Optical fiber differential protection decision method and system based on adaptive multi-transmission line

Abstract

The application provides an optical fiber differential protection decision method and system based on an adaptive multi-transmission line, which comprises the following steps: collecting real-time optical signal waveform data of each line in the multi-transmission line, extracting a differential current feature set corresponding to each line, inputting a joint feature sequence into a pre-trained adaptive multi-transmission line model to obtain a protection priority parameter corresponding to each line; matching a differential protection threshold corresponding to each line based on the protection priority parameter, and generating an adaptive trigger condition of each line; monitoring an actual differential current value of each line in real time, and if it is detected that the actual differential current value of a target line reaches the adaptive trigger condition corresponding to the target line, activating a differential protection action for the target line; and adjusting a monitoring frequency of the lines that have not triggered the protection action according to the priority parameter output by the adaptive multi-transmission line model. The application can realize the dual improvement of fault positioning accuracy and system operation stability during optical fiber differential protection.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of data processing, and in particular, to a differential protection decision method and system based on adaptive multi-transmission lines. BACKGROUND

[0002] As a core means of key line fault detection in power systems, the optical fiber differential protection technology mainly judges the occurrence of faults by comparing the amplitude and phase difference of current signals at both ends of the transmission line. The traditional scheme relies on a pre-set fixed differential threshold value, and triggers a protection action when the real-time differential current exceeds the threshold value. The protection criterion is based on the isolated analysis of the current characteristics of a single line, and lacks the ability to globally evaluate the dynamic coupling effect under the cooperative operation of multiple lines. The existing methods usually use static parameter configuration, such as setting a unified protection threshold value according to historical fault data, without considering factors such as the shift of optical fiber transmission characteristics caused by line load fluctuations and environmental temperature changes, which interfere with the authenticity of the differential current, resulting in false actions of high-load lines due to transient current fluctuations, and the possibility of masking real fault characteristics due to signal attenuation in low-load lines. Especially in the complex networking scene of multiple transmission lines, the existing technology cannot effectively identify the abnormal propagation path of differential current caused by energy transfer between lines, and the monitoring resource allocation mechanism is rigid, with all lines using the same sampling frequency to collect data, resulting in the transient abnormal characteristics of high-fault-risk lines being missed due to too long sampling intervals, and low-risk lines causing waste of hardware resources due to continuous high-frequency monitoring. In addition, the traditional differential protection model lacks dynamic filtering capability for interference components such as harmonic distortion and transient pulses in signal waveforms, making it difficult to distinguish between real fault currents and noise interference, further exacerbating the contradiction between protection sensitivity and anti-interference performance, and there is an urgent need for a differential protection decision method that can integrate multiple line dynamic characteristics, adaptively optimize the protection threshold value, and realize intelligent allocation of monitoring resources. SUMMARY

[0003] The application aims to provide an optical fiber differential protection decision method and system based on adaptive multi-transmission lines. The application is implemented in the following manner: in a first aspect, the embodiments of the application provide an optical fiber differential protection decision method based on adaptive multi-transmission lines, which comprises: collecting real-time optical signal waveform data of each line in the multi-transmission lines, and extracting a differential current feature set corresponding to each line; generating a joint feature sequence based on the differential current feature set of all lines, and inputting the joint feature sequence into a pre-trained adaptive multi-transmission line model to obtain a protection priority parameter corresponding to each line; dynamically matching a differential protection threshold corresponding to each line according to the protection priority parameter to generate an adaptive trigger condition of each line; monitoring an actual differential current value of each line in real time, and if it is detected that the actual differential current value of a target line reaches the adaptive trigger condition corresponding thereto, activating a differential protection action for the target line; and adjusting a monitoring frequency of the lines that have not triggered the protection action according to the priority parameter output by the adaptive multi-transmission line model.

[0004] In another aspect, the application provides an optical fiber differential protection decision system, which comprises: one or more processors; a memory; and one or more computer programs; wherein the one or more computer programs are stored in the memory and configured to be executed by the one or more processors, and when the one or more computer programs are executed by the processors, the method described above is implemented.

[0005] The adaptive multi-transmission line-based optical fiber differential protection decision method provided by the application extracts a differential current feature set and generates a joint feature sequence by collecting real-time optical signal waveform data of the multi-transmission line, dynamically outputs protection priority parameters of each line by using a pre-trained adaptive multi-transmission line model, generates an adaptive trigger condition according to the priority parameters by dynamically matching a differential protection threshold value, monitors a differential current value in real time and triggers a protection action, and simultaneously adjusts the monitoring frequency of a non-triggered line based on the priority parameters to form a closed-loop feedback, which can accurately capture line abnormal fluctuations and cross-line cascading failure risks through the synergistic effect of multi-dimensional data fusion (optical signal waveform, differential current feature, and signal transmission delay parameter) and dynamic model decision, effectively improves the response sensitivity and anti-interference ability of differential protection, and avoids the problems of false triggering or protection delay caused by traditional fixed thresholds. Through the threshold dynamic compensation mechanism driven by the priority parameters and the closed-loop adjustment strategy of the monitoring frequency, the adaptive balance between efficient allocation of protection resources and abnormal detection sensitivity is realized, which can not only preferentially identify transient fault features in high-load lines, but also reduce resource consumption caused by redundant monitoring in stable lines, thereby enhancing the overall robustness of the differential protection system in complex power grid environments. The dynamic weight superposition calculation based on the joint feature sequence and real-time load parameters ensures that the protection priority parameters can simultaneously reflect the correlation between the current operating state and the historical load trend of the line, and the multi-level protection trigger logic combining harmonic distortion rate detection and buffer interval verification further suppresses the interference of environmental noise on the differential current criterion, forming a stereoscopic protection decision system driven by data and coordinated with physical rules, and finally realizing the dual improvement of fault positioning accuracy and system operation stability. BRIEF DESCRIPTION OF DRAWINGS

[0006] Figure 1 is a flowchart of an adaptive multi-transmission line-based optical fiber differential protection decision method provided by an embodiment of the application;

[0007] Figure 2 is a composition schematic diagram of an optical fiber differential protection decision system provided by an embodiment of the application. DETAILED DESCRIPTION

[0008] The embodiments of the application are described below with reference to the accompanying drawings. The terms used in the embodiment part of the application are only used to explain the specific embodiments of the application, and are not intended to limit the application.

[0009] The execution subject of the adaptive multi-transmission line-based optical fiber differential protection decision method in the embodiment of the application is an optical fiber differential protection decision system, including but not limited to a server, a personal computer, a notebook computer, a tablet computer, etc. Figure 1As shown, the method comprises: step S100: collecting real-time optical signal waveform data of each line in the multi-transmission line, and extracting a differential current feature set corresponding to each line.

[0010] The real-time optical signal waveform data refers to the waveform information of the optical signal transmitted by each line in the multi-transmission line changing with time at the current moment, which reflects the real-time state of the optical signal. The differential current feature set is a collection of a series of features related to the differential current extracted from the real-time optical signal waveform data, which can be used for subsequent analysis and judgment of the line state. In the embodiments of the present application, the real-time optical signal waveform data of each line in the multi-transmission line can be collected by a special collection device, and then the signal processing and analysis algorithm is used to extract the differential current feature set corresponding to each line from the collected real-time optical signal waveform data. For example, in a system including three transmission lines, the real-time optical signal waveform data of the three lines is collected by an optical signal collector, and then the differential current feature set of each line is extracted by algorithms such as Fourier transform, including current amplitude, frequency, phase and the like.

[0011] Step S200: generating a joint feature sequence based on the differential current feature sets of all lines, and inputting the joint feature sequence into a pre-trained adaptive multi-transmission line model to obtain protection priority parameters corresponding to each line.

[0012] The joint feature sequence is a sequence containing feature information of each line generated by integrating and processing the differential current feature sets of all lines, which can comprehensively reflect the overall characteristics of the multi-transmission line. The pre-trained adaptive multi-transmission line model is trained by a large amount of historical data, which can output protection priority parameters corresponding to each line according to the input joint feature sequence. The protection priority parameter is used to measure the importance of each line in the protection decision. In the embodiments of the present application, first, the joint feature sequence is generated according to the differential current feature sets of all lines, and the specific generation method can refer to the subsequent embodiments; then the generated joint feature sequence is input into the pre-trained adaptive multi-transmission line model, and the model outputs the protection priority parameters corresponding to each line after calculation and analysis. For example, in a scenario with five transmission lines, a joint feature sequence is generated according to the differential current feature sets of the five lines, and the sequence is input into the pre-trained adaptive multi-transmission line model, and the model outputs the protection priority parameters corresponding to each of the five lines to determine the order of each line in the protection decision.

[0013] As an implementation form, in step S200, the joint feature sequence is generated based on the differential current feature sets of all lines, and specifically can include: in step S210, a feature subset with an amplitude fluctuation rate exceeding a preset fluctuation threshold is filtered out from each differential current feature set, and each feature subset is aligned according to a time stamp.

[0014] The amplitude fluctuation rate refers to the fluctuation degree of the amplitude of the feature value in the differential current feature set with time, and the preset fluctuation threshold is a standard value preset for judging whether the amplitude fluctuation rate is too large. The feature subset is a collection of part of the features filtered out from the differential current feature set and meeting a specific condition (such as the amplitude fluctuation rate exceeding the preset fluctuation threshold). The time stamp is information for marking the time when the feature data is generated. In the embodiments of the present application, the differential current feature set of each line is filtered to determine the feature subset with the amplitude fluctuation rate exceeding the preset fluctuation threshold, and then the feature subsets of different lines are aligned according to the time stamp, so that the feature subsets of different lines have a corresponding relationship in time. For example, in a system including four transmission lines, for the differential current feature set of each line, a feature subset with an amplitude fluctuation rate exceeding a preset fluctuation threshold (such as 5%) is filtered out, and then the feature subsets of the four lines are aligned according to the time stamp, facilitating subsequent analysis and processing.

[0015] In step S220, the energy distribution parameters of the feature subset of each line in a preset time window are calculated, and the maximum energy gradient value in the energy distribution parameters of each line is extracted.

[0016] The preset time window is a time interval preset for segmented analysis of the feature subset. The energy distribution parameter is a parameter describing the energy distribution of the feature subset in the preset time window, which reflects the energy change characteristics of the feature subset. The maximum energy gradient value is the gradient value of the place where the energy changes most quickly in the energy distribution parameter, which can reflect the degree of energy change of the feature subset. In the embodiments of the present application, for the feature subset of each line, the energy distribution parameters in the preset time window are calculated, and then the maximum energy gradient value is extracted from the energy distribution parameters. For example, for the feature subset of a transmission line, the preset time window is set to 10 seconds, the energy distribution parameters of the feature subset in the 10 seconds are calculated, and then the maximum energy gradient value is determined to understand the degree of energy change of the line in this time period.

[0017] In an implementation form, in step S220, the energy distribution parameters of the feature subset of each line in a preset time window are calculated, and the maximum energy gradient value in the energy distribution parameters of each line is extracted, and specifically can include: in step S221, the feature subset is divided into a plurality of continuous time segments according to the length of the preset time window, and each time segment contains amplitude fluctuation rate data of a fixed time length.

[0018] The time segment is a small time interval obtained by dividing the feature subset according to the length of the preset time window, and each time segment contains fixed-length amplitude fluctuation rate data, which can reflect the amplitude fluctuation of the differential current feature in the time period. In the embodiments of the present application, the feature subset is divided according to the length of the preset time window to obtain a plurality of continuous time segments, and each time segment has a fixed length and contains corresponding amplitude fluctuation rate data. For example, the preset time window is 5 seconds, the feature subset of a line is divided according to the length of 5 seconds to obtain a plurality of continuous 5-second time segments, and each time segment contains the amplitude fluctuation rate data in the 5 seconds.

