Power transformation line alarm device based on real-time monitoring

Abstract

The invention relates to the technical field of power distribution lines, and particularly discloses a power transformation line alarm device based on real-time monitoring, which is characterized in that a current signal and a voltage signal of a power distribution feed-out loop are acquired in real time, and after signal denoising processing is performed on the current signal and the voltage signal, a spectrogram of the current signal and a spectrogram of the voltage signal are acquired through fast Fourier transform; furthermore, a data processing algorithm based on deep learning is introduced to learn frequency domain characteristic modes of the current signal and the voltage signal, so that characteristic phase search alignment and joint sensing are performed on the current signal and the voltage signal; the intelligent monitoring of whether the distribution line has a private change behavior is realized based on the frequency domain correlation characteristic between the current signal and the voltage signal. Through the mode, the characteristic difference between natural load change and artificial change can be effectively distinguished, the detection reliability of privately-changed distribution lines is improved, and false alarms are reduced.

Description

Technical Field

[0001] The present application relates to the technical field of distribution lines, and more specifically, to a substation line alarm device based on real-time monitoring. Background Art

[0002] With the expansion of the scale of the power system and the improvement of the intelligentization demand, the safe and stable operation of distribution lines has become an important issue in the field of power management. Among them, privately modifying distribution lines (such as illegal wiring and unauthorized line modification) is one of the main reasons for power loss, equipment failure, and even safety accidents. Traditional monitoring methods usually rely on manual inspections or simple current threshold alarms, which have problems such as lagging response, high false alarm rate, and difficulty in accurately identifying hidden modification behaviors.

[0003] In recent years, intelligent monitoring technologies based on data analysis have gradually become a research hotspot. For example, the invention patent with the publication number CN117833479A proposes a monitoring and alarm device for unauthorized modification of distribution lines. By collecting the current data of each distribution feed circuit, the data before the detection moment is composed of the first sample set, and the data after that is composed of the second sample set. Jump current data sets are extracted from the two sample sets respectively. Based on the number of different samples in the two jump current data sets, it is judged whether there is unauthorized wiring or line modification through threshold comparison, so as to realize the monitoring of unauthorized modification behaviors of distribution lines.

[0004] However, in the prior art, the detection logic for unauthorized wiring behaviors depends on the number of different samples of current jumps. Both normal load fluctuations (such as equipment startup and shutdown) and unauthorized line modification behaviors can cause current jumps. The traditional static threshold detection method lacks adaptability, is easily affected by random noise and generates false alarms, and cannot distinguish the characteristic differences between natural load changes and human modifications. For example, frequent startup and shutdown of factory equipment may cause the number of different samples to exceed the standard and trigger false alarms.

[0005] Therefore, an optimized substation line alarm device based on real-time monitoring is expected. Summary of the Invention

[0006] To solve the above technical problems, the present application is proposed. An embodiment of the present application provides a substation line alarm device based on real-time monitoring, which collects current signals and voltage signals of a distribution feed-out loop in real time, performs signal denoising processing on both, obtains spectrograms of the current signal and the voltage signal through fast Fourier transform, and further introduces a data processing algorithm based on deep learning to learn the frequency-domain feature patterns of the current signal and the voltage signal. Then, by performing feature phase search alignment and joint perception on both, intelligent monitoring of whether there is unauthorized modification behavior in the distribution line is realized based on the frequency-domain correlation characteristics between the current signal and the voltage signal. In this way, the characteristic differences between natural load changes and human modifications can be effectively distinguished, the detection reliability of unauthorized modification of the distribution line can be improved, and false alarms can be reduced.

[0007] Correspondingly, according to one aspect of the present application, there is provided a substation line alarm device based on real-time monitoring, which includes:

[0008] A distribution line signal monitoring module for collecting current signals and voltage signals of a distribution feed-out loop in real time;

[0009] A signal spectrum analysis module for performing spectrum feature analysis on the current signal and the voltage signal to obtain a current signal frequency-domain feature map and a voltage signal frequency-domain feature map;

[0010] A joint perception module for performing feature joint perception on the current signal frequency-domain feature map and the voltage signal frequency-domain feature map to obtain a current-voltage signal frequency-domain pattern joint perception feature map, wherein performing feature joint perception on the current signal frequency-domain feature map and the voltage signal frequency-domain feature map includes: performing feature phase dynamic search alignment and joint perception on the current signal frequency-domain feature map and the voltage signal frequency-domain feature map to obtain the current-voltage signal frequency-domain pattern joint perception feature map;

[0011] A line modification alarm module for determining whether to send a line modification alarm message to the management personnel based on the current-voltage signal frequency-domain pattern joint perception feature map.

[0012] Compared with the prior art, the substation line alarm device based on real-time monitoring provided by the present application collects the current signal and voltage signal of the distribution feed-out loop in real time. After denoising the two signals, the spectrograms of the current signal and voltage signal are obtained through fast Fourier transform. Furthermore, a data processing algorithm based on deep learning is introduced to learn the frequency domain feature patterns of the current signal and voltage signal. Then, through feature phase search alignment and joint perception of the two, intelligent monitoring of whether there is any unauthorized modification behavior in the distribution line is realized based on the frequency domain correlation characteristics between the current signal and voltage signal. In this way, the characteristic differences between natural load changes and human modifications can be effectively distinguished, the detection reliability of unauthorized modification of the distribution line can be improved, and false alarms can be reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other objects, features, and advantages of the present application will become more obvious. The accompanying drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application, and do not constitute a limitation to the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0014] Figure 1 It is a block diagram of the substation line alarm device based on real-time monitoring according to an embodiment of the present application.

[0015] Figure 2 It is a schematic diagram of the data flow of the substation line alarm device based on real-time monitoring according to an embodiment of the present application.

[0016] Figure 3 It is a block diagram of the signal spectrum analysis module in the substation line alarm device based on real-time monitoring according to an embodiment of the present application.

[0017] Figure 4 It is a block diagram of the joint perception module in the substation line alarm device based on real-time monitoring according to an embodiment of the present application.

[0018] Figure 5 It is a block diagram of the feature dynamic search alignment unit in the substation line alarm device based on real-time monitoring according to an embodiment of the present application.

