High-precision waveform acquisition and time-stamping synchronization system for power grid fault location

Abstract

This invention discloses a high-precision waveform acquisition and time-stamping synchronization system for power grid fault location, relating to the field of power grid waveform synchronization technology. Step one involves synchronously outputting three levels of sampling under the same Coordinated Universal Time (UTC) time axis, and generating upper bounds for time uncertainty and group delay signatures for channels and segments. Step two involves near-lossless quality locking of high-energy segments, followed by encoding and verification of sample location and peak amplitude using a common-source detection kernel; failure to meet the criteria results in lossless rollback. Step three involves writing segment indexes, quality metadata, fragment verification structures, and full-file signatures into a compatible single-file extended container. Step four involves progressive retrieval by index, breakpoint continuation, and cross-site deduplication, along with cold and hot layered storage. Step five involves calculating reference times and generating acceptance lists based on group delay signature compensation and quality-weighted loading. This solution achieves cross-device alignment, reproducibility of key features, peak reduction during transmission and storage, unified playback and evidence collection standards, and considers tool compatibility and auditing.

Description

Technical Field

[0001] This invention relates to the field of power grid waveform synchronization technology, specifically a high-precision waveform acquisition and time-stamping synchronization system for power grid fault location. Background Technology

[0002] The main power grid and distribution network are deeply interconnected, with long-distance overhead lines, submarine cables, flexible DC converters, and wind and solar power electronic devices operating in parallel. Fault transients are characterized by high frequency, short duration, and strong nonlinearity. To achieve rapid location and post-event verification, the process-level sampling and merging units, protection and waveform recording devices within the station generally have ultra-high-speed event recording capabilities of one million times per second, while also generating longer time window records of ten thousand to one hundred thousand times per second for analysis and archiving. Events are transmitted back to the centralized platform via the station's redundant network and external dedicated network for centralized processing and archiving; disturbance data exchange mostly adopts the single-file encapsulation and variable sampling rate mechanism of the 2013 standard.

[0003] In the current network, protection triggers often occur back-to-back, with simultaneous recording at both or even multiple ends, leading to a surge in event volume and concurrency. Full data transfer and centralized playback create "floods" in links and storage, increasing retention and disaster recovery costs. More critically, existing common formats only provide record-level time quality codes, failing to express time uncertainties at the channel or segment level. Differences in the analog front-ends of different devices and their anti-aliasing filters, as well as channel link differences, introduce group delay biases that vary with frequency. This bias lacks clear description and transmission in event data, making it difficult to reproduce the arrival times and phases across devices and sites at the playback end, thus limiting the consistency of traveling wave ranging and transient criteria.

[0004] While variable sampling rates are permitted, in practice, either high-sampled and low-sampled data are separated into two file categories, or clear segment guidance is lacking within a single file, making it difficult for playback tools to unambiguously splice and retrieve data on a unified timeline. Multi-terminal repeated recordings lack segment-level verification and deduplication, and repeated uploading and storage further amplify resource consumption. Compression techniques are mostly used for continuous streaming data, but are less common in event domains and lack fidelity constraints related to critical protection quantities, posing risks of smoothing out spikes and time-of-arrival drift after compression. If these problems persist, they will lead to delayed location decisions, congestion on centralized platforms, increased storage costs, and time-difference disputes in evidence collection and cross-device verification, extending maintenance and power restoration time and affecting power supply reliability and user-side loss control.

[0005] Therefore, the current challenge is to meet the following constraints simultaneously in the event-level data organization and quality representation under the scenario of one million events per second: multiple sampled data files are consistently carried on a unified time axis; the time uncertainty of channels and segments, the group delay deviation that varies with frequency, and the repetitive segments are accurately and searchably represented; and the key characteristics of high-energy segments, such as traveling waves and spikes, are maintained with clear constraints so as to meet the synchronization requirements of reducing file size, reproducibility, and alignment under the conditions of multi-station concurrency and large-scale backhaul. Summary of the Invention

[0006] (a) Technical problems to be solved

[0007] To address the shortcomings of existing technologies, this invention provides a high-precision waveform acquisition and time-stamping synchronization system for power grid fault location. This system includes near-lossless quality locking of high-energy segments, encoding followed by verification of sample location and peak amplitude using a common-source detection kernel, with lossless rollback for non-compliant segments; writing segment indexes, quality metadata, fragment verification structures, and full file signatures into a compatible single-file extended container; progressive index retrieval, breakpoint resumption, and cross-site deduplication; and implementing hot and cold layered storage; and obtaining reference times and generating acceptance lists based on group delay signature compensation and quality-weighted loading. This system achieves cross-device alignment, reproducible key features, peak smoothing during transmission and storage, unified playback and evidence collection standards, and tool compatibility, solving the technical problems described in the background art.

[0008] (II) Technical Solution

[0009] To achieve the above objectives, the present invention is implemented through the following technical solution: a high-precision waveform acquisition and time-stamping synchronization system for power grid fault location, comprising: synchronously acquiring three levels of sampling at one million times per second, one hundred thousand times per second, and ten thousand times per second under the same Coordinated Universal Time time axis; generating an upper bound of time uncertainty and a group delay signature for each channel and each segment and binding them to the segment.

[0010] Near-lossless compression is performed only on high-energy segments that are processed one million times per second. After encoding, the position of the sample and the peak amplitude are calculated using a homologous detection kernel. If the threshold is exceeded, the segment is rolled back to lossless. The threshold is linked to the upper bound of time.

[0011] Write the fragment index, upper bound of time uncertainty, and quality metadata of group delay signature into a compatible single-file extended container, and write the fragment verification structure and whole file signature to make the fragments correspond one-to-one with the metadata;

[0012] The central side retrieves high-energy fragments based on the fragment index and performs verification and breakpoint resumption according to the fragment verification structure. It performs cross-site deduplication based on the digest matching and allocates them to cold and hot tiered storage according to priority.

[0013] The playback end reads the group delay signature to perform time and phase compensation on the channel, performs quality weighted loading based on the upper bound of time uncertainty, and renders the segment and the estimated arrival sample position on a unified time axis.

[0014] Furthermore, a unified time axis is constructed using whole seconds and nanoseconds. For each segment, an upper bound for time uncertainty is generated based on the timing residual, asymmetric error, drift boundary, and the composite aperture of the marker quantization. Group delay signatures and their version numbers are established and bound to the segments through frequency sweep calibration and in-operation self-calibration.

[0015] Furthermore, channel-independent prediction-residual and entropy encoding processing is performed on high-energy fragments, and a near-lossless threshold is determined based on the equivalent group delay derived from the self-group delay signature and the upper bound of the time uncertainty. After encoding, the arrival sample position and peak amplitude are extracted using the same-source detection kernel for verification.

[0016] Furthermore, when the sample location or peak amplitude does not meet the preset criteria, the corresponding segment is rolled back to lossless or the abnormal short window is rolled back to lossless, and the audit record of passing or rolling back is written along with the segment. The audit record includes the coding category, threshold, sample location, and error field.

