A method and system for compressing fault recording files and transmitting them to low-Earth orbit satellites with packet loss resistance.

By using streaming parsing and feature-driven data block processing, combined with FEC groups and RS coding, the problem of fault waveform file transmission under FTU resource constraints and weak links of low-Earth orbit satellites was solved, achieving efficient and reliable transmission and recovery of critical fault data.

CN123124013APending Publication Date: 2026-11-17STATE GRID HUNAN ELECTRIC POWER COMPANY LIMITED +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611167342.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-03
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies struggle to efficiently process fault recording files under FTU resource constraints and to reliably transmit and recover critical fault data in low-Earth orbit satellite weak link environments. In particular, they struggle to guarantee the accuracy of the time correspondence between analog and digital data in the event of communication interruption and packet loss.

Method used

By streaming fault recording files, structured data blocks are generated. Feature indexes are used to select decorrelation mode, the prediction frame length is dynamically determined, and analog prediction residuals and switch events are co-coded. Combined with FEC groups and RS coding, data is fragmented and transmitted to ensure reliability and recovery efficiency when the quality of low-Earth orbit satellite links fluctuates.

Benefits of technology

It effectively reduces the consumption of FTU computing and storage resources by fault recording file processing, improves transmission reliability and packet loss recovery capability in the weak link environment of low-orbit satellites, and maintains the accuracy of the time correspondence between analog and digital data.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123124013A_ABST
    Figure CN123124013A_ABST
Patent Text Reader

Abstract

This invention discloses a method and system for compressing fault waveform files and transmitting them with low-Earth orbit (LEO) satellites to combat packet loss. The method includes: streaming parsing of fault waveform files to generate structured data blocks; calculating characteristic indicators of three-phase voltage or current data in the structured data blocks and selecting appropriate decorrelation modes for the data blocks; determining the prediction frame length of the data blocks, adaptively selecting a prediction model, and generating analog prediction residuals; co-coding the analog prediction residuals and switching events to obtain independent compressed frame data; encapsulating the independent compressed frame data into transmission fragments and scheduling their transmission according to the LEO satellite link quality and fragment service priority; reassembling and independently decoding the received transmission fragments, retransmitting unrecovered missing fragments, and concatenating them in chronological order to obtain a complete fault waveform file. This invention can improve the transmission reliability, timeliness, and packet loss recovery capability of critical fault data in weak LEO satellite link environments.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the fields of fault waveform data processing and satellite communication technology, specifically to a method and system for compressing fault waveform files and transmitting them to low-orbit satellites with packet loss resistance. Background Technology

[0002] With the continuous development of distribution automation systems, feeder terminal units (FTUs) are widely used for monitoring the operational status and handling faults in distribution lines. FTUs can collect information such as three-phase voltage, three-phase current, and switch status in the line, and generate fault recording files when short circuits, grounding faults, or other abnormal conditions are detected. Fault recording files typically use the COMTRADE format to record the sampling data of each analog channel, the status changes of the digital channels, and related configuration information before and after the fault occurs, providing data support for fault location, fault cause analysis, and distribution network operation and maintenance.

[0003] COMTRADE fault waveform files typically include configuration files and data files, involving multiple analog and digital input channels. Due to the high sampling frequency of analog signals, the large number of channels, and the need to retain data before and after the fault for a certain period, the amount of waveform data generated by a single fault is substantial. When an FTU detects multiple faults or abnormal events consecutively within a short period, it may also accumulate multiple fault waveform files awaiting transmission. For FTUs with limited processor performance, RAM, and storage capacity, reading, caching, compressing, and transmitting waveform files consumes significant computing and storage resources, thus affecting the normal operation of other power distribution automation functions.

[0004] In remote mountainous areas, islands, plateaus, and other power distribution network areas with insufficient terrestrial communication network coverage, FTUs and power distribution master stations may need to rely on low-Earth orbit (LEO) satellite links to transmit fault waveform files. The relative speed between LEO satellites and ground terminals is relatively high, the duration of a single communication window is limited, and link quality is easily affected by obstructions, weather, terminal attitude, and the wireless propagation environment. Therefore, LEO satellite links may experience limited visibility time, high bit error rates, packet loss, communication interruptions, and transmission delay fluctuations. When the fault waveform file data volume is large, it may not be possible to complete the transmission within a single communication window; if a link interruption or data loss occurs during transmission, significant retransmission overhead may also occur.

[0005] Existing methods for processing fault waveform recordings typically employ general lossless compression algorithms or directly perform differential coding, predictive coding, or other processing on the recorded data. General lossless compression algorithms primarily compress fault waveform recordings as ordinary byte sequences, failing to adequately utilize the correlation between three-phase voltages and currents, as well as the changing characteristics of power fault waveforms. Therefore, their compression effectiveness is limited for certain types of waveform data. Furthermore, some compression processes require pre-reading or caching large amounts of data, placing high demands on the FTU's memory capacity and processor computing power.

[0006] Existing power waveform compression methods typically employ fixed data segmentation or fixed encoding parameters. During steady-state operation, voltage and current waveforms exhibit strong periodicity and correlation; however, during faults or transient events such as short circuits, grounding, and switching operations, waveforms may change significantly within a short period. Fixed processing methods struggle to simultaneously adapt to both steady-state and transient waveforms, easily leading to increased prediction errors, increased encoded data volume, or unstable compression performance when waveform characteristics change. Furthermore, while analog and digital data in fault recordings share a unified time base, existing technologies often process them separately. When switch states change or partial data loss occurs, a discrepancy may arise between the recovered analog sampling points and the timing of digital changes, affecting the determination of the fault occurrence time and subsequent fault analysis.

[0007] In satellite transmission of fault recording files, existing technologies typically employ either whole-file transmission or ordinary data packet transmission. Whole-file transmission requires waiting for the complete file to be received before parsing and fault analysis can begin. When the link is interrupted or parts of the file contain errors, it often necessitates retransmitting a large portion of the data. While ordinary data packet transmission allows for the splitting of the file, there are usually strong dependencies between different data packets, and the loss of some packets can render subsequent data unusable. Furthermore, existing transmission methods often do not adequately distinguish between data from critical fault periods and data from general periods. When the communication window is short or link quality deteriorates, critical fault data may fail to reach the main station in a timely manner.

[0008] Therefore, existing technologies still struggle to simultaneously meet the demands for efficient and lossless processing of fault recording files under FTU resource constraints, as well as the requirements for low-latency and reliable transmission under weak low-Earth orbit satellite link conditions. How to reduce the computational and storage resource consumption of the FTU during fault recording file processing, reduce the amount of data to be transmitted while maintaining the accuracy of analog and digital quantities and their time correspondences, and improve the availability and recovery efficiency of fault recording data under conditions of link packet loss, interruption, and latency fluctuations are pressing technical problems that need to be solved in this field. Summary of the Invention

[0009] The technical problem to be solved by this invention is to provide a method and system for compressing fault recording files and transmitting them to low-Earth orbit satellites with packet loss resistance, thereby improving the transmission reliability, timeliness, and packet loss recovery capability of critical fault data in low-Earth orbit satellite weak link environments.

[0010] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A method for compressing fault recording files and transmitting them via low-Earth orbit satellites with packet loss resistance includes the following steps: S1, stream analyzes the fault recording file, demultiplexes the analog and digital quantities in it, and generates structured data blocks; S2, calculate the characteristic indicators of the three-phase voltage data or three-phase current data in the structured data block, and select the corresponding decorrelation mode for the data block in the direct mode, phase-to-phase differential mode and integer reversible Clarke transform mode according to the characteristic indicators and preset criteria; wherein the characteristic indicators include one or more of the following: three-phase balance, zero-sequence component significance and transient change intensity. S3, based on the power frequency cycle, determines the prediction frame length of the data block according to the waveform stability and switching events, and adaptively selects the prediction model and generates analog prediction residuals based on the data characteristics within each prediction frame and the computing power of the FTU terminal. S4. Co-encode the analog prediction residual and the digital event. Determine the encoding parameters based on the statistical characteristics of the analog prediction residual and execute the encoding. Use the digital event aligned with the analog time axis to control the analog prediction model reset or the encoding parameters switch to obtain an independent compressed frame that corresponds one-to-one with the corresponding time window and does not depend on the previous compressed frame. S5, each independent compressed frame is padded and divided into multiple equal-length source fragments, multiple check fragments are generated using system RS encoding, multiple fragments corresponding to the same independent compressed frame are grouped into an FEC group, and each source fragment or check fragment is encapsulated into a transmission fragment, and then the transmission is scheduled according to the low-orbit satellite link quality and fragment service priority. S6, the master station reassembles the received transmission fragments according to the FEC group, and marks the transmission fragments that failed the verification as erased; when any number of valid fragments in the same FEC group are obtained, the corresponding independent compressed frames are recovered and verified, and then the independent compressed frames are independently decoded without relying on the previous compressed frame; when the number of valid fragments in the same FEC group is less than the preset number, the unrecovered fragments are selectively retransmitted according to the group identifier and fragment sequence number; the decoding results are spliced ​​together in chronological order, the lossless objects are restored item by item, and the fault waveform dataset is reconstructed.