[0019] Figure 6 It is a block diagram of the feature joint perception aggregation unit in the substation line alarm device based on real-time monitoring according to an embodiment of the present application.

[0020] Figure 7 It is a block diagram of the line modification alarm module in the substation line alarm device based on real-time monitoring according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] As shown in this application and the claims, unless the context clearly indicates otherwise, words such as "a", "an", "one" and / or "the" are not specifically singular and may also include the plural. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0022] Although this application makes various references to certain modules in the system according to the embodiments of this application, however, any number of different modules can be used and run on the user terminal and / or server. The modules are merely illustrative, and different aspects of the system and method can use different modules.

[0023] Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of this application. It should be understood that the operations before or below are not necessarily executed precisely in order. On the contrary, according to needs, various steps can be executed in reverse order or simultaneously. At the same time, other operations can also be added to these processes, or one or more steps can be removed from these processes.

[0024] Next, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments of this application. It should be understood that this application is not limited by the exemplary embodiments described here.

[0025] It is worth noting that in this application, all actions of obtaining data are carried out on the premise of complying with the corresponding data protection regulations and policies of the country where it is located and obtaining the authorization given by the owner of the corresponding device.

[0026] As described in the above background art, Patent CN117833479A proposed a monitoring and alarm device for unauthorized modification of distribution lines. By collecting the current data of each distribution feed circuit, the data before the detection moment is composed of the first sample set, and the data after that is composed of the second sample set. The jump current data sets are extracted from the two sample sets respectively. Based on the number of different samples in the two jump current data sets, it is judged whether there is unauthorized wiring or line modification through threshold comparison, so as to realize the monitoring of unauthorized modification of distribution lines. In the prior art, the detection of unauthorized wiring depends on the number of different samples of current jumps, but both normal load fluctuations (such as equipment startup and shutdown) and unauthorized line modification will cause current jumps. The traditional static threshold detection lacks adaptability, is easily affected by noise and generates false alarms, and cannot distinguish the characteristics of natural load changes and human modifications. To solve the above technical problems, the present application proposes an optimized substation line alarm device based on real-time monitoring. It collects the current signal and voltage signal of the distribution feed circuit in real time, performs signal denoising processing on the two, and obtains the spectrograms of the current signal and voltage signal through fast Fourier transform. Furthermore, a data processing algorithm based on deep learning is introduced to learn the frequency domain feature patterns of the current signal and voltage signal, and then through feature phase search alignment and joint perception of the two, intelligent monitoring of whether there is unauthorized modification of the distribution line is realized based on the frequency domain correlation characteristics between the current signal and voltage signal. In this way, the characteristic differences between natural load changes and human modifications can be effectively distinguished, the detection reliability of unauthorized modification of distribution lines can be improved, and false alarms can be reduced.

[0027] Figure 1 It is a block diagram of a substation line alarm device based on real-time monitoring according to an embodiment of the present application. Figure 2 It is a schematic diagram of data flow of a substation line alarm device based on real-time monitoring according to an embodiment of the present application. As Figure 1 and Figure 2As shown, the substation line alarm device 100 based on real-time monitoring includes: a distribution line signal monitoring module 110, which is used to collect the current signal and voltage signal of the distribution feeder circuit in real time; a signal spectrum analysis module 120, which is used to perform spectrum feature analysis on the current signal and the voltage signal to obtain a current signal frequency domain feature diagram and a voltage signal frequency domain feature diagram; a joint perception module 130, which is used to perform feature joint perception on the current signal frequency domain feature diagram and the voltage signal frequency domain feature diagram to obtain a current-voltage signal frequency domain mode joint perception feature diagram, wherein the feature joint perception on the current signal frequency domain feature diagram and the voltage signal frequency domain feature diagram includes: performing feature phase dynamic search alignment and joint perception on the current signal frequency domain feature diagram and the voltage signal frequency domain feature diagram to obtain the current-voltage signal frequency domain mode joint perception feature diagram; a line change alarm module 140, which is used to determine whether to send a line change alarm message to a management personnel based on the current-voltage signal frequency domain mode joint perception feature diagram.

[0028] In the above-mentioned substation alarm device based on real-time monitoring, the distribution line signal monitoring module 110 is used to collect the current signal and voltage signal of the distribution feeder circuit in real time. It should be understood that unauthorized modification of the distribution line (such as illegal wiring) will change the current and voltage characteristics of the line at the same time. The present application can use the physical coupling characteristics of current and voltage in the power system by synchronously collecting the current signal and voltage signal of the distribution feeder circuit to provide multi-dimensional complementary information for the detection of unauthorized line modification behavior. Specifically, although the current signal can directly reflect the change of the line load, when relying solely on the current data, normal load fluctuations (such as the start and stop or periodic operation of large equipment) and illegal modifications (such as unauthorized wiring) may cause similar current jumps or harmonic distortions, making it difficult for the system to distinguish the essential differences between the two. For example, when a motor in a factory starts, the instantaneous surge in current may be misjudged as an abnormal behavior of illegal wiring, and the change pattern of the voltage signal is often different from that of the current. Under normal load fluctuations, the voltage may remain relatively stable due to the grid regulation mechanism (such as reactive power compensation), while illegal changes (such as asymmetric wiring or ground faults) may directly destroy the symmetry of the voltage waveform, causing fundamental phase shift or abnormal enhancement of specific harmonics. Therefore, this application can more effectively identify whether the distribution line has been privately modified by combining current and voltage signals for joint analysis.

[0029] During the specific operation process, it is necessary to deploy high-precision sensors to measure the current signal and voltage signal respectively. For the current signal, Hall effect sensors or Rogowski coils are usually used for measurement. Based on the Hall effect principle, Hall effect sensors can measure current without contacting the circuit. This non-invasive characteristic makes their installation more convenient and avoids interfering with the original circuit structure. Rogowski coils, on the other hand, are widely used in the measurement of alternating current due to their small size, light weight, fast response speed, etc., and are especially suitable for capturing high-frequency current signals. Both of these sensors can effectively convert current information into electrical signals for further processing.