[0017] Furthermore, fragment indexes and quality metadata are written to the extended area of ​​the single-file extended container. The fragment indexes include at least the start and end times, sampling rate, number of samples, file offset, encoding method, and digest value. The quality metadata includes at least the upper bound of time uncertainty, group delay signature, arrival sample location, and audit record entry.

[0018] Furthermore, a fragment verification structure is constructed and leaf summary list and hierarchical information are written into it. A package list containing fragment index summary, quality metadata summary and root summary is generated. A full file signature is generated based on the package list. The package list and full file signature are written into the extension area and a one-to-one correspondence is established with the fragments.

[0019] Furthermore, the central side calculates fragment priorities based on fragment indexes and quality metadata, prioritizes the retrieval plans for high-energy fragments and higher-priority fragments, initiates requests for continuous windows aggregated by file offset and length, and verifies and registers each page at the receiving end based on the fragment verification structure.

[0020] Furthermore, in the scenario of resuming interrupted downloads, only pages that fail the verification are retrieved again; when importing data across sites, the target storage is queried using the leaf summary as the key. If the target storage already exists, a reference relationship is established and the writing is skipped. If the target storage does not exist, the corresponding leaf summary is added to the summary set after the writing is completed to support subsequent deduplication.

[0021] Furthermore, based on the fragment priority score and tag attributes, fragments are written to the super-hot layer, hot layer, or archive layer, and the object identifier, the level to which they belong, and the retention policy are registered in the directory service, so that the playback end can retrieve and load them according to the object identifier and level and maintain the reference relationship consistent with the scope retrieval.

[0022] Furthermore, the playback end reads the group delay signature and performs frequency domain phase pre-compensation or equivalent fractional delay filtering compensation on each channel, and reconstructs the multi-channel segment sequence on a unified time axis, while maintaining a consistent reference relationship between the arrival sample position and the segment index during the rendering process.

[0023] Furthermore, loading weights are generated based on the upper bound of time uncertainty and the equivalent group delay. The segments are then loaded according to the weights, and the upper bound of time uncertainty and the group delay signature version number of each channel are marked on the playback interface. The loading order is consistent with the aforementioned hierarchical records, and the references and numbers correspond.

[0024] Furthermore, a weighted absolute deviation target is constructed based on the error between the arrival sample position and arrival time of each channel, a reference time is obtained, and each channel is aligned on a unified time axis. Subsequently, an acceptance list containing segment number, reference time, deviation, and half-width fields is generated and archived according to the object identifier.

[0025] Furthermore, the extended area is placed at the end of the information segment and serialized using normalized encoding. A magic number segment, version number, and checksum are set for identification, enabling older players that do not recognize the extended area to ignore it and read the basic segment. Fields are labeled with Chinese names and units.

[0026] (III) Beneficial Effects

[0027] This invention provides a high-precision waveform acquisition and time-stamping synchronization system for power grid fault location, which has the following advantages:

[0028] Based on multi-level unified sampling, upper bound of time uncertainty, and group delay signature, a unified time semantic is established across multiple devices and sites. The quality metadata is written along with channels and segments, providing an entry point for subsequent threshold linkage, compensation, and quality-based usage, reducing alignment ambiguity. A near-lossless quality locking and post-encoding feature verification → automatic rollback is adopted, compressing only high-energy segments and verifying the sample position and peak amplitude with a common-source detection kernel. If the criteria are not met, rollback is performed on-site and audit entries are recorded, constraining key quantity offsets from the source. Segment indexes, quality metadata, fragment verification structures, and full-file signatures are written into a compatible single-file extended container, forming a one-to-one correspondence between segments and metadata. Older players can ignore the extended area, while newer tools can read the index and quality fields, ensuring cross-system flow.

[0029] The central side performs progressive retrieval based on fragment indexes and quality metadata, first loading high-energy fragments and necessary indexes; then, it completes breakpoint resumption and page-level verification according to the fragment verification structure, and performs cross-site deduplication using digests; finally, it writes to the cold and hot layers according to priority, unifying scheduling and disk placement. The playback end reads the group delay signature to execute time and phase compensation, reconstructs multi-channel fragments on a unified timeline; it performs quality-weighted loading based on the upper bound of time uncertainty, obtains the reference time and aligns each channel, and outputs an acceptance list containing fragment numbers and audit fields. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the high-precision waveform acquisition and time-stamping synchronization system for power grid fault location according to the present invention. Detailed Implementation

[0031] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0032] Please see Figure 1 This invention provides a high-precision waveform acquisition and time-stamping synchronization system for power grid fault location, comprising:

[0033] Step 1: Establish a unified time reference for three sampling levels under the same Coordinated Universal Time (UTC) time axis, and generate quality metadata for each channel and segment that can be used throughout the entire process; specifically, form an upper bound for time uncertainty to define the worst-case error of timestamp, and form a group delay signature to characterize the frequency-related time deviation of the front end; use whole seconds plus nanoseconds to count in the unified clock domain, and label the segment with signature version number and generation time to ensure that different devices and sites can directly reference and splice the same field.

[0034] For power grid scenarios involving long cross-regional lines, submarine cables, and power electronic devices, fault transients are characterized by high frequency, steep edges, and short duration. The merging unit, protection and waveform recording device in the station needs to simultaneously save three waveform levels: one million times per second, one hundred thousand times per second, and ten thousand times per second, in order to take into account traveling wave location, transient interpretation, and post-event verification.

[0035] In current networks, single-file encapsulation and variable sampling rate writing are mostly used for file encapsulation and transmission, but only record-level time quality prompts are provided; the arrival delay that varies with frequency introduced by the analog front-end and anti-aliasing filtering of different devices is not expressed in the event data, which makes it difficult to align the playback end across devices and sites; back-to-back triggering and superimposed dual-end recording cause data surges in the link and storage.

[0036] Faults often occur back-to-back, and time comparisons across devices and sites must be based on the same time reference; otherwise, any subsequent compression, encapsulation, and playback will amplify the time error.

[0037] Therefore, before the three sampling levels are implemented, a unified clock domain mapping needs to be constructed, and upper bounds for channel-level and segment-level time uncertainties need to be given to clarify the worst-case error boundary of the timestamp. The in-station time synchronization link, satellite time synchronization status, device local oscillator drift, and link asymmetry are quantized into several synthesizable error components. First, a mapping from local counting to Coordinated Universal Time is established, and then the segment-level time uncertainty upper bound is synthesized using a measurable upper bound operation and written along with the segment. The three sampling levels share the same mapping, so that different sampling intervals can be unambiguously spliced ​​on a unified time axis.

[0038] Furthermore, a linear mapping is established between local sample counts and Coordinated World Time, and aligned counts are introduced to constrain the zeros:

[0039]

[0040] Where: Coordinated Universal Time Real numbers, used for unified timeline calibration of three segments; local counting. : A non-negative integer, corresponding to the sampling point number; sampling interval : Positive real number, with values ​​corresponding to the reciprocals of one million, one hundred thousand, and ten thousand samples respectively; clock skew : Real number, estimated by the alignment process; clock drift : Real number, obtained by fitting the drift within the alignment period; Alignment count : Integer, the reference count for this alignment.