[0011] Further, step S1 includes: The COMTRADE configuration file is read sequentially, and the original byte sequence of the configuration file, as well as the data file type, byte order and record format, are preserved as global configuration metadata. The sampling rate, the scaling factor, offset, unit and first or second value identifier of the analog channel and the normal status logic value of the digital channel are parsed in the working copy of the configuration file, and a channel mapping table including channel index, channel type, name and conversion parameters is established. The data block buffer length is determined based on the correspondence between the sampling rate and the power frequency cycle, and the COMTRADE data file is read in blocks using a read-and-press method. The analog and digital signals at each sampling point of the data block are demultiplexed according to the channel mapping table. The original analog code values ​​are retained to form an analog original code value matrix. An engineering unit value working copy, used only for feature extraction and threshold judgment, is calculated according to the scaling factor and offset. The original codewords of the digital signals are retained. At the same time, a logical state working copy for event detection is generated according to the channel mapping table. Neither the engineering unit value working copy nor the logical state working copy overwrites the corresponding original code value or original codeword. The system detects and marks analog quantity exceeding limits, abnormal transients in digital signals, and missing channels, generating a structured data block that includes global configuration metadata references, block start timestamps, number of sampling points, sampling rate, sample number and timestamp sequence, original analog quantity code value matrix, original digital signal codeword and logic state matrix, anomaly marker bitmap, and missing channel information. The original code values ​​of abnormal sampling points are retained in the structured data block, and the interpolation results are only used as optional analysis results and not as lossless encoding objects.

[0012] Further, in step S2, the three-phase balance is calculated by the ratio of the difference between the maximum and minimum values ​​of the three-phase effective values ​​to the average value of the three-phase effective values, expressed as:

[0013]

[0014] in, Indicates the three-phase balance; This represents the valid value of the k-th phase within the current data block; The significance of the zero-sequence component is calculated by the ratio of the zero-sequence component energy to the total three-phase energy within the data block, expressed as:

[0015]

[0016] in, Indicates the significance of the zero-order component; This represents the zero-order component of the nth sampling point; These represent the voltage or current sample values ​​of phases a, b, and c at the nth sampling point, respectively. The transient change intensity is calculated by the ratio of the maximum absolute value of the three-phase first-order difference to the average absolute value of the three-phase first-order difference at the sampling points within the data block, expressed as:

[0017]

[0018] in, Indicates the intensity of transient mutations; This represents the first-order difference of the k-th phase at the n-th sampling point; n represents the sampling point number; L represents the number of sampling points in the current structured data block.

[0019] Further, in step S2, based on characteristic indicators and preset criteria, the appropriate decorrelation mode is selected for the data block among the direct-through mode, phase-to-phase differential mode, and integer invertible Clarke transform mode, specifically including: The decorrelation mode is selected based on a multi-level criterion, prioritizing transient change intensity and secondarily considering three-phase balance or zero-sequence component significance. Specifically: when the transient change intensity is higher than a high threshold, the direct pass mode is selected; when the transient change intensity is lower than a high threshold and the three-phase balance is higher than the corresponding threshold or the zero-sequence component significance is higher than the corresponding threshold, the phase difference mode is selected, using the difference between the sampled values ​​of two adjacent phases as the difference component and retaining the third phase as the reference phase; when none of the above conditions are met, the integer reversible Clarke transform mode is selected, using a fixed-point linear combination of the three-phase sampled values ​​and rounding down to transform the three-phase stationary coordinate system to a two-phase stationary coordinate system and the zero-sequence component; finally, the mode identifier of the selected decorrelation mode and the necessary parameters required for the inverse transform are written into the data block header.

[0020] Furthermore, in step S3, when determining the predicted frame length of the data block based on waveform stability and switching events, the predicted frame length is dynamically selected between short frames and long frames based on the waveform differential variance, zero-crossing density, and switching event flags within the current data window. Specifically, a short frame is selected when the waveform differential variance exceeds the corresponding threshold, the zero-crossing density exceeds the corresponding threshold, or a switching event is detected; otherwise, a long frame is selected. Based on the data characteristics of each prediction frame and the computing power of the FTU terminal, the prediction model is adaptively selected and the analog prediction residuals are generated, specifically including: Based on the autocorrelation characteristics of the current intra-frame analog data and the available computing power of the FTU terminal, the prediction order of the linear prediction model is selected and the prediction coefficients are solved. The prediction coefficients are then subjected to fixed-point quantization, and the prediction residuals are calculated. The autocorrelation characteristics are obtained by summing the products of different delayed sample values ​​after removing the intra-frame mean. The prediction coefficients of each order are obtained by solving the Yule-Walker equations formed by these autocorrelation characteristics, expressed as follows:

[0021]

[0022] in, This represents the autocorrelation coefficient with a delay order of k, where k represents the autocorrelation delay order. p represents the average value of the analog sampled values ​​within the current prediction frame, and p represents the prediction order determined based on the autocorrelation characteristics and the available computing power of the FTU terminal. This represents the sampled value of the i-th analog quantity within the current prediction frame; Indicates the i-th element in the current prediction frame k analog sample values; Indicates the current predicted frame length; Indicates a delay order of m The autocorrelation coefficient of n; The fixed-point quantization of the prediction coefficients and the calculation of the prediction residuals include: scaling and rounding the prediction coefficients according to the decimal width of the fixed-point number to obtain the fixed-point quantized prediction coefficients; calculating the prediction value based on the fixed-point quantized prediction coefficients and the historical analog sample values ​​before the current sampling point; and using the difference between the actual analog sample value and the prediction value as the analog prediction residual. When the moving average of the absolute value of the analog prediction residual exceeds the residual threshold, or the compression gain of the current prediction frame is lower than a preset multiple of the compression gain of the previous steady-state prediction frame, the prediction model is degraded sequentially to second-order difference prediction, first-order difference prediction, and unmodified encoding. Among them, second-order difference prediction, first-order difference prediction, and unmodified encoding respectively use the first two analog sample values ​​for linear extrapolation, use the previous analog sample value as the prediction value, and set the prediction value to zero. Finally, the system outputs the analog prediction residual sequence, prediction frame length, prediction mode identifier, fixed-point quantization prediction coefficients, and prediction reset point position for each prediction frame.

[0023] Further, step S4 includes: The analog prediction residuals are sign-mapped to convert signed analog prediction residuals into non-negative integers, and the zero-value density, variance, and maximum amplitude of the sign-mapped analog prediction residuals are calculated. Specifically, the sign mapping is represented by the difference between the absolute value of the analog prediction residual and the binary sign identifier. The zero-value density is calculated as the proportion of analog prediction residual samples with absolute values ​​less than a preset zero-value threshold to the total number of analog prediction residual samples in the current prediction frame or subframe. The variance is calculated as the average of the squared deviations of the analog prediction residuals relative to the residual mean. The maximum amplitude is the largest absolute value of the non-negative integer after sign mapping. Rice coding parameters are adaptively determined based on the zero-value density and variance. The Rice encoding parameters The relationship between the expected value of the absolute value of non-negative integers and the base-2 logarithm is determined by using a sign mapping. Using the Rice encoding parameters The non-negative integer after sign mapping is decomposed into a quotient and a remainder, wherein the quotient is the non-negative integer after sign mapping divided by... The result is that the remainder is a pair of non-negative integers after symbol mapping. The result of the modulo operation; the quotient is encoded using unary code, and the remainder is a fixed-length remainder. Bit-binary code encoding; For non-negative integers after symbol mapping that exceed a preset threshold T, escape coding is used, wherein the escape coding word width is the minimum number of binary bits required for the maximum magnitude of the non-negative integer after symbol mapping in the current prediction frame or subframe. Run-length encoding is performed on the switch state sequence corresponding to the switch event, the switch state value and continuous length are recorded, and the analog sampling point with the smallest time distance from the switch change time is determined as the corresponding analog sampling point, so as to align the switch change event with the analog sampling time sequence. The analog quantity prediction model is controlled according to the positional relationship between the switching quantity change time and the analog quantity prediction window. The analog quantity prediction window is determined by the current prediction processing time and the prediction history length. When the switching quantity change time is within the analog quantity prediction window, at least one of the following operations is performed: terminating the current prediction subframe and starting a new prediction subframe, clearing or resetting the history buffer of the analog quantity prediction model, reducing the prediction order, or increasing the Rice coding parameter. The Rice encoding result and the escape encoding result are used to form an analog residual bitstream. The digital run-length encoding result and the analog sampling point index, edge direction and duration corresponding to the digital change time are used to form a digital event bitstream. The frame identifier, time window start point, number of sampling points, prediction subframe boundary, prediction reset flag, decorrelation mode, prediction mode, prediction order, fixed-point quantization prediction coefficients, Rice parameters, initial raw code values ​​of each channel, bitstream length and frame-level CRC are organized together with the analog residual bitstream, digital event bitstream and anomaly marker bitstream into an independent compressed frame, so that the prediction state or encoding parameters of the previous compressed frame do not need to be referenced when decoding the independent compressed frame.

[0024] Furthermore, the expression for the symbol mapping is:

[0025]

[0026] in, This represents the analog prediction residual at the t-th analog sampling point. This represents the actual sampled value of the t-th analog sampling point. This represents the predicted value of the t-th analog sample point. This represents the non-negative integer obtained after sign mapping of the analog prediction residual. Represents a binary symbolic function; The Rice encoding parameters The expression is:

[0027] in, Represents Rice encoding parameters. Represents the expected value of the absolute value of non-negative integers after sign mapping, and predicts the residual variance based on the analog quantity. Make an estimate. Indicates rounding down; When determining the analog sampling point with the smallest time distance from the moment of the switch change as the corresponding analog sampling point, the specific expression for mapping the moment of the switch change to the nearest analog sampling point index is as follows:

[0028] in, This represents the sampling time of the i-th analog quantity sampling point. Indicates the moment when the switch quantity changes position. This indicates the index of the analog sampling point corresponding to the moment of change of the switch quantity.