[0030] For the acquisition of voltage signals, voltage dividers or voltage transformers can be used. Voltage dividers convert high voltage into low voltage signals through a series resistance network and are suitable for the measurement of DC and low-frequency AC voltages; while voltage transformers are devices based on the principle of electromagnetic induction that can safely reduce the high voltage in the power grid to a standard low voltage level and are especially suitable for the measurement of high-voltage AC voltages. Whichever method is chosen, it is required that the devices used have good linearity and stability to ensure the authenticity and accuracy of the measurement results.

[0031] After completing the hardware selection, the next step is to consider how to arrange these sensors to achieve the best acquisition effect. Considering the inherent correlation between the current and voltage signals, the synchronization between the two should be maintained as much as possible during the layout. This means that when choosing the installation location, it is necessary to ensure that the current sensor and voltage sensor are as close as possible so as to obtain the current and voltage data at the same moment. In addition, it should also be noted that due to the complexity and variability of the power system, environmental factors such as temperature, humidity, and electromagnetic interference will all affect signal acquisition. Therefore, when deploying sensors, necessary protection measures need to be taken, such as using shielded cables to reduce external electromagnetic field interference, or setting up temperature control devices around the sensors to maintain a stable operating temperature range.

[0032] To further improve the quality of data acquisition, digital technology can also be introduced. For example, a digital signal processor (DSP) is used to quickly sample and quantize the received analog signal. In this way, not only can the resolution of the signal be greatly improved, but the influence of noise can also be effectively reduced. Specifically, the DSP chip integrates a high-speed ADC (analog-to-digital converter) inside, which can convert continuous analog signals into discrete digital signals, facilitating subsequent data processing. Moreover, with the help of advanced filtering algorithms, the false components caused by external interference can be removed without affecting the characteristics of the original signal, ensuring that the final obtained current and voltage signals are pure and reliable.

[0033] In addition, during the entire acquisition process, time synchronization is also a crucial aspect that cannot be overlooked. Since it is necessary to monitor the current and voltage information at multiple points simultaneously, if there are deviations in the time bases among the various sensors, even if the data itself is highly accurate, it will be difficult to achieve effective joint analysis. Therefore, a GPS timing module can be used to provide a unified time reference for all devices participating in data acquisition. The advantage of this approach is that no matter how far apart the sensors are, they can record their respective observed values based on the same time stamp, thereby laying a solid foundation for subsequent feature extraction and pattern recognition.

[0034] During the entire acquisition stage, attention should also be paid to the issues of data storage and transmission. On the one hand, as the monitoring density increases, the amount of data generated will also expand, which requires an efficient data management system to be responsible for the real-time storage and backup of data. On the other hand, considering the limitations of the on-site environment, sometimes it may not be possible to directly process the acquired information on-site. In this case, wireless communication technology needs to be used to transmit the data to a remote server in a timely manner. Whether it is Wi-Fi, ZigBee, or 4G / 5G networks, they can be flexibly selected according to actual needs to ensure the security and reliability of data transmission.

[0035] In the above-mentioned substation line alarm device based on real-time monitoring, the signal spectrum analysis module 120 is used to perform spectrum feature analysis on the current signal and the voltage signal to obtain a current signal frequency-domain feature map and a voltage signal frequency-domain feature map. Among them, Figure 3 is a block diagram of the signal spectrum analysis module in the substation line alarm device based on real-time monitoring according to an embodiment of the present application. As Figure 3 shown, the signal spectrum analysis module 120 includes: a signal denoising unit 121, which is used to perform signal denoising processing based on wavelet transform on the current signal and the voltage signal to obtain a filtered current signal and a filtered voltage signal; a Fourier transform unit 122, which is used to perform fast Fourier transform on the filtered current signal and the filtered voltage signal to obtain a current signal spectrum map and a voltage signal spectrum map; a signal spectrum feature extraction unit 123, which is used to extract signal spectrum features from the current signal spectrum map and the voltage signal spectrum map respectively to obtain the current signal frequency-domain feature map and the voltage signal frequency-domain feature map.

[0036] Specifically, the signal denoising unit 121 is configured to perform signal denoising processing based on wavelet transform on the current signal and the voltage signal to obtain a filtered current signal and a filtered voltage signal. It should be understood that since the distribution line is vulnerable to environmental noise (such as electromagnetic interference and instantaneous pulses), random noise is mixed into the current and voltage signals, which in turn affects the accuracy of subsequent identification of unauthorized wiring changes. Therefore, the present application further uses the wavelet transform method to denoise the current signal and the voltage signal. Specifically, wavelet transform is a time-frequency analysis method that can decompose a signal into sub-band signals of different frequencies. By selecting appropriate wavelet basis functions and thresholds, the wavelet coefficients of the signal are processed. For the wavelet coefficients corresponding to noise, their amplitudes are usually small, and they are suppressed by setting a threshold; while the amplitudes of the effective coefficients of the signal are large and are retained. Then, through inverse wavelet transform, the processed wavelet coefficients are reconstructed into a filtered signal, which can effectively remove the noise interference in the signal and improve the purity of the signal. Moreover, traditional filtering methods (such as mean filtering) may blur the characteristics of the real signal, while wavelet transform can adaptively separate noise from the effective signal and retain high-frequency mutation details (such as the instantaneous impact of illegal wiring), providing a reliable data basis for subsequent signal analysis.

[0037] Specifically, the Fourier transform unit 122 is configured to perform fast Fourier transform on the filtered current signal and the filtered voltage signal to obtain a current signal spectrogram and a voltage signal spectrogram. It should be understood that the present application considers that it is difficult to directly distinguish normal load fluctuations from illegal changes in the time domain signal (both may cause current jumps), while frequency domain analysis can reveal deep features such as harmonic distortion and phase shift, which is more helpful for accurately identifying whether the distribution line has been unauthorizedly changed. Therefore, the present application uses the fast Fourier transform (FFT) method to transform the filtered current signal and the filtered voltage signal from the time domain to the frequency domain, respectively obtaining a current signal spectrogram and a voltage signal spectrogram. Specifically, the fast Fourier transform (FFT) is an efficient discrete Fourier transform (DFT) algorithm that reveals the spectral characteristics of a signal by decomposing the original signal into multiple sine wave components of different frequencies, quickly calculating the amplitude and phase information of the signal at different frequencies, and then plotting a spectrogram containing the amplitude-frequency distribution. In the current signal spectrogram and the voltage signal spectrogram, the intensity distribution of each frequency component in the signal is intuitively shown, which helps to observe the frequency characteristic changes caused by unauthorized changes in the distribution line, such as abnormal increases in harmonic components and shifts in the fundamental frequency.