[0041] First, capture a set of pulses when the pulse arrives at the full second. The clock offset and alignment count are obtained, and then the sampling points are drift-fitted with multiple timestamps within an alignment period to obtain the clock drift. This mapping is applied to each sampling point of the three sampling levels simultaneously to uniformly generate nanosecond-level time stamps aligned to whole seconds. In application, the timestamps of the three segments have strict consistency, and time comparisons across devices have a traceable source under the unified mapping. The introduction of the alignment count ensures that the mapping zero point is determined, avoiding systematic deviations caused by zero-point drift after long-term operation.

[0042] The timing residual, link asymmetry, short-term drift boundary, and time stamp quantization error are combined by upper bound:

[0043]

[0044] Where: Upper bound of time uncertainty : A non-negative real number representing the worst-case error of a segment timestamp; timing residual : Real number, read the local clock offset between the local and master clocks from the station's time synchronization protocol stack and take the maximum absolute value; asymmetric error : Real number, given by the link round-trip asymmetric assessment, and the delay is measured using round-trip packets. ,remember ;

[0045] Drift boundary : A non-negative real number, an upper bound derived from the extrapolation of the drift trend within the alignment period, based on clock drift. Multiply by the fragment duration count range The upper realm, Marking error : A non-negative real number, given by the quantization precision bound of the sampling marker.

[0046] First, read the residuals from timing and link monitoring, calculate the round-trip delay and symmetry index to obtain the asymmetric error, then extrapolate the short-term drift bound based on the drift fitting coefficients of the previous alignment period, and finally add the marked quantization accuracy bound to form the upper bound of the segment-level time uncertainty; then, apply the upper bound of the time uncertainty. Quality metadata is written to each segment and displayed on the device interface with text and color. In application, the upper bound of the time stamp is clearly expressed at the segment level, and subsequent compression and playback can be based on this for threshold linkage and confidence presentation; different channels of the same event are under a unified clock domain, and the time difference between them is limited to a clear boundary after compensation.

[0047] Different analog front-ends and anti-aliasing filters in different devices produce different arrival delays in different frequency bands. If these delays are not expressed and compensated for at the event layer, they will create systematic biases in traveling wave localization and transient criteria. Therefore, it is necessary to generate frequency-dependent group delay signatures and maintain their validity during operation.

[0048] During the commissioning phase, a sweep frequency signal is used to pass through the analog front end to measure the phase-frequency relationship, derive the group delay curve, and solidify it into a group delay signature. During the operation phase, key frequency points are refreshed in a short-time self-calibration mode to prevent offset caused by temperature drift and aging, and the equivalent group delay is calculated on each event segment for time and phase compensation at the playback end.

[0049] Furthermore, a frequency sweep signal with known amplitude and phase is passed through the analog front end to calculate the frequency domain group delay and solidify it as a signature. At the same time, the equivalent group delay of the segment is defined:

[0050]

[0051] In the formula: equivalent group extension : Real number, representing the energy weighted to time delay of a segment within a frequency band; lower frequency limit Frequency limit : Real number, defined by the frequency band segment; spectral energy function Unit arbitrary amplitude squared / Hertz, non-negative real function, estimated from the segment spectrum; segment spectrum The frequency band was estimated using the Welch method (overlapping window + averaging). Derived from the main band of fragment energy; group extension function : Real function, calculated from the frequency sweep calibration results; The acquisition method is as follows: the sweep frequency signal is injected point by point at the factory, and the phase is measured. And approximate the group delay using the central difference. Discrete points are interpolated into continuous functions using cubic splines.

[0052] Furthermore, a swept-frequency signal source is used to inject the signal point by point from low to high frequency, and the difference between the output phase and the input phase is recorded. The group delay function is obtained by differentiating the signal with respect to frequency. Cubic spline interpolation is used to form a continuous representation between discrete frequency points, making it a group delay signature; for each event segment, based on the segment spectral energy function... Calculate equivalent group delay It is also attached to the fragment metadata. In application, the group delay signature fixes the frequency-related arrival deviation into a stable, callable quantity, and the equivalent group delay provides a clear basis for fragment-level time compensation; interpolation ensures smoothness and computability between frequency points.

[0053] To ensure that the group-delayed signature does not drift during runtime and to ensure that the three segments can be directly spliced ​​on a unified timeline, the segment timeline expansion is defined as follows:

[0054]

[0055] Where: fragment time : Real number, representing the first The first segment The time of each sample on a unified timeline; the start time of the segment. : Real number, the starting point obtained from the mapping calculation; segment sampling interval : Positive real number, corresponding to the sampling level of this segment; Sample number : Non-negative integer.

[0056] On a monthly basis, during low-load windows, key frequency points of group delay are checked using short-duration narrowband pulses. If the group delay at a key frequency point deviates from a set threshold, the interpolation coefficients are automatically updated and a new version of the group delay signature is generated. When writing segments, the unified timeline time for each sample is calculated using the above formula, and the equivalent group delay and signature version number of the segment are attached. The three segments are directly spliced ​​together on a unified timeline at the playback end without further alignment. In application, the group delay signature is maintainable, the time monotonicity of segment splicing is guaranteed by the formula, and the signature version number ensures the traceability of the compensation basis in the evidence collection process.

[0057] The frequency-dependent time error is made public as a callable signature and bound to the segment timeline, providing direct input for subsequent compression threshold linkage, encapsulation quality metadata and playback compensation.

[0058] Step 2: Reduce the size of the traveling wave without compromising the reproducibility of key quantities. Perform near-lossless quality locking only on high-energy segments, and link the threshold with the upper bound of the time uncertainty and the equivalent group delay from Step 1. After encoding, recalculate the sample position and peak amplitude using a common-source detection kernel. If the criteria are not met, trigger a segment or local lossless rollback, generating an audit record for either pass or rollback. Simultaneously, record the threshold trajectory and detection kernel configuration, retaining the encoding and verification criteria to ensure consistent reading and traceability of the processing history during subsequent encapsulation and playback.

[0059] High-energy segments carry the leading edge of traveling waves and high-frequency details. If compressed with only a fixed threshold, uncontrollable timing deviations may occur due to timing degradation or front-end group delay offset. Therefore, it is necessary to embed the upper bound of time uncertainty and the equivalent group delay into the threshold decision, so that the allowable near-lossless amplitude adaptively converges with the quality boundary, thereby actively reducing the compression aggressiveness when the input quality deteriorates.

[0060] First, the high-energy segments of each channel are sliced ​​along a unified time axis, and then an easily entropy-evolved integer sequence is formed using a prediction-residual approach. Simultaneously, the upper bound of the time uncertainty and the equivalent group delay of the segment are read, and an effective threshold is given through a threshold linkage function. This threshold controls the residual quantization and bit plane truncation. Finally, a near-lossless bitstream and threshold strategy trajectory are generated, providing a basis for subsequent feature verification and backoff determination.