[0019] Step S222: The differential current values corresponding to all amplitude fluctuation rates in each time segment are accumulated to generate the energy accumulation value corresponding to each time segment.

[0020] The energy accumulation value is obtained by accumulating the differential current values corresponding to all amplitude fluctuation rates in each time segment, which reflects the energy accumulation of the differential current in the time segment. In the embodiments of the present application, the differential current values corresponding to the amplitude fluctuation rates in each time segment are accumulated to obtain the energy accumulation value corresponding to each time segment. For example, in a time segment, there are a plurality of differential current values corresponding to the amplitude fluctuation rates, and the values are added to obtain the energy accumulation value of the time segment, so as to measure the energy accumulation degree in the time segment.

[0021] Step S223: The boundary energy change rate of each time segment is calculated according to the difference of the energy accumulation values of adjacent time segments, and all the boundary energy change rates are arranged in time sequence as an energy difference sequence.

[0022] The boundary energy change rate refers to the difference of the energy accumulation values of adjacent time segments, which reflects the rate of energy change between adjacent time segments. The energy difference sequence is a sequence obtained by arranging the boundary energy change rates of all time segments in time sequence, which can show the energy change of the feature subset in the entire time range. In the embodiments of the present application, the difference of the energy accumulation values of adjacent time segments is calculated to obtain the boundary energy change rate of each time segment, and then the boundary energy change rates are arranged in time sequence to form the energy difference sequence. For example, there are three continuous time segments, the difference of the energy accumulation values of the second time segment and the first time segment, and the third time segment and the second time segment are calculated to obtain two boundary energy change rates, and the two boundary energy change rates are arranged in time sequence to form part of the energy difference sequence.

[0023] Step S224: screening all positive transition difference values from the energy difference sequence, and performing weighted summation on the screened difference values according to a preset weight coefficient to generate a dynamic energy distribution parameter in a preset time window.

[0024] The positive transition difference value refers to a difference value with a positive value and a clear growth trend in the energy difference sequence, which indicates that the energy has a positive and dramatic change in this time period. The preset weight coefficient is a coefficient preset for weighting the screened positive transition difference value. Different positive transition difference values can correspond to different weight coefficients. The dynamic energy distribution parameter is a parameter obtained by performing weighted summation on the screened positive transition difference value according to the preset weight coefficient, which can more accurately reflect the dynamic distribution of the energy in the preset time window. In the embodiments of the present application, all positive transition difference values are determined from the energy difference sequence, and then weighted summation is performed on these difference values according to the preset weight coefficient to obtain the dynamic energy distribution parameter in the preset time window. For example, several positive transition difference values are screened from the energy difference sequence, and weighted summation is performed on these difference values according to the preset weight coefficient (for example, the weight of the first difference value is 0.3, the weight of the second difference value is 0.5, etc.) to obtain the dynamic energy distribution parameter.

[0025] Step S225: traversing the gradient change trend of each difference value in the dynamic energy distribution parameter, identifying an interval segment with continuously increasing difference values, and extracting the maximum rising slope of the difference values in the interval segment as a maximum energy gradient value.

[0026] The gradient change trend refers to the change slope of each difference value in the dynamic energy distribution parameter, which reflects the speed and direction of energy change. The interval segment with continuously increasing difference values refers to a time period in which the difference values in the dynamic energy distribution parameter continuously increase, and the energy presents an upward trend in this interval segment. The maximum rising slope refers to the slope of the fastest rising place of the difference values in the interval segment with continuously increasing difference values, which is taken as the maximum energy gradient value and can reflect the degree of energy change in this interval segment. In the embodiments of the present application, the gradient change trend of each difference value in the dynamic energy distribution parameter is traversed to determine the interval segment with continuously increasing difference values, and then the maximum rising slope of the difference values in these interval segments is extracted as the maximum energy gradient value. For example, in the dynamic energy distribution parameter, it is found that a segment of continuous difference values is increasing, the rising slope of the difference values in this interval segment is calculated, and the maximum rising slope is determined as the maximum energy gradient value.

[0027] Step S226: superimposing the maximum energy gradient value and the average change rate in the dynamic energy distribution parameter to generate a corrected energy gradient parameter, and outputting the corrected energy gradient parameter as the maximum energy gradient value in the energy distribution parameter.

[0028] The average change rate is the average change speed of all difference values in the dynamic energy distribution parameter, which reflects the average change of energy in the entire time range. The corrected energy gradient parameter is a parameter obtained by superimposing the maximum energy gradient value and the average change rate in the dynamic energy distribution parameter, which comprehensively considers the intensity and average change of energy change. In the embodiments of the present application, the maximum energy gradient value is added to the average change rate in the dynamic energy distribution parameter to obtain the corrected energy gradient parameter, which is output as the maximum energy gradient value in the energy distribution parameter.

[0029] Step S230: splice the maximum energy gradient values of each line into an initial joint sequence according to the line number order, and normalize the initial joint sequence.

[0030] The initial joint sequence is a sequence obtained by splicing the maximum energy gradient values of each line in turn according to the line number order, which contains the key information of the energy change of each line. The normalization processing is to process the data in the initial joint sequence so that its value range is within a preset interval (usually [0, 1]), so that the dimensional difference between the maximum energy gradient values of different lines can be eliminated, and subsequent analysis and processing can be facilitated. In the embodiments of the present application, the maximum energy gradient values of each line are spliced into an initial joint sequence according to the line number order, and then the sequence is normalized. For example, there are three lines, and their maximum energy gradient values are 5, 8 and 3, respectively. The initial joint sequence [5, 8, 3] is spliced according to the line number order, and then the sequence is normalized so that its value range is between [0, 1].

[0031] Step S240: match the normalized initial joint sequence with the standard mode in the historical joint feature library, and eliminate the abnormal feature segments in the initial joint sequence that have a coincidence degree exceeding a preset coincidence degree with the historical abnormal mode.

[0032] The historical joint feature library is a database for storing joint feature sequences of multiple transmission lines in history, which contains various normal and abnormal standard patterns. The standard pattern is a representative feature pattern in the historical joint feature library, which is used to match the normalized initial joint sequence. The historical abnormal pattern is a feature pattern marked as abnormal in the historical joint feature library. The preset coincidence degree is a preset standard value for judging the similarity between the current sequence and the historical abnormal pattern. The abnormal feature segment is the part of the normalized initial joint sequence that has a coincidence degree exceeding the preset coincidence degree with the historical abnormal pattern. In the embodiments of the present application, the normalized initial joint sequence is compared with the standard patterns in the historical joint feature library to determine the abnormal feature segment having a coincidence degree exceeding the preset coincidence degree with the historical abnormal pattern, and the abnormal feature segment is removed from the initial joint sequence. For example, the preset coincidence degree is 80%, the normalized initial joint sequence is matched with the standard patterns in the historical joint feature library, it is determined that a segment in the normalized initial joint sequence has a coincidence degree of 85% with a historical abnormal pattern, and the segment is removed from the initial joint sequence as the abnormal feature segment.

[0033] Step S250: re-arranging the remaining feature segments in time sequence to generate a final joint feature sequence.

[0034] The remaining feature segment is the part of the initial joint sequence after removing the abnormal feature segment. The final joint feature sequence is the sequence obtained by re-arranging the remaining feature segments in time sequence, which more accurately reflects the normal feature information of the multiple transmission lines. In the embodiments of the present application, the remaining feature segments after removing the abnormal feature segment are re-arranged in time sequence to generate the final joint feature sequence. For example, three feature segments are left after removing the abnormal feature segment from the initial joint sequence, and the three feature segments are re-arranged in time sequence to obtain the final joint feature sequence.

[0035] As an implementation manner, in the step S200, the joint feature sequence is input into the pre-trained adaptive multi-transmission line model to obtain the protection priority parameters corresponding to each line, which can specifically include the following steps: step S260: dividing the final joint feature sequence into a plurality of continuous time segments, each time segment containing a fixed number of feature points.

[0036] The feature point is each data point in the final joint feature sequence, which represents the feature information at a certain time. In the embodiments of the present application, the final joint feature sequence is divided into a plurality of continuous time segments according to certain rules, and each time segment contains a fixed number of feature points. For example, the final joint feature sequence is divided into time segments each containing 10 feature points, which facilitates subsequent individual analysis and processing of each time segment.

[0037] Step S270: input each time segment into a feature fusion layer in the adaptive multi-transmission line model respectively, and output line state weight coefficients corresponding to each time segment.

[0038] The feature fusion layer is a network layer in the adaptive multi-transmission line model, which can perform feature fusion processing on the input time segments and comprehensively consider the relationship between the feature points. The line state weight coefficient is a coefficient output by the feature fusion layer according to the input time segments, which is used to measure the line state and reflects the importance and state information of the line in the time segment. In the embodiments of the present application, each time segment is sequentially input into the feature fusion layer of the adaptive multi-transmission line model, and the feature fusion layer outputs the line state weight coefficients corresponding to each time segment after calculation and analysis. For example, a time segment containing 10 feature points is input into the feature fusion layer, and the feature fusion layer outputs a line state weight coefficient corresponding to the time segment as 0.8, indicating that the state of the line in the time segment is important.

[0039] Step S280: calculate the cumulative weight value of each line in the whole monitoring period according to the line state weight coefficients of each time segment, and sort the lines according to the cumulative weight values from high to low.

[0040] The cumulative weight value is obtained by adding the line state weight coefficients of each line in each time segment, which comprehensively reflects the importance of the line in the whole monitoring period. In the embodiments of the present application, the line state weight coefficients of each line in each time segment are added to obtain the cumulative weight value of each line, and then the lines are sorted according to the cumulative weight values from high to low. For example, the line state weight coefficients of a line in three time segments are 0.7, 0.8 and 0.6 respectively, which are added to obtain a cumulative weight value of 2.1, and then all lines are arranged according to the cumulative weight values from high to low.

[0041] Step S290: map the sorted line serial number to a preset priority interval to generate a dynamic protection priority parameter corresponding to each line.

[0042] The preset priority interval is a preset interval range used to divide the line priority, and the dynamic protection priority parameter is a parameter obtained by mapping the sorted line serial number to the preset priority interval, which can dynamically reflect the priority of each line in the protection decision. In the embodiments of the present application, the sorted line serial number is correspondingly mapped to the preset priority interval to obtain the dynamic protection priority parameter corresponding to each line. For example, the preset priority interval is [1, 5], and the sorted line serial number is 3, which is mapped to the priority interval to obtain a dynamic protection priority parameter of 3 for the line.

[0043] Step S2100: Weighted superposition of the dynamic protection priority parameter and the real-time load parameter to obtain a final protection priority parameter.

[0044] The real-time load parameter represents the load size of the line at the current time, which reflects the actual working state of the line. The weighted superposition is an operation of adding the dynamic protection priority parameter and the real-time load parameter according to a certain weight, and the final protection priority parameter is a parameter obtained by weighted superposition, which comprehensively considers the priority of the line and the real-time load condition, and more accurately determines the importance of each line in the protection decision. In the embodiment of the application, the dynamic protection priority parameter and the real-time load parameter are weighted and superimposed according to a preset weight to obtain the final protection priority parameter.

[0045] In an embodiment, the training process of the pre-trained adaptive multi-transmission line model can include the following steps: step S10: collecting a plurality of groups of historical transmission line optical signal waveform sample data, and extracting a differential current feature vector corresponding to each group of sample data.