[0038] Specifically, in a specific example of the present application, the signal spectrum feature extraction unit 123 is configured to: perform spectrum feature extraction on the current signal spectrogram and the voltage signal spectrogram respectively based on the ASPP model to obtain the current signal frequency domain feature map and the voltage signal frequency domain feature map. It should be understood that considering the limited ability of the traditional CNN network to mine signal features, it is difficult to simultaneously capture the local harmonic details and the global spectrum morphology in the signal spectrogram. Therefore, the present application further adopts the ASPP (Atrous Spatial Pyramid Pooling) model to perform spectrum feature extraction on the current signal spectrogram and the voltage signal spectrogram respectively. Specifically, the ASPP model increases the receptive field of the convolutional kernel without increasing the computational amount by introducing a dilation rate in the standard convolutional kernel, so as to be able to capture multi-scale context information. In addition, the ASPP model includes multiple parallel atrous convolutional layers with different dilation rates, and each atrous convolutional layer can capture spectrum features in different ranges, such as local harmonic peaks, spectrum envelope shapes, etc. Furthermore, by fusing multi-scale spectrum features to more comprehensively understand the complex patterns of current and voltage signals in the frequency domain, generating the current signal frequency domain feature map and the voltage signal frequency domain feature map, which helps to provide richer and more effective feature information for subsequent line change detection.

[0039] In the above-mentioned substation line alarm device based on real-time monitoring, the joint perception module 130 is used to perform feature joint perception on the current signal frequency-domain feature map and the voltage signal frequency-domain feature map to obtain a joint perception feature map of the current-voltage signal frequency-domain pattern. Among them, performing feature joint perception on the current signal frequency-domain feature map and the voltage signal frequency-domain feature map includes: performing feature phase dynamic search alignment and joint perception on the current signal frequency-domain feature map and the voltage signal frequency-domain feature map to obtain the joint perception feature map of the current-voltage signal frequency-domain pattern. Specifically, since illegal wiring will destroy the inherent coupling relationship between current and voltage in the frequency domain, resulting in abnormal phenomena such as phase out-of-step and frequency mismatch between the current signal and the voltage signal. In addition, the phase out-of-step of illegal wiring is usually accompanied by abnormal enhancement of specific harmonic groups or sudden changes in the spectral energy distribution. Although normal load fluctuations may cause phase changes, their harmonic characteristics remain stable. Based on this, in order to effectively capture the abnormal current-voltage frequency-domain pattern caused by unauthorized wiring changes, the present application proposes a feature joint perception method, which determines the best correspondence between the two by dynamically searching for local regions with similar phase and frequency characteristics in the current signal frequency-domain feature map and the voltage signal frequency-domain feature map, and then performs multi-dimensional interactive analysis on the two to capture the co-variation of the two in terms of harmonic distribution, energy ratio, and phase relationship, thereby revealing the change in the coupling relationship between the current signal and the voltage signal, generating a joint perception feature map of the current-voltage signal frequency-domain pattern, and providing a strong feature basis for the subsequent intelligent identification of unauthorized wiring changes. Among them, Figure 4 is a block diagram of the joint perception module in the substation line alarm device based on real-time monitoring according to an embodiment of the present application. As Figure 4 shown, the joint perception module 130 includes: a feature decoupling unit 131, which is used to perform feature decoupling on the current signal frequency-domain feature map and the voltage signal frequency-domain feature map to obtain a set of current signal local frequency-domain feature coding matrices and a set of voltage signal local frequency-domain feature coding matrices; a feature dynamic search alignment unit 132, which is used to perform feature dynamic search alignment on the set of current signal local frequency-domain feature coding matrices and the set of voltage signal local frequency-domain feature coding matrices to obtain a set of phase-aligned {current signal local frequency-domain feature coding matrix, voltage signal local frequency-domain feature coding matrix} feature pairs; a feature joint perception aggregation unit 133, which is used to perform feature joint perception aggregation based on the attention mechanism on the set of phase-aligned {current signal local frequency-domain feature coding matrix, voltage signal local frequency-domain feature coding matrix} feature pairs to obtain the joint perception feature map of the current-voltage signal frequency-domain pattern.

[0040] Specifically, the feature decoupling unit 131 is expressed by the formula:

[0041] Decouple(F1) = {M 11 , M 12 ,..., M 1i ,..., M 1n}

[0042] Decouple(F2) = {M 21 , M 22 ,..., M 2j ,..., M 2n}

[0043] Wherein, F1 represents the frequency-domain feature map of the current signal, F2 represents the frequency-domain feature map of the voltage signal, Decouple(·) represents feature decoupling, M 11 , M 12 , M 1i and M 1n respectively represent the 1st, 2nd, ith, and nth current signal local frequency-domain feature coding matrices in the set of current signal local frequency-domain feature coding matrices, n is the number of the current signal local frequency-domain feature coding matrices, M 21 , M 22 , M 2j and M 2n respectively represent the 1st, 2nd, jth, and nth voltage signal local frequency-domain feature coding matrices in the set of voltage signal local frequency-domain feature coding matrices.

[0044] That is, by feature decoupling, the global feature is decomposed into the set of current signal local frequency-domain feature coding matrices and the set of voltage signal local frequency-domain feature coding matrices, which can strip the originally aliased global semantics into distinguishable local feature units, so as to establish a more refined local alignment benchmark in the frequency-domain correlation analysis of current and voltage. Through this decoupling operation, the model can focus on local features, enhance the robustness of phase dynamic search alignment, so that each current signal local frequency-domain feature coding matrix and voltage signal local frequency-domain feature coding matrix only focuses on the feature distribution of a specific frequency band or time segment, avoiding the deformation sensitivity problem of the global feature caused by noise interference, and at the same time retaining the differential feature patterns of the current and voltage signals in the frequency-domain phase.