[0061] To separate high-frequency mutations from low-frequency smoothing components, autoregressive prediction is first performed on the sequence within each channel, and then the residual sequence is used for entropy encoding.

[0062]

[0063] Where: input sample : Integer, representing the current channel's position on the unified timeline. One sample; residual sample Integer, used as input for entropy coding; prediction order : Positive integers, the order of which can be finitely small to balance complexity and adaptability; prediction coefficients : Real number, obtained by solving the minimum absolute error of the short window or bounded least squares, to ensure that the peaks are not excessively smoothed.

[0064] Furthermore, estimation is performed within the segment using a sliding window. The window length is coordinated with the equivalent group delay to avoid phase distortion in the main frequency band of the group delay; the beginning of each window is written without prediction, and the residual is formed by the above prediction within the window. Then, amplitude limiting and entropy coding are performed. In application, residual shaping reduces energy redundancy, allowing subsequent entropy coding to focus more on local changes; window and group delay are set in tandem to reduce the impact of long trends on compression ratio while maintaining the discernibility of peak positions; independent channel processing ensures traceability of different physical channels.

[0065] To avoid the superposition of time error and compression error, a function is constructed that maps the upper bound of time uncertainty and the equivalent group delay to an effective threshold, thus limiting the upper limits of residual quantization and bit plane truncation:

[0066]

[0067] Where: effective threshold : A positive real number that determines the maximum allowable amplitude of residual quantization and bit plane truncation; the least significant bit of the quantization. The inherent resolving power of the device; the upper bound of the time uncertainty. : Non-negative real number, from the fragment-level quality annotation in step one; equivalent group extension : Real number, from the frequency-group delay signature weighted result of step one; mapping coefficient : Positive real number, used to map time units to amplitude tolerance; weighting coefficient : Non-negative real number, used to adjust the effect of group delay on the threshold; regularization term : Positive real numbers, to avoid the denominator approaching zero.

[0068] Furthermore, the upper bound of the step response slope of the device channel is used. To bridge this gap, we map the upper bound of the time error to the amplitude tolerance, and provide an alternative expression:

[0069]

[0070] in It can be estimated from the maximum rising edge during factory frequency sweep and step test; , which is the safety factor; Time-amplitude mapping coefficient;

[0071] Furthermore, the upper bound of the time uncertainty of reading the segment. With equivalent group extension Calculate the effective threshold This value limits the residual quantization step size and truncation depth; when processing method: Increase or As the threshold increases, it automatically tightens, thus becoming more conservative when input quality deteriorates; processing method: and Within a good range, the threshold does not exceed the lowest quantized bit. To avoid exceeding the physical resolution of the device. In application, threshold linkage transforms the input quality-compression tolerance into a monotonic relationship, reducing arbitrary parameter settings; different channels within the same event due to... and Differentiation is achieved through differentiated tolerance, improving overall usability; the boundedness of the linkage function ensures that the device's resolution limit is not exceeded even in the worst case.

[0072] Near-lossless design allows for small amplitude deviations, but the arrival time of the traveling wave and the peak amplitude are key quantities for positioning and identification. They must be recalculated using the same algorithm after encoding and compared with the reference value. If they exceed the limits, they should be rolled back in place to avoid the error being propagated to the sealing-core and application-certification processes.

[0073] The near-lossless bitstream is decoded into a verification sequence. The arrival time detection kernel, which is of the same origin as the device, is used to calculate the position of the arriving sample and to verify the peak amplitude. The arrival time threshold and amplitude threshold are used as criteria to form a backoff decision, triggering segment-level backoff or partial backoff, and generating pass / backoff audit records, recording the threshold, error and judgment result.

[0074] Furthermore, peak search is performed on the calibration sequence using convolutional detection kernels to obtain the estimated arrival sample location:

[0075]

[0076] Where: Arrival sample location estimation : A non-negative integer representing the sample index with the largest detection kernel response amplitude; verification sequence : Integer or fixed-point, a segment after decoding a near-lossless bitstream;

[0077] Detection nucleus The real sequence employs a finite-supported, band-limited leading-edge template with a window function to make it sensitive to the rising edge of the traveling wave, wherein:

[0078]

[0079] In the formula: It is a unit step discrete form. The kernel length is an odd number, chosen based on the sampling rate and desired rising edge width. This kernel is sensitive to positive steep edges, supporting convolution peak localization; if the scene is dominated by negative steep edges, a different kernel can be selected. ; Hanning window weight, real number; Core length, odd number;

[0080] check sequence With detection nucleus A finite-length convolution is performed to obtain the response sequence; parabolic interpolation is then used to refine the sample location estimation. The amplitude is compared with the reference arrival sample position, which is obtained from before encoding or estimated by the adjacent uncompressed window; at the same time, the reference peak amplitude is read and compared with the current peak amplitude to obtain the amplitude deviation.

[0081] In application, this process compares the arrival sample position and peak amplitude before and after encoding on the same time axis, which can directly reveal the perturbation of key quantities by compression; parabolic interpolation improves the precision of subsample-level positioning and reduces the rounding deviation caused by sampling interval; and the use of homogeneous detection kernels ensures the consistency of verification caliber.

[0082] Furthermore, by combining the arrival time error and amplitude error with the threshold, the backoff criterion is obtained:

[0083]

[0084] Among them: Criterion Instruction : Values Or 1, Indicates passage, Indicates rollback; arrival time error : Real number, estimated from the location of arrival at the sample. The amplitude error is obtained by multiplying the difference between the arrival sample position and the reference position by the sampling interval. : Real number, representing the difference in peak amplitude; arrival time threshold : Positive real number, the upper limit set by the engineering, used as the conjunctive criterion for arrival time error, taken from the self-inspection and type test limits of the protection device; amplitude threshold. : Positive real number, upper limit set by the project; indicator function Boolean mapping: 1 if the condition is met, 0 otherwise.

[0085] If the criterion indicates If the segment passes, the encoding category, threshold, arrival time error, and whether it passes are recorded; if the criterion indicates... If the error is concentrated within a short window, a fragment-level lossless rollback is triggered, and the fragment is recoded losslessly. If the error is concentrated within a short window, a local lossless rollback is performed on that short window, while the remaining parts remain near-lossless encoded. All rollback actions and the final state are written to the pass / rollback audit log.

[0086] In application, the conjunctive criterion simultaneously constrains both time and amplitude, avoiding allowing passage based on a single dimension; segment-level rollback and partial rollback provide granular layering, controlling distortion propagation while also considering volume; audit records ensure traceability and accountability. Recalculated key quantities and thresholds are merged into explicit rollback actions, forming a closed loop of compression-verification-rollback upon exceeding limits, ensuring that compression does not alter the core information used as the basis for positioning.

[0087] Step 3: Without disrupting the existing ecosystem, solidify the structural and quality information generated in Steps 1 and 2 into a compatible single-file extended container. Establish a fragment index, write the upper bound of the time uncertainty and the group delay signature, record the arrival position of the sample and audit information, and configure the fragment verification structure and the entire file signature to ensure that fragments and metadata correspond one-to-one, are searchable, and verifiable. The extended area does not change the semantics of the basic segments, uses a fixed field order and Chinese unit annotations, and older players can ignore the extended area, while new tools use it to implement the location, extraction, and verification processes.

