to the time sequence using the time period as a sliding window to construct an energy coupling matrix sequence; extracting the energy coupling degree corresponding to any pair of multimodal monitoring signal channels in all sliding time periods within the energy coupling matrix sequence to form a channel pair energy coupling degree time series; performing continuous time difference processing on the channel pair energy coupling degree time series to calculate the change in energy coupling degree between adjacent sliding time periods to obtain a channel pair energy change sequence; performing sliding time window smoothing and averaging operations on the channel pair energy change sequence to calculate the average value of the channel pair energy change amplitude to obtain the entropy change rate of the channel pair; summing up the entropy change rates of all channel pairs within the same sliding time period to obtain the average entropy change rate of that time period; arranging the average entropy change rates of each time period according to the sliding time period sequence to generate an average entropy change rate time series; and plotting the energy stability monitoring curve with the center time of the sliding time period as the horizontal axis and the corresponding average entropy change rate as the vertical axis.

8. The intelligent switchgear safety monitoring method based on data analysis according to claim 1, characterized in that, The generation of the corresponding component identifier specifically includes: performing trend analysis on the energy stability monitoring curve, calculating the average slope of the sliding sampling time period of the energy stability monitoring curve within a continuous sampling time period, and determining that the intelligent switch cabinet is in a latent energy disturbance state when the average slope is lower than a preset descent threshold for several consecutive sampling time periods; after determining the latent energy disturbance state, calculating the energy anomaly impact index of each multimodal monitoring signal channel based on the energy coupling degree change rate of each multimodal monitoring signal channel pair in the energy coupling matrix sequence; sorting all multimodal monitoring signal channels in descending order according to the energy anomaly impact index, and selecting the multimodal monitoring signal channel with the highest energy anomaly impact index as the anomaly monitoring signal channel; determining the physical component location corresponding to the anomaly monitoring signal channel based on the channel number of the anomaly monitoring signal channel in the energy coupling matrix and its energy coupling relationship with other channel pairs, and generating a component identifier corresponding to the physical component.

9. The intelligent switchgear safety monitoring method based on data analysis according to claim 1, characterized in that, The generation of the monitoring and diagnostic results specifically includes: based on the channel number of the abnormal monitoring signal channel in the multimodal monitoring feature matrix, extracting the standardized multimodal monitoring signal sequence of the abnormal monitoring signal channel within a continuous sampling time period, and calculating the information entropy value sequence corresponding to each sampling time period; within the same sampling time period, calculating the overall information entropy distribution based on the information entropy value sequence of all multimodal monitoring signal channels, determining the entropy contribution rate of each multimodal monitoring signal channel in the overall information entropy distribution, and constructing an entropy contribution rate vector; based on the entropy contribution rate of the abnormal monitoring signal channel, performing entropy feature inversion from the entropy contribution rate vector space to recover the energy distribution characteristics of the abnormal monitoring signal channel in the energy coupling structure, and generating the energy contribution inversion result of the abnormal monitoring signal channel; performing feature decomposition on the energy contribution inversion result, extracting the main feature components characterizing the energy disturbance properties, forming the energy disturbance feature set of the abnormal monitoring signal channel; and generating the monitoring and diagnostic results based on the energy disturbance feature set.

10. A data analysis-based intelligent switchgear safety monitoring system, executing the data analysis-based intelligent switchgear safety monitoring method according to any one of claims 1 to 9, characterized in that, include: The data acquisition and preprocessing module is used to acquire multimodal monitoring signal data during the operation of the intelligent switchgear and generate standardized monitoring signal sequences; The feature extraction module is used to extract the temporal fluctuation features, frequency domain energy features, and phase consistency features of each monitoring signal channel to form a multimodal monitoring feature matrix; The entropy spectrum construction module is used to calculate the information entropy value of the monitoring signal channel and map it to a unified entropy spectrum space to generate a multimodal entropy spectrum sequence. The energy coupling analysis module is used to calculate the cross-modal energy coupling degree and generate the energy coupling matrix. The energy stability monitoring module is used to calculate the entropy change rate of the channel pair and generate energy stability monitoring curves. Anomaly identification module is used to locate the anomaly monitoring signal channel and generate component identification when the energy stability monitoring curve shows a downward trend; The energy inversion and diagnosis module is used to perform energy inversion and eigenvalue decomposition based on the entropy contribution rate to generate monitoring and diagnosis results; the monitoring center sending module is used to send the abnormal monitoring signal channel, the corresponding component identifier, and the monitoring and diagnosis results to the monitoring center.

Images