A multi-channel energy data acquisition method and device for power dispatch data network

Abstract

The application discloses a kind of power dispatching data network multi-channel energy data acquisition method and device, first according to the topology associated relationship of each channel belonging to it is allocated staggered sampling time, avoid mutual interference caused by adjacent channel simultaneous acquisition, obtain original energy data;Again, according to the operating margin deviation law, the risk level of equipment is evaluated and the channel that needs to be focused on is filtered to form a priority sequence;Further, the sending rhythm and real-time electrical quantity direction on the channel are tracked, and whether it is consistent with the preset connection relationship is verified, from which the connection degradation and numerical freezing channel are identified, and according to the degree of independence from the main channel physical path, the standby link is switched to the optimal one;Finally, according to the risk level of channel, the allowable range of abnormal judgment and the allocation ratio of collection resources are dynamically adjusted, the fusion data stream considering the collection accuracy and resource utilization efficiency is formed, and the anti-interference ability of multi-channel energy data collection and the timeliness of abnormal channel disposal are improved.

Description

Technical Field

[0001] This invention relates to the field of power dispatching technology, and in particular to a method and apparatus for acquiring multi-channel energy data in a power dispatching data network. Background Technology

[0002] In the power dispatch data network, there are a large number of data acquisition terminals. Multiple channels often share the same communication link or the same data acquisition platform processing resources. If the sampling time is concentrated at the same moment, it can easily cause link congestion or mutual interference. Information such as line connection direction and equipment importance is usually only registered in the ledger when the equipment is put into operation and is rarely checked afterward. However, the equipment itself may show signs of deterioration such as loose connection and stagnant metering output as it has been in operation for many years, which is difficult to detect in time.

[0003] Existing multi-channel acquisition schemes often schedule sampling according to a fixed cycle without considering mutual interference between channels. Information such as connection direction is often directly based on ledgers and lacks independent verification based on actual operating data. When anomalies occur, it is difficult to distinguish between channel-specific problems and common link problems. When activating backup channels, their physical independence from the main channel is rarely examined, resulting in inaccurate allocation of acquisition resources and delays in anomaly handling. Summary of the Invention

[0004] This invention discloses a method and apparatus for multi-channel energy data acquisition in a power dispatching data network. The method aims to stagger the sampling times of each channel according to the topology to reduce mutual interference, identify the risk level of the equipment and abnormal connection direction based on the operating data, preferentially switch the deteriorated and stagnant channels to physically independent backup links, and dynamically adjust the judgment threshold and acquisition resource allocation according to the risk level.

[0005] The first aspect of this invention proposes a method for acquiring multi-channel energy data in a power dispatching data network, comprising the following steps:

[0006] Obtain the channel configuration parameters of the acquisition terminal, and perform interleaved acquisition based on the topological correlation phase difference according to the channel configuration parameters to generate raw energy data;

[0007] The original energy data is used to locate the margin deviation interval to determine the equipment criticality level. Low load records below a preset proportion are selected for the equipment criticality level to obtain a set of hidden danger channels. Based on the set of hidden danger channels, a channel priority sequence is established by weighting frequency and duration.

[0008] Based on the channel priority sequence, track the channel uplink records to obtain the abnormal uplink interval duration. Verify the abnormal uplink interval duration against the window change to determine the associated response status. From the associated response status, verify the consistency of the sending and receiving end channel directions with deviation tolerance and output the true connection status.

[0009] Based on the actual connection status, the degraded channels and numerically frozen channels are identified as abnormal channel groups. For the abnormal channel groups, the primary and backup switching is performed with maximum heterogeneity to obtain switching records. The connectivity and homogeneity are verified through the switching records to generate channel health records.

[0010] Based on the channel health records, the criticality of abnormal channels is graded and corrected to form a channel risk classification. Based on the channel risk classification, a window with abnormally stable margin and fluctuating load is determined to obtain a dynamic dead zone threshold. The bandwidth weights of the numerically frozen channels are allocated inversely proportional to the variance using the dynamic dead zone threshold to generate a fused data stream.

[0011] A second aspect of this invention provides a multi-channel energy data acquisition device for a power dispatch data network, comprising:

[0012] The data acquisition unit is used to acquire the channel configuration parameters of the acquisition terminal, and to perform interleaved acquisition based on the topological correlation phase difference according to the channel configuration parameters to generate raw energy data;

[0013] The hidden danger screening unit is used to determine the criticality level of equipment by locating the margin deviation range of the original energy data, to filter low load records below a preset proportion for the criticality level of the equipment to obtain a set of hidden danger channels, and to establish a channel priority sequence based on the set of hidden danger channels by frequency and duration.

[0014] The connection diagnostic unit is used to track the channel uplink records according to the channel priority sequence to obtain the abnormal uplink interval duration, verify the abnormal uplink interval duration with the same window change to determine the associated response status, and verify the consistency of the sending and receiving end channel direction from the associated response status with deviation tolerance to output the true connection status.

[0015] The redundant switching unit is used to locate the degraded channels and numerically frozen channels as abnormal channel groups based on the actual connection status, and to perform primary and backup switching for the abnormal channel groups with maximum heterogeneity to obtain switching records. The switching records are used to verify connectivity and homogeneity to generate channel health records.

[0016] The risk allocation unit is used to correct the criticality of abnormal channels based on the channel health records to form a channel risk classification, determine a window with abnormally stable margin and fluctuating load based on the channel risk classification to obtain a dynamic dead zone threshold, and allocate bandwidth weights to the numerically frozen channels according to the inverse variance ratio using the dynamic dead zone threshold to generate a fused data stream.