[0088] With three sampling levels coexisting on a unified timeline, if the one-to-one correspondence between the segment start and end, sampling interval, number of samples, byte offset, and encoding method is not clearly defined, subsequent incremental retrieval of only high-energy segments will lose its positioning basis. Simultaneously, the upper bound of quality-related time uncertainty, equivalent group delay, estimated arrival sample position, and pass / back audit records need to be strongly bound to the segments to avoid mismatch during cross-system transfer. Therefore, the structure and quality boundaries must be isomorphized and documented together.

[0089] First, read the segment start time, segment sampling interval, number of segment samples, near-lossless or lossless bitstream, upper bound of time uncertainty, equivalent group delay, estimated arrival sample position, and pass / back audit records output from steps one and two. Then, establish a segment index and quality metadata in the extended area, and perform byte alignment on the segment data area to give the segments fixed offset and alignment granularity, which facilitates segment retrieval and cross-site deduplication. Finally, generate an integrated orchestration of segment index, quality metadata, and segment data.

[0090] The fragment index needs to provide the fragment offset to the file, as well as the encoding method and length, which can be randomly located. To ensure that the fragment retrieval is aligned with the storage system pages, offset planning is performed using alignment granularity:

[0091]

[0092] The parameters have the following meanings: fragment offset : A non-negative integer, representing the first... The starting offset of each data fragment in the container file; starting offset : A non-negative integer representing the starting position of the fragment data region; fragment length in bytes. : A positive integer, representing the first... Actual bitstream length of each segment; alignment granularity : A positive integer representing the suggested page size for the file system or transport layer.

[0093] The segments are sorted according to their natural order on a unified timeline, the aligned occupancy of each segment is calculated, and the final occupancy is determined accordingly. Record the segment number, start and end times, sampling interval, number of samples, encoding method, segment byte length, and segment offset in the segment index, and mark the index checksum to prevent index corruption; make the arrival sample position estimate an optional field of the index to facilitate upper-level sorting and retrieval by arrival time.

[0094] When applied, after offset alignment, fragments can be located and extracted in whole pages within the entire file, significantly reducing the addressing overhead of range downloads; the estimated arrival sample position is entered into the index, and with the unified timeline, key fragments are retrieved first with a clear entry point; index self-checking improves robustness during cross-system dumps.

[0095] The upper bound of time uncertainty, equivalent group delay, arrival sample location estimation, pass / back audit records, coding category and threshold strategy trajectory are essential for playback and quality control and should be inextricably linked to the segment bitstream.

[0096] Construct a quality metadata object for each segment, with fields including: channel set, segment start and end times, segment sampling interval, and upper bound of time uncertainty. Equivalent group extension Arrival of sample location estimation Arrival time error Amplitude error The encoding category (near lossless or lossless), effective threshold trajectory, and pass / back audit records are used to serialize the object and write it into the extension area in a one-to-one correspondence with the fragment data, and save its offset position in the index. To prevent bitstream-metadata mismatch, a metadata digest is calculated for the metadata object and the digest is written into the index.

[0097] When applied, the quality metadata has a precise location and summary verification within the container, and any data movement that deviates from the fragment will be identified; when the host computer reads, it can directly pull the quality boundary + bitstream by fragment, avoiding secondary association; the threshold strategy trajectory and pass / rollback audit records are written to disk together, providing audit-oriented traceability.

[0098] Container files need to have segment verification, breakpoint resumption, cross-site deduplication, and source verification in cross-site transmission and multi-copy storage. A single whole file digest cannot support page-level retransmission and deduplication. Therefore, it is necessary to build a segment verification tree with the fragment data area and index area as input, and sign the whole file based on the list.

[0099] The container is divided into byte blocks of fixed page size. A leaf digest is calculated for each byte block, and these are synthesized layer by layer in a binary manner until the root digest is obtained. Simultaneously, a packing manifest is constructed, containing a fragment index, quality metadata digest, root digest, and version information. Finally, the manifest is digitally signed. The fragment verification tree is used for page-level verification and deduplication, while the signature is used for source verification and handover verification.

[0100] To support range downloading of high-energy segments first and cross-site deduplication, a binary concatenated fragmented checksum tree is adopted:

[0101]

[0102] The parameters have the following meanings: Leaf Summary : Fixed-length byte string, for the first A digest of a block of bytes; a hash function : SHA-256, a collision-resistant hash function with fixed output length, is selected; byte block : Length is a fixed page size; parent node summary A fixed-length byte string, obtained by concatenating and hashing the child node digests; child node digests : Fixed-length byte strings, each representing a summary of a pair of sibling nodes; concatenation symbols : Indicates byte-level concatenation.

[0103] Taking the file view, which is a combination of fragment data region, index, and quality metadata, as input, it is split into pages and calculated page by page. Then synthesize from the bottom up. The root digest is obtained; the page size, page number, root digest, leaf digest list, and hierarchy information are written to the extended area; during range download, the returned page is compared with the corresponding... The verification can then be completed; when importing data from other sites, the corresponding data is compared first. If the results are identical, the page is determined to already exist, thus achieving page-level deduplication. In application, the fragmented verification tree performs full file verification down to the page level, ensuring local integrity in scenarios with interrupted resume downloads and providing clear criteria for cross-site deduplication; because the root digest can be derived from the leaf digests, the overall consistency of the container can also be quickly verified.

[0104] Furthermore, to ensure that who encapsulated what and who encapsulated it in the handover chain can be verified by a third party, the encapsulation list needs to be digitally signed:

[0105]

[0106] The parameters have the following meanings: signature value Fixed-length byte string used for signature verification; signature operations ECDSAP-256 is selected as the signature algorithm on the private key side; list digest. : Regarding the list The hash result; list A structured object containing a fragment index, quality metadata digest, fragment checksum root digest, and extended region version number. List The field order is fixed, and the data is CBORcanonical encoded before hashing. The public key is included with the container or issued by the directory service; if signature verification fails, the application is rejected.

[0107] Furthermore, generate a list. Then, its hash is calculated, and the signature value is calculated using the private key of the encapsulated entity. , list With signature value It is written to the extension area; at the receiving end, the signed value is verified using the publicly available signature verification key. Summary of the list Verification is performed; if successful, it confirms that the source and content have not been tampered with. In application, digital signatures enable containers to prove their origin when transferred across organizations; the coexistence of manifests and signatures ensures that structure, quality, and verification are endorsed by the same entity; verification failure can immediately prevent erroneous data from entering the analysis process.

[0108] Step 4: Driven by the fragment index and quality metadata in the container, implement event-oriented network and storage orchestration at the central side. First, retrieve high-energy fragments and necessary indexes incrementally according to priority, and complete page-level verification and breakpoint resumption according to the fragment verification structure; then, perform cross-site deduplication using leaf summaries, and write them to the cold and hot tiered storage and directory service according to priority to form a traceable inbound list; register the object identifier and level, retention period and replication policy in the directory service, keeping it consistent with the fragment index number, to facilitate subsequent layered retrieval and batch playback loading.