[0045] Figure 5 is the block diagram of the feature dynamic search alignment unit in the substation line alarm device based on real-time monitoring according to the embodiment of the present application. As Figure 5As shown, the feature dynamic search alignment unit 132 includes: a feature alignment degree calculation sub-unit 1321, configured to calculate the feature phase alignment degree between the i-th local frequency domain feature encoding matrix in the set of local frequency domain feature encoding matrices of the current signal and each local frequency domain feature encoding matrix in the set of local frequency domain feature encoding matrices of the voltage signal to obtain a set of current signal-voltage signal frequency domain feature phase alignment degrees; a feature pair construction sub-unit 1322, configured to extract the local frequency domain feature encoding matrix of the voltage signal corresponding to the maximum value in the set of current signal-voltage signal frequency domain feature phase alignment degrees, and form a phase alignment feature pair with the i-th local frequency domain feature encoding matrix of the current signal to obtain the phase alignment {local frequency domain feature encoding matrix of the current signal, local frequency domain feature encoding matrix of the voltage signal} feature pair.

[0046] More specifically, in a specific example of the present application, the feature alignment degree calculation sub-unit 1321 is configured to: calculate the product between the transpose matrix of the i-th local frequency domain feature encoding matrix of the current signal and the local frequency domain feature encoding matrix of the voltage signal to obtain a current signal-voltage signal local frequency domain feature joint perception encoding matrix; calculate the Frobenius norm of the current signal-voltage signal local frequency domain feature joint perception encoding matrix to obtain the current signal-voltage signal frequency domain feature phase alignment degree, which is expressed by the formula:

[0047]

[0048] where d P (·,·) represents the phase alignment degree measurement function, (·) T represents the transpose of the matrix, ‖·‖ F represents the Frobenius norm of the matrix.

[0049] That is, by performing a product operation on the transpose matrix of the local frequency domain feature encoding matrix of the current signal and the local frequency domain feature encoding matrix of the voltage signal, the cross-correlation between the two in the frequency domain feature space is extracted, and this correlation reflects the energy distribution synchronization of the signal in the time-frequency dimension. At the same time, the essence of calculating the Frobenius norm is to convert this cross-correlation into a quantifiable numerical index for measuring the phase alignment degree of the local feature analysis units of the current signal and the voltage signal in the frequency domain feature space. Specifically, a larger Frobenius norm value means a higher phase alignment degree between the local frequency domain feature encoding matrix of the current signal and the local frequency domain feature encoding matrix of the voltage signal in the frequency domain feature space, that is, the energy distribution of the two in the time-frequency dimension is more synchronized, thereby enhancing the analysis accuracy of the correlation between the current and voltage signals in the frequency domain phase.

[0050] More specifically, the feature pair construction subunit 1322 is expressed by the formula:

[0051]

[0052] where arg max j (·) represents the index corresponding to the maximum value, k represents the index of the local frequency domain feature coding matrix of the voltage signal with the highest alignment degree with the feature phase of the local frequency domain feature coding matrix of the i-th current signal, and M 2k is the local three-dimensional space feature coding matrix of the fire scene that forms a phase-aligned feature pair with the local feature coding matrix of the i-th infrared thermal imaging.

[0053] That is, by selecting the local frequency domain feature coding matrix of the voltage signal corresponding to the maximum value from the set of phase alignment degrees, that is, M 2k , the feature segment with the strongest synergy between the current signal and the voltage signal in the frequency domain feature space can be extracted. This synergy reflects the physical coupling law between the two under normal operating conditions. And forming a phase-aligned {local frequency domain feature coding matrix of current signal, local frequency domain feature coding matrix of voltage signal} feature pair with the corresponding local frequency domain feature coding matrix of the current signal, that is, {M 1i , M 2k} pair , essentially constructs a dynamic reference benchmark based on the synergy of frequency domain features. When the line is privately modified, the harmonic distortion caused by the change of line parameters will destroy the phase synchronization between the current and the voltage, resulting in the inability to form a feature pair with high synergy. Although the natural load fluctuation may cause amplitude changes, the statistical characteristics of its phase coupling relationship still maintain a certain continuity with the normal operating conditions.

[0054] Figure 6 It is a block diagram of the feature joint perception aggregation unit in the substation line alarm device based on real-time monitoring according to an embodiment of the present application. As Figure 6 shown, the feature joint perception aggregation unit 133 includes: an attention weight calculation subunit 1331, configured to input each phase-aligned {local frequency domain feature coding matrix of current signal, local frequency domain feature coding matrix of voltage signal} feature pair in the set of phase-aligned {local frequency domain feature coding matrix of current signal, local frequency domain feature coding matrix of voltage signal} feature pairs into the feature joint perception attention network to obtain a set of current signal-voltage signal frequency domain feature joint perception attention weights; a feature aggregation subunit 1332, configured to perform attention-driven significant aggregation on the set of phase-aligned {local frequency domain feature coding matrix of current signal, local frequency domain feature coding matrix of voltage signal} feature pairs based on the set of current signal-voltage signal frequency domain feature joint perception attention weights to obtain the current-voltage signal frequency domain pattern joint perception feature map.

[0055] Specifically, in a preferred example of the present application, the attention weight calculation sub-unit 1331 is configured to: calculate the current signal-voltage signal local frequency domain feature joint perception coding matrix between the current signal local frequency domain feature coding matrix and the voltage signal local frequency domain feature coding matrix in the phase-aligned {current signal local frequency domain feature coding matrix, voltage signal local frequency domain feature coding matrix} feature pair; based on the feature distribution stationary quantity of the current signal local frequency domain feature coding matrix and the voltage signal local frequency domain feature coding matrix in the phase-aligned {current signal local frequency domain feature coding matrix, voltage signal local frequency domain feature coding matrix} feature pair, perform embedding optimization on the matrix trace of the current signal-voltage signal local frequency domain feature joint perception coding matrix to obtain the current signal-voltage signal frequency domain feature joint perception attention factor, which is expressed by the formula:

[0056] a i = Tr(M 1i T M 2k )

[0057] M 1i T ' = exp[-(α - β + a i ) ⊙ M 1i T )

[0058] M 2k ' = exp[-(β - α + a i ) ⊙ M 2k )

[0059] a i ' = Tr(M 1i T 'M 2k ')