[0046] The optical signal waveform sample data of the historical transmission line refers to a set of waveform information of the optical signal of the multi-transmission line changing with time in the past period of time, which contains rich line state information. The differential current feature vector is a vector composed of features related to the differential current extracted from the optical signal waveform sample data, which can be used to describe the current characteristics of the line. In the embodiment of the application, a plurality of groups of historical transmission line optical signal waveform sample data are collected by a data acquisition device, and then a preset signal processing and analysis algorithm is used to extract a corresponding differential current feature vector from each group of sample data. For example, 100 groups of historical transmission line optical signal waveform sample data are collected, each group of data is processed, and a corresponding differential current feature vector is extracted, each vector containing current amplitude, frequency, phase and other characteristics.

[0047] Step S20: Time domain and frequency domain joint analysis of the differential current feature vector to generate a feature label sequence of each group of sample data.

[0048] The time-frequency joint analysis is an analysis method that comprehensively considers the characteristics of a signal in both time and frequency dimensions, and can more comprehensively understand the characteristics of the signal. The feature marker sequence is a sequence containing line feature marker information generated by performing time-frequency joint analysis on the differential current feature vector, and can more accurately describe the state of the line. In the embodiments of the present application, the extracted differential current feature vector is subjected to time-frequency joint analysis, and the specific analysis method can refer to the subsequent embodiments. Then, the feature marker sequence of each group of sample data is generated according to the analysis result. For example, a group of differential current feature vectors is subjected to time-frequency joint analysis to obtain the feature marker sequence of the group of sample data, and each marker in the sequence represents a preset line feature.

[0049] As an implementation manner, in step S20, the differential current feature vector is subjected to time-frequency joint analysis to generate the feature marker sequence of each group of sample data, which can specifically include: in step S21, the differential current feature vector is divided into a plurality of time domain segments, and the average amplitude and phase shift of each time domain segment are calculated.

[0050] The time domain segment is a small time period obtained by dividing the differential current feature vector in time sequence, the average amplitude is the average value of the differential current amplitude in each time domain segment, and the phase shift is the shift of the differential current phase relative to the reference phase in each time domain segment. In the embodiments of the present application, the differential current feature vector is divided into a plurality of time domain segments according to a certain time interval, and then the differential current amplitude and phase in each time domain segment are calculated to obtain the average amplitude and phase shift. For example, a differential current feature vector is divided into 10 time domain segments according to a time interval of 1 second, and the average amplitude and phase shift of each time domain segment are calculated respectively.

[0051] In step S22, a fast Fourier transform is performed on each time domain segment to extract the dominant frequency component in the frequency energy distribution of each time domain segment.

[0052] The fast Fourier transform is used to convert the time domain signal into the frequency domain signal, and each time domain segment can be converted into the frequency domain by the fast Fourier transform to obtain its frequency energy distribution. The dominant frequency component is the frequency component with the largest energy proportion in the frequency energy distribution, which reflects the main frequency characteristics of the time domain segment. In the embodiments of the present application, the fast Fourier transform is performed on each time domain segment to convert it from the time domain to the frequency domain, and then the dominant frequency component is determined from the frequency energy distribution. For example, the fast Fourier transform is performed on a time domain segment to obtain its frequency energy distribution, and it is found that the component with a frequency of 50 Hz has the largest energy proportion, so 50 Hz is taken as the dominant frequency component of the time domain segment.

[0053] Step S23: match the dominant frequency component with a preset typical fault frequency library, and mark the potential fault type corresponding to each time domain segment.

[0054] The preset typical fault frequency library is a database containing frequency information corresponding to various typical faults, and by matching the dominant frequency component with the frequency information in the library, the potential fault type corresponding to each time domain segment can be determined. In the embodiments of the present application, the dominant frequency component of each time domain segment is compared with the preset typical fault frequency library to determine the matched fault type, and the potential fault type corresponding to each time domain segment is marked. For example, the preset typical fault frequency library records that a frequency of 50 Hz may correspond to a line short circuit fault, and the dominant frequency component of a certain time domain segment is 50 Hz, so the time domain segment is marked as possibly existing a line short circuit fault.

[0055] Step S24: weight and fuse the average amplitude and phase shift of the time domain segment according to the potential fault type to generate a time-frequency joint feature parameter.

[0056] It can be understood that the weighting and fusion is a process of assigning different weights to the average amplitude and phase shift according to the potential fault type, and then comprehensively calculating them. The time-frequency joint feature parameter is a parameter obtained by weighting and fusion, which comprehensively considers the time domain and frequency domain features, and can more comprehensively describe the state of the line. In the embodiments of the present application, different weights are assigned to the average amplitude and phase shift of each time domain segment according to the potential fault type corresponding to the time domain segment, and then they are weighted and fused to generate a time-frequency joint feature parameter. For example, a certain time domain segment is marked as possibly existing a line short circuit fault, a weight of 0.6 is assigned to the average amplitude of the time domain segment, and a weight of 0.4 is assigned to the phase shift, and then they are weighted and fused to obtain a time-frequency joint feature parameter.

[0057] Step S25: arrange all time-frequency joint feature parameters in the same group of sample data in time sequence to form a feature mark sequence.

[0058] In the embodiments of the present application, the time-frequency joint feature parameters of each time domain segment in the same group of sample data are arranged in time sequence to form a feature mark sequence. This sequence contains comprehensive feature information of the group of sample data in time domain and frequency domain, and provides more accurate data for subsequent model training. For example, a group of sample data contains 10 time domain segments, and the time-frequency joint feature parameters of the 10 time domain segments are arranged in time sequence to form a feature mark sequence containing 10 elements.

[0059] Step S30: input the feature mark sequence into the convolution kernel alignment layer of the initial adaptive multi-transmission line model to output the alignment weight of each feature mark in the time dimension.

[0060] The convolution kernel alignment layer is a network layer in the initial adaptive multi-transmission line model, which can process the input feature mark sequence to determine the alignment relationship of each feature mark in the time dimension. The alignment weight is a weight value output by the convolution kernel alignment layer according to the feature mark sequence, which is used to represent the alignment degree of each feature mark in the time dimension, and can make the feature mark sequence more orderly and unified in time. In the embodiment of the present application, the generated feature mark sequence is input into the convolution kernel alignment layer of the initial adaptive multi-transmission line model, and the convolution kernel alignment layer outputs the alignment weight of each feature mark in the time dimension after calculation and analysis. For example, a feature mark sequence containing 10 feature marks is input into the convolution kernel alignment layer, and the alignment weight of each feature mark in the time dimension is obtained, such as the alignment weight of the first feature mark is 0.8, the alignment weight of the second feature mark is 0.7, and so on.

[0061] As an implementation manner, in step S30, the feature mark sequence is input into the convolution kernel alignment layer in the initial adaptive multi-transmission line model, and the alignment weight of each feature mark in the time dimension is output. Specifically, it can include the following steps.

[0062] The sliding window is a window in the convolution kernel alignment layer for local analysis of the feature mark sequence, and sliding windows of different scales can capture local patterns of different sizes in the feature mark sequence. The local pattern is a local segment with certain rules and characteristics in the feature mark sequence, and by capturing these local patterns, the structure and characteristics of the feature mark sequence can be better understood. In the embodiment of the present application, a plurality of sliding windows of different scales are set in the convolution kernel alignment layer, for example, sliding windows of scales 3, 5 and 7 are set, and these sliding windows are allowed to slide on the feature mark sequence to capture local patterns therein.

[0063] In step S32, the similarity score between the feature mark covered by each sliding window and the preset reference pattern is calculated.

[0064] The preset reference pattern is a representative feature pattern preset in advance, and the similarity score is a score for measuring the similarity between the feature mark covered by each sliding window and the preset reference pattern. The higher the score, the higher the similarity. In the embodiment of the present application, for the feature mark covered by each sliding window, the similarity score between the feature mark and the preset reference pattern is calculated. For example, the cosine similarity algorithm is used to calculate the similarity score between the feature mark covered by the sliding window and the preset reference pattern, and a score between 0 and 1 is obtained.

[0065] Step S33: dynamically adjusting the step size and coverage range of each sliding window according to the similarity score, to generate multi-scale alignment parameters.

[0066] The step size refers to the distance of each sliding window, and the coverage range refers to the length of the sliding window covering the feature marker sequence. The multi-scale alignment parameters are obtained by dynamically adjusting the step size and coverage range of the sliding window according to the similarity score, which can make the sliding window more accurately capture the local pattern in the feature marker sequence. In the embodiments of the present application, the step size and coverage range of each sliding window are dynamically adjusted according to the similarity score of the feature marker covered by each sliding window and the preset reference pattern, to generate multi-scale alignment parameters. For example, if the similarity score of the feature marker covered by a certain sliding window and the preset reference pattern is low, the step size is appropriately reduced and the coverage range is appropriately expanded, to improve the accuracy of capturing the local pattern.

[0067] Step S34: inputting the multi-scale alignment parameters into the weight allocation network, and outputting the position weight of each feature marker on the time axis.

[0068] The weight allocation network is a pre-trained network that can allocate the position weight of each feature marker on the time axis according to the input multi-scale alignment parameters. The position weight is used to represent the importance and relative position relationship of each feature marker on the time axis. In the embodiments of the present application, the generated multi-scale alignment parameters are input into the weight allocation network, and the network outputs the position weight of each feature marker on the time axis after calculation. For example, the input multi-scale alignment parameters contain adjustment information of different sliding windows, and the weight allocation network allocates the position weight of each feature marker in the feature marker sequence according to these information, such as the position weight of the first feature marker is 0.6, the position weight of the second feature marker is 0.8, and so on. These weights reflect the importance difference of each feature marker on the time axis.

[0069] Exemplarily, the weight allocation network can adopt a fully connected neural network architecture, which is composed of an input layer, multiple hidden layers and an output layer. The number of neurons in the input layer is the same as the dimension of the multi-scale alignment parameters, and each element in the multi-scale alignment parameters corresponds to a neuron in the input layer, which is responsible for receiving the input multi-scale alignment parameter data. The number of neurons in the hidden layer can be flexibly adjusted according to actual situation, to adapt to different complexity requirements of the task. In each hidden layer, the neurons perform weighted summation on the output of the previous layer, and perform nonlinear transformation through an activation function (such as ReLU function). The ReLU function can solve the gradient vanishing problem, enhance the expression ability of the network, and make the network learn more complex features and patterns.

[0070] The number of neurons in the output layer is the same as the number of feature markers in the feature marker sequence, and the output value of each neuron is the position weight of the corresponding feature marker on the time axis. In the training phase, a large amount of historical data can be used to train the weight distribution network. The training process uses the general backpropagation algorithm to continuously adjust the weights and biases of the network to minimize the error between the network output position weight and the true expected weight. Specifically, the multi-scale alignment parameters can be input into the network first to obtain the output result of the network, then the loss function (such as mean square error loss function) between the output result and the true label is calculated, then the partial derivative of the network weight and bias is calculated according to the loss function, and finally the network weight and bias are updated according to the partial derivative using an optimization algorithm (such as stochastic gradient descent algorithm). After multiple iterations of training, the weight distribution network can gradually learn the mapping relationship between the multi-scale alignment parameters and the position weight, so as to accurately output reasonable position weight according to the input multi-scale alignment parameters. The position weight can provide more accurate basis for subsequent nonlinear interpolation of the feature marker sequence, improve the accuracy and effectiveness of model training, and enable the model to better capture the information of the feature marker in the time dimension when processing related data of multiple transmission lines, providing more reliable support for line protection priority judgment and other tasks.

[0071] Step S35: Nonlinear interpolation of the feature marker sequence according to the position weight to generate a standardized feature sequence aligned in the time dimension.

[0072] Nonlinear interpolation is a method of processing the feature marker sequence according to the position weight to make the sequence more smooth and orderly in the time dimension. The standardized feature sequence is a feature sequence aligned in the time dimension after nonlinear interpolation, which eliminates the inconsistency of the feature markers in time and facilitates subsequent model training. In the embodiments of the present application, the feature marker sequence is subjected to nonlinear interpolation operation according to the position weight of each feature marker. For example, for two adjacent feature markers in the feature marker sequence, the interpolation points in between are calculated according to their position weights, so that the sequence is more continuous and regular in the time dimension, and finally a standardized feature sequence aligned in the time dimension is generated.