[0029] Further, step S5 includes: Each independent compressed frame corresponding to a time window is treated as a unique source object for an FEC group. For the i-th independent compressed frame CF i After filling the end of the frame with a predetermined padding value, it is divided into k. i Each source is divided into equal-length fragments, and the system RS(n) is used. i ,k i Encoding to generate n i -k i Each equal-length check fragment makes the independent compressed frame CF i With FEC Group G i One-to-one correspondence; Each source fragment or check fragment is encapsulated into a transmission fragment. The fixed header of the transmission fragment includes at least a synchronization word, a global waveform recording task identifier, an FEC group identifier, an independent compressed frame identifier, and n. i k i The fragment sequence number within the group, the source fragment or check fragment identifier, the fragment length, the actual length of the independent compressed frame, the absolute timestamp, the retransmission flag, the fragment CRC, and the frame-level CRC; The link quality factor of the low-Earth orbit satellite link is determined by weighting the signal-to-noise ratio, bit error rate, and remaining visibility time. Then, the transmission priority is determined by weighting the service priority of the transmission fragment, the fragment number, and the ratio of the link quality requirement to the current link quality factor. Each transmission fragment is then sent from high to low according to the transmission priority.

[0030] Furthermore, the calculation expression for the link quality factor is as follows:

[0031] in, For link quality factor, For normalized signal-to-noise ratio, To normalize the bit error rate, To normalize the remaining visible time, , and These are the weighting coefficients; The expression for calculating the transmission priority is:

[0032] in, For transmission fragmentation Transmission priority, For business priority, To meet the link quality requirements for transmitting fragmented data, This is the current fragment number. This represents the total number of transmission fragments. This represents the current link quality factor. , and These are the weighting coefficients.

[0033] A fault recording file compression and low-Earth orbit satellite packet loss-resistant transmission system includes a microprocessor and a memory interconnected, wherein the microprocessor is programmed or configured to perform the steps of the fault recording file compression and low-Earth orbit satellite packet loss-resistant transmission method.

[0034] Compared with the prior art, the advantages of the present invention are as follows: This invention generates structured data blocks by streaming fault recording files, eliminating the need for full file loading. It simultaneously identifies the original configuration metadata, original analog code values, original switch states, and sample time correspondences as lossless objects, ensuring that engineering quantity conversions and feature calculations do not alter the original data to be recovered. By selecting a decorrelation mode based on three-phase balance, zero-sequence component significance, and transient change intensity, it adapts to different fault conditions and reduces data redundancy. The prediction frame length is dynamically determined based on the power frequency cycle, and the prediction model is selected by combining intra-frame features and terminal computing power, balancing compression efficiency and transient waveform fidelity. Through co-coding of analog prediction residuals and switch events, and writing complete restart states in each independent compressed frame, the timing consistency of the two types of data is maintained, and cross-frame error propagation is avoided. By mapping independent compressed frames one-to-one with FEC groups and recovering and independently decoding the frame after obtaining any k valid fragments of the same group, the reliability and streaming availability of critical data under weak links are improved. Selective retransmission by FEC group reduces redundant transmissions. Attached Figure Description

[0035] Figure 1 This is a flowchart illustrating the fault recording file compression and low-orbit satellite packet loss-resistant transmission method according to an embodiment of the present invention. Detailed Implementation

[0036] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0037] like Figure 1 As shown, the fault recording file compression and low-orbit satellite packet loss-resistant transmission method of this embodiment includes the following steps: S1, Streaming parsing of COMTRADE files and demultiplexing data: Streaming parsing of fault recording files, demultiplexing and preprocessing the analog and digital quantities in them, and retaining the configuration metadata byte sequence, the original analog code value matrix, the original digital state matrix, and the sample number and timestamp sequence as lossless objects in addition to the engineering quantity working copy used for feature calculation, generating structured data blocks.

[0038] In a specific application embodiment, step S1 includes: S11 sequentially reads the COMTRADE configuration file (.cfg), preserving the original byte sequence of the configuration file, as well as the data file type, byte order, and record format as global configuration metadata. It then parses key metadata from the working copy of the configuration file, such as the sampling rate, scale factor of analog channels, offset, unit, and primary or secondary value identifiers, and the normal state logic values ​​of digital channels. The sampling rate is among these parameters. Obtain directly from the configuration line. For the first... Each analog signal channel is analyzed to determine its scaling factor. Offset Units and primary / secondary value identifiers. For the first... Each switch channel is analyzed to determine its normal state logic value.

[0039] S12, Establish a channel mapping table including channel index, channel type (analog A / switch D), name, and corresponding conversion parameters. This process maintains only the metadata structure and a small circular buffer in memory, avoiding full file residency.

[0040] S13, determine the data block buffer length based on the correspondence between the sampling rate and the power frequency cycle, and read the COMTRADE data file in blocks using a read-and-press method, for example, by opening the data file (.dat) in streaming mode. Set a fixed-size data block buffer. Its capacity The number of sampling points is dynamically adjusted based on the available memory in the terminal, with a typical value being the number of sampling points corresponding to several power frequency cycles. ,in For the number of periods (e.g.) ), The sampling rate. Read from the file stream. Row data to cache Each row of data includes a timestamp. And the original sampled values ​​of all channels. After reading the corresponding row data into the cache, it is parsed in real time. After completing the parsing of the current cache block, it outputs a structured data block. Only the metadata structure and a small circular buffer are retained throughout the process. The memory usage is constant, which can adapt to the limited computing power and storage resources of FTU. Compared with the traditional whole file loading method, the resource usage is reduced by more than 60%.

[0041] S14, demultiplex the analog and digital signals at each sampling point of the data block according to the channel mapping table. For analog channels, directly read and retain the original code value C from the data file. in This forms the original analog code value matrix X. raw Simultaneously, within an independent work buffer zone, the scaling factor 'a' in the configuration file is applied. i and offset b i Calculate the unit value V of the projectin The engineering unit value is only used for feature extraction, limit violation detection, and mode selection, and is not written back to X. raw Furthermore, it is not used as a comparison object for lossless recovery; its conversion formula is as follows:

[0042] in, and These are the scaling factor and offset of the i-th analog channel, respectively.

[0043] When an FTU needs to calculate characteristic parameters in a fixed-point manner, it can multiply a working copy of the engineering unit value by a scaling factor S and round it to form a temporary characteristic value X. feat Scaling factor S, rounding method and X feat All of these only affect feature calculation and do not change the original analog code value C used in subsequent reversible decorrelation and prediction residual coding. in The out-of-limit detection result is written to the anomaly flag bitmap F. err The original code value corresponding to the anomaly point is still retained in X. raw In this case, skipping or replacing the original code value with an interpolated value is not performed.

[0044] For digital input channels, the original codewords in the data file are retained to form D. raw And generate the logic state working matrix D separately based on the channel mapping table. logic When an illegal transient occurs between consecutive sampling points, it is only recorded in the anomaly marker bitmap and D is not modified. raw The original codeword in.

[0045] S15 detects and marks analog quantity exceeding limits, abnormal transients in digital quantities, and missing channels. During parsing, missing channels are detected and marked synchronously. If a channel exists in the configuration file but lacks a corresponding data column in a certain line of the data file, the value of that channel at that sampling point is set to a predefined missing identifier (such as a specific integer maximum value), and the missing channel description field is updated. Records in the middle.

[0046] S16, Generate structured data block D out D out Including global configuration metadata references M ref Block start timestamp t start Number of sampling points L B Sampling rate f s Sample number and timestamp sequence T S Analog signal original code value matrix X raw , Switch quantity raw codeword matrix D raw , Switch logic state working matrix D logic , Anomaly marker bitmap F errAnd missing channel information F miss Upon initial output, the original byte sequence of the COMTRADE configuration file, data file type, byte order, and record format are combined to form global configuration metadata M, which is then transmitted as an independent compressed frame. The lossless object U referred to in this embodiment consists of M and X. raw D raw T S F err and F miss Composition; Lossless recovery refers to the fact that the M recovered at the decoding end is consistent with the input metadata byte by byte, and the recovered X... raw D raw and T S The sequence is consistent element-wise with the input sequence, and the correspondence of quality markers remains unchanged. Engineering unit values, working copies of logical states, and optional interpolation results are used only for feature calculation or analysis display and do not replace the original data in U. Output D after parsing all lines of the current cache block. out This is to keep memory usage bounded during streaming processing.

[0047] As can be understood, step S1 reads the waveform recording configuration and data file, extracts metadata such as sampling rate, channel type, and scaling factor, and establishes a mapping between analog and digital channels. A read-while-compress mechanism is employed, buffering the data stream in small blocks to avoid loading the entire file. During parsing, the sampled values ​​are simultaneously integerized and normalized, abnormal sampling points or missing channels are identified and marked, and corresponding descriptive information is generated. These preprocessed data blocks and metadata are directly output, providing a clearly structured input suitable for streaming processing for subsequent compression encoding.

[0048] S2, based on waveform characteristics, adaptively selects the reversible decorrelation mode: It calculates the characteristic indices of three-phase voltage or three-phase current data in the structured data block, and selects the appropriate decorrelation mode for the data block from the direct-through mode, phase-to-phase differential mode, and integer reversible Clarke transform mode according to the characteristic indices and preset criteria; the characteristic indices include one or more of three-phase balance, zero-sequence component significance, and transient change intensity. This enables efficient and lossless decorrelation under different fault conditions, solving the problem of poor decorrelation performance and even the introduction of redundancy in unbalanced faults using fixed transform.