[0017] The beneficial effects of this invention are reflected in the following points: First, the sampling time of each channel is not uniformly set at the beginning of the cycle or staggered based on experience. Instead, the sampling cycle is evenly divided according to the number of channels in its topology association group. Each channel occupies a segment of non-overlapping phase, and the timing of adjacent channels triggering sampling is naturally dispersed and no longer concentrated at the same instant. Whether a device needs special attention is not simply based on its type or installation location as preset, but is evaluated based on the actual number and magnitude of deviations in the operational margin. The evaluation results, combined with the frequency and duration of occurrence, form a unified priority ranking, so that monitoring resources are prioritized for channels that truly show signs of risk, rather than inspecting them one by one according to a fixed list. Second, the verification of channel connection status is split into two independent layers of evidence: one layer is whether the anomalies of multiple channels on the same physical link occur synchronously, used to distinguish between common link problems and individual problems of a single channel; the other layer is to directly compare the direction of real-time electrical quantities with the pre-registered design direction of the line. The comparison tolerance itself will be dynamically widened or tightened according to the dispersion of the historical measurement results of the line, rather than uniformly applying the same fixed tolerance to the entire line. The two layers of evidence are independent and recorded separately. The final determination of the connection status is based on the joint support of both, and no conclusion is drawn directly from the result of a single layer. Finally, the selection of the backup channel does not rely on the superficial judgment of "whether a backup exists", but rather quantifies the actual overlap between the primary and backup channels on the acquisition terminal and transmission path, and combines this with whether there is a temporal correlation between their historical failures, comprehensively assesses the degree of physical independence, and then selects the best option for switching. The risk level of the equipment will be dynamically adjusted up or down according to its historical abnormal performance. The higher the risk of the channel, the tighter the tolerance range for abnormality judgment and the more detection resources are allocated accordingly. This ensures that the monitoring investment is adjusted in real time according to the actual risk status of the equipment, rather than being evenly distributed or configured according to a fixed ratio. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating a multi-channel energy data acquisition method for a power dispatch data network according to the present invention.

[0019] Figure 2 This is a schematic diagram comparing the overlap of heterogeneous paths in the primary and backup channels of this invention.

[0020] Figure 3 This is a schematic diagram of the power flow direction verification at the sending and receiving ends of the present invention.

[0021] Figure 4 This is a structural block diagram of a multi-channel energy data acquisition device for a power dispatching data network according to the present invention.

[0022] Wherein: 1-Main channel acquisition terminal; 2-Main channel transmission path; 3-Backup channel A transmission path; 4-Path overlap section; 5-Backup channel B acquisition terminal; 6-Backup channel B transmission path; 7-Dispatch center; 8-Optimal switching target marker; 9-Sending substation; 10-Receiving substation; 11-Transmission line; 12-Calibrated power flow direction; 13-Measured power flow direction; 14-Deviation angle; 15-Deviation tolerance range. Detailed Implementation

[0023] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0024] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0025] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0026] The technical solutions of the embodiments of this application will be described below.

[0027] like Figure 1 As shown, this embodiment of the invention provides a method for acquiring multi-channel energy data in a power dispatching data network, including the following steps S101-S105:

[0028] Step S101: Obtain the channel configuration parameters of the acquisition terminal, and perform interleaved acquisition based on the topological correlation phase difference according to the channel configuration parameters to generate raw energy data.

[0029] Specifically, the channel configuration parameters of the acquisition terminal are obtained. The acquisition terminal periodically sends registration messages containing the number of channels, topology association group number, and measurement point device identifier via fieldbus. The messages use fixed-length field encoding according to the terminal's factory protocol, and each field is arranged sequentially with a fixed byte width. The acquisition management platform reads the data one by one according to the field offset address and generates an initial configuration record, without relying on separator parsing to avoid field misalignment. Message verification uses a cyclic redundancy check code, which covers all message content except for the check field itself. Messages that fail verification are discarded as a whole without any salvage parsing of partial fields. After being discarded, they wait for the acquisition terminal to re-send in the next cycle. Terminals that fail verification for three consecutive cycles are marked as having communication abnormalities and their sampling tasks are suspended. For channels where the topology association group number field in the channel configuration parameters is empty, the acquisition management platform sends a supplementary recording instruction to the corresponding acquisition terminal. The instruction carries the physical slot number of the channel in the terminal chassis. After receiving the instruction, the terminal fills in the field in its local configuration table and re-sends in the next cycle. The interval between the issuance of supplementary recording instructions is not less than one complete transmission cycle of the terminal. If two channels under the same acquisition terminal send identical measurement point device identifiers, it is considered duplicate wiring. The acquisition management platform will suspend the sampling task for the channel with the larger channel number until the measurement point identifier is verified and corrected on-site, after which the suspension will be manually lifted. Each time the channel configuration parameters are updated, the number of message transmission cycles between the current update and the previous update is also recorded. Updates with an interval shorter than the minimum configuration validity period for this terminal model will not take effect temporarily, and the sampling task will continue to be executed using the channel configuration parameters before the update.

[0030] In some embodiments, the step of generating raw energy data by performing interleaved acquisition based on the channel configuration parameters and the topological associated phase difference includes: determining a phase offset by evenly distributing a preset period according to the number of channels within the topological associated group based on the channel configuration parameters; generating a comb-shaped sampling sequence by allocating interleaved sampling times for the channels based on the phase offset; determining the interleaving effectiveness by calculating the overlap ratio of sampling times of adjacent channels based on the comb-shaped sampling sequence; and reallocating the phase offset for channels whose interleaving effectiveness is lower than a preset threshold and outputting raw energy data.