[0109] The single-file extended container already provides fragment offset, fragment byte length, encoding method, estimated arrival sample location, upper bound of time uncertainty, equivalent group delay, and leaf summary list. If the central side still fetches the entire file, it will lose the timeliness advantage of retrieving key fragments first. Therefore, a clear priority function is needed to solidify the following priorities into a scheduling sequence: smaller upper bound of time uncertainty and equivalent group delay, smaller fragment byte length, and items marked as high-energy fragments are retrieved first. Based on this, range requests are generated. After retrieval, each page is verified using leaf summaries, so that only missing pages are retrieved when resuming from a breakpoint.

[0110] First, the fragment index is parsed to obtain the fragment offset, fragment byte length, sampling interval, and estimated arrival sample position for each fragment. Then, the quality metadata is read to obtain the upper bound of time uncertainty and equivalent group delay. Based on this, the fragment priority is calculated and the request plan is output. A continuous page window is generated according to the request plan, a range request is sent, the leaf digest is compared page by page and the verification results are recorded. If the process is interrupted, the pages that have passed are retained and the unretrieved window is rebuilt. After all pages have passed, the fragment page list is output and handed over to the next sub-step for disk write.

[0111] Following the principles of prioritizing segments with higher interpretability and smaller values, a segment priority scoring function is constructed after removing dimensions. The scheduling order is determined by both quality boundaries and size.

[0112]

[0113] Among them: fragment priority score Real numbers, used for sorting; larger values ​​have higher priority; upper bound of time uncertainty. : Non-negative real number, from the quality metadata object in step three; equivalent group extension : Real number, from the quality metadata object in step three; fragment byte length : Unit: byte, positive integer, from fragment index; weight Non-negative real numbers, used to establish trade-offs between different dimensions; lower limit term. Positive real numbers, avoiding denominators approaching zero; high-energy fragment set : A set of indices containing the numbers of segments labeled as one million per second; indicator functions : If the condition is met, take Not satisfied with taking .

[0114] Or, as an alternative:

[0115]

[0116] Among them, the normalized time upper bound , For engineering reference, such as Normalized equivalent group extension , For engineering reference, such as Normalized scale , For engineering reference, such as ;constant To avoid the lower bound term where the denominator approaches zero.

[0117] Furthermore, based on the container extension area read... Combined with high-energy markers, calculate fragment priority scores. The segments are sorted in descending order to generate a priority queue. For the segments at the head of the queue, they are divided into page windows based on the segment offset and segment byte length, and a range request plan is output. The mapping relationship between the plan number and the segment number is recorded for breakpoint resume. In application, the priority function unifies the quality boundary and scale to the same sorting criterion, reducing human intervention; the indicator naturally prioritizes high-energy segments, ensuring that critical moments are judged first; the planned output provides a direct recovery anchor point for breakpoint resume.

[0118] To reduce the management overhead of multiple range requests, consecutive window aggregation is needed on the page list of the same segment, minimizing the request sequence while maximizing coverage. A window aggregation degree is defined to evaluate the continuity of window grouping:

[0119]

[0120] Where: window aggregation degree : Values ​​are in the range The closer The better the continuity; while maintaining Without decreasing the number of pages, prioritize merging adjacent page segments with the smallest distance; prohibit merging pages that span multiple segments; number of windows. : A positive integer representing the number of windows after merging; the starting page of the window. Termination Page : A non-negative integer, representing the first... Page range of a window; extreme value function Take the larger of the two values ​​to ensure that the denominator is not zero.

[0121] Merge adjacent pages in ascending order of page number on the fragment page list to generate a small number of consecutive windows, thus increasing window aggregation. Approximately 1; Send a range request for each window, and after receiving the page content, compare it with the leaf summary. The comparison process involves matching pages; if a match is successful, the page is marked as retrieved; otherwise, it is added to the re-retrieval queue. If the network is interrupted, the list of retrieved pages is retained, and upon recovery, requests are only made to the re-retrieval queue. In application, window aggregation reduces the number of request splits, lowering handshake costs; leaf summary bit-by-bit comparison provides page-level verification, independent of full file validation; and breakpoint resumption resumes on a page-by-page basis, avoiding repeated retrieval of already successful pages.

[0122] Multi-terminal synchronous recording generates a large number of duplicate pages. If these are not identified and skipped before being stored, it will directly increase storage costs and extend storage time. Furthermore, different segments have different requirements for subsequent interpretation. High-energy segments should be placed in a quickly accessible media layer, while other segments should be placed in an archiving layer. The layering results and subsequent retention strategies should be bound to object identifiers for retrieval and auditing.

[0123] Using the leaf summary set as the deduplication key, the summary index on the object storage side is queried page by page of the retrieved page list. If it exists, the writing is skipped; otherwise, it is written and the summary is registered. Then, based on the fragment priority score and quality boundary, the stratification determination is calculated. High-energy fragments and high-scoring fragments are sent to the hot layer, and the rest are sent to the archive layer. At the same time, a stratified entry list and retention plan are generated, and the object identifier is backfilled to the upstream.

[0124] Furthermore, the leaf summary of the extended region is written in step three. Using the key, construct a page-level deduplication check:

[0125]

[0126] Among them: deduplication determination The value can be 0 or , This indicates that the page already exists and does not need to be written; Leaf Summary : Fixed-length byte string, from the fragmented checksum tree; digest set : Set, representing the set of leaf summaries registered at the storage end; Indicator function If the condition is met, take 1; otherwise, take 0. .

[0127] For each retrieved page, a query is performed. If deduplication is determined If the record is skipped and a reference relationship is established, a soft reference is created to the page for the current segment; if Then write it to that page and add the leaf summary. Add to summary collection After writing is complete, rewrite the page using... Perform verification to ensure consistency in disk persistence. In application, page-level deduplication uses the leaf digest as the unique key, and duplicate content is uniformly referenced, saving physical resources; disk persistence consistency verification ensures that the hash values ​​of the storage layer and network layer are consistent, eliminating potential problems during transmission and disk persistence; reference relationships can be used for dependency cleanup during subsequent object migration.

[0128] To align the network-side target of prioritizing high-energy fragments and rapid rendering with the storage-side media selection, a clear hierarchical determination is needed based on fragment priority scores and quality boundaries. A hierarchical determination function is defined, using two thresholds to divide the three-level destinations:

[0129]

[0130] Among them: stratification : Values These are mapped to the archiving layer, the thermal layer, and the superheated layer, respectively; the first threshold. Second threshold : Real number, and satisfying Indicator functions If the condition is met, select 1; otherwise, select 1. .

[0131] Furthermore, a hierarchical determination is calculated for each segment. ,like Then write to the superheated layer and set a longer retention period and a higher number of replicas; if Then write to the hot layer and set a medium retention period; if If an object is found to be missing, it is written to the archive layer and a longer retention period but a lower number of copies is set. All object identifiers, levels, and retention policies are written to the hierarchical inbound list and registered in the container's central-side directory service for easy retrieval by layer.