[0049] In a specific application embodiment, step S2 includes: S21, Sample the three-phase voltage or current sequence after demultiplexing. (or Process data in blocks, block length Typically, these values ​​are integer multiples of the power frequency cycle. To assess waveform characteristics, three key indicators need to be calculated in real time. Three-phase balance index... It is obtained by calculating the relative deviation of the three-phase effective values ​​within a data block, specifically by calculating the ratio of the difference between the maximum and minimum values ​​of the three-phase effective values ​​to the average value of the three-phase effective values. The expression is:

[0050]

[0051] in, Indicates the three-phase balance; This represents the valid value of the k-th phase within the current data block.

[0052] The significance of the zero-sequence component is calculated by the ratio of the zero-sequence component energy to the total three-phase energy within the data block, expressed as:

[0053]

[0054] in, Indicates the significance of the zero-order component; This represents the zero-order component of the nth sampling point; These represent the voltage sample values ​​of phase a, phase b, and phase c at the nth sampling point, respectively.

[0055] The transient change intensity is calculated by the ratio of the maximum absolute value of the three-phase first-order difference to the average absolute value of the three-phase first-order difference at the sampling points within the data block, expressed as:

[0056]

[0057] in, Indicates the intensity of transient mutations; This represents the first-order difference of the k-th phase at the n-th sampling point; n represents the sampling point number; L represents the number of sampling points in the current structured data block.

[0058] S22, based on characteristic indicators and preset criteria, selects the appropriate decorrelation mode for the data block among the direct-through mode, phase-to-phase differential mode, and integer invertible Clarke transform mode, specifically including: The decorrelation mode is selected based on a multi-level criterion that prioritizes transient mutation intensity, followed by three-phase balance or zero-order component significance. The first-level criterion targets transiently severe regions: when the transient mutation intensity exceeds a high threshold (…), the decorrelation mode is selected. , When the threshold is high, the direct-through mode is selected. In this mode, the three-phase data is output as is, avoiding the introduction of transformation distortion or increased computational complexity during periods of drastic change, and helping to maintain the original integrity of the fault transient impact waveform. The second-level criterion targets imbalance and zero-order significant operating conditions: when the transient mutation intensity is below the high threshold and the three-phase balance is above the corresponding threshold ( ) or the zero-order component is significantly higher than the corresponding threshold ( When selecting the phase-to-phase differential mode, the difference between the sampled values ​​of two adjacent phases is used as the differential component (e.g., and And retain the third phase As a reference phase, this treatment is robust to single-phase grounding or unbalanced fault waveforms because the phase-to-phase differential effectively suppresses the zero-sequence component of the common mode. The third-level criterion is the default case: when none of the above conditions are met, it indicates that the waveform is in a normal equilibrium state. In this case, the integer reversible Clarke transform mode is selected, using a fixed-point linear combination of the three-phase sampled values ​​and rounding down to transform the three-phase stationary coordinate system to a two-phase stationary coordinate system and the zero-sequence component. Specifically, this mode uses a fixed-point integer approximation orthogonal transformation matrix to transform the three-phase stationary coordinate system... Transform to a two-phase stationary coordinate system and zero-sequence components Its integer invertible transformation formula is:

[0059]

[0060]

[0061] in This indicates a rounding operation towards negative infinity, and is used in conjunction with a scaling factor to ensure the reversibility of the transformation. , , These represent the α-axis component, β-axis component, and zero-sequence component of the nth sampling point obtained after performing an integer reversible Clarke transform on the three-phase voltage samples, respectively. This transform can fully utilize the strong spatial correlation of the three-phase waveforms, concentrating energy to... and This component significantly improves the efficiency of subsequent prediction compression.

[0062] S23. The mode identifier of the selected decorrelation mode and the necessary parameters required for the inverse transform are written into the header of the data block. The identifier of the selected mode (e.g., 2-bit encoding: 00-pass, 01-phase differential, 10-integer Clarke transform) and the necessary parameters required for the inverse transform (e.g., reference phase index in phase differential mode, scaling factor correction value for integer transform) are encoded and embedded into the header metadata area of ​​the data block. Based on the header information, the decoding end can accurately perform the corresponding inverse operation, thereby recovering the original three-phase sample values ​​block by block without loss. This provides an accurate data foundation for subsequent steps.

[0063] It is understandable that step S2 calculates characteristic indicators such as balance, zero-sequence component significance, and transient change intensity in real time for the demultiplexed three-phase voltage and current data. Based on preset criteria, an integer reversible Clarke transform, phase-to-phase differential, or direct-through mode is dynamically selected for each data block. Normal balanced waveforms use spatial transformation to improve correlation, while single-phase grounding or unbalanced waveforms switch to phase-to-phase differential to maintain robustness. Direct-through mode can be temporarily used for sections with severe transients. The selected mode and the parameters required for the inverse transform are recorded and embedded in the data block header, ensuring that the master station can recover the original waveform block by block without loss.

[0064] S3, Dynamic Frame Segmentation and Adaptive Prediction to Extract Residuals: Based on the power frequency cycle, the prediction frame length of the data block is determined according to the waveform stability and switching events. The prediction model is adaptively selected based on the data characteristics within each prediction frame and the computing power of the FTU terminal, and analog prediction residuals are generated.

[0065] In a specific application embodiment, step S3 includes: S31 determines the predicted frame length of the data block based on waveform stability and switching events. Specifically, it uses the power frequency cycle. Reference frame length The actual frame length is dynamically adjusted based on the waveform characteristics calculated in real time. Calculate the waveform difference variance within the current data window. Zero-crossing density and switch event flags The prediction frame length is dynamically selected between short and long frames; whereby, when the waveform difference variance exceeds the corresponding threshold ( ), zero-crossing density exceeds the corresponding threshold ( ) or a switch event is detected ( Select short frame when (like (one cycle), otherwise choose long frame. (like (each cycle). Among them, the waveform difference variance... The zero-crossing density is calculated using the variance of the first-order difference sequence of adjacent sampling points within the current data window. Calculated by the ratio of the number of times the waveform crosses zero within the current data window to the length of the data window; Switch event flag. Set to 1 if a change in the status of a switch quantity is detected within the current data window; otherwise, set to 0. and The threshold is set based on historical steady-state statistics. This dynamic framing strategy ensures that short frames are used during sudden fault changes or switching actions to effectively isolate the spread of prediction errors, while long frames are used in the steady-state segment to reduce frame header overhead and improve overall compression efficiency.

[0066] S32, based on the data characteristics of each prediction frame and the computing power of the FTU terminal, adaptively selects the prediction model and generates analog prediction residuals, specifically including: For each frame of data: Based on the autocorrelation characteristics of the current intra-frame analog data Available computing power of FTU terminals Choosing the prediction order of a linear prediction model The prediction coefficients are then solved, and the prediction coefficients are quantized at fixed points and the prediction residuals are calculated.

[0067] The autocorrelation characteristic is obtained by summing the products of different delayed sample values ​​after removing the intra-frame mean. To limit the calculation range of the autocorrelation coefficient, a maximum prediction order is preset based on the available computing power and storage resources of the FTU terminal, i.e., k=1,2,… Before calculation The expression for each autocorrelation coefficient is:

[0068] in, This represents the autocorrelation coefficient with a delay order of k, where k represents the autocorrelation delay order. This represents the average value of the analog sampled values ​​within the current prediction frame. This represents the sampled value of the i-th analog quantity within the current prediction frame; Indicates the i-th element in the current prediction frame k analog sample values.

[0069] The prediction coefficients of each order are obtained by solving the Yule-Walker equation, which is based on autocorrelation characteristics. The expression is as follows:

[0070] In the above formula, p represents the prediction order determined based on the autocorrelation characteristics and the available computing power of the FTU terminal. Specifically, it is taken as the smaller of the minimum delay order when the correlation coefficient decays to below a preset ratio and the maximum order supported by the terminal computing power. Represents the nth-order prediction coefficient of a linear prediction model; Indicates the current predicted frame length; Indicates a delay order of m The autocorrelation coefficient of n. The prediction order p is determined by the autocorrelation coefficient decay rate and the available computing power of the FTU terminal. The common determination is as follows: p = min(p opt , p cap ), where p opt To minimize the delay order at which the autocorrelation coefficient R(k) first decays to below a predetermined proportion of R(0), pcap The smaller of the two values ​​is taken as the actual prediction order, which is the maximum prediction order supported by the available computing power of the FTU terminal.

[0071] S33, perform fixed-point quantization on the prediction coefficients and calculate the prediction residuals.

[0072] To adapt to the fixed-point processor of the FTU, the prediction coefficients Scaling and rounding according to the decimal width of the fixed-point number yields the fixed-point quantization prediction coefficients, and converting the floating-point coefficients... Quantified as Fixed-point number in format The quantification formula is ,in The decimal places are determined based on the dynamic range and precision requirements of the coefficients. (This refers to the quantization coefficients.) Write it into the frame header for use by the decoding end.

[0073] The predicted value for each sampling point within the frame is calculated based on the fixed-point quantization prediction coefficients and the historical analog sample values ​​preceding the current sampling point. The expression is as follows:

[0074] in, For fixed-point quantitative prediction coefficients, For predicted values, This is the nth historical analog sample value before the current sampling point.