[0031] The phase offset is determined based on the channel configuration parameters, with the number of channels within a topology association group evenly distributed across a preset period. The acquisition management platform reads the channel list under the same topology association group from the channel configuration parameters, sorts the channels in the list in ascending order of their numerical identifiers, and assigns a sequential number starting from zero to each channel. This number, along with the number of the associated topology association group, is written to a temporary allocation table instead of being directly written back to the channel configuration parameters themselves, thus avoiding intermediate states from contaminating the final configuration during the allocation process. The phase offset is calculated using the formula φ_k = k × (T / N), where T is the preset sampling period in milliseconds, N is the total number of channels in the topology association group, k is the sequential number of the channels within the group, and φ_k is the offset of the k-th channel relative to the start of the period, also in milliseconds. The calculation is performed line by line based on the temporary allocation table. The calculation results are first written to a temporary storage field in the allocation table, and only become effective after verifying that adjacent φ_k intervals are evenly distributed across the period without gaps. If a channel within a group goes offline for more than a set time due to a fault, that channel is removed from the N-value statistics of the current allocation round. The remaining channels are renumbered according to the new N-value after removal, and the phase offset is recalculated. During renumbering, the original ascending order is retained, and only the position of the offline channel is skipped. After the offline channel recovers, it is re-entered into the group according to the next round in which the recovery time occurs, without retroactively calculating the offset of previous rounds. When the total number of channels in two adjacent topology-related groups is exactly the same, each group reads its own independent channel configuration parameters and calculates the phase offset separately. The two sets of calculations use their own independent temporary allocation table instances and do not reference each other's data, avoiding cross-group channels being mistakenly identified as being in the same group, which would cause the offsets to overlap.

[0032] A comb-like sampling sequence is generated by allocating interleaved sampling times for each channel based on the phase offset. The local timer of the acquisition terminal triggers a baseline timing at the beginning of each cycle. Subsequently, delay trigger points are set one by one according to the phase offset corresponding to each channel. When the delay arrives, a sampling action is performed on that channel, and the actual trigger timestamp is recorded. The timestamp accuracy is down to the microsecond level. The trigger point setting uses a hardware comparison-matching timer interrupt instead of software polling to reduce the impact of delay jitter on the actual execution effect of the phase offset. The interrupt priority is set higher than other regular tasks on the terminal to ensure that the timer interrupt is not blocked by ordinary data transmission and reception tasks. The trigger times for each channel are staggered within the cycle, distributed like comb teeth, hence the name comb-like sampling sequence. Trigger timestamp lists are stored separately according to channel numbers. If, within a certain cycle, the trigger delay exceeds the timing accuracy limit of the terminal model due to excessive terminal load, the timestamp for that cycle is recorded according to the actual trigger time, without backfilling the theoretical value. The delay amplitude is separately marked in an additional field. Channels with zero phase offset are sampled directly at the start of each cycle without a delayed trigger. While the trigger path differs from other channels, the timestamp field format for writing the comb sampling sequence remains consistent, and its length is exactly the same as other channels, without shortening the record length due to the simplified trigger path. After the terminal timer restarts, the comb sampling sequence recorded before the restart is retained in the original storage area. The trigger timestamps after the restart are connected to a separate list, with the restart time serving as the dividing point. No numerical smoothing is applied to the two lists, and a restart identifier is marked at the connection point for device ledger verification. This identifier field occupies a fixed one byte.

[0033] The effectiveness of interleaving is determined by calculating the overlap ratio of sampling times of adjacent channels based on the comb sampling sequence. The acquisition management platform reads the actual trigger timestamps of each channel within the same topological association group in the comb sampling sequence cycle by cycle. After sorting by value, it calculates the difference between the timestamps of adjacent channels. If the difference is less than the duration of a single sampling for that terminal model, it is considered an overlap. Two samples overlapping in time constitute interference. Sorting and difference calculation are completed within a single cycle and are not compared across cycles. The sorting results are only stored in memory for use in the current cycle. The overlap ratio is calculated by dividing the number of overlapping pairs in the cycle by the total number of adjacent channel pairs. The total number of adjacent channel pairs equals the number of channels in the group minus one. After calculation cycle by cycle, the arithmetic mean of the ratios for all cycles within the current acquisition period is taken. During the averaging process, comb sampling sequences with fewer than ten cycles do not have their ratio results temporarily provided until the number of cycles is sufficient. During the period when the ratio is unavailable, the interleaving effectiveness of that group uses the judgment value from the previous batch. The interleaving validity is compared one by one with the overlap tolerance measured in each field deployment verification of this type of terminal. The verification data comes from the actual operation records of the same type of terminal in different field environments. If the average value does not exceed the tolerance, it is set as qualified; if it exceeds the tolerance, it is set as unqualified. The comparison result, along with the average value of the current round, is written into the judgment record of this round. The version number of the comb sampling sequence on which the judgment is based is also recorded in the record field. If the interleaving validity of the same topology association group is unqualified in three consecutive rounds of judgment, the comparison process of this step is suspended and transferred to the step of reallocating phase offset. During the suspension, the group uses the judgment result of the most recent round and the existing historical judgment records are not cleared due to the suspension.