[0060] where, ⊙ represents dot product, Tr(·) is the matrix trace metric function, a i is the matrix trace of the current signal-voltage signal local frequency domain feature joint perception coding matrix of the i-th phase-aligned {current signal local frequency domain feature coding matrix, voltage signal local frequency domain feature coding matrix} feature pair, α and β represent the number of biconnectivity between M 1i T and M 2k , exp(·) represents the exponential function with base e, M 1i T ' represents the optimized M 1i T , M 2k ' represents the optimized M 2k , a i' represents the current signal-voltage signal frequency domain feature joint perception attention factor of the feature pair {M 1i , M 2k}. pair Here, let M

[0061] = M a = M 1i T and M b = M 2k , and m ai ∈ M a , m bi ∈ M b . By calculating the number of eigenvalues that satisfy the L1 distance d ai (m bi , m L1 ) < ε and the L2 distance d ai (m bi , m L2 ) < ε between the corresponding m ai and m bi , the number of biconnected components between the matrices M a and M b is obtained, that is, the distribution map smoothness between the local frequency domain feature encoding matrix of the current signal representing phase alignment and the local frequency domain feature encoding matrix of the voltage signal. Then, in the discrete manifold representation of the local frequency domain feature encoding matrix of the current signal and the local frequency domain feature encoding matrix of the voltage signal, the trace metric operation of the local frequency domain feature encoding matrix of the current signal and the local frequency domain feature encoding matrix of the voltage signal is embedded and optimized through embedding manifold compactification based on the biconnected number, so as to improve the calculation accuracy of the joint matrix trace metric on the basis of enhancing the long-range correlation connectivity robustness.

[0062] More specifically, in a specific example of the present application, the attention weight calculation sub-unit 1331 is further configured to: input the current signal-voltage signal frequency domain feature joint perception attention factor into the sigmoid activation function to obtain the current signal-voltage signal frequency domain feature joint perception attention weight, which is expressed by the formula:

[0063] w i = A net ({M 1i , M 2k} pair ) = sigmoid(a i ')

[0064] Where sigmoid is the sigmoid activation function, {·,·} pair represents the phase-aligned feature pair, A net is the feature joint perception attention network, wi Represents {M 1i , M 2k} pair The joint perception attention weight of the current signal - voltage signal in the frequency domain feature.

[0065] That is, by introducing the Sigmoid activation function, the original distribution of the joint perception attention factor of the current signal - voltage signal in the frequency domain feature can be transformed into a weight assignment with clear physical meaning. That is, the high - weight area corresponds to the key discriminant basis with strong distinguishability in the frequency domain feature, and the low - weight area corresponds to noise or redundant information. In this way, by strengthening the model's ability to focus on the features of sensitive frequency bands, the frequency domain fluctuation interference caused by environmental noise and conventional load changes can be effectively suppressed. The joint perception attention weight of the current signal - voltage signal in the frequency domain feature generated can accurately point to the unique feature expression of the man - made alteration behavior in the frequency domain.

[0066] More specifically, the feature aggregation sub - unit 1332 is expressed by the formula:

[0067]

[0068] F f = [M′1, M′2..., M′ i ,..., M′ n

[0069] Where Represents subtraction by position points, Represents addition by position points, W1 and W2 are different weight parameter matrices, M i ′ represents the local alignment and fusion feature matrix of the current signal - voltage signal, F f Represents the joint perception feature map of the current - voltage signal in the frequency domain pattern, M1′, M2′, M i ′ and M′ n Respectively represent the 1st, 2nd, ith, and nth local alignment and fusion feature matrices of the current signal - voltage signal in the joint perception feature map of the current - voltage signal in the frequency domain pattern.

[0070] That is, attention - driven significant aggregation focuses on the key areas in the interaction of the current and voltage frequency domain features by dynamically assigning weights, avoiding information dilution caused by simple averaging or splicing. In this way, the frequency domain feature patterns extracted by the deep learning algorithm, combined with the screening and strengthening of complementary information by the attention mechanism, construct a more robust joint feature expression, so as to accurately capture the unique frequency domain correlation anomalies of the unauthorized wiring behavior, while suppressing the common noise generated by natural load changes, making the finally generated joint perception feature map of the current - voltage signal in the frequency domain pattern significantly improve the feature distinguishability, reduce the false alarm rate, enhance the detection reliability and optimize the alarm accuracy.​

[0071] In the above substation line alarm device based on real-time monitoring, the line change alarm module 140 is used to determine whether to send a line change alarm message to the management personnel based on the current-voltage signal frequency-domain mode joint perception feature map. Among them, Figure 7 It is a block diagram of the line change alarm module in the substation line alarm device based on real-time monitoring according to an embodiment of the present application. As Figure 7 shown, the line change alarm module 140 includes: a line change detection unit 141, which is used to input the current-voltage signal frequency-domain mode joint perception feature map into a line change detection module based on a classifier to obtain a detection result, and the detection result is used to indicate whether there is an unauthorized change behavior of the power distribution line; an alarm unit 142, which is used to send a line change alarm message to the management personnel in response to the detection result indicating that there is an unauthorized change behavior of the power distribution line.

[0072] Specifically, the line change detection unit 141 is used to input the current-voltage signal frequency-domain mode joint perception feature map into a line change detection module based on a classifier to obtain a detection result, and the detection result is used to indicate whether there is an unauthorized change behavior of the power distribution line. That is, a classification algorithm is used to perform feature analysis on the current-voltage signal frequency-domain mode joint perception feature map, so as to judge whether there is an unauthorized change behavior of the power distribution line. Specifically, the classifier is based on a neural network architecture, and through supervised learning, it distinguishes the current-voltage frequency-domain mode features under normal load fluctuations and the state of unauthorized changes to the power distribution line, establishes a decision boundary, and realizes the intelligent identification of unauthorized line change behaviors. In the detection stage, the classifier receives the current-voltage signal frequency-domain mode joint perception feature map as input, and through feature extraction and non-linear transformation of multiple hidden layers, gradually abstracts high-level feature representations, excavates the frequency-domain correlation mode between the current signal and the voltage signal, and calculates the probability that it belongs to the categories of normal load fluctuations and unauthorized changes to the power distribution line accordingly, so as to give a detection result and realize the intelligent detection of unauthorized change behaviors of the power distribution line.