[0075] The difference between the actual sampled value and the predicted value is then used as the analog quantity prediction residual. The residual sequence is expressed as follows:

[0076] .

[0077] S34, Real-time monitoring of prediction performance, when the moving average of the absolute value of the analog prediction residual is... (The moving average passes through the current sampling point and its preceding points) The result is calculated using the arithmetic mean of the absolute values ​​of the prediction residuals from one sampling point. ,in (The sliding window length) exceeds the residual threshold. Or the compression gain of the current predicted frame ( When the compression gain is lower than a preset multiple of the previous steady-state predicted frame, it is determined that the prediction gain has decreased or diverged. At this time, the prediction model degradation mechanism is triggered: the prediction model is degraded sequentially to second-order difference prediction, first-order difference prediction, and unmodified encoding; wherein, second-order difference prediction, first-order difference prediction, and unmodified encoding respectively use the first two analog sample values ​​for linear extrapolation, use the first analog sample value as the prediction value, and set the prediction value to zero. Specifically, it first degrades to Second-order difference prediction If the returns still do not meet the requirements, it will further degenerate into... First-order difference Ultimately, it can degenerate into the original encoding mode, that is... When degradation occurs, a reset point marker is set within the frame. Upon receiving this marker, the decoder clears the prediction history to prevent error accumulation. This adaptive prediction and degradation mechanism ensures that the coding system can still operate stably and output controllable residuals during fault transient phases where waveform characteristics change drastically.

[0078] S35, output the analog prediction residual sequence corresponding to each prediction frame. And accompanying frame description information, including predicted frame length. Prediction pattern identifier (e.g., higher-order linear prediction, second-order difference, first-order difference, original encoding), fixed-point quantization prediction coefficients (If applicable) and predict the location of the reset point. This information provides a highly decorrelational and statistically concentrated data source for subsequent entropy coding.

[0079] It is understandable that step S3 uses the power frequency cycle as a benchmark, combined with real-time detected fault abrupt changes, switching event triggers, and waveform stability, to dynamically determine the frame length of the data block. Longer frames are used in the steady-state segment to improve efficiency, while shorter frames are used in the transient abrupt change segment to control prediction errors. Based on the data characteristics within the frame and the terminal's computing power, the order of the prediction model is adaptively selected and fixed-point quantization is performed. If prediction residual divergence or a decrease in returns is detected, the model will automatically degenerate to a lower-order difference or even unmodulated encoding, and a reset point will be set within the frame. Step S3 can output a residual sequence with prediction mode information.

[0080] S4 performs co-coding on analog prediction residuals and digital events. It determines the encoding parameters based on the statistical characteristics of the analog prediction residuals and performs the encoding. It also uses digital events aligned with the analog time axis to control the resetting of the analog prediction model or the switching of encoding parameters, resulting in independent compressed frames that correspond one-to-one with the corresponding time windows and do not depend on the previous compressed frame.

[0081] In a specific application embodiment, step S4 includes: S41, Prediction residuals of analog quantities Perform sign mapping to convert the signed analog prediction residuals into a non-negative integer sequence. And calculate statistical characteristics for each data block (usually a prediction frame or subframe): calculate the zero density of the analog prediction residuals after symbol mapping. ,variance and maximum amplitude The symbol mapping is represented by the difference between the absolute value of the analog prediction residual and the binary symbol identifier.

[0082]

[0083] in, This represents the analog prediction residual at the t-th analog sampling point. This represents the actual sampled value of the t-th analog sampling point. This represents the predicted value of the t-th analog sample point. This represents the non-negative integer obtained after sign mapping of the analog prediction residual. Represents a binary symbolic function. .

[0084] Zero-value density predicts the number of residual samples by using analog quantities whose absolute values ​​are less than a preset zero-value threshold. Percentage of the total number of analog quantity prediction residual samples in the current prediction frame or subframe The proportions were calculated to obtain: ;in yes The number of samples.

[0085] Variance is calculated by averaging the squared deviations of the predicted residuals relative to the mean of the residuals:

[0086] in, This represents the sample variance of the analog quantity prediction residual within the current prediction frame or prediction subframe, used to reflect the degree of dispersion of the analog quantity prediction residual relative to its mean; t represents the analog quantity sampling point index. It represents the mean value of the analog prediction residuals within the current prediction frame or prediction subframe.

[0087] The maximum amplitude is the largest absolute value of a non-negative integer after sign mapping: .

[0088] S42, based on zero-value density and variance adaptive Determine Rice encoding parameters Rice encoding parameters The Rice coding parameters are determined by the relationship between the expected value of the absolute value of non-negative integers after sign mapping and the base-2 logarithm. The expression is:

[0089] in, Represents Rice encoding parameters. Represents the expected value of the absolute value of non-negative integers after sign mapping, and predicts the residual variance based on the analog quantity. To make an estimate, specifically, for prediction residuals that follow a Laplace distribution, pass Make an estimate. This indicates rounding down. This adaptive mechanism can dynamically adjust coding efficiency based on drastic changes in residual statistical characteristics before and after a fault, using a smaller rounding factor in the steady-state high-zero region. To achieve the maximum compression ratio, a larger value is used in the transient high variance region. The values ​​cover a wider range of numerical values.

[0090] S43, for parameters within the specified range Perform efficient Rice coding: using Rice coding parameters Decompose the non-negative integer after sign mapping into a quotient and a remainder, where the quotient is the non-negative integer after sign mapping divided by... The results obtained ( The remainder is the pair of non-negative integers after symbol mapping. The result of taking the modulus ( The merchant uses unary code encoding (); (One '1' followed by one '0'), the remainder is of fixed length. Bit-based binary code encoding.

[0091] For non-negative integers after symbol mapping exceeding the preset threshold T Escape coding is employed, where the escape coding word width is the minimum number of binary bits required to achieve the maximum magnitude of a non-negative integer after symbol mapping within the current prediction frame or subframe. Specifically, a special all-'1' unary code is first output as an escape flag, and then a fixed-length... Bit direct encoding This hybrid encoding strategy ensures that the encoded stream does not drastically expand due to a single Rice parameter mismatch when large residuals are generated by sudden failures, thus maintaining the robustness of compression.

[0092] S44, perform run-length encoding on the switch state sequence corresponding to the switch event, record the switch state value and continuous length, and determine the analog sampling point with the smallest time distance from the switch change time as the corresponding analog sampling point, so as to align the switch change event with the analog sampling time sequence.

[0093] Let the analog quantity sampling time series be The switch state sequence is First of all, Perform run-length encoding and record state values. and its continuous length More importantly, it accurately records each state transition event: the moment of transition. It needs to be mapped to the index of the nearest analog sampling point. ; Edge direction ; and the duration of the new state. (In terms of the number of sampling points). These event parameters, along with the run-length encoding results, are written into the bitstream to ensure accurate reconstruction of the switching event timing information at the decoding end.

[0094] S45, control the analog quantity prediction model according to the positional relationship between the switch quantity change time and the analog quantity prediction window, wherein the analog quantity prediction window is determined by the current prediction processing time and the prediction history length, and when the switch quantity change time is within the analog quantity prediction window, perform at least one of the following operations: terminate the current prediction subframe and start a new prediction subframe, clear or reset the history buffer of the analog quantity prediction model, reduce the prediction order, or increase the Rice coding parameter.

[0095] Specifically, the core of cooperative coding lies in the feedback of switching events to the analog signal encoding process. An analog signal prediction window is defined. ,in This refers to the historical length corresponding to the prediction order. When a switch change is detected... satisfy Upon this, the analog encoder is immediately triggered to reset or switch, including: terminating the encoding of the current subframe and immediately starting a new subframe; clearing or resetting the predictor history buffer to its default initial value; or switching to a more conservative set of encoding parameters (such as a lower prediction order and a larger Rice parameter). This mechanism effectively isolates the mismatch impact on the prediction model caused by sudden changes in electrical quantities due to switching actions (such as circuit breaker tripping), preventing errors from spreading in subsequent samples. This ensures the consistency of the encoding and decoding process near the fault time and improves the reliable recovery capability of data during critical periods.

[0096] S46, the Rice encoding result and escape encoding result are combined to form an analog residual bitstream, and the digital run-length encoding result and the analog sampling point index, edge direction and duration corresponding to the digital change time are combined to form a digital event bitstream; the frame identifier FrameID, time window start point, number of sampling points, prediction subframe boundary, prediction reset flag, decorrelation mode, prediction mode, prediction order, fixed-point quantization prediction coefficients, Rice parameters, initial raw code values ​​of each channel, bitstream length and frame-level CRC are combined with the analog residual bitstream, digital event bitstream and anomaly marker bitstream to form an independent compressed frame CF. i Each CF i It corresponds to a continuous and non-overlapping time window, and carries all intra-frame states required to recover the original code values ​​of that window; in addition to the initially obtained global configuration metadata, the decoded CF... i Not dependent on CF i-1 Or the prediction history and coding parameters of earlier compressed frames.

[0097] It is understandable that step S4 analyzes the statistical characteristics of the analog residual sequence, such as zero-value density and variance, and adaptively assigns Rice coding parameters to different data blocks. Residuals within the parameter range are efficiently Rice coded, while outliers are handled using escape coding. Simultaneously, for switching quantities, not only is state run-length coding performed, but the change time, edge direction, and duration are also precisely recorded and strictly aligned with the analog time axis. When a switching event occurs within the analog prediction window, the analog prediction model is reset or coding parameters are switched, ensuring coding consistency and decoding reliability of data near the fault time.