[0132] In application, the hierarchical decision function extends the priority of the network scheduling phase to the storage medium selection, maintaining semantic consistency before and after; the dual threshold structure provides hysteresis space, reducing the frequent migration of fragments between levels; and list backfilling enables upstream to perform direct access and auditing based on object identifiers.

[0133] Step 5: At the playback end, perform time and phase compensation on multiple channels based on the group delay signature, and render the segments on a unified timeline; perform quality-weighted loading based on the upper bound of time uncertainty and equivalent group delay, calculate the reference time and align each channel, and generate an acceptance list containing segment number, reference time, deviation and half-width fields to ensure consistency in playback, interpretation and evidence collection; at the same time, only load high-energy segments for initial interpretation, and load other segments as needed. The entire process uses the index, signature and verification fields in the container to ensure consistent references at the file level and session level.

[0134] Cross-device and cross-site time alignment requires the playback end to remove frequency-dependent group delay introduced by the analog front-end and reconstruct the waveform on a unified time axis. Simultaneously, the loading order should prioritize displaying content with small upper time bounds, small equivalent group delays, moderate segment sizes, and those marked as high-energy segments, so that operators can quickly identify critical moments. Therefore, group delay compensation needs to be implemented in the frequency domain, and loading weights derived from quality boundaries to form a first-come, first-served playback sequence.

[0135] Therefore, the group delay function and the equivalent group delay of the segment are first read from the extended region of the container, and the frequency domain representation of each channel is phase-corrected according to the group delay function to obtain the compensated channel sequence. Then, the loading weight is constructed with the upper bound of time uncertainty and the equivalent group delay, and the segments are sorted and loaded in batches. The compensated multi-channel waveform is rendered on a unified time axis to provide input for the consistency criterion and acceptance process.

[0136] Furthermore, to eliminate the arrival time differences of channels at different frequencies, each channel is phase-pre-shifted in the frequency domain according to the group delay function, and then inversely transformed into a time-domain sequence on a unified time axis:

[0137]

[0138] Among them: the spectrum after frequency domain compensation : for the first Frequency domain representation after channel compensation; original frequency domain spectrum : for the first The frequency domain representation before channel compensation is obtained by transforming the time domain sequence; imaginary unit. :satisfy ;frequency Unit Hertz, real number, operating frequency band of the coverage device; group delay function : Real function, derived from the group extension signature of step one / step three.

[0139] Furthermore, a frequency domain transformation is performed on each channel to obtain the original frequency domain spectrum. Calculate the frequency domain compensated spectrum according to the formula. The inverse transform yields the compensated sequence, which is then used with the unified time axis function from step one. Alignment to the same time base; the original sampling point index is retained for high-energy segments contained in the compensated sequence to ensure that the estimated arrival sample position is consistent with the subsequent consistency criteria. In application, frequency domain phase pre-positioning externalizes the frequency-dependent time deviation at the playback end, and the unified time axis ensures that cross-channel and cross-station splicing is unambiguous; the steep edge of the compensated sequence at the moment of arrival is closer to the actual channel arrival time, which facilitates consistency verification.

[0140] Furthermore, to prioritize the presentation of high-quality segments during playback, loading weights are mapped based on the temporal upper bound and the equivalent group delay:

[0141]

[0142] Where: Loading weight : Positive real number, determining the order of fragment loading and rendering; upper bound of time uncertainty : Non-negative real number, derived from mass metadata; equivalent group extension : Real number, from quality metadata; weighting factor : A non-negative real number used to adjust the strength of the effect of group delay on weights; scaling factor : Positive real numbers that ensure the weights fall within a range that facilitates sorting; lower bound constants : Positive real numbers, to avoid the denominator approaching zero.

[0143] Furthermore, read each segment and Calculate the loading weight The sequences are then sorted in descending order to generate playback sequences; the first batch of sequences contains only high-energy fragments and loaded weights. Larger, medium-speed segments allow for pre-processing; during rendering, weights and quality indicators are labeled for each channel to facilitate handover. In application, the weighting function translates the time and front-end quality boundaries into a clear loading order, reducing manual selection; numerical monotonicity ensures that higher quality segments are loaded earlier, concentrating playback resources on key segments. Measurable effects and measurement conditions: Recording the weight-loading order curve and loading batch logs verifies the monotonicity and boundary behavior of the weight calculation; parameters... Record it in the acceptance checklist.

[0144] Multi-terminal event interpretation requires a verifiable arrival reference time, and acceptance is based on the deviation of each channel at that reference time and its own confidence band. Traditional methods describing events using mean and variance are not robust to spike-type data and are difficult to directly connect to the time upper bound and group delay signature. Therefore, it is necessary to obtain a robust arrival reference time for outliers by minimizing the weighted absolute deviation, and then construct a confidence band using the half-width synthesized from the upper bound to generate a deliverable list.

[0145] For each channel, the arrival sample position and unified time axis time of the compensated sequence are extracted. A robust objective is constructed by combining the arrival time error and the loading weight, and the reference time is obtained by solving. Then, based on the upper bound of time and the equivalent group extension, the confidence band width is synthesized, and the judgment and audit field links for whether each channel falls into the confidence band are given. The playback report and acceptance list are output.

[0146] To obtain a robust reference time, a weighted average method is used. The problem of finding the extreme value of an objective:

[0147]

[0148] Among them: reference time : Real number, representing a consistent time reference across multiple endpoints; unified timeline moment. : Real number, the value of which is obtained by the unified time axis function in step one at the estimated location of the sample position;

[0149] Arrival at Sample Location Estimation : Non-negative integer, estimated from the records in steps two and three; arrival time error : Real number, from the audit record in step two.

[0150] Furthermore, leveraging the properties of the piecewise convex structure, a weighted median iteration or semi-convergent method is used to solve the problem, minimizing the objective function in one dimension. After solving, the reference time is... The network reference time is written into the report, and the deviation of each channel is calculated for subsequent confidence band determination.

[0151] When applying, weighting The solution corresponding to the objective is a weighted median, which is less affected by abnormal channels; the weights are consistent with the previous step, forming a consistent standard across steps; the output reference time can be directly recalculated and verified.

[0152] Furthermore, to transform the quality boundary into a visible confidence band, the half-width of each channel is synthesized and a band drop determination is performed:

[0153]

[0154] Where: confidence band half width : A non-negative real number, representing the first... Allowable offset of the channel around the reference time; upper bound of time uncertainty. : Non-negative real number, derived from mass metadata; equivalent group extension Real number, derived from mass metadata; composite coefficient : Non-negative real numbers, adjusting the contribution of the group extension term to the half-width; residual upper bound : A non-negative real number representing the safety margin of the playback link residual error. It is the upper bound of the sum of the transformation reconstruction quantization error, numerical rounding error, and player rendering sampling error. In engineering, it can be set as a fixed constant plus a term inversely proportional to the window length.

[0155] Calculating the confidence band half-width Then, the deviation of each channel and The comparison process involves labeling and marking areas of failure / exit. The acceptance checklist lists each channel identifier, reference time, deviation value, half-width, area failure determination, and audit field link. The playback report provides the reference time source, weight configuration, and convergence log. In application, the confidence band visually synthesizes the time upper bound, equivalent group delay, and residual safety edge, avoiding abstract statistics. The determination criteria directly correspond to quality metadata, facilitating auditing and review. After providing a robust reference time, the quality upper bound is synthesized into an intuitive confidence band and determination list. These two elements unify the calculation conclusions and quality boundaries at the presentation layer, forming a deliverable and auditable result.

[0156] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0157] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0158] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0159] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0160] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A high-precision recording wave acquisition and time scale synchronization system for power grid fault location, characterized in that: Comprising, Synchronously acquire 1 million samples per second, 100,000 samples per second, and 10,000 samples per second three grades of sampling under the same coordinated universal time axis, generate a time uncertainty upper bound representing the worst error of the timestamp of each segment, and bind the time uncertainty upper bound as segment-level quality metadata with the segment; at the same time, for each channel included in the segment, quote the channel group delay signature obtained based on the analog front-end sweep frequency calibration and in-service calibration, and record the group delay signature of each channel and its version number together with the segment; Only perform near-lossless compression on high-energy segments of 1 million samples per second, and after encoding, calculate the arrival sample position and peak amplitude with the homologous detection kernel, if the threshold is exceeded, the segment is rolled back to lossless, and the threshold is linked with the time upper bound; wherein the homologous detection kernel is a convolution detection kernel homologous to the arrival time detection algorithm used in the device traveling wave protection, which is used to convolve the calibration sequence to obtain the response sequence; the arrival sample position is the sample index where the response sequence reaches the maximum value, and the peak amplitude is the instantaneous amplitude of the waveform corresponding to the arrival sample position; Write the segment index, time uncertainty upper bound, and group delay signature quality metadata in a compatible single file extension container, and write the fragmented check structure and the whole file signature, so that the segment corresponds to the metadata one by one; The center side preferentially retrieves high-energy segments according to the segment index and performs calibration and breakpoint continuation according to the fragmented check structure, cross-site deduplication based on digest matching, and priority allocation to cold and hot tiered storage; The playback end reads the group delay signature to compensate the time and phase of the channel, loads the quality according to the time uncertainty upper bound, and renders the segment and the arrival sample position estimation on a unified time axis.

2. The high-precision recording wave acquisition and time scale synchronization system according to claim 1, wherein: A unified time axis is constructed with whole seconds and nanosecond counts, and a time uncertainty upper bound is generated for each segment according to the combined aperture of the timing residual error, asymmetric error, drift bound, and marker quantization error; And by calibrating the sweep frequency signal through the analog front-end in the commissioning phase, the phase-frequency relationship between the output and the input is measured, the derived group delay function is solidified as the channel group delay signature, and in the running phase, the key frequency points of the group delay function are calibrated and updated in a short-time self-calibration manner to form a group delay signature with a version number, which is recorded together with the segment.

3. The high-precision recording wave acquisition and time scale synchronization system according to claim 1, wherein: For high-energy segments, perform channel-independent predictive-residual and entropy encoding processing, and determine the near-lossless threshold based on the equivalent group delay derived from the group delay signature and the time uncertainty upper bound, and after encoding, extract the arrival sample position and peak amplitude with the homologous detection kernel for calibration.

4. The high-precision recording wave acquisition and time scale synchronization system according to claim 3, wherein: When the arrival sample position or peak amplitude does not meet the preset criteria, the corresponding segment is rolled back to lossless or the abnormal short window is rolled back to lossless, and the audit record of passing or rolling back is written together with the segment, including encoding category, threshold, arrival sample position, and error fields.

5. The high-precision recording acquisition and time scale synchronization system of claim 1, wherein: The segment index and quality metadata are written in the extension area of the single file extension container, and the segment index at least contains start and end time, sampling rate, sample number, file offset, encoding mode and digest value, and the quality metadata at least contains time uncertainty upper bound, group delay signature, arrival sample position and audit record entry.

6. The high-precision recording acquisition and time scale synchronization system of claim 5, wherein: The sharding verification structure is constructed, and the leaf digest list and hierarchical information are written, the encapsulation manifest containing segment index digest, quality metadata digest and root digest is generated, the whole file signature is generated based on the encapsulation manifest, and the encapsulation manifest and the whole file signature are written in the extension area and correspond to the segments one by one.

7. The high-precision recording acquisition and time scale synchronization system of claim 6, wherein: The segment priority is calculated according to the segment index and quality metadata on the center side, the range retrieval plan is generated for high-energy segments and segments with higher priority, the request is initiated for the aggregated continuous window according to the file offset and length, and each page is checked and registered according to the sharding verification structure on the receiving end.

8. The high-precision recording acquisition and time scale synchronization system of claim 7, wherein: In the breakpoint continuation scenario, only the pages that do not pass the check are retrieved; in the cross-site storage, the target storage is queried with the leaf digest as the key, the reference relationship is established and the writing is skipped when it exists, and the corresponding leaf digest is added to the digest set to support subsequent deduplication when it does not exist.

9. The high-precision recording acquisition and time scale synchronization system of claim 8, wherein: According to the segment priority score and the marking attribute, the segments are written into the super-hot layer, the hot layer or the archive layer, and the object identifier, the belonging hierarchy and the retention policy are registered in the directory service, so that the playback end can perform retrieval and loading according to the object identifier and the hierarchy and maintain consistent reference relationship with the range retrieval.

10. The high-precision recording acquisition and time scale synchronization system of claim 9, wherein: The playback end reads the group delay signature to perform frequency domain phase pre-compensation or equivalent fractional delay filter compensation on each channel, and reconstructs the multi-channel segment sequence on the unified time axis while maintaining consistent reference relationship of the arrival sample position and the segment index in the rendering process.

11. The high-precision recording acquisition and time scale synchronization system of claim 10, wherein: The loading weight is generated according to the time uncertainty upper bound and the equivalent group delay, the segments are loaded in order according to the weight, and the time uncertainty upper bound and the group delay signature version number of each channel are marked on the playback interface, and the loading order is consistent with the reference and numbering correspondence recorded in the foregoing hierarchical record.

12. The high-precision recording acquisition and time scale synchronization system of claim 10, wherein: The weighted absolute deviation target is constructed based on the arrival sample position and the arrival time error of each channel, the reference time is obtained, and each channel is aligned on the unified time axis, and then the acceptance manifest containing the segment number, the reference time, the deviation and the half-width field is generated and archived according to the object identifier.

13. The high-precision recording wave acquisition and time scale synchronization system of claim 5, characterized in that: The extended area is placed at the end of the information section and serialized with normalized encoding, a magic number field and a version number and a check value are set for identification, so that the old version player that does not identify the extended area can ignore the extended area and read the basic section, and the field is marked with Chinese name and unit.

Images