[0073] In a specific example of the present application, the line change detection unit 141 is further configured to: expand the current-voltage signal frequency-domain pattern joint perception feature map into a current-voltage signal frequency-domain pattern joint perception feature vector; use the fully connected layer of the line change detection module to perform fully connected encoding on the current-voltage signal frequency-domain pattern joint perception feature vector to obtain a current-voltage signal frequency-domain pattern joint perception fully connected encoding vector; input the current-voltage signal frequency-domain pattern joint perception fully connected encoding vector into the Softmax classification function of the line change detection module to obtain the probability values of the current-voltage signal frequency-domain pattern joint perception feature map belonging to each classification label, where the classification labels include the existence of unauthorized modification behavior and the non-existence of unauthorized modification behavior; determine the classification label corresponding to the largest probability value as the detection result.

[0074] Specifically, the alarm unit 142 is configured to send a line change alarm message to the management personnel in response to the detection result that there is unauthorized modification behavior in the power distribution line. Specifically, when it is detected that there is unauthorized line modification behavior in the power distribution line, the alarm mechanism will be immediately triggered, and a line change alarm message will be sent to the management personnel through a preset communication channel, so that the management personnel can take corresponding measures in time to avoid safety accidents and economic losses.

[0075] In the specific implementation process, in order to quickly and accurately convey the emergency situation to the relevant management personnel, multiple communication channels can be used for message transmission, including but not limited to short message service (SMS), email, instant messaging software, and dedicated application notifications, etc. Each communication method has its unique advantages and applicable scenarios, and in actual applications, the most suitable method can be selected according to the specific situation or a combination of methods can be used to improve the success rate and coverage of the notification.

[0076] For the short message service, this is a traditional means widely used in emergency notifications. SMS has the characteristic of high arrival rate and can be successfully sent even in poor network conditions. In addition, the short message service supports receiving messages globally with any number without installing any additional applications, making this notification method extremely convenient. However, there is a limit to the length of the SMS content, and there may be a delay in delivery. Therefore, when using SMS as an alarm channel, the SMS content needs to be carefully designed to ensure that the core information can be clearly conveyed to the recipient. For example, key information such as the specific line number, the expected affected range, and the recommended preliminary action measures can be included in the SMS so that the management personnel can make a quick response.

[0077] Email is also a commonly used communication method, especially suitable for transmitting notification information containing rich details. Compared with text messages, emails allow a larger text capacity and can attach various formats of attachments such as pictures, documents, and even videos, making emails an ideal choice for delivering complex information. However, emails rely on an internet connection, and unstable network conditions may cause email sending failures. Therefore, when deciding to use email as one of the alarm channels, it is necessary to ensure the stability of the network environment and consider setting up backup plans in case of emergencies. Additionally, since modern people receive a large number of emails daily, to avoid the alarm information being overlooked, the word "urgent" should be clearly marked in the email subject line and a prominent font color should be used to attract attention.

[0078] In summary, the substation line alarm device based on real-time monitoring according to the embodiments of the present application is elucidated. It collects the current signal and voltage signal of the distribution feed-out loop in real time, obtains the spectrograms of the current signal and voltage signal through fast Fourier transform after signal denoising processing of the two, and further introduces a data processing algorithm based on deep learning to learn the frequency domain characteristic patterns of the current signal and voltage signal. Then, through feature phase search alignment and joint perception of the two, intelligent monitoring of whether there is unauthorized modification behavior in the distribution line is realized based on the frequency domain correlation characteristics between the current signal and voltage signal. In this way, the characteristic differences between natural load changes and human modifications can be effectively distinguished, the detection reliability of unauthorized modification of the distribution line can be improved, and false alarms can be reduced.

[0079] The basic principles of the present invention have been described above in combination with specific embodiments. However, it should be noted that the advantages, benefits, effects, etc. mentioned in the present invention are only examples and not limitations. It cannot be considered that these advantages, benefits, effects, etc. are essential for each embodiment of the present invention. Additionally, the specific details of the above embodiments are only for the purposes of illustration and easy understanding, and not for limitation. The above details do not limit the present invention to necessarily adopt the above specific details for implementation.

[0080] In the above embodiments, the descriptions of each embodiment have their own emphases. For parts not detailed or recorded in a certain embodiment, reference can be made to the relevant descriptions of other embodiments. In the several embodiments provided by the present invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are only illustrative. For example, the unit division is only a logical function division, and there can be other division methods in actual implementation. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0081] It is obvious to those skilled in the art that the present invention is not limited to the details of the above-described exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention. Therefore, in any respect, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be embraced within the present invention. Any reference signs in the claims should not be construed as limiting the claims involved.

[0082] In addition, it is obvious that the term "comprising" does not exclude other elements or steps, and the singular does not exclude the plural. The multiple elements recited in the system claims can also be implemented by one element through software or hardware.

[0083] Finally, it should be noted that the above description has been given for purposes of illustration and description. In addition, the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A substation line alarm device based on real-time monitoring, characterized in that, Including: A power distribution line signal monitoring module, configured to collect current signals and voltage signals of power distribution outgoing loops in real time; A signal spectrum analysis module, configured to perform spectrum feature analysis on the current signals and the voltage signals to obtain a current signal frequency domain feature map and a voltage signal frequency domain feature map; A joint sensing module, configured to perform feature joint sensing on the current signal frequency domain feature map and the voltage signal frequency domain feature map to obtain a current-voltage signal frequency domain pattern joint sensing feature map, wherein performing feature joint sensing on the current signal frequency domain feature map and the voltage signal frequency domain feature map includes: performing feature phase dynamic search alignment and joint sensing on the current signal frequency domain feature map and the voltage signal frequency domain feature map to obtain the current-voltage signal frequency domain pattern joint sensing feature map; A line change alarm module, configured to determine whether to send a line change alarm message to a management staff based on the current-voltage signal frequency domain pattern joint sensing feature map.

2. The alarm device for a power transmission line based on real-time monitoring according to claim 1, wherein The signal spectrum analysis module includes: A signal denoising unit, configured to perform signal denoising processing based on wavelet transform on the current signals and the voltage signals to obtain filtered current signals and filtered voltage signals; A Fourier transform unit, configured to perform fast Fourier transform on the filtered current signals and the filtered voltage signals to obtain a current signal spectrum map and a voltage signal spectrum map; A signal spectrum feature extraction unit, configured to extract signal spectrum features from the current signal spectrum map and the voltage signal spectrum map respectively to obtain the current signal frequency domain feature map and the voltage signal frequency domain feature map.