[0098] S5, Encapsulated anti-packet loss transmission: Each independent compressed frame is treated as a unique source object of an FEC group. Each independent compressed frame is padded and divided into k source fragments. The system RS(n,k) encoding is used to generate nk check fragments. The n fragments corresponding to the same independent compressed frame are combined into an FEC group. Each source fragment or check fragment is encapsulated into a transmission fragment. Then, the transmission is scheduled according to the low-orbit satellite link quality and fragment service priority.

[0099] In a specific application embodiment, step S5 includes: S51, the i-th independent compressed frame CF output in step S4 i As the i-th FEC group G i The only source object. According to the predetermined fragmented payload length Lshard in CF i The tail padding is a predefined padding value that can be removed, and it is divided into k. i Equal-length source fragments D i,0 To D i,ki-1 The system RS(n) is adopted. i ,k iEncoding to generate n i -k i One equal-length parity segment P i,ki To P i,ni-1 By k i Individual source fragmentation and n i -k i Each verification segment constitutes an FEC group G i and satisfy CF i With G i One-to-one correspondence. Global configuration metadata is compressed into a separate metadata frame CF0, and forms an FEC group G0 in the same way. Each source fragment or check fragment is treated as a transport fragment P. i,j The payload is such that a transmission fragment carries only one fragment in the same FEC group and does not carry a complete independent compressed frame on its own.

[0100] Specifically, each transmission fragment P i,j The fixed header may include: a 16-bit synchronization word (Sync), a 32-bit global waveform recording task identifier (TaskID), a 32-bit FEC group identifier (GroupID), a 32-bit independent compressed frame identifier (FrameID), and an 8-bit parameter n. i 8-bit parameter k i The data structure includes: a 16-bit fragment index (ShardIndex), a 1-bit source fragment or check fragment identifier (ShardType), a 24-bit fragment payload length (Lshard), a 32-bit independent compressed frame actual length (Lframe), a 64-bit absolute timestamp (Tabs), an 8-bit retransmission flag (Flagretx), a 32-bit fragment CRC value (CRCshard), and a 32-bit frame-level CRC value (CRCframe). The receiver first marks fragments that fail checksum verification as erased using CRCshard; when any k data is obtained from the same FEC group... i When there are valid fragments, according to the system RS(n) i ,k i Generate matrix recovery k i Each source is fragmented, concatenated according to its group number, and frame trailing padding is removed to obtain the CF. i Then, the frame integrity is verified using CRC frames. For a known erasure, a maximum of n frames can be recovered. i -k i There are n erase fragments; when there are v unknown errors and e known erases simultaneously, 2v + e ≤ n is satisfied. i -k i This ensures error correction. The independent decoding referred to in this embodiment refers to CF... i After recovery by the FEC group, lossless objects in the corresponding time window can be restored without referencing the state of the previous compressed frame, provided that global configuration metadata is obtained. This is not because a single transport fragment can be decoded independently.

[0101] S52 determines the link quality factor of the low-Earth orbit satellite link based on the weighted result of the signal-to-noise ratio, bit error rate, and remaining visibility time. Then, it determines the transmission priority based on the weighted result of the service priority of the transmission fragment, the fragment sequence number, and the ratio of the link quality requirement to the current link quality factor. The fragment scheduler then sends each transmission fragment in descending order of transmission priority.

[0102] The calculation expression for the link quality factor is as follows:

[0103] in, For link quality factor, For normalized signal-to-noise ratio, To normalize the bit error rate, This normalizes the remaining visible time, where each normalization parameter is the ratio of its corresponding metric to its preset upper limit: =SNR / SNR max , =BER / BER max , =T vis / T vis,max SNR max BER max and T vis,max These are the upper limits for signal-to-noise ratio, bit error rate, and visible time per transit, preset according to the low-Earth orbit satellite link specifications. , and These are weighting coefficients, which are preset based on the degree of influence of each indicator on the link availability, satisfying α+β+γ=1; The expression for calculating transmission priority is:

[0104] in, For transmission fragmentation Transmission priority, The priority of the service is determined based on whether the recording period corresponding to the transmission segment contains the moment of fault change or the moment of switch change. Segments containing the critical fault period are set as high priority (data of the critical fault period is set as high priority). The link quality requirements for transmission fragmentation are determined based on the minimum demodulation signal-to-noise ratio required by the modulation and coding scheme used in the transmission fragmentation. This is the current fragment sequence number (affecting decoding continuity). This represents the total number of transmission fragments. This represents the current link quality factor. , and These are weighting coefficients, preset according to the degree of influence of each factor on transmission timeliness, to meet... + + =1. The scheduler sends first. High-value fragmentation ensures that critical data before and after a sudden fault are transmitted first within the high-quality link window, thus enabling the most important fault waveform segments to be recovered earliest on the main station side, significantly reducing the transmission latency of critical information.

[0105] It is understandable that determining transmission priority by combining business priority, fragment sequence number, and link quality requirements, and prioritizing the transmission of data during critical fault periods, can ensure that core fault data is transmitted first within the high-quality link window, reducing the latency of critical information by more than 80% compared to ordinary no-priority transmission methods.

[0106] S6, the master station reassembles the received transmission fragments according to the FEC group, and marks the transmission fragments that fail to be verified as erased; when any k valid fragments in the same FEC group are obtained, the corresponding independent compressed frames are recovered and verified, and then the independent compressed frames are independently decoded without relying on the previous compressed frame; when there are fewer than k valid fragments in the same FEC group, the unrecovered fragments are selectively retransmitted according to the group identifier and fragment sequence number; the decoding results are spliced ​​together in chronological order, the lossless objects are restored item by item, and the fault waveform dataset is reconstructed.

[0107] In a specific application embodiment, the master station receiver maintains a reassembly buffer using TaskID, GroupID, and ShardIndex as the union key. Upon receiving P... i,j Next, CRC shards are checked first; fragments that fail the check are erased as known fragments, while fragments that pass the check are assigned to the corresponding FEC group according to GroupID and deduplicated according to ShardIndex. If a certain FEC group has obtained at least k i If there are 10 valid partitions, then select any k partitions from them. i Each segment is used to recover k from the corresponding RS generation matrix submatrix. i Each source is fragmented, concatenated according to ShardIndex, padding is removed based on Lframe, and the CF is recovered using CRCframe verification. i If at the retransmission decision time, the number of valid fragments r of a certain FEC group i Less than k i Then it is determined that at least k is still needed. i -r i For each valid fragment, priority is given to requesting the retransmission of source fragments that have not yet been received by the group; if necessary, unreceived check fragments may also be requested to be sent. Retransmission requests are uniquely identified by TaskID, GroupID, and ShardIndex, and FEC groups that have already been successfully recovered will not be retransmitted.

[0108] For independent compressed frames CF that have been recovered via FEC and passed CRC frame verification i The master station decoder reads the decorrelation mode, prediction mode, prediction order, fixed-point quantization prediction coefficients, Rice parameters, initial raw code values ​​for each channel, and length of each bitstream from the frame header. It then sequentially performs entropy decoding, prediction residual reconstruction, and inverse transform to recover the analog raw code value matrix X for that time window. raw,i , Switch quantity raw codeword matrix D raw,i Sample number and timestamp sequence TS i And quality markings. Because each CF i It carries the complete intra-frame restart state, and its decoding does not depend on the previous frame. Each time window can be streamed into the waveform analysis software immediately after the corresponding FEC group is restored; after all data FEC groups and metadata FEC groups have been restored, the COMTRADE configuration data and waveform data are reconstructed according to the original recording format and time order, so that the reconstructed lossless object is consistent with the original byte-by-byte or element-by-element data.

[0109] It is understandable that the transmission fragments are reassembled out of order, deduplicated, and checked using TaskID, GroupID, and ShardIndex. The lost source fragments are recovered using RS check fragments within the same FEC group. When there are fewer than k valid fragments, only fragments that have not yet been obtained in the FEC group are retransmitted. After the FEC group is recovered, the independent compressed frames corresponding to it are independently decoded and local waveforms are output, thereby supporting streaming recovery under intermittent visibility of low-orbit satellites and high-error links.

[0110] Compared with the prior art, the present invention has the following beneficial effects: This invention performs streaming parsing on fault recording files, demultiplexing and preprocessing the analog and digital quantities to generate structured data blocks. This allows FTU terminals to read, process, and compress data simultaneously without loading the complete COMTRADE fault recording file into memory all at once. This reduces the impact of recording file processing on RAM, storage space, and processor computing power, making it more suitable for resource-constrained FTU terminals.

[0111] This invention calculates characteristic indicators such as three-phase balance, zero-sequence component significance, and transient change intensity based on a working copy of the engineering quantity obtained from the original analog code value. It selects the appropriate decorrelation mode from among the direct-through mode, phase-to-phase differential mode, and integer reversible Clarke transform mode; the working copy does not overwrite the original analog code value. By performing reversible decorrelation and predictive residual coding on the original code value, the redundancy of the three-phase data space is reduced, while the decoding end can recover the original analog code value element by element, rather than only recovering the scaled and rounded engineering unit value.