[0073] Step S40: Dynamic sampling of the feature marker sequence based on the alignment weight to generate a standardized feature set for training.

[0074] Dynamic sampling is a targeted sampling of the feature label sequence according to the alignment weight to obtain more representative and effective data. The standardized feature set is a feature set generated after dynamic sampling for training the model, which contains key feature information selected from the feature label sequence. In the embodiments of the present application, the feature label sequence is dynamically sampled according to the alignment weight output by the convolution kernel alignment layer, and the specific sampling method can refer to the subsequent implementation manner, and finally the standardized feature set for training is generated. For example, the part with higher weight in the feature label sequence is selected for sampling according to the alignment weight to form the standardized feature set, so as to improve the efficiency and accuracy of model training.

[0075] As an implementation manner, in step S40, the feature label sequence is dynamically sampled based on the alignment weight to generate a standardized feature set for training, which can specifically include: in step S41, filtering out the feature labels with weight values higher than a preset weight threshold as key feature points according to the position weight.

[0076] The preset weight threshold is a standard value for judging the importance of the feature label, and the key feature point is the feature label in the feature label sequence with a position weight higher than the preset weight threshold, which represents important information in the sequence. In the embodiments of the present application, the position weight of each feature label in the feature label sequence is compared, and the feature label with a weight value higher than the preset weight threshold is selected as a key feature point. For example, the preset weight threshold is 0.7, and there are 10 feature labels in the feature label sequence, of which 3 feature labels have position weights of 0.8, 0.9 and 0.8 respectively, which are higher than the preset weight threshold, and the 3 feature labels are taken as key feature points.

[0077] In step S42, the interval region between the key feature points is filled with mean value to generate a continuous feature trajectory curve.

[0078] The mean value filling refers to filling the interval region between the key feature points with the mean value between them to form a continuous curve. The feature trajectory curve is a continuous curve reflecting the change trend of the feature label after mean value filling, which can make the feature information more complete and coherent. In the embodiments of the present application, for the selected key feature points, the mean value of the interval region between them is calculated and filled into the interval region to generate a continuous feature trajectory curve. For example, there are two key feature points, and the interval region between them contains 5 data points. The mean value of the two key feature points is calculated and filled into the positions of the 5 data points to form a continuous feature trajectory curve.

[0079] In step S43, the feature trajectory curve is resampled at equal time intervals to obtain uniformly distributed feature sampling points.

[0080] The equal-time-interval resampling refers to selecting data points on the feature trajectory curve according to a fixed time interval to obtain uniformly distributed feature sampling points. These feature sampling points can more accurately reflect the feature information of the feature marker sequence, facilitating subsequent model training. In the embodiment of the present application, the generated feature trajectory curve is resampled according to a pre-set equal-time-interval to obtain uniformly distributed feature sampling points. For example, the equal-time-interval is set to 0.1 second, and a data point is selected every 0.1 second on the feature trajectory curve to obtain a series of uniformly distributed feature sampling points.

[0081] Step S44: binding the feature sampling points with the corresponding potential fault type labels to generate standardized training samples.

[0082] The potential fault type label is obtained according to the matching of the dominant frequency component with the pre-set typical fault frequency library in the foregoing step, and is used to represent the fault type that each feature sampling point may correspond to. The standardized training sample is a sample obtained after binding the feature sampling points with the corresponding potential fault type labels, and contains feature information and corresponding fault type information, and is a basic data unit for model training. In the embodiment of the present application, the feature sampling points obtained by resampling are one-to-one bound with the corresponding potential fault type labels to generate standardized training samples. For example, the potential fault type corresponding to a feature sampling point is a line short-circuit fault, and the feature sampling point is bound with the label “line short-circuit fault” to form a standardized training sample.

[0083] Step S45: randomly shuffling and batch dividing the standardized training samples of all historical transmission lines to generate a standardized feature set for final training.

[0084] When randomly shuffling, the order of all standardized training samples is randomly disturbed to avoid the influence of sample order on the model in the training process. Batch division is to group the standardized training samples after random shuffling according to a certain number, and each group is called a batch to facilitate batch training of the model. The standardized feature set for final training is a feature set for model training obtained after random shuffling and batch division. In the embodiment of the present application, the standardized training samples of all historical transmission lines are randomly shuffled, and then batch divided according to a pre-set batch size to generate a standardized feature set for final training. For example, there are 1000 standardized training samples, and the pre-set batch size is 100. After random shuffling of these samples, they are divided into 10 batches to form a standardized feature set for final training.

[0085] Step S50: iteratively train the initial adaptive multi-transmission line model using the standardized feature set until an error rate of a protection priority parameter output by the initial adaptive multi-transmission line model and a preset reference parameter is lower than a preset error threshold, to obtain an adaptive multi-transmission line model.

[0086] The iterative training refers to a process of inputting the standardized feature set into the initial adaptive multi-transmission line model multiple times, continuously adjusting the parameters of the model, and gradually making the output result of the model close to the preset reference parameter. The error rate refers to the difference between the protection priority parameter output by the model and the preset reference parameter, and the preset error threshold is a standard value preset for judging whether the model training is qualified. The adaptive multi-transmission line model is obtained after iterative training, and the error rate between the protection priority parameter output by the model and the preset reference parameter is lower than the preset error threshold, and the model can more accurately perform line protection priority judgment. The preset reference parameter is a standard value determined based on a large amount of historical data and actual operation experience, and can be adaptively adjusted according to actual conditions, and represents the reasonable protection priority parameter of each line under different line states. For example, the value can be determined after statistical analysis and expert evaluation of actual fault discrimination results, or can be an experienced value summarized according to industry standards and long-term operation. In the embodiments of the present application, the standardized feature set for final training is input into the initial adaptive multi-transmission line model, and multiple iterative training is performed. In each iteration, the predicted protection priority parameter is calculated and output according to the input standardized feature set, and then the error rate between the predicted parameter and the preset reference parameter is calculated. According to the error rate, the parameters of the model are updated using a preset optimization algorithm (such as a stochastic gradient descent algorithm), so that the output result of the model gradually approaches the preset reference parameter. For example, for a set of standardized feature sets, the initial adaptive multi-transmission line model outputs the protection priority parameters of each line, which are compared with the preset reference parameters to calculate the error rate. If the error rate is high, it means that the output of the model is far from the expected result, and the parameters of the model need to be adjusted. After multiple iterative training, when the error rate is lower than the preset error threshold, it is considered that the model training is qualified, and the adaptive multi-transmission line model is obtained. Exemplarily, the initial adaptive multi-transmission line model can adopt a multi-layer perception (MLP) architecture, which is composed of an input layer, multiple hidden layers and an output layer. The number of neurons in the input layer is the same as the dimension of the standardized feature set, and is responsible for receiving the data of the standardized feature set. The number of neurons in the hidden layer can be adjusted according to actual conditions, and each hidden layer performs nonlinear transformation on the input through an activation function (such as a ReLU function) to enhance the expression ability of the model. The number of neurons in the output layer is the same as the number of lines, and the output of each neuron is the protection priority parameter of the corresponding line. In the training process, the standardized feature set is used as the input, and the preset reference parameter is used as the training label. The error is calculated and the weights and biases of the model are updated through the back propagation algorithm, so that the model can learn the mapping relationship between the standardized feature set and the protection priority parameter, and finally realize accurate line protection priority judgment function, improve the accuracy and reliability of line protection decision.

[0087] As an implementation, in step S50, the initial adaptive multi-transmission line model is iteratively trained using the standardized feature set, which can specifically include: in step S51, in each iteration process, the standardized training samples in the current batch are input into the initial adaptive multi-transmission line model, and the predicted protection priority parameters are output.

[0088] In each iteration training, a batch of standardized training samples are selected from the standardized feature set for final training, and are input into the initial adaptive multi-transmission line model. The model calculates and analyzes according to the input sample data, and outputs the predicted protection priority parameters. For example, the current batch contains 100 standardized training samples, which are input into the initial adaptive multi-transmission line model, and the model outputs the predicted protection priority parameters corresponding to the 100 samples.

[0089] In step S52, the mean square error between the predicted protection priority parameters and the corresponding reference parameters is calculated, and the weights of the initial adaptive multi-transmission line model are updated according to the mean square error.

[0090] The mean square error is an index for measuring the difference between the predicted value and the true value, which is obtained by calculating the average of the square of the difference between the predicted protection priority parameters and the corresponding reference parameters. The weights of the model are the parameters in the model, and the performance of the model can be adjusted by updating the weights. In the embodiments of the present application, the mean square error between the predicted protection priority parameters of the model and the corresponding reference parameters is calculated, and then the weights of the initial adaptive multi-transmission line model are updated using a preset optimization algorithm (such as a stochastic gradient descent algorithm) according to the mean square error, so that the output result of the model is closer to the reference parameters. For example, the mean square error between the predicted protection priority parameters and the reference parameters is calculated to be 0.05, and the weights of the model are updated according to this error value, so that the model can output more accurate results in the next iteration.

[0091] In step S53, after each iteration ends, samples are randomly selected from the validation set for model performance evaluation, and the fault detection accuracy and the false trigger rate are calculated.

[0092] The validation set is a group of data pre-divided for evaluating the performance of the model in the training process. The fault detection accuracy refers to the proportion of the model correctly detecting the fault line, and the false trigger rate refers to the proportion of the model incorrectly triggering the protection action. After each iteration training ends, a certain number of samples are randomly selected from the validation set and input into the current trained model, and the fault detection accuracy and the false trigger rate are calculated according to the output result of the model to evaluate the performance of the model. For example, 200 samples are randomly selected from the validation set, the model correctly detects 180 fault lines and incorrectly triggers the protection action 10 times, and the fault detection accuracy is calculated to be 90% and the false trigger rate is calculated to be 5%.

[0093] Step S54: If the fault detection accuracy rate does not improve and the false trigger rate does not decrease in continuous multiple iterations, triggering the early stopping mechanism and saving the current optimal model parameters.

[0094] The early stopping mechanism is a strategy for preventing model overfitting. When the fault detection accuracy rate of the model does not improve and the false trigger rate does not decrease in continuous multiple iterations, it indicates that the model may have reached an optimal state or begun to overfit. At this time, the early stopping mechanism is triggered, the training process is stopped, and the current optimal model parameters are saved. For example, set 5 consecutive iterations as the judgment standard. In 5 consecutive iterations, the fault detection accuracy rate of the model remains at 90%, and the false trigger rate remains at 5%. The early stopping mechanism is triggered, and the parameters of the current model are saved.

[0095] Step S55: Load the optimal model parameters into the initial adaptive multi-transmission line model to generate a pre-trained adaptive multi-transmission line model.

[0096] After triggering the early stopping mechanism and saving the optimal model parameters, these parameters are loaded into the initial adaptive multi-transmission line model to replace the original parameters, obtaining a pre-trained adaptive multi-transmission line model. This model has been trained and optimized and can more accurately output the protection priority parameters corresponding to each line according to the input joint feature sequence, providing more reliable support for subsequent optical fiber differential protection decision-making. For example, load the saved optimal model parameters into the initial adaptive multi-transmission line model to complete parameter replacement and generate a pre-trained adaptive multi-transmission line model. This model can be used in actual line protection decision-making processes.

[0097] Step S300: Dynamically match the differential protection threshold values corresponding to each line according to the protection priority parameters to generate adaptive trigger conditions for each line.