3. The alarm device for a power transformation line based on real-time monitoring according to claim 2, wherein, The signal spectrum feature extraction unit is configured to: Perform spectrum feature extraction based on the ASPP model on the current signal spectrum map and the voltage signal spectrum map respectively to obtain the current signal frequency domain feature map and the voltage signal frequency domain feature map.

4. The alarm device for a power transmission line based on real-time monitoring according to claim 3, characterized in that, The joint sensing module includes: A feature decoupling unit, configured to perform feature decoupling on the current signal frequency domain feature map and the voltage signal frequency domain feature map to obtain a set of current signal local frequency domain feature encoding matrices and a set of voltage signal local frequency domain feature encoding matrices; A feature dynamic search alignment unit, configured to perform feature dynamic search alignment on the set of current signal local frequency domain feature encoding matrices and the set of voltage signal local frequency domain feature encoding matrices to obtain a set of phase-aligned {current signal local frequency domain feature encoding matrix, voltage signal local frequency domain feature encoding matrix} feature pairs; A feature joint sensing aggregation unit, configured to perform feature joint sensing aggregation based on an attention mechanism on the set of phase-aligned {current signal local frequency domain feature encoding matrix, voltage signal local frequency domain feature encoding matrix} feature pairs to obtain the current-voltage signal frequency domain pattern joint sensing feature map.

5. The alarm device for a power transformation line based on real-time monitoring according to claim 4, wherein, The feature dynamic search alignment unit includes: A feature alignment degree calculation subunit, configured to calculate the feature phase alignment degree between the i-th current signal local frequency domain feature encoding matrix in the set of current signal local frequency domain feature encoding matrices and each voltage signal local frequency domain feature encoding matrix in the set of voltage signal local frequency domain feature encoding matrices to obtain a set of current signal-voltage signal frequency domain feature phase alignment degrees; A feature pair construction subunit, configured to extract the voltage signal local frequency domain feature encoding matrix corresponding to the maximum value in the set of current signal-voltage signal frequency domain feature phase alignment degrees, and form a phase alignment feature pair with the i-th current signal local frequency domain feature encoding matrix to obtain the phase alignment {current signal local frequency domain feature encoding matrix, voltage signal local frequency domain feature encoding matrix} feature pair.

6. The alarm device for a power transformation line based on real-time monitoring according to claim 5, wherein The feature alignment degree calculation subunit is configured to: calculate the product between the transpose matrix of the i-th current signal local frequency domain feature encoding matrix and the voltage signal local frequency domain feature encoding matrix to obtain a current signal-voltage signal local frequency domain feature joint perception encoding matrix; calculate the Frobenius norm of the current signal-voltage signal local frequency domain feature joint perception encoding matrix to obtain the current signal-voltage signal frequency domain feature phase alignment degree.

7. The alarm device for a power transmission line based on real-time monitoring according to claim 6, wherein, The feature joint perception aggregation unit includes: An attention weight calculation subunit, configured to input each phase alignment {current signal local frequency domain feature encoding matrix, voltage signal local frequency domain feature encoding matrix} feature pair in the set of phase alignment {current signal local frequency domain feature encoding matrix, voltage signal local frequency domain feature encoding matrix} feature pairs into a feature joint perception attention network to obtain a set of current signal-voltage signal frequency domain feature joint perception attention weights; A feature aggregation subunit, configured to perform attention-driven significant aggregation on the set of phase alignment {current signal local frequency domain feature encoding matrix, voltage signal local frequency domain feature encoding matrix} feature pairs based on the set of current signal-voltage signal frequency domain feature joint perception attention weights to obtain the current-voltage signal frequency domain pattern joint perception feature map.

8. The alarm device for a power transmission line based on real-time monitoring according to claim 7, wherein, The attention weight calculation subunit is configured to: calculate the current signal-voltage signal local frequency domain feature joint perception encoding matrix between the current signal local frequency domain feature encoding matrix and the voltage signal local frequency domain feature encoding matrix in the phase alignment {current signal local frequency domain feature encoding matrix, voltage signal local frequency domain feature encoding matrix} feature pair; perform embedding optimization on the matrix trace of the current signal-voltage signal local frequency domain feature joint perception encoding matrix based on the feature distribution stationary quantity of the current signal local frequency domain feature encoding matrix and the voltage signal local frequency domain feature encoding matrix in the phase alignment {current signal local frequency domain feature encoding matrix, voltage signal local frequency domain feature encoding matrix} feature pair to obtain a current signal-voltage signal frequency domain feature joint perception attention factor; Input the joint perception attention factor of the current signal-voltage signal frequency domain characteristics into the sigmoid activation function to obtain the joint perception attention weight of the current signal-voltage signal frequency domain characteristics.

9. The alarm device for a power transformation line based on real-time monitoring according to claim 8, characterized in that The line modification alarm module includes: A line modification detection unit, configured to input the joint perception feature map of the current-voltage signal frequency domain pattern into a line modification detection module based on a classifier to obtain a detection result, where the detection result is used to indicate whether there is an unauthorized modification behavior in the distribution line; An alarm unit, configured to send a line modification alarm message to the management personnel in response to the detection result that there is an unauthorized modification behavior in the distribution line.

10. The alarm device for a power transformation line based on real-time monitoring according to claim 9, characterized in that, The line modification detection unit is configured to: Unfold the joint perception feature map of the current-voltage signal frequency domain pattern into a joint perception feature vector of the current-voltage signal frequency domain pattern; Use the fully connected layer of the line modification detection module to perform fully connected encoding on the joint perception feature vector of the current-voltage signal frequency domain pattern to obtain a fully connected encoded vector of the current-voltage signal frequency domain pattern joint perception: Input the fully connected encoded vector of the current-voltage signal frequency domain pattern joint perception into the Softmax classification function of the line modification detection module to obtain the probability values of the joint perception feature map of the current-voltage signal frequency domain pattern belonging to each classification label, where the classification labels include the existence of unauthorized modification behavior and the non-existence of unauthorized modification behavior; Determine the classification label corresponding to the largest of the probability values as the detection result.

Images