[0112] This invention uses the power frequency cycle as a benchmark, determines the prediction frame length of data blocks based on waveform stability and switching events, and adaptively selects the prediction model according to the data characteristics within each prediction frame and the computing power of the FTU terminal. Longer prediction frames can be used for steady-state waveforms to reduce frame header and model description overhead, while shorter prediction frames can be used during transient abrupt changes to limit the propagation range of prediction errors, thus balancing compression efficiency and transient waveform recovery accuracy. When prediction performance degrades, processing such as prediction model degradation and resetting can reduce the increase in analog prediction residuals and error accumulation caused by model mismatch, improving the stability of compression performance under different fault stages.

[0113] This invention performs co-coding of analog quantity prediction residuals and digital quantity events, and keeps the digital quantity events aligned with the analog quantity time axis. When a digital quantity changes position, the analog quantity prediction model can be reset or the encoding parameters can be switched, so that the moment of digital quantity change accurately corresponds to the corresponding analog quantity sampling point, avoiding the fault timing deviation caused by separate processing of analog and digital quantities, and improving the accuracy of fault occurrence identification and fault judgment results.

[0114] This invention determines Rice coding parameters based on the statistical characteristics of analog prediction residuals, performs Rice coding on the analog prediction residuals, and performs escape coding on data exceeding a preset threshold. This allows the coding method to adapt to the different residual distributions exhibited by fault recording data in the steady-state and transient phases. Therefore, it can improve compression efficiency by utilizing the relatively concentrated nature of small residuals, while avoiding abnormal increases in code length caused by a small number of large residuals, thus enhancing the robustness of the compression process.

[0115] This invention ensures that each independent compressed frame corresponds one-to-one with a FEC group. The independent compressed frame is divided into k source fragments and n check fragments are generated. Each transmission fragment carries only one source fragment or check fragment. After obtaining any k valid fragments from the same group, the master station first recovers and checks the independent compressed frame, and then independently decodes the corresponding time window using the complete state within the frame. Therefore, it can obtain the fault waveform corresponding to the recovered FEC group without waiting for the complete fault recording task to finish transmitting.

[0116] This invention schedules the transmission of data fragments based on the quality of the low-Earth orbit satellite link and the priority of fragmented services, enabling fragments containing critical information such as fault abrupt changes and switch quantity changes to be transmitted with priority. In situations where low-Earth orbit satellite visibility is limited, the bit error rate is high, or latency fluctuates, this invention can increase the probability of critical fault data reaching the master station within a limited communication window, thus shortening the available time of critical fault waveforms.

[0117] The master station reassembles and corrects errors in the received transmission fragments according to FEC groups. For FEC groups with fewer than k valid fragments, selective retransmission is performed based on GroupID and ShardIndex. FEC groups that have already been recovered are not retransmitted. Independent compressed frames that have been recovered and passed frame-level CRC checks are independently decoded. The original analog code values, original switch states, sample time relationships, and quality markers of each time window are sequentially spliced ​​together, and the fault waveform dataset is reconstructed by combining it with global configuration metadata.

[0118] This invention integrates a processing flow through structured parsing of fault recording files, adaptive decorrelation, adaptive prediction, co-coding of analog prediction residuals and digital events, one-to-one mapping of independent compressed frames and FEC groups, and master station recovery. This flow identifies global configuration metadata, original analog code values, original digital states, sample time relationships, and quality markers as lossless objects, and ensures that engineering quantity conversion serves only feature calculations. Each independent compressed frame carries the complete intra-frame decoding state, and any k valid fragments from the same FEC group can recover the frame. This approach balances FTU terminal resource consumption, lossless fault recording data, satellite link transmission efficiency, and recovery capabilities after packet loss, making it suitable for remote operation and maintenance, fault monitoring, and emergency fault handling in distribution networks in areas with insufficient terrestrial communication coverage.

[0119] The present invention further provides a fault recording file compression and low-Earth orbit satellite packet loss-resistant transmission system, including a microprocessor and a memory interconnected thereto, wherein the microprocessor is programmed or configured to perform the steps of the fault recording file compression and low-Earth orbit satellite packet loss-resistant transmission method.

[0120] The system of the present invention corresponds to the method described above and has the same advantages as the method described above.

[0121] The present invention can implement all or part of the processes in the methods of the above embodiments, or it can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, it can implement the steps of the above method embodiments. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. Computer-readable media include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. The memory is used to store computer programs and / or modules. The processor implements various functions by running or executing the computer programs and / or modules stored in the memory, and by calling data stored in the memory. The memory may include high-speed random access memory, as well as non-volatile memory, such as hard disks, RAM, plug-in hard disks, smart media cards (SMC), secure digital (SD) cards, flash cards, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0122] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A fault recording file compression and low-orbit satellite anti-packet loss transmission method, characterized in that, Including the following steps: S1, stream analyzes the fault recording file, demultiplexes the analog and digital quantities in it, and generates structured data blocks; S2, calculate the characteristic indicators of the three-phase voltage data or three-phase current data in the structured data block, and select the corresponding decorrelation mode for the data block in the direct mode, phase-to-phase differential mode and integer reversible Clarke transform mode according to the characteristic indicators and preset criteria; wherein the characteristic indicators include one or more of the following: three-phase balance, zero-sequence component significance and transient change intensity. S3, based on the power frequency cycle, determines the prediction frame length of the data block according to the waveform stability and switching events, and adaptively selects the prediction model and generates analog prediction residuals based on the data characteristics within each prediction frame and the computing power of the FTU terminal. S4. Co-encode the analog prediction residual and the digital event. Determine the encoding parameters based on the statistical characteristics of the analog prediction residual and execute the encoding. Use the digital event aligned with the analog time axis to control the analog prediction model reset or the encoding parameters switch to obtain an independent compressed frame that corresponds one-to-one with the corresponding time window and does not depend on the previous compressed frame. S5, each independent compressed frame is padded and divided into multiple equal-length source fragments, multiple check fragments are generated using system RS encoding, multiple fragments corresponding to the same independent compressed frame are grouped into an FEC group, and each source fragment or check fragment is encapsulated into a transmission fragment, and then the transmission is scheduled according to the low-orbit satellite link quality and fragment service priority. S6, the master station reassembles the received transmission fragments according to the FEC group and marks the transmission fragments that failed the verification as to be erased; When any number of valid fragments in the same FEC group are obtained, the corresponding independent compressed frames are recovered and verified. Then, the independent compressed frames are independently decoded without relying on the previous compressed frame. When the number of valid fragments in the same FEC group is less than the preset number, the unrecovered fragments are selectively retransmitted according to the group identifier and fragment sequence number. The decoding results are spliced ​​together in chronological order to restore the lossless objects item by item and reconstruct the fault waveform dataset.

2. The fault recording file compression and low-orbit satellite anti-packet loss transmission method according to claim 1, characterized in that, Step S1 includes: The COMTRADE configuration file is read sequentially, and the original byte sequence of the configuration file, as well as the data file type, byte order and record format, are preserved as global configuration metadata. The sampling rate, the scaling factor, offset, unit and first or second value identifier of the analog channel and the normal status logic value of the digital channel are parsed in the working copy of the configuration file, and a channel mapping table including channel index, channel type, name and conversion parameters is established. The data block buffer length is determined based on the correspondence between the sampling rate and the power frequency cycle, and the COMTRADE data file is read in blocks using a read-and-press method. The analog and digital signals at each sampling point of the data block are demultiplexed according to the channel mapping table. The original analog code values ​​are retained to form an analog original code value matrix. An engineering unit value working copy, used only for feature extraction and threshold judgment, is calculated according to the scaling factor and offset. The original codewords of the digital signals are retained. At the same time, a logical state working copy for event detection is generated according to the channel mapping table. Neither the engineering unit value working copy nor the logical state working copy overwrites the corresponding original code value or original codeword. The system detects and marks analog quantity exceeding limits, abnormal transients in digital signals, and missing channels, generating a structured data block that includes global configuration metadata references, block start timestamps, number of sampling points, sampling rate, sample number and timestamp sequence, original analog quantity code value matrix, original digital signal codeword and logic state matrix, anomaly marker bitmap, and missing channel information. The original code values ​​of abnormal sampling points are retained in the structured data block, and the interpolation results are only used as optional analysis results and not as lossless encoding objects.

3. The fault recording file compression and low-orbit satellite anti-packet loss transmission method according to claim 1, characterized in that, In step S2, the three-phase balance is calculated by the ratio of the difference between the maximum and minimum values ​​of the three-phase effective values ​​to the average value of the three-phase effective values, expressed as: wherein, represents the three-phase balance degree; represents the effective value of the kth phase within the current data block; The significance of the zero-sequence component is calculated by the ratio of the zero-sequence component energy to the total three-phase energy within the data block, expressed as: in, Indicates the significance of the zero-order component; This represents the zero-order component of the nth sampling point; These represent the voltage or current sample values ​​of phases a, b, and c at the nth sampling point, respectively. The transient change intensity is calculated by the ratio of the maximum absolute value of the three-phase first-order difference to the average absolute value of the three-phase first-order difference at the sampling points within the data block, expressed as: in, Indicates the intensity of transient mutations; This represents the first-order difference of the k-th phase at the n-th sampling point; n represents the sampling point number; L represents the number of sampling points in the current structured data block.