[0098] The differential protection threshold value is a current value standard for determining whether a line needs to trigger a differential protection action. Different lines may correspond to different differential protection threshold values. The adaptive trigger condition is generated based on the differential protection threshold values dynamically matched according to the protection priority parameters. It can adaptively adjust the conditions for triggering protection actions according to the actual situation of the line. In the embodiments of the present application, the corresponding differential protection threshold values are dynamically matched from the pre-set threshold value range according to the protection priority parameters of each line, and then the adaptive trigger conditions for each line are generated based on these threshold values. For example, for a line with a high protection priority, the matched differential protection threshold value may be low, and the corresponding adaptive trigger condition is also more stringent. For a line with a low protection priority, the matched differential protection threshold value may be high, and the adaptive trigger condition is relatively loose.

[0099] As an implementation, in step S300, the differential protection threshold corresponding to each line is dynamically matched according to the protection priority parameter, and the adaptive triggering condition of each line is generated, which can specifically include: in step S310, the historical differential current peak value data of each line is obtained, and a reference peak value range matching the current environmental parameter is extracted from the historical differential current peak value data.

[0100] The historical differential current peak value data refers to the maximum value record of the differential current of each line in a period of time in the past, which reflects the current peak value condition of the line under different conditions. The current environmental parameter refers to the environmental condition of the line at present, such as temperature, humidity, voltage, etc. Different environmental parameters can affect the current characteristics of the line. The reference peak value range is the current peak value interval matching the current environmental parameter selected from the historical differential current peak value data, which provides a reference basis for subsequent determination of the differential protection threshold. In the embodiment of the application, the historical differential current peak value data of each line is obtained through the data storage system, and then the reference peak value range matching the current environmental parameter is selected from these data according to the current environmental parameter. For example, the current environmental temperature is 25℃, and the current peak value range when the temperature is around 25℃ is selected from the historical differential current peak value data as the reference peak value range.

[0101] In step S320, the upper limit value of the reference peak value range is adjusted according to the final protection priority parameter to generate an initial differential protection threshold.

[0102] The initial differential protection threshold is a threshold obtained by adjusting the upper limit value of the reference peak value range according to the final protection priority parameter, which preliminarily determines the current standard for triggering the differential protection action of the line. In the embodiment of the application, the upper limit value of the reference peak value range is adjusted according to the final protection priority parameter of each line to obtain the initial differential protection threshold. For example, for the line with a higher final protection priority, the upper limit value of the reference peak value range is appropriately reduced, and the generated initial differential protection threshold is also lower; for the line with a lower final protection priority, the upper limit value of the reference peak value range is appropriately increased, and the generated initial differential protection threshold is also higher.

[0103] In step S330, the signal transmission delay parameter between adjacent nodes of each line is collected in real time, and the initial differential protection threshold is dynamically compensated according to the signal transmission delay parameter.

[0104] The signal transmission delay parameter refers to the time required for signal transmission between adjacent nodes in the line, which reflects the delay of the signal in the transmission process. Since the signal transmission delay can cause errors in current measurement, the initial differential protection threshold needs to be dynamically compensated to improve the accuracy of protection. In the embodiments of the present application, a special measuring device is used to collect the signal transmission delay parameters between adjacent nodes of each line in real time, and then the initial differential protection threshold is adjusted according to these parameters to obtain the compensated differential protection threshold. For example, when the signal transmission delay is large, the initial differential protection threshold is appropriately increased to avoid false triggering caused by delay; when the signal transmission delay is small, the initial differential protection threshold can be appropriately reduced to improve the sensitivity of protection.

[0105] Step S340: If the compensated differential protection threshold of the target line is lower than the preset minimum protection threshold, the preset minimum protection threshold is taken as the final triggering condition of the target line; otherwise, the deviation coefficient of the real-time differential current value is multiplied by the compensated differential protection threshold to generate the adaptive triggering condition of the target line.

[0106] The preset minimum protection threshold is a preset minimum differential protection threshold that ensures the basic safety of the line. When the compensated differential protection threshold is lower than the preset minimum protection threshold, the preset minimum protection threshold is taken as the final triggering condition to ensure the safety of the line. The deviation coefficient is a coefficient calculated according to the difference between the real-time differential current value and the compensated differential protection threshold, which is used to adjust the compensated differential protection threshold to generate an adaptive triggering condition that is more in line with the actual situation. In the embodiments of the present application, the compensated differential protection threshold of the target line is judged. If it is lower than the preset minimum protection threshold, the preset minimum protection threshold is taken as the final triggering condition; otherwise, the compensated differential protection threshold is multiplied by the deviation coefficient to obtain the adaptive triggering condition of the target line. For example, the compensated differential protection threshold of the target line is 5A, the preset minimum protection threshold is 3A, and the deviation coefficient of the real-time differential current value and the compensated differential protection threshold is 0.8. Since 5A is greater than 3A, 5A is multiplied by 0.8 to obtain the adaptive triggering condition of 4A.

[0107] Step S400: Real-time monitoring of the actual differential current value of each line, if the actual differential current value of the target line reaches its corresponding adaptive triggering condition, the differential protection action for the target line is activated.

[0108] The actual differential current value refers to the actual differential current value of the line at the current time, which is obtained in real time by the current measuring device. The differential protection action is a series of protection measures, such as tripping and alarming, which are started in order to protect the line when the actual differential current value of the line reaches the adaptive trigger condition. In the embodiment of the present application, the actual differential current value of each line is monitored in real time by the current monitoring device, and when it is detected that the actual differential current value of the target line reaches the adaptive trigger condition corresponding thereto, the differential protection action for the target line is immediately activated. For example, the adaptive trigger condition of the target line is 4A, and when the actual differential current value of the line is monitored in real time and reaches 4A, the differential protection action is started.

[0109] As an implementation manner, in step S400, if it is detected that the actual differential current value of the target line reaches the adaptive trigger condition corresponding thereto, the differential protection action for the target line is activated, which can specifically include the following step S410: when the actual differential current value of the target line exceeds the adaptive trigger condition for the first time, a protection action delay counter is started, and the harmonic distortion rate of the target line is continuously collected during the delay period.

[0110] The protection action delay counter is a device for recording the delay time from the first time when the actual differential current value exceeds the adaptive trigger condition to the triggering of the protection action, which can avoid the false triggering caused by instantaneous current fluctuation. The harmonic distortion rate refers to the ratio of the harmonic content to the fundamental content of the current in the target line, which reflects the distortion degree of the current waveform, and an excessively high harmonic distortion rate may indicate that the line has a fault. In the embodiment of the present application, when the actual differential current value of the target line exceeds the adaptive trigger condition for the first time, the protection action delay counter is immediately started, and the harmonic distortion rate of the target line is continuously collected during the delay period by a special measuring device. For example, the adaptive trigger condition of the target line is 5A, when the actual differential current value reaches 5.1A, the protection action delay counter is started, and the harmonic distortion rate of the line is continuously collected within 10 seconds of the delay.

[0111] Step S420: if the harmonic distortion rate continuously exceeds the preset distortion threshold during the delay period, the differential protection tripping instruction is triggered.

[0112] The preset distortion threshold is a standard value preset for judging whether the harmonic distortion rate is too high. When the harmonic distortion rate exceeds the threshold, it indicates that the line may have a serious fault, and protection measures need to be taken immediately. The differential protection tripping instruction is an instruction for cutting off the power supply of the line, which can prevent the fault from further expanding. In the embodiments of the present application, if the harmonic distortion rate of the target line continuously exceeds the preset distortion threshold during the protection action delay time, the differential protection tripping instruction is triggered immediately. For example, the preset distortion threshold is 10%, and the harmonic distortion rate of the target line remains above 12% within the 10-second delay time. Then, the differential protection tripping instruction is triggered, and the power supply of the line is cut off.

[0113] Step S430: If the harmonic distortion rate does not exceed the preset distortion threshold, the deviation of the actual differential current value of the target line from the adaptive trigger condition is re-verified after the delay time ends.

[0114] The deviation refers to the difference between the actual differential current value of the target line and the adaptive trigger condition, which reflects the degree of deviation of the actual current value from the trigger condition. In the embodiments of the present application, if the harmonic distortion rate of the target line does not exceed the preset distortion threshold during the protection action delay time, the actual differential current value of the target line is measured again, and the deviation from the adaptive trigger condition is calculated. For example, the adaptive trigger condition of the target line is 6A, and the actual differential current value measured after the delay time ends is 6.2A. The calculated deviation is 0.2A.

[0115] Step S440: When the deviation is greater than the preset buffer interval, the differential protection alarm signal is triggered and the standby line switching protocol is started.

[0116] The preset buffer interval is a range preset for allowing the actual differential current value to fluctuate around the adaptive trigger condition. When the deviation exceeds the range, it indicates that the line may have a potential fault. The differential protection alarm signal is a signal for reminding the operation and maintenance personnel that the line may have a problem. The standby line switching protocol is a series of operating rules for switching the load to the standby line when the main line fails. In the embodiments of the present application, when the deviation of the actual differential current value of the target line from the adaptive trigger condition is greater than the preset buffer interval, the differential protection alarm signal is triggered, and the standby line switching protocol is started. For example, the preset buffer interval is 0.1A, and the deviation of the target line is 0.2A, which is greater than the preset buffer interval. Therefore, the differential protection alarm signal is triggered, and the standby line switching protocol is started to switch the load to the standby line.

[0117] Step S450: When the deviation is less than or equal to the preset buffer interval, the protection action delay counter is reset and the actual differential current value of the target line is continued to be monitored.

[0118] In the embodiments of the present application, if the deviation of the actual differential current value of the target line from the adaptive triggering condition is less than or equal to the preset buffer interval, it indicates that the current fluctuation of the line is within an acceptable range, at which time the protection action delay counter is reset, so as to restart timing when the current anomaly is monitored again next time, and the actual differential current value of the target line is continuously monitored in real time. For example, the preset buffer interval is 0.1 A, and the deviation of the target line is 0.05 A, which is less than the preset buffer interval. The protection action delay counter is reset, and the actual differential current value of the line is continuously monitored.

[0119] Step S500: adjusting the monitoring frequency of the lines that have not triggered the protection action according to the priority parameters output by the adaptive multi-transmission line model.

[0120] The monitoring frequency refers to the time interval of data acquisition and state monitoring of the line. Adjusting the monitoring frequency can reasonably allocate monitoring resources according to the importance and state of the line. In the embodiments of the present application, the monitoring frequency of the lines that have not triggered the protection action is adjusted according to the priority parameters of the lines output by the adaptive multi-transmission line model. For the lines with high priority, the monitoring frequency is increased, so as to timely discover potential problems. For the lines with low priority, the monitoring frequency is appropriately reduced, so as to reduce the waste of monitoring resources. For example, the monitoring frequency of a line is increased from once every hour to once every half an hour, because the priority of the line is high according to the adaptive multi-transmission line model.

[0121] As an implementation manner, in step S500, the monitoring frequency of the lines that have not triggered the protection action is adjusted according to the priority parameters output by the adaptive multi-transmission line model, which can specifically include: step S510: obtaining the final protection priority parameters corresponding to the lines output by the adaptive multi-transmission line model, and dividing the final protection priority parameters into a high priority interval, a medium priority interval and a low priority interval.

[0122] The final protection priority parameter is a parameter for measuring the protection priority of the line obtained by comprehensively considering various factors of the line. The high priority interval, the medium priority interval and the low priority interval are ranges for dividing the priority of the line that are preset. In the embodiments of the present application, the final protection priority parameters corresponding to the lines are obtained from the adaptive multi-transmission line model, and then the parameters are divided into the high priority interval, the medium priority interval and the low priority interval according to the preset ranges. For example, the final protection priority parameter range [1, 10] is divided into the high priority interval [7, 10], the medium priority interval [3, 6] and the low priority interval [1, 2], and the lines are divided into the corresponding intervals according to the final protection priority parameters of the lines.