[0034] For channels with interleaving validity below a preset threshold, phase offsets are reallocated to output raw energy data. For topology groups where interleaving validity is deemed insufficient, the sequential numbering of channels within the group is cyclically shifted by one position in the temporary allocation table. That is, the channel originally numbered N-1 is moved to number zero, and the remaining channels are shifted one position to the next. After the shift, the phase offset is regenerated using an equal-sharing formula, and the trigger time point of the corresponding channel in the comb sampling sequence is replaced. The shift action takes effect synchronously on all channels within the group, without shifting channels individually. After the shift, the overlap ratio is recalculated, and the interleaving validity is reassessed. If it still does not meet the threshold, the shift continues. When the number of shifts reaches the total number of channels in the group, automatic shifting terminates, and maintenance personnel conduct on-site verification of the actual wiring of the topology group to check for any abnormalities. During the verification period, the group continues sampling with the phase offset before the shift, without interrupting the continuous generation of raw energy data. Each channel reads the instantaneous energy value output by the local metering chip through the serial interface at the corresponding trigger time according to the final effective comb sampling sequence. The reading action and the trigger signal are completed within the same interrupt cycle, without delay across interrupt cycles. The reading timeout is set to twice the duration of a single sampling. After the timeout, the current reading attempt is terminated instead of being extended. The raw energy data is recorded one by one by the channel, along with the read energy value and the corresponding trigger timestamp. The moment when the reading fails, i.e., the metering chip does not respond, is recorded as a null value at the corresponding position of the channel. This is distinguished from the case of a true reading of zero by different flag bits. Null values ​​are not included in this round of statistics, but the timestamp position is retained for identification and skipping during subsequent margin calculations to avoid null values ​​being mixed into the continuous sampling sequence.

[0035] Step S102: Use the original energy data to locate the margin deviation interval to determine the equipment criticality level, select low load records below the preset proportion for the equipment criticality level to obtain the hidden danger channel set, and establish the channel priority sequence based on the hidden danger channel set by frequency and duration weighting.

[0036] Specifically, the criticality level of equipment is determined by locating the margin deviation interval using raw energy data. Each channel of the acquisition terminal corresponds to one transformer outgoing line or feeder switch in the substation. The platform reads the load value from the raw energy data for each channel, and combines it with the rated capacity registered in the equipment ledger for that outgoing line. The operating margin is calculated using the formula M=(CL) / C, where C is the rated capacity, L is the measured load, and M is the margin, which is generally between zero and one. The lower the value, the closer the outgoing line load is to full load. The raw energy data used for the calculation must be obtained from continuous sampling of the same channel, without any missing values ​​left by reading failures. When the margin value at multiple consecutive moments in the margin sequence deviates from the historical moving average of that outgoing line by more than a set range, for example, when a standby feeder that should have been operating at a low level for a long time suddenly becomes fully loaded for a long time, or when the load of a main transformer outgoing line suddenly drops sharply, the corresponding start and end time intervals are marked as a margin deviation interval. The start and end times of the interval are the two moments when the deviation begins and when the load recovers to near the mean. Equipment criticality classification is based on a comprehensive assessment of two indicators: the number of margin deviation intervals that occur within the statistical window for that outgoing line, and the maximum deviation magnitude of a single interval. For example, if a main transformer outgoing line supplying power to a hospital or data center experiences a significant margin deviation, even if the number of deviations is not high, it should be classified as a high-level line. However, ordinary residential branch lines, even if they experience a high number of deviations, are not necessarily classified as the same level. The two indicators have independent boundaries and are not weighted and combined. If multiple outgoing lines of the same transformer have inconsistent equipment criticality classifications, the one with the higher classification is used as the final equipment criticality classification result, rather than averaging, to avoid the risk of an important outgoing line being lowered by the stable performance of other branches.

[0037] In some embodiments, the step of selecting low-load records below a preset proportion based on the equipment criticality classification to obtain a set of potential hazards includes: establishing a low-load judgment benchmark according to the equipment criticality classification and a rated capacity proportion range; calculating the continuous satisfaction duration of the low-load judgment benchmark to obtain candidate potential hazard periods; calculating the margin recovery rate of adjacent normal periods from the candidate potential hazard periods to form a potential hazard confidence level; and selecting channels with a confidence level higher than a preset confidence threshold based on the potential hazard confidence level to obtain a set of potential hazard channels.

[0038] A low-load judgment benchmark is established based on the criticality classification of equipment and the rated capacity ratio range. Taking a main transformer outgoing line and a residential branch line in a substation as an example, when the criticality classifications of the two equipment are different, the platform looks up the upper and lower limits of the rated capacity ratio range corresponding to their respective levels from a pre-maintained ratio range comparison table. The comparison table is maintained by equipment type, and similar equipment uses the same table. The higher the level, the higher the upper limit of the ratio range is set. That is, the main transformer outgoing line may still be judged as low-load even if the load accounts for a higher proportion of the rated capacity, because for this type of channel, even if the load is not very low, it is already abnormally relaxed relative to its expected sustained high load characteristics. The low-load judgment benchmark consists of two values: the upper and lower limits of this ratio range. All outgoing lines under the same level share the same benchmark and are not adjusted individually due to different installation locations. The benchmark value is accurate to one percent of the rated capacity, and the decimal part exceeding this precision is rounded to the nearest integer. When a newly commissioned substation has not yet completed its classification calibration, resulting in a temporary lack of a corresponding interval for the criticality classification of a certain equipment, a proportional interval of an adjacent lower level is temporarily borrowed as a temporary benchmark. This temporary borrowing is marked in the low-load judgment benchmark. Once the formal interval is completed, it will be replaced. Both the old and new benchmarks are retained for traceability. The replacement action is triggered when the classification calibration is completed. When a branch line is re-evaluated to a higher level due to added load and the equipment criticality classification changes, the corresponding low-load judgment benchmark is simultaneously retrieved according to the new level. The judgment benchmark generated under the old level is immediately discontinued. The discontinuation time is consistent with the effective time of the level change, and the change process is simultaneously recorded in the level history of that outgoing line.

[0039] Candidate potential hazard periods are identified by statistically analyzing the duration of continuous compliance with the low-load judgment benchmark. The ratio of the measured load of a given outgoing line to its rated capacity is compared hourly with the upper and lower limits of its applicable low-load judgment benchmark. This comparison is performed incrementally with each new sampling point and is not repeated for all historical moments. If an industrial dedicated line that should be stably carrying a medium load consistently falls within the low-load judgment benchmark range, it indicates that its load has been consistently below its expected level. The number of consecutively compliant moments is converted into duration based on the sampling period, with the conversion result retained to the second level of accuracy. Candidate potential hazard periods are recorded as consecutive compliance intervals whose duration exceeds the preset minimum judgment duration. The minimum judgment duration is related to the criticality level of the equipment to which the outgoing line belongs; the higher the level, the shorter the corresponding minimum judgment duration. This means that important outgoing lines are included in the candidate list earlier once low-load signs appear, rather than waiting for long-term observation. The judgment duration is maintained in an independent configuration table according to the level classification. If, during the continuous satisfaction process, a single sampling point's proportion jumps out of the low-load judgment benchmark range and then immediately falls back, this is an example of instantaneous fluctuations caused by short-term peak electricity consumption. The cumulative duration of this sampling point is not interrupted; the calculation is still based on the overall range before and after the jump. The tolerable number of single-point jumps within a candidate time period is no more than two. If it exceeds two, the period is truncated at that point, and a new independent time period begins. Candidate potential hazard time periods are recorded according to the outgoing line channel number and start and end times. Multiple non-adjacent candidate potential hazard time periods occurring on the same channel within the statistical period are independently entered into the margin recovery calculation process, and are not merged due to proximity. Intermediate calculation data is not shared between different segments.

[0040] The confidence level of a potential hazard is determined by calculating the margin recovery rate of adjacent normal periods during the candidate hazard period. If a feeder with a consistently low load shows a low load record, for example, an outgoing line that should be supplying stable power only carries a very light load for several days, it could be due to concentrated maintenance shutdowns of downstream users, or it could be due to loose wiring in the metering circuit causing low readings. Although the two symptoms are similar, they need to be differentiated. The platform uses the end time of the candidate hazard period as a dividing line and extracts adjacent normal periods of equal length. The difference between the average margin during the normal period and the average margin during the candidate hazard period is calculated. This difference represents the margin recovery rate, expressed as a percentage of rated capacity. If the load recovers significantly after the low load ends, it indicates that the previous fluctuations were likely seasonal or planned rather than equipment problems. If the load remains low after the low load ends, it is more likely that there is a hazard in the metering or wiring. The confidence level of potential hazards is obtained by comparing the recovery rate with the percentile of the historical recovery rate distribution of the same type for that channel. The higher the recovery rate is in the historical distribution, the more it resembles a real load fluctuation, and the lower the confidence level of the hazard. Conversely, the lower the recovery rate is in the historical distribution, the less it resembles a normal fluctuation, and the higher the confidence level of the hazard. This inverse correlation is the core basis for judgment in this step and is the key difference from simply looking at the absolute value of the recovery rate. The comparison table is maintained separately according to the criticality level of the equipment to which the channel belongs. The historical distribution of different levels of outgoing lines does not share the same comparison table. The comparison table is refitted every time a certain number of new samples are accumulated, and the parameters before and after the fitting are retained for retrospective comparison. If a complete adjacent normal time period cannot be captured due to the channel being offline, the hazard confidence level of that time period is temporarily suspended from calculation and marked as pending evaluation. Incomplete data is not used for forced calculation, nor are values ​​from other time periods of the same channel used as substitutes.

[0041] Based on the screening of channels with hazard confidence scores exceeding a preset threshold, a hazard channel set is obtained. For each channel, the hazard confidence scores for all candidate hazard time periods are aggregated. The highest value is taken as the representative hazard confidence score for that channel, and compared one by one with the preset threshold. The threshold is the lower limit of the confidence score distribution in all verified cases of actual faults for that equipment model in the past. Current transformer disconnections or metering chip failures are considered such actual faults; that is, channels that have historically experienced problems generally have confidence scores no lower than this value. The threshold is reviewed every six months in conjunction with newly verified cases. The review is organized by the equipment management department, and written records are kept. The hazard channel set only includes channels whose representative hazard confidence scores exceed the threshold. If the representative value is exactly equal to the threshold, it is treated as exceeding it. Even if a channel has multiple candidate hazard time periods, it will not be included in the hazard channel set if the representative confidence score does not exceed the threshold, thus avoiding misjudging a simple low load as equipment abnormality and causing unnecessary on-site investigation costs. If all candidate hazard periods for a certain channel are marked as pending evaluation and a representative hazard confidence level cannot be obtained, that channel will not participate in this round of screening. It will be re-evaluated in the next statistical cycle after the data for adjacent normal periods are supplemented. Temporary exclusion is not equivalent to being judged as having no hazards; the two must be marked separately to avoid confusion and missed reporting. The hazard channel set is sorted by representative confidence level value for subsequent steps to verify in order of priority. The sorting results are refreshed after each round of screening, overwriting the results of the previous round, and the old sorting is no longer retained.

[0042] A channel priority sequence is established based on a frequency- and duration-weighted approach using a set of potential hazard channels. For each channel in the set, two indicators are calculated: the total number of times a candidate hazard occurs and the cumulative total duration. A channel experiencing low load only once a year for several months is analogous to a line slowly revealing a wiring hazard due to prolonged light load. This contrasts with another channel experiencing frequent, brief low loads more than ten times a day, perhaps due to a regular work schedule. The nature of the risks they reflect is different; one requires immediate investigation, while the other is mostly normal fluctuation. Therefore, both indicators are normalized to the zero-to-one range and then added together using preset weighting coefficients to obtain a priority score, rather than directly comparing the original frequency or duration. The channel priority sequence arranges all channels in the potential hazard channel set from highest to lowest score. In cases of identical scores, the channel with the longer cumulative total duration is prioritized; if the durations are also identical, the channel with the smaller channel number in the priority sequence is prioritized. These three criteria are compared sequentially without weighting, and the comparison order is fixed and not adjusted based on data conditions or the results of a particular round. The weighting coefficient is used to balance the impact of frequency and duration on the score. Its value is derived from a statistical fit of the influence of two indicators on the actual handling priority in the fault response records of this equipment model, rather than a fixed proportion set based on experience. Field experience shows that a longer duration is a better predictor of a real fault than the number of occurrences; therefore, the weight corresponding to duration is higher. The coefficient is reviewed annually based on newly added records. If a channel is manually verified as a false alarm and removed from the priority sequence, the corresponding position is simultaneously removed without retaining historical score placeholders. The removal action is recorded in the current round of false alarm statistics for equipment management department review and reference.

[0043] Step S103: Based on the channel priority sequence, track the channel uplink records to obtain the abnormal uplink interval duration, verify the abnormal uplink interval duration against the window change to determine the associated response status, and from the associated response status, verify the consistency of the sending and receiving end channel directions with deviation tolerance to output the true connection status.

[0044] Specifically, the abnormal upload interval duration is obtained by tracing the upload records of each channel according to the channel priority sequence. The order in which the priority sequence determines the order of retrieval. All recent data upload records for each channel are retrieved one by one. The record content is the actual timestamp of each data arrival at the platform side. This is compared with the fixed upload cycle specified in the channel configuration parameters. If the difference between two adjacent actual arrival timestamps exceeds a certain percentage of the set cycle, it is counted as an abnormal upload interval. The percentage is determined by referring to the upper limit of normal fluctuation obtained from previous network latency tests of this communication module model. The abnormal upload interval duration is the cumulative value of all abnormal intervals exceeding the set cycle within the statistical window of this channel. The accumulation process is performed sequentially without peak truncation. The larger the value, the more serious the deviation of the overall upload rhythm of the channel from the set cycle. The accumulation unit is the same as the upload cycle, which is seconds. The accumulation result is fixed once after each statistical window is completed and is not retrospectively modified by new data arriving after the window ends. The tracking scope only covers the top-ranked channels in the channel priority sequence. The specific coverage ratio is adjusted by on-site maintenance personnel based on the current inspection resource configuration. Channels ranked lower are not tracked in this round and will be included in the tracking scope of the next round based on the ranking changes. If the number of uploaded records in a channel's statistics window is too small, below the set number, the abnormal upload interval duration will not be calculated in this round and will be marked as insufficient sample. It will be included in the statistics after sufficient records are accumulated. During the period of insufficient sample, the channel will remain in its original ranking in the channel priority sequence. The abnormal upload interval duration is counted separately for each channel. When accumulating, the abnormality rate is calculated based on the set upload cycle of the channel. The rate is updated successively as the window scrolls, allowing for direct comparison of the deviation degree of each channel when verifying changes in the same window.

[0045] Verify the abnormal transmission interval duration by checking the changes within the same window to determine the associated response status. Once a statistical window showing a significantly higher abnormal transmission interval duration for a certain channel is confirmed, use this window as a baseline to identify other channels belonging to the same acquisition terminal or the same field communication link. Verify whether the abnormal transmission interval duration of these channels increases synchronously within the same window. The scope of the identification is based on the physical link affiliation recorded in the field wiring diagram, not on temporary allocations based on topology association group numbers. If the two affiliation methods are inconsistent in individual field settings, the wiring diagram shall prevail. If multiple channels on the same link show a synchronous increase in abnormal transmission interval duration within the same window, it is determined that the abnormality within the window is a common fluctuation at the link level rather than an individual problem of a single channel. If only the channel is abnormal while the other channels on the same link remain normal, it is determined to be an individual abnormality of that channel. Individual abnormalities are more likely to point to problems with the channel's own equipment or wiring rather than common factors on the network side. The associated response status record records the judgment result, with values ​​categorized into two types: link-level commonality and individual anomaly. Each type of judgment is associated with the start and end times of a specific window and a list of involved channel numbers. Link-level commonality records additionally carry the total number of involved channels; this total number of channels field is used to verify whether the link scope itself is fully registered. If the same channel is judged as link-level commonality and individual anomaly in different statistical windows, the two associated response status judgments are recorded independently, and all historical judgments are retained for traceability. For windows with a link-level commonality associated response status, the sending and receiving end direction verification directly reuses the result of any channel that has already completed verification within that link, without repeating the direction calculation channel by channel. Windows with individual anomalies still require separate and complete verification.

[0046] In some embodiments, the step of verifying the consistency of the sending and receiving channel directions from the associated response state with deviation tolerance and outputting the true connection state includes: using the associated response state to identify preset power flow direction labels between the sending and receiving channels to obtain a calibration direction relationship; measuring the deviation angle value of the real-time direction based on the calibration direction relationship; verifying whether the deviation angle value falls within a preset deviation tolerance range to obtain a direction health mark; and using the direction health mark to statistically analyze the consistency ratio of the sending and receiving channel groups and output the true connection state.

[0047] The calibration direction relationship is obtained by using the associated response status to identify preset power flow direction labels between the sending and receiving channels. For example... Figure 3As shown, for channels whose associated response status is determined to be individual anomalies, along with representative channels sampled from the link-level commonality, the power flow direction tags pre-registered at both ends of the corresponding line are retrieved from the equipment ledger. These tags indicate whether the current design flow direction is from the sending-end substation 9 to the receiving-end substation 10 or vice versa, i.e., the power flow direction 12. These tags are registered by the construction party when the line is put into operation and do not automatically change due to adjustments in operating mode. If modifications are necessary, on-site verification and re-registration are required for them to take effect. The registration record also retains the old tags before modification for comparison. The calibrated direction relationship record records the roles of the two channels at the sending and receiving ends of the line (sending-end substation 9 or receiving-end substation 10) and the calibrated power flow direction 12. The two channels at the sending and receiving ends of the same line share the same calibrated direction relationship record, rather than creating separate records for each end, to avoid contradictory direction definitions at both ends. The determination of the roles at both ends is based on the initial construction registration results. For lines where the current direction label is missing from the equipment log, the directional relationship is temporarily treated as undetermined. The entire directional verification process for this line is skipped; only the routine upload interval and associated response status verification results are retained. A log entry reminder is generated simultaneously for missing labels, and the corresponding responsible work team is notified according to the associated response status type. The reminder record is retained until the missing label is completed. When the directional relationship is re-registered due to label modification, the deviation angle values ​​previously measured based on the old directional relationship are not retrospectively corrected. Only new measurements from the time the update takes effect use the new directional relationship. The directional change record for this line is simultaneously recorded at the time of the transition between old and new directional relationships to avoid unexplained jumps in the deviation angle value sequence caused by the mixing of old and new calibers.

[0048] The deviation angle of the real-time direction is determined based on the calibration direction relationship. After the roles of the two ends of the line are determined by the calibration direction relationship, for example, if the current of a line should flow from the sending end to the receiving end in the design, but the actual power flow is reversed for a long time, it often means that the wiring at both ends is reversed or the operation mode has been changed and the ledger is not synchronized. Read the voltage and current phasor data of the sending and receiving ends of the transmission line 11 at the same time as synchronously measured by the acquisition terminal, and calculate the phase angle θ_m corresponding to the measured power flow direction 13 according to the phasor method. The phase angle is based on the zero degree reference of the sending end pointing to the receiving end. The sampling time of the phasor data shares the same clock reference with the regular sampling time of the channel to which the line belongs. The measurement accuracy is limited by the rated resolution of the phasor measurement unit of this model. The resolution value is also recorded in the metadata of this measurement. The deviation angle 14 is calculated using the formula Δθ = θ_m - θ_0, where θ_m is the phase angle measured in real time, θ_0 is the design direction angle recorded in the calibration direction relationship, and Δθ is the deviation angle 14 measured in this instance, in degrees. Δθ close to zero indicates that the real-time direction is consistent with the calibration direction, while close to 180 degrees indicates that the directions are opposite. When Δθ is between these two values, it is temporarily recorded as the angle itself without qualitative directional classification. If the phasor data acquisition time differs significantly from the center time of the current statistical window for that line, the deviation angle value is recalculated using the phasor data closest to the center time within the window. The calibration direction relationship upon which θ_0 is based remains unchanged and is not adjusted with the recalculation. The reason for the recalculation is simultaneously recorded in the remarks field of this measurement. The deviation angle value is recorded sequentially according to the line number. Multiple measurement results for the same line within the statistical period are retained separately without time averaging. The original sequence serves as the input basis for aggregation analysis, which can be directly retrieved when establishing the angle distribution interval.

[0049] For example, the step of verifying whether the deviation angle value falls within a preset deviation tolerance range to obtain a directional health mark includes: performing numerical aggregation on the deviation angle value to establish an angle distribution range; screening samples whose deviation angle from the center exceeds a preset multiple of the standard deviation from the angle distribution range as abnormal angle points; calculating the proportion of occurrence of the abnormal angle points within a preset period to obtain an interval correction coefficient; and using the interval correction coefficient to adjust the preset deviation tolerance range to obtain a directional health mark.

[0050] An angle distribution interval is established by numerically aggregating the deviation angle values. This interval covers the complete angle range from 0 degrees to 360 degrees, and is divided into several equally wide intervals with a fixed width. The interval width is an integer multiple of the angle measurement accuracy of the metering equipment. Within the statistical period for this line, all deviation angle values ​​are assigned to their corresponding intervals based on their numerical values. The number of measurements falling into each interval is accumulated. The assignment operation uniquely identifies the interval based on the numerical value, preventing duplicate counting across intervals. After assignment, the original deviation angle value sequence is retained separately and not deleted due to interval assignment. These point-by-point records are retained for recalculating interval divisions when needed. The total number of intervals covers the complete angle range. Even if some intervals have zero accumulated measurements, they are not omitted, and adjacent intervals are not merged due to empty intervals. Empty intervals also occupy a fixed storage location, facilitating observation of whether the angle distribution interval pattern of the line gradually concentrates towards a certain angle over operating time. When the number of deviation angle measurements for the same route is less than the minimum set sample size, the angle distribution interval will not be established temporarily. The angle verification for this route will be postponed until the sample size is sufficient. Insufficient samples will prevent the interval from being statistically insignificant and weakening the reliability of abnormal angle point screening. When the model of the metering equipment for this route is changed, and the angle measurement accuracy changes accordingly, the interval width of the angle distribution interval will be redefined according to the new accuracy. Previously defined intervals based on the old accuracy will not be merged or split for adaptation. Instead, a new set of intervals will be created and statistically analyzed independently at the time of the change. Subsequent calculations of the mean and dispersion based on the intervals will only use data from the new interval after the change and will not mix data across different equipment models.

[0051] Anomaly angle points are identified by screening samples whose deviations from the center angle exceed a preset multiple of the standard deviation within the angle distribution interval. The number of measurements in each interval is used to calculate the mean μ and standard deviation σ of all deviation angle values ​​for the line. The mean is used as the center angle, and the standard deviation reflects the dispersion of the angle measurement results. Both are calculated by weighting the values ​​at the midpoints of the intervals, avoiding direct calculations using raw point-by-point data to save storage costs. The accuracy loss of the weighted calculation is controlled within one-tenth of the angle measurement accuracy; if it exceeds this range, point-by-point calculation is reverted. Anomaly angle points are screened according to the criterion |Δθ_i-μ|>κ·σ, where Δθ_i is the result of the i-th deviation angle measurement, and κ is a preset multiple. Measurement results that satisfy this inequality are recorded as an anomaly angle point. The κ value is determined with reference to the statistical upper limit of the angle measurement noise level of this type of metering equipment; the higher the noise level, the larger the κ value. The κ values ​​for different equipment models are maintained separately and do not share the same value. If the center angle μ within the angular distribution range is close to 180 degrees, it indicates that the actual direction of the line is mostly opposite to the calibrated direction. In this case, the screening of abnormal angle points is changed to recalculate the deviation using 180 degrees as a reference, avoiding the screening of a large number of false anomalies based on incorrect center angles and masking the few truly noteworthy deviations. Abnormal angle points are recorded according to the line number and the corresponding original measurement time. Multiple abnormal angle points on the same line are retained independently for counting their occurrence rate point by point, and the counting process is carried out in chronological order of measurement time.

[0052] The interval correction coefficient is obtained by statistically analyzing the proportion of abnormal angle points occurring within a preset period. The proportion of abnormal angle points within a preset period relative to the total number of deviation angle measurements in the same period is first calculated. A higher proportion indicates greater fluctuation in the angle measurement results of the line, similar to a measurement loop experiencing severe interference and fluctuating direction readings. In such cases, applying a fixed tolerance range would cause even normal fluctuations to be interpreted as directional anomalies, leading to misjudgments. The length of the statistical period is related to the criticality level of the equipment on the line; higher-level equipment has shorter periods for faster problem detection, while lower-level equipment has longer periods to reduce unnecessary frequent calculations. The period length is maintained in the configuration table according to each level, and the configuration table is reviewed annually based on operational experience. The interval correction coefficient is the adjustment multiple obtained by converting the occurrence proportion using a preset conversion relationship. A higher proportion corresponds to a larger correction coefficient. The conversion relationship is derived from statistical fitting of past field angle fluctuation cases for this model of metering equipment, without using a uniform linear ratio. The fitting relationship is reviewed annually with new cases, and the relationships before and after the review are retained for comparison. When the proportion of abnormal angle points is lower than the set lower limit, the interval correction coefficient takes the default value of 1. That is, the preset deviation tolerance range is not widened or tightened in any way. The default value of 1 is also included in the record of interval correction coefficients in each cycle and is not omitted from the cycle entry due to the low proportion. The interval correction coefficient is calculated cycle by cycle according to the line number. Each cycle is calculated based on the actual measured proportion of the current cycle. The jump amplitude of the proportion between adjacent cycles is also recorded. If the jump is too large, it indicates that the direction measurement of the line has been unstable recently and is for manual key investigation.

[0053] The directional health label is obtained by adjusting the preset deviation tolerance range using an interval correction coefficient. The upper and lower limits of the deviation tolerance range 15 are multiplied by the interval correction coefficient to obtain the adjusted deviation tolerance range 15 actually applicable to the line in this cycle. A correction coefficient greater than 1 indicates a wider range, less than 1 indicates a tighter range, and equal to 1 indicates the range remains unchanged. This adjustment is performed once at the beginning of each statistical cycle and remains unchanged within the cycle. The directional health label is determined based on whether the current deviation angle value of the line falls within the adjusted deviation tolerance range 15. Values ​​within the range are marked as healthy, while values ​​outside the range are marked as unhealthy. The specific extent of the deviation is recorded in an additional field of this label for manual review, providing a reference for the degree of deviation rather than simply considering the binary result of health or unhealthiness. The unit of the excess extent is consistent with the deviation angle value, in degrees. If the interval correction coefficient is greater than the preset upper limit for three consecutive cycles, it is determined that there may be an equipment-level problem with the angle measurement of the line. This continuously excessive interval correction coefficient is also recorded in the equipment file for verification during equipment inspection. After the inspection conclusion confirms that there is no abnormality in the equipment, the tolerance interval benchmark value of the line can be manually adjusted upwards. The directional health mark is uniformly marked as pending review until then. After the tolerance interval benchmark value is manually adjusted upwards, the consecutive over-limit count of the interval correction coefficient in the next statistical cycle will start from zero again. The historical counts that caused the three consecutive cycles to be excessive will not be cleared but will not participate in the new round of continuous judgment. This avoids the misjudgment of equipment problems triggered again by using the old count immediately after the benchmark is adjusted. The directional health mark will also be removed from the pending review status and will resume normal judgment based on whether it falls within the interval.

[0054] The true connection status is output by statistically analyzing the consistency ratio of the sending and receiving end channel groups using directional health tags. Within the statistical period, all directional health tags for the sending and receiving end channel groups are first aggregated by group, and the proportion of tags marked as healthy is calculated out of all valid tags. Tags awaiting verification are