4. The fault recording file compression and low-orbit satellite anti-packet loss transmission method according to claim 1, characterized in that, In step S2, based on characteristic indicators and preset criteria, the appropriate decorrelation mode is selected for the data block among the straight-through mode, phase-to-phase differential mode, and integer invertible Clarke transform mode. Specifically, this includes: The decorrelation mode is selected based on a multi-level criterion, prioritizing transient change intensity and secondarily considering three-phase balance or zero-sequence component significance. Specifically: when the transient change intensity is higher than a high threshold, the direct pass mode is selected; when the transient change intensity is lower than a high threshold and the three-phase balance is higher than the corresponding threshold or the zero-sequence component significance is higher than the corresponding threshold, the phase difference mode is selected, using the difference between the sampled values ​​of two adjacent phases as the difference component and retaining the third phase as the reference phase; when none of the above conditions are met, the integer reversible Clarke transform mode is selected, using a fixed-point linear combination of the three-phase sampled values ​​and rounding down to transform the three-phase stationary coordinate system to a two-phase stationary coordinate system and the zero-sequence component; finally, the mode identifier of the selected decorrelation mode and the necessary parameters required for the inverse transform are written into the data block header.

5. The fault recording file compression and low-orbit satellite anti-packet loss transmission method according to claim 1, characterized in that, In step S3, when determining the predicted frame length of the data block based on waveform stability and switching events, the predicted frame length is dynamically selected between short and long frames based on the waveform differential variance, zero-crossing density, and switching event flags within the current data window. Specifically, a short frame is selected when the waveform differential variance exceeds the corresponding threshold, the zero-crossing density exceeds the corresponding threshold, or a switching event is detected; otherwise, a long frame is selected. Based on the data characteristics of each prediction frame and the computing power of the FTU terminal, the prediction model is adaptively selected and the analog prediction residuals are generated, specifically including: Based on the autocorrelation characteristics of the current intra-frame analog data and the available computing power of the FTU terminal, the prediction order of the linear prediction model is selected and the prediction coefficients are solved. The prediction coefficients are then subjected to fixed-point quantization, and the prediction residuals are calculated. The autocorrelation characteristics are obtained by summing the products of different delayed sample values ​​after removing the intra-frame mean. The prediction coefficients of each order are obtained by solving the Yule-Walker equations formed by these autocorrelation characteristics, expressed as follows: in, This represents the autocorrelation coefficient with a delay order of k, where k represents the autocorrelation delay order. This represents the average value of the analog sampled values ​​within the current prediction frame, and p represents the prediction order determined based on the autocorrelation characteristics and the available computing power of the FTU terminal. This represents the sampled value of the i-th analog quantity within the current prediction frame; Indicates the i-th element in the current prediction frame k analog sample values; Indicates the current predicted frame length; Indicates a delay order of m The autocorrelation coefficient of n; The fixed-point quantization of the prediction coefficients and the calculation of the prediction residuals include: scaling and rounding the prediction coefficients according to the decimal width of the fixed-point number to obtain the fixed-point quantized prediction coefficients; calculating the prediction value based on the fixed-point quantized prediction coefficients and the historical analog sample values ​​before the current sampling point; and using the difference between the actual analog sample value and the prediction value as the analog prediction residual. When the moving average of the absolute value of the analog prediction residual exceeds the residual threshold, or the compression gain of the current prediction frame is lower than a preset multiple of the compression gain of the previous steady-state prediction frame, the prediction model is degraded sequentially to second-order difference prediction, first-order difference prediction, and unmodified encoding. Among them, second-order difference prediction, first-order difference prediction, and unmodified encoding respectively use the first two analog sample values ​​for linear extrapolation, use the previous analog sample value as the prediction value, and set the prediction value to zero. Finally, the system outputs the analog prediction residual sequence, prediction frame length, prediction mode identifier, fixed-point quantization prediction coefficients, and prediction reset point position for each prediction frame.

6. The method for compressing fault recording files and transmitting them to low-orbit satellites with packet loss resistance according to claim 1, characterized in that, Step S4 includes: The analog prediction residuals are sign-mapped to convert signed analog prediction residuals into non-negative integers, and the zero-value density, variance, and maximum amplitude of the sign-mapped analog prediction residuals are calculated. Specifically, the sign mapping is represented by the difference between the absolute value of the analog prediction residual and the binary sign identifier. The zero-value density is calculated as the proportion of analog prediction residual samples with absolute values ​​less than a preset zero-value threshold to the total number of analog prediction residual samples in the current prediction frame or subframe. The variance is calculated as the average of the squared deviations of the analog prediction residuals relative to the residual mean. The maximum amplitude is the largest absolute value of the non-negative integer after sign mapping. Rice coding parameters are adaptively determined based on the zero-value density and variance. The Rice encoding parameters The relationship between the expected value of the absolute value of non-negative integers and the base-2 logarithm is determined by using a sign mapping. Using the Rice encoding parameters The non-negative integer after sign mapping is decomposed into a quotient and a remainder, wherein the quotient is the non-negative integer after sign mapping divided by... The result is that the remainder is a pair of non-negative integers after symbol mapping. The result of the modulo operation; the quotient is encoded using unary code, and the remainder is a fixed-length remainder. Bit-binary code encoding; For non-negative integers after symbol mapping that exceed a preset threshold T, escape coding is used, wherein the escape coding word width is the minimum number of binary bits required for the maximum magnitude of the non-negative integer after symbol mapping in the current prediction frame or subframe. Run-length encoding is performed on the switch state sequence corresponding to the switch event, the switch state value and continuous length are recorded, and the analog sampling point with the smallest time distance from the switch change time is determined as the corresponding analog sampling point, so as to align the switch change event with the analog sampling time sequence. The analog quantity prediction model is controlled according to the positional relationship between the switching quantity change time and the analog quantity prediction window. The analog quantity prediction window is determined by the current prediction processing time and the prediction history length. When the switching quantity change time is within the analog quantity prediction window, at least one of the following operations is performed: terminating the current prediction subframe and starting a new prediction subframe, clearing or resetting the history buffer of the analog quantity prediction model, reducing the prediction order, or increasing the Rice coding parameter. The Rice encoding result and the escape encoding result are used to form an analog residual bitstream. The digital run-length encoding result and the analog sampling point index, edge direction and duration corresponding to the digital change time are used to form a digital event bitstream. The frame identifier, time window start point, number of sampling points, prediction subframe boundary, prediction reset flag, decorrelation mode, prediction mode, prediction order, fixed-point quantization prediction coefficients, Rice parameters, initial raw code values ​​of each channel, bitstream length and frame-level CRC are organized together with the analog residual bitstream, digital event bitstream and anomaly marker bitstream into an independent compressed frame, so that the prediction state or encoding parameters of the previous compressed frame do not need to be referenced when decoding the independent compressed frame.

7. The method for compressing fault recording files and transmitting them to low-orbit satellites with packet loss resistance according to claim 6, characterized in that, The expression for the symbol mapping is: in, This represents the analog prediction residual at the t-th analog sampling point. This represents the actual sampled value of the t-th analog sampling point. This represents the predicted value of the t-th analog sample point. This represents the non-negative integer obtained after sign mapping of the analog prediction residual. Represents a binary symbolic function; The Rice coding parameter The expression is: wherein denotes the Rice encoding parameter, denotes the expected value of the absolute value of the non-negative integer after symbol mapping and predicts the residual variance from the analog quantity is estimated, denotes the floor function; When determining the analog sampling point with the smallest time distance from the moment of the switch change as the corresponding analog sampling point, the specific expression for mapping the moment of the switch change to the nearest analog sampling point index is as follows: wherein, denotes the sampling time of the i-th analog quantity sampling point, denotes the switching quantity displacement time, denotes the analog quantity sampling point index corresponding to the switching quantity displacement time.

8. The fault recording file compression and low-orbit satellite anti-packet loss transmission method according to claim 1, characterized in that, Step S5 includes: Each independent compressed frame corresponding to a time window is treated as a unique source object for an FEC group. For the i-th independent compressed frame CF i After filling the end of the frame with a predetermined padding value, it is divided into k. i Each source is divided into equal-length fragments, and the system RS(n) is used. i ,k i Encoding to generate n i -k i Each equal-length check fragment makes the independent compressed frame CF i With FEC Group G i One-to-one correspondence; Each source fragment or check fragment is encapsulated into a transmission fragment. The fixed header of the transmission fragment includes at least a synchronization word, a global waveform recording task identifier, an FEC group identifier, an independent compressed frame identifier, and n. i k i The fragment sequence number within the group, the source fragment or check fragment identifier, the fragment length, the actual length of the independent compressed frame, the absolute timestamp, the retransmission flag, the fragment CRC, and the frame-level CRC; The link quality factor of the low-Earth orbit satellite link is determined by weighting the signal-to-noise ratio, bit error rate, and remaining visibility time. Then, the transmission priority is determined by weighting the service priority of the transmission fragment, the fragment number, and the ratio of the link quality requirement to the current link quality factor. Each transmission fragment is then sent from high to low according to the transmission priority.

9. The fault recording file compression and low-orbit satellite anti-packet loss transmission method according to claim 8, characterized in that, The calculation expression for the link quality factor is as follows: wherein, is a link quality factor, is a normalized signal-to-noise ratio, is a normalized bit error rate, is a normalized residual visible time, , and are weight coefficients, respectively. The expression for calculating the transmission priority is: in, For transmission fragmentation Transmission priority, For business priority, To meet the link quality requirements for transmitting fragmented data, This is the current fragment number. This represents the total number of transmission fragments. This represents the current link quality factor. , and These are the weighting coefficients.

10. A fault recording file compression and low-Earth orbit satellite packet loss-resistant transmission system, comprising a microprocessor and a memory interconnected, characterized in that, The microprocessor is programmed or configured to perform the steps of the fault recording file compression and low-orbit satellite packet loss-resistant transmission method according to any one of claims 1 to 9.