[0123] Step S520: generating an initial monitoring frequency adjustment factor corresponding to each priority interval to which the line belongs, wherein the first adjustment factor corresponds to the high priority interval, the second adjustment factor corresponds to the medium priority interval, and the third adjustment factor corresponds to the low priority interval.

[0124] The initial monitoring frequency adjustment factor is a coefficient for adjusting the initial monitoring frequency of the line, and different priority intervals correspond to different adjustment factors. The first adjustment factor, the second adjustment factor, and the third adjustment factor are adjustment factors corresponding to the high priority interval, the medium priority interval, and the low priority interval, respectively, and have different values for different priority intervals. In the embodiment of the present application, the initial monitoring frequency adjustment factor corresponding to each priority interval to which the line belongs is generated. For example, the first adjustment factor corresponding to the high priority interval is 1.5, the second adjustment factor corresponding to the medium priority interval is 1, and the third adjustment factor corresponding to the low priority interval is 0.5.

[0125] Step S530: collecting the signal interference intensity and the line load fluctuation rate between adjacent nodes of the line that has not triggered the protection action in real time, and inputting the signal interference intensity and the load fluctuation rate into the dynamic frequency compensation model to output the real-time dynamic adjustment parameter of each line.

[0126] The signal interference intensity refers to the degree of interference of the signal between adjacent nodes of the line, which affects the transmission quality of the signal and the accuracy of the current measurement. The line load fluctuation rate refers to the fluctuation degree of the line load over time, which reflects the stability of the line load. The dynamic frequency compensation model is a pre-trained model that can output the real-time dynamic adjustment parameter of each line according to the input signal interference intensity and line load fluctuation rate, which is used to further adjust the monitoring frequency of the line. In the embodiment of the present application, the signal interference intensity and the line load fluctuation rate between adjacent nodes of the line that has not triggered the protection action are collected in real time by a special measuring device, for example, a signal strength tester is used to measure the signal interference intensity between adjacent nodes, and the line load fluctuation rate is calculated by statistical analysis of the line load data. Then, the collected signal interference intensity and load fluctuation rate are input as input data into the dynamic frequency compensation model. The model outputs the real-time dynamic adjustment parameter corresponding to each line after internal calculation and analysis.

[0127] Exemplarily, the dynamic frequency compensation model can adopt a long short-term memory network (LSTM) including an input layer, an LSTM layer, and an output layer. The input layer receives the two sequence data of the signal interference intensity and the line load fluctuation rate, and converts them into a format suitable for processing by the LSTM layer. The LSTM layer is composed of a plurality of LSTM units, each of which includes an input gate, a forgetting gate, and an output gate, and controls the flow and memory of information through these gate mechanisms, thereby learning the patterns and rules of changes of the signal interference intensity and the line load fluctuation rate over time. The output layer outputs the real-time dynamic adjustment parameter of each line through linear transformation and other operations based on the output of the LSTM layer. In the training stage, a large amount of historical signal interference intensity, line load fluctuation rate data, and corresponding actual monitoring frequency adjustment situations can be used as training samples, and the parameters of the model are continuously adjusted through a back propagation algorithm, so that the model can accurately predict the appropriate real-time dynamic adjustment parameter according to the input signal interference intensity and line load fluctuation rate, thereby improving the accuracy and effectiveness of the line monitoring frequency adjustment and better adapting to the real-time operation state of the line.

[0128] Step S540: superimposing the initial monitoring frequency adjustment factor and the corresponding real-time dynamic adjustment parameter to generate the final monitoring frequency coefficient of each line that has not triggered the protection action, and adjusting the sampling interval of the data acquisition equipment thereof according to the final monitoring frequency coefficient.

[0129] The final monitoring frequency coefficient is a coefficient obtained by superimposing the initial monitoring frequency adjustment factor and the real-time dynamic adjustment parameter, which comprehensively considers the priority and real-time state of the line and is used to determine the final monitoring frequency of the line. The sampling interval of the data acquisition equipment refers to the time interval for one data acquisition of the data acquisition equipment, and the adjustment of the sampling interval can realize the adjustment of the line monitoring frequency. In the embodiment of the present application, the initial monitoring frequency adjustment factor of each line that has not triggered the protection action is added to the corresponding real-time dynamic adjustment parameter to obtain the final monitoring frequency coefficient, and then the sampling interval of the data acquisition equipment is adjusted according to the coefficient. For example, the initial monitoring frequency adjustment factor of a certain line is 1.2, the real-time dynamic adjustment parameter is 0.8, the sum of which is 2, and the sampling interval of the data acquisition equipment is shortened from the original 10 minutes to 5 minutes according to the coefficient.

[0130] Step S550: continuously acquiring the differential current feature set of each line under the adjusted sampling interval, and if it is detected that the differential current feature set of the same line does not appear a feature subset exceeding the preset fluctuation threshold in a plurality of continuous sampling periods, the final monitoring frequency coefficient thereof is reduced by a preset attenuation ratio.

[0131] The preset fluctuation threshold is a preset standard value for determining whether the differential current feature set is abnormally fluctuated, and the preset attenuation ratio is a preset ratio for reducing the final monitoring frequency coefficient. In the embodiment of the application, the differential current feature set of each line is continuously collected under the adjusted sampling interval. If it is found that the differential current feature set of the same line does not have a feature subset exceeding the preset fluctuation threshold in a plurality of continuous sampling periods, it is indicated that the line state is relatively stable, and the final monitoring frequency coefficient thereof is reduced according to the preset attenuation ratio. For example, the preset fluctuation threshold is 5%, the preset attenuation ratio is 0.1, and the amplitude fluctuation rate of the differential current feature set of a line does not exceed 5% in five continuous sampling periods, and then the final monitoring frequency coefficient of the line is reduced from 2 to 1.8.

[0132] Step S560: feeding back the reduced final monitoring frequency coefficient to the dynamic frequency compensation model to update the weight distribution relationship of the real-time dynamic adjustment parameter, and forming a closed-loop monitoring frequency regulation mechanism based on the linkage of the priority parameter and the line state.

[0133] The weight distribution relationship is the weight proportion of each factor for calculating the real-time dynamic adjustment parameter in the dynamic frequency compensation model. In the embodiment of the application, the reduced final monitoring frequency coefficient is fed back to the dynamic frequency compensation model, and the model updates the weight distribution relationship of the real-time dynamic adjustment parameter according to the new coefficient, thereby forming a closed-loop monitoring frequency regulation mechanism based on the linkage of the priority parameter and the line state. This mechanism can automatically adjust the monitoring frequency according to the priority and real-time state of the line, and realize the optimized allocation of monitoring resources. For example, the reduced final monitoring frequency coefficient is 1.8, which is fed back to the dynamic frequency compensation model, and the model updates the weight distribution relationship of the signal interference intensity and the line load fluctuation rate, so that the subsequently calculated real-time dynamic adjustment parameter is more consistent with the actual state of the line.

[0134] As an implementation mode, after the closed-loop monitoring frequency regulation mechanism based on the linkage of the priority parameter and the line state is formed in step S560, the method provided by the application can further include: step S570: generating a dynamic monitoring priority queue of each line based on the final monitoring frequency coefficient of each line in the closed-loop monitoring frequency regulation mechanism, and allocating the hardware resource occupation proportion of the data acquisition equipment according to the dynamic monitoring priority queue.

[0135] The dynamic monitoring priority queue is generated according to the final monitoring frequency coefficient of each line, and the lines in the queue are arranged in descending order of monitoring priority. The hardware resource occupation ratio refers to the proportion of the hardware resources allocated by the data acquisition device when monitoring different lines. Reasonable allocation of the hardware resource occupation ratio can improve the monitoring efficiency. In the embodiment of the present application, the dynamic monitoring priority queue of each line is generated according to the final monitoring frequency coefficient of each line in the closed-loop monitoring frequency regulation mechanism, and then the hardware resource occupation ratio of the data acquisition device is allocated according to the queue. For example, the line with a higher final monitoring frequency coefficient ranks higher in the dynamic monitoring priority queue, and the hardware resource occupation ratio of the data acquisition device allocated to the line is also higher.

[0136] Step S580: Real-time acquisition of the differential current feature set collected by each line under the adjusted sampling interval, and extraction of the potential abnormal segment in the differential current feature set that matches the historical abnormal waveform.

[0137] The potential abnormal segment refers to the part in the differential current feature set that is similar to the historical abnormal waveform. These segments may indicate potential faults in the line. In the embodiment of the present application, the differential current feature set of each line is continuously collected under the adjusted sampling interval, and then the collected differential current feature set is compared with the historical abnormal waveform to extract the potential abnormal segment that matches the historical abnormal waveform. For example, the historical abnormal waveform shows a sudden increase in current amplitude, and a segment of the collected differential current feature set also shows a sudden increase in current amplitude. The segment is extracted as a potential abnormal segment.

[0138] Step S590: Inputting the potential abnormal segment into the abnormal prediction module in the adaptive multi-transmission line model, and outputting the real-time abnormal confidence parameter of each line.

[0139] The abnormal prediction module is a module in the adaptive multi-transmission line model for analyzing and judging the potential abnormal segment, and the real-time abnormal confidence parameter is a parameter output by the abnormal prediction module according to the input potential abnormal segment, which represents the possibility of the line appearing abnormal. In the embodiment of the present application, the extracted potential abnormal segment is input into the abnormal prediction module of the adaptive multi-transmission line model, and the module outputs the real-time abnormal confidence parameter of each line after analysis and calculation. For example, the potential abnormal segment of a certain line is input into the abnormal prediction module, and the module outputs the real-time abnormal confidence parameter of the line as 0.8, indicating that the line has a high possibility of appearing abnormal.

[0140] Exemplarily, the abnormality prediction module can adopt a convolutional neural network (CNN) architecture, including an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer. The input layer receives potential abnormal segment data and converts it into a format suitable for processing by the convolutional layer. The convolutional layer is composed of multiple convolutional kernels that extract local features in the potential abnormal segment through convolution operations. Each convolutional kernel learns different feature patterns, such as trends in current amplitude and fluctuations in frequency. The pooling layer down-samples the output of the convolutional layer, reducing the data dimension while preserving important feature information and enhancing the robustness of the model. The fully connected layer integrates the features output by the pooling layer, establishing global connections between the features. The output layer is processed by a linear transformation and an activation function (such as a Sigmoid function) to output real-time abnormal confidence parameters for each line. The Sigmoid function maps the output values to a range of 0 to 1 to represent the likelihood of line abnormalities. During the training phase, historical potential abnormal segment data and corresponding actual abnormal conditions can be used as training samples to continuously adjust the model parameters through a general backpropagation algorithm. This enables the abnormality prediction module to accurately output reasonable real-time abnormal confidence parameters based on input potential abnormal segments, improving the accuracy of line abnormality prediction.

[0141] Step S5100: Dynamically adjust the sorting weight of the dynamic monitoring priority queue according to the real-time abnormal confidence parameters, and reallocate the hardware resource occupation proportion to prioritize processing of lines with real-time abnormal confidence parameters higher than a preset threshold.

[0142] The sorting weight is a weight value used to determine the order of lines in the dynamic monitoring priority queue, and the preset threshold is a standard value used to determine whether the likelihood of line abnormalities is high. In the embodiments of the present application, the sorting weight of the dynamic monitoring priority queue is dynamically adjusted according to the real-time abnormal confidence parameters of each line, and lines with real-time abnormal confidence parameters higher than the preset threshold are placed at the front. Then, the hardware resource occupation proportion of the data acquisition device is reallocated according to the new queue to prioritize monitoring and processing of these lines. For example, if the preset threshold is 0.7 and the real-time abnormal confidence parameter of a line is 0.8, the sorting weight of the line in the dynamic monitoring priority queue is increased, and the hardware resource occupation proportion of the data acquisition device allocated to the line is correspondingly increased.

[0143] Step S5110: Coupling the reallocated hardware resource occupation proportion with the final monitoring frequency coefficient to generate an optimized monitoring strategy for each line, and updating the frequency adjustment rule in the closed-loop monitoring frequency regulation mechanism according to the optimized monitoring strategy.

[0144] The coupling is an operation of comprehensively considering and combining the re-allocated hardware resource occupation ratio and the final monitoring frequency coefficient, and the optimized monitoring strategy is a more reasonable and effective monitoring strategy obtained through the coupling, which comprehensively considers the priority of the line, the abnormal possibility and the hardware resource allocation. The frequency adjustment rule is a rule for adjusting the line monitoring frequency in the closed-loop monitoring frequency regulation mechanism, and updating the frequency adjustment rule in the closed-loop monitoring frequency regulation mechanism according to the optimized monitoring strategy can make the closed-loop monitoring frequency regulation mechanism more accurate and effective. In the embodiment of the application, the re-allocated hardware resource occupation ratio and the final monitoring frequency coefficient are coupled to generate the optimized monitoring strategy of each line, and then the frequency adjustment rule in the closed-loop monitoring frequency regulation mechanism is updated according to the strategy. For example, the hardware resource occupation ratio of a certain line after re-allocation increases, and the final monitoring frequency coefficient is also high, and the optimized monitoring strategy generated after coupling is to further improve the monitoring frequency of the line, and the frequency adjustment rule in the closed-loop monitoring frequency regulation mechanism is updated according to the strategy, so that the line can be paid more attention in subsequent monitoring.

[0145] As an implementation manner, after the optimized monitoring strategy of each line is generated in step S5110, the method provided by the application can further include: step S5120, constructing an abnormal association network across lines according to the real-time abnormal confidence parameters of the lines in the optimized monitoring strategy, wherein each node represents a line, and the connection weight between the nodes represents the synchronous change rate of the abnormal confidence parameters of the two lines.

[0146] The abnormal association network is a network for representing the abnormal association relationship between the lines, the node represents the line, and the connection weight represents the synchronous change rate of the abnormal confidence parameters of the two lines, and the synchronous change rate reflects the correlation of the abnormal conditions of the two lines. In the embodiment of the application, the abnormal association network across lines is constructed according to the real-time abnormal confidence parameters of the lines in the optimized monitoring strategy. For example, there are three lines A, B and C, the synchronous change rate of the abnormal confidence parameters of line A and line B is 0.6, the synchronous change rate of line A and line C is 0.3, and the synchronous change rate of line B and line C is 0.4, and the abnormal association network is constructed according to these data, and the connection weights between nodes A, B and C are 0.6, 0.3 and 0.4 respectively.

[0147] Step S5130: traversing the node pairs in the abnormal association network whose connection weights exceed a preset synchronization threshold, and marking the node pairs as an associated abnormal group.

[0148] The preset synchronization threshold is a standard value preset for judging whether abnormal conditions of two lines are highly related. The associated abnormal group is a group composed of node pairs in the abnormal association network whose connection weights exceed the preset synchronization threshold. The abnormal conditions of these lines can be mutually influenced. In the embodiments of the present application, the connection weights of all node pairs in the abnormal association network are traversed to determine the node pairs whose connection weights exceed the preset synchronization threshold, and these node pairs are marked as the associated abnormal group. For example, the preset synchronization threshold is 0.5, and in the abnormal association network, the connection weight of line A and line B is 0.6, which exceeds the preset synchronization threshold. Line A and line B are marked as an associated abnormal group.

[0149] Step S5140: For each line in the associated abnormal group, the final monitoring frequency coefficient of the line is synchronously adjusted so that the lines in the same associated abnormal group use the same sampling interval for data collection.

[0150] In the embodiments of the present application, for each line group marked as an associated abnormal group, the final monitoring frequency coefficients of the lines in the group are synchronously adjusted so that the lines use the same sampling interval for data collection. In this way, the monitoring consistency of the lines in the associated abnormal group can be ensured, and cross-line abnormal conditions can be found and handled. For example, an associated abnormal group includes line A and line B. The final monitoring frequency coefficients of line A and line B are adjusted to the same value so that the data collection devices of line A and line B use the same sampling interval for data collection.

[0151] Step S5150: Under the synchronously adjusted sampling interval, the differential current feature sets of all lines in the associated abnormal group are collected, and it is detected whether there is a cross-line cascading abnormal waveform mode.

[0152] The cascading abnormal waveform mode refers to an abnormal waveform mode that simultaneously occurs in multiple lines in the associated abnormal group and has correlation, which can indicate that there is a cross-line fault. In the embodiments of the present application, under the synchronously adjusted sampling interval, the differential current feature sets of all lines in the associated abnormal group are collected, and then the collected feature sets are analyzed to detect whether there is a cross-line cascading abnormal waveform mode. For example, in two lines in an associated abnormal group, a waveform mode in which the current amplitude suddenly increases and the phase changes similarly is simultaneously detected, which is regarded as a cross-line cascading abnormal waveform mode.

[0153] Step S5160: If the cascading abnormal waveform mode is detected, a global protection coordination protocol is triggered, the final monitoring frequency coefficients of all lines in the associated abnormal group are forcibly increased to the preset highest monitoring frequency, and the validity duration of the global protection coordination protocol is locked.

[0154] The global protection coordination protocol is a protocol for coordinating the handling of cross-line abnormal conditions. The preset maximum monitoring frequency is a preset maximum monitoring frequency that ensures timely detection and handling of abnormal conditions. The effective duration is the effective time range of the global protection coordination protocol. In the embodiments of the present application, if a cross-line cascading abnormal waveform mode is detected in the associated abnormal group, the global protection coordination protocol is triggered immediately, the final monitoring frequency coefficient of all lines in the associated abnormal group is forced to be raised to the preset maximum monitoring frequency, and the effective duration of the global protection coordination protocol is locked. For example, the preset maximum monitoring frequency is once per minute, and after the cascading abnormal waveform mode is detected, the final monitoring frequency coefficient of all lines in the associated abnormal group is adjusted to the monitoring frequency corresponding to once per minute, and the effective duration of the global protection coordination protocol is locked for 30 minutes.

[0155] Step S5170: After the effective duration ends, the global protection coordination protocol is unlocked, and the final monitoring frequency coefficient of each line based on the closed-loop monitoring frequency regulation mechanism is restored.

[0156] After the effective duration of the global protection coordination protocol ends, the protocol is unlocked, and the preset maximum monitoring frequency is no longer forced to be maintained. Instead, the final monitoring frequency coefficient of each line based on the closed-loop monitoring frequency regulation mechanism is restored, so that the monitoring frequency of the line returns to the normal regulation state. For example, the effective duration of the global protection coordination protocol is 30 minutes, and after 30 minutes, the protocol is unlocked, and the final monitoring frequency coefficient of each line is restored to the value determined by the closed-loop monitoring frequency regulation mechanism.

[0157] As an implementation manner, after the final monitoring frequency coefficient of each line based on the closed-loop monitoring frequency regulation mechanism is restored in step S5170, the method provided by the present application can further include: step S5180: acquiring the differential current feature set of all lines in the associated abnormal group collected during the effective duration of the global protection coordination protocol, and extracting abnormal waveform recovery parameters in the differential current feature set.

[0158] The abnormal waveform recovery parameter is a parameter for describing the recovery of the abnormal waveform in the differential current feature set, which can reflect the recovery degree of the abnormal state of the line. In the embodiments of the present application, the differential current feature set of all lines in the associated abnormal group collected during the effective duration of the global protection coordination protocol is acquired from the data storage system, and then the feature set is analyzed to extract the abnormal waveform recovery parameters. For example, by analyzing the changes of current amplitude, frequency, phase and other parameters, abnormal waveform recovery parameters such as the time when the current amplitude returns to the normal range and the convergence degree of frequency fluctuation are extracted.

[0159] Step S5190: judging whether the abnormal state of each line has been eliminated according to the abnormal waveform recovery parameter, if yes, generating a corresponding abnormal recovery mark, otherwise generating an abnormal persistence mark.

[0160] The abnormal recovery mark is a mark for indicating that the abnormal state of the line has been eliminated, and the abnormal persistence mark is a mark for indicating that the abnormal state of the line still exists. In the embodiment of the application, according to the extracted abnormal waveform recovery parameter, the abnormal state of each line is judged, if the abnormal state has been eliminated, a corresponding abnormal recovery mark is generated, and if the abnormal state still exists, an abnormal persistence mark is generated. For example, the abnormal waveform recovery parameter of a certain line shows that the current amplitude and frequency have been recovered to the normal range, it is judged that the abnormal state of the line has been eliminated, and an abnormal recovery mark is generated; the abnormal waveform recovery parameter of another line shows that the current amplitude still fluctuates greatly, it is judged that the abnormal state of the line still exists, and an abnormal persistence mark is generated.

[0161] Step S5200: inputting the abnormal recovery mark into the closed-loop monitoring frequency regulation mechanism, triggering the gradient attenuation mechanism of the final monitoring frequency coefficient of the corresponding line, and gradually reducing the sampling interval thereof until reaching the initial set value in the closed-loop monitoring frequency regulation mechanism.

[0162] The gradient attenuation mechanism is a mechanism for gradually reducing the final monitoring frequency coefficient in the closed-loop monitoring frequency regulation mechanism, which can make the monitoring frequency of the line gradually recover to the normal level after the abnormal state is eliminated. In the embodiment of the application, the abnormal recovery mark is input into the closed-loop monitoring frequency regulation mechanism, the gradient attenuation mechanism of the final monitoring frequency coefficient of the corresponding line is triggered, the final monitoring frequency coefficient of the line is gradually reduced according to a certain gradient, so as to gradually increase the sampling interval thereof, until reaching the initial set value in the closed-loop monitoring frequency regulation mechanism. For example, after the abnormal state of a certain line is eliminated, the gradient attenuation mechanism is triggered, the final monitoring frequency coefficient is reduced by 0.1 every time interval, until reaching the initial set value.

[0163] Step S5210: inputting the abnormal persistence mark into the priority reevaluation module in the adaptive multi-transmission line model, recalculating the final protection priority parameter of the corresponding line, and updating the final monitoring frequency coefficient thereof according to the recalculated parameter.

[0164] The priority reevaluation module is a module in the adaptive multi-transmission line model for reevaluating the line priority, which can recalculate the final protection priority parameter according to the latest state of the line. In the embodiments of the present application, the abnormal duration mark is input into the priority reevaluation module of the adaptive multi-transmission line model, which takes the abnormal duration mark as input, and combines the historical operation data of the line, the previous protection priority parameter and other related line state information to recalculate the final protection priority parameter of the corresponding line. For example, the time of abnormal duration, the severity of the abnormality and other factors are analyzed, and these factors are comprehensively considered and calculated according to the preset weight and rules.

[0165] Exemplarily, the priority reevaluation module can be implemented by adopting an architecture combining rules and machine learning. The rule part contains a series of pre-set rules, for example, if the abnormal duration time exceeds the preset time length, the final protection priority parameter of the line is appropriately increased; if the severity of the abnormality reaches a certain level, the priority parameter is also adjusted accordingly. The machine learning can adopt a decision tree algorithm. The decision tree takes the abnormal duration mark and other related line state information as input features, and constructs a decision tree model by learning from historical data. Each internal node in the decision tree model is a test on a feature, each branch is a test output, and each leaf node is a class (which can be understood as different levels of final protection priority parameters here). When a new abnormal duration mark and related information are input, the decision tree makes a judgment according to the internal decision rules and outputs the corresponding final protection priority parameter. After obtaining the recalculated final protection priority parameter, the priority reevaluation module updates the final monitoring frequency coefficient of the line according to the pre-set corresponding relationship. For example, if the final protection priority parameter is increased, the final monitoring frequency coefficient is correspondingly increased, so as to increase the monitoring frequency of the line, so as to more timely discover the further abnormal situation of the line; on the contrary, if the final protection priority parameter is decreased, the final monitoring frequency coefficient is appropriately decreased, so as to reduce unnecessary monitoring resource consumption. Based on this, the priority reevaluation module can dynamically adjust the priority and monitoring frequency of the line when the line has an abnormal duration, so that the whole line monitoring system is more flexible and efficient, and the response capability to the abnormal situation of the line is enhanced.

[0166] Step S5220: synchronizing the updated final monitoring frequency coefficient to the dynamic monitoring priority queue and the connection weight calculation of the associated abnormal group, forming a dynamic frequency backtracking mechanism based on the abnormal recovery state, so as to adaptively balance the resource allocation and the abnormal detection sensitivity in the subsequent monitoring period.

[0167] The dynamic frequency backtracking mechanism is a mechanism that dynamically adjusts the monitoring frequency based on the abnormal recovery status of a line. It can adaptively balance the allocation of monitoring resources and the sensitivity of anomaly detection in subsequent monitoring cycles. In this embodiment, the updated final monitoring frequency coefficient is synchronized to the dynamic monitoring priority queue and the connection weight calculation of associated anomaly groups. This allows these mechanisms to adjust according to the latest state of the line, forming a dynamic frequency backtracking mechanism based on the anomaly recovery status. For example, after the final monitoring frequency coefficient of a line is updated, it is synchronized to the dynamic monitoring priority queue, adjusting the line's position in the queue. Simultaneously, the new state of the line is considered in the connection weight calculation of associated anomaly groups, thereby more rationally allocating resources and improving the sensitivity of anomaly detection in subsequent monitoring.

[0168] This invention provides an optical fiber differential protection decision system, such as... Figure 2 As shown, the fiber optic differential protection decision system 100 includes a processor 101 and a memory 103. The processor 101 and the memory 103 are connected, for example, via a bus 102. Optionally, the fiber optic differential protection decision system 100 may further include a transceiver 104. It should be noted that in practical applications, the transceiver 104 is not limited to one, and the structure of this fiber optic differential protection decision system 100 does not constitute a limitation on the embodiments of the present invention.

[0169] Processor 101 may be a CPU, a general-purpose processor, a GPU, a DSP, an ASIC, an FPGA, or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in connection with this disclosure. Processor 101 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0170] Bus 102 may include a pathway for transmitting information between the aforementioned components. Bus 102 may be a PCI bus or an EISA bus, etc. Bus 102 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 2 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0171] The memory 103 can be a ROM, or other type of static storage devices that can store static information and instructions, a RAM, or other type of dynamic storage device that can store information and instructions, an EEPROM, a CD-ROM or other optical disk storage, a magneto-optical disk storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program codes in the form of instructions or data structures and that can be accessed by a computer, but not limited to this.

[0172] The memory 103 is used to store application program codes for implementing the solutions of the present application, and is controlled by the processor 101 to perform. The processor 101 is used to execute the application program codes stored in the memory 103 to realize the content shown in any of the foregoing method embodiments.

[0173] The embodiment of the present application provides a fiber differential protection decision system, the fiber differential protection decision system in the embodiment of the present application includes: one or more processors;Memory;One or more computer programs, wherein the one or more computer programs are stored in the memory and are configured to be executed by the one or more processors, and when the one or more programs are executed by the processor, the above method is realized.

[0174] The embodiment of the present application provides a computer readable storage medium, and the computer readable storage medium stores a computer program, and when the computer program runs on the processor, the processor can execute the corresponding content in the foregoing method embodiments.

[0175] It should be understood that, although each step in the flowchart of the accompanying drawings is shown in sequence according to the direction of the arrow, these steps are not necessarily executed in sequence according to the direction of the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and they can be executed in other sequences. Moreover, at least part of the steps in the flowchart of the accompanying drawings can include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution sequence is not necessarily sequential, but can be executed alternately or alternately with at least part of other steps or sub-steps or stages of other steps.

[0176] The above only describes some embodiments of the present application, and it should be pointed out that, for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present application.

Claims

1. A method for fiber differential protection decision based on adaptive multi-transmission line, characterized in that, The method comprises: collecting real-time optical signal waveform data of each line in the multi-transmission line, and extracting a differential current feature set corresponding to each line; generating a joint feature sequence based on the differential current feature sets of all lines, and inputting the joint feature sequence into a pre-trained adaptive multi-transmission line model to obtain a protection priority parameter corresponding to each line; specifically including: selecting a feature subset with an amplitude fluctuation rate exceeding a preset fluctuation threshold from the differential current feature set of each line, and aligning each feature subset according to the time stamp; calculating the energy distribution parameter of the feature subset of each line within a preset time window, and extracting the maximum energy gradient value in the energy distribution parameter of each line; splicing the maximum energy gradient values of each line into an initial joint sequence according to the line number order, and performing normalization processing on the initial joint sequence; matching the normalized initial joint sequence with a standard mode in a historical joint feature library, and eliminating abnormal feature segments in the initial joint sequence with a coincidence degree exceeding a preset coincidence degree with historical abnormal modes; rearranging the remaining feature segments in time sequence to generate a final joint feature sequence; dividing the final joint feature sequence into a plurality of continuous time segments, each time segment containing a fixed number of feature points; inputting each time segment into a feature fusion layer in the adaptive multi-transmission line model respectively, and outputting a line state weight coefficient corresponding to each time segment; calculating the cumulative weight value of each line within the overall monitoring period according to the line state weight coefficient of each time segment, and sorting each line according to the cumulative weight value from high to low; mapping the sorted line number to a preset priority interval to generate a dynamic protection priority parameter corresponding to each line; weighting and superimposing the dynamic protection priority parameter and a real-time load parameter to obtain a final protection priority parameter; dynamically matching the differential protection threshold corresponding to each line according to the protection priority parameter to generate an adaptive trigger condition for each line; monitoring the actual differential current value of each line in real time, and if the actual differential current value of the target line reaches its corresponding adaptive trigger condition, activating the differential protection action for the target line; adjusting the monitoring frequency of the lines without triggering the protection action according to the priority parameter output by the adaptive multi-transmission line model.

2. The method of claim 1, wherein, The adaptive triggering condition of each line is generated by dynamically matching the differential protection threshold of each line according to the protection priority parameter, including: obtaining historical differential current peak value data of each line, and extracting a reference peak value range matching the current environmental parameter from the historical differential current peak value data; adjusting the upper limit value of the reference peak value range according to the final protection priority parameter to generate an initial differential protection threshold; collecting the signal transmission delay parameter between adjacent nodes of each line in real time, and dynamically compensating the initial differential protection threshold according to the signal transmission delay parameter; if it is detected that the compensated differential protection threshold of the target line is lower than the preset minimum protection threshold, the preset minimum protection threshold is taken as the final triggering condition of the target line; otherwise, the deviation coefficient of the compensated differential protection threshold and the real-time differential current value is multiplied to generate the adaptive triggering condition of the target line.

3. The method of claim 2, wherein, If it is detected that the actual differential current value of the target line reaches its corresponding adaptive triggering condition, the differential protection action for the target line is activated, including: when the actual differential current value of the target line exceeds its adaptive triggering condition for the first time, a protection action delay counter is started, and the harmonic distortion rate of the target line is continuously collected during the delay period; if the harmonic distortion rate continuously exceeds the preset distortion threshold during the delay period, a differential protection tripping instruction is triggered; if the harmonic distortion rate does not exceed the preset distortion threshold, the deviation amount of the actual differential current value of the target line and the adaptive triggering condition is re-verified after the delay ends; when the deviation amount is greater than a preset buffer interval, a differential protection alarm signal is triggered and a backup line switching protocol is started; when the deviation amount is less than or equal to the preset buffer interval, the protection action delay counter is reset and the actual differential current value of the target line is continuously monitored.

4. The method of claim 1, wherein, The training process of the pre-trained adaptive multi-transmission line model includes: collecting a plurality of groups of historical transmission line optical signal waveform sample data, and extracting a differential current feature vector corresponding to each group of sample data; performing joint time domain and frequency domain analysis on the differential current feature vector to generate a feature label sequence of each group of sample data; inputting the feature label sequence into a convolution kernel alignment layer in the initial adaptive multi-transmission line model to output alignment weights of each feature label in the time dimension; dynamically sampling the feature label sequence based on the alignment weights to generate a standardized feature set for training; iteratively training the initial adaptive multi-transmission line model using the standardized feature set until the error rate of the protection priority parameter output by the initial adaptive multi-transmission line model and the preset reference parameter is lower than a preset error threshold, and obtaining the adaptive multi-transmission line model.

5. The method of claim 4, wherein, The time domain and frequency domain joint analysis on the differential current feature vector is performed to generate a feature label sequence of each group of sample data, including: segmenting the differential current feature vector into multiple time domain segments, and calculating the average amplitude and phase offset of each time domain segment; performing fast Fourier transform on each time domain segment to extract a dominant frequency component in a frequency energy distribution of the each time domain segment; matching the dominant frequency component with a preset typical fault frequency library to label a potential fault type corresponding to each time domain segment; weighting and fusing the average amplitude and phase offset of the time domain segment according to the potential fault type to generate a time-frequency joint feature parameter; and arranging all time-frequency joint feature parameters in the same group of sample data in time sequence to form the feature label sequence.

6. The method of claim 5, wherein, The convolution kernel alignment layer in the initial adaptive multi-transmission line model is input with the feature label sequence to output alignment weights of each feature label in the time dimension, including: setting multiple sliding windows of different scales in the convolution kernel alignment layer, each sliding window being used to capture a local mode in the feature label sequence; calculating a similarity score between feature labels covered by each sliding window and a preset reference mode; dynamically adjusting a step size and a coverage range of each sliding window according to the similarity score to generate multi-scale alignment parameters; inputting the multi-scale alignment parameters into a weight distribution network to output a position weight of each feature label on a time axis; and performing nonlinear interpolation on the feature label sequence according to the position weight to generate a standardized feature sequence after time dimension alignment.

7. The method of claim 6, wherein, The dynamic sampling of the feature mark sequence based on the alignment weight generates a standardized feature set for training, including: screening feature marks with weight values higher than a preset weight threshold as key feature points according to the position weight; performing mean value filling on the interval regions between the key feature points to generate a continuous feature trajectory curve; resampling the feature trajectory curve at equal time intervals to obtain uniformly distributed feature sampling points; binding the feature sampling points with corresponding potential fault type labels to generate standardized training samples; randomly shuffling and batch dividing the standardized training samples of all historical transmission lines to generate a standardized feature set for final training; and the iterative training of the initial adaptive multi-transmission line model using the standardized feature set, including: in each iteration process, inputting the standardized training samples of the current batch into the initial adaptive multi-transmission line model to output predicted protection priority parameters; calculating the mean square error between the predicted protection priority parameters and corresponding reference parameters, and updating the weight of the initial adaptive multi-transmission line model according to the mean square error; after each iteration, randomly selecting samples from the validation set for model performance evaluation, calculating the fault detection accuracy and false trigger rate thereof; if the fault detection accuracy of continuous multiple iterations does not improve and the false trigger rate does not decrease, triggering an early stop mechanism and saving the current optimal model parameters; loading the optimal model parameters into the initial adaptive multi-transmission line model to generate the pre-trained adaptive multi-transmission line model.

8. An optical fiber differential protection decision system, characterized in that, Comprise: one or more processors; a memory; one or more computer programs; wherein the one or more computer programs are stored in the memory and are configured to be executed by the one or more processors, when the one or more computer programs are executed by the processor, the method as claimed in any one of claims 1-7 is realized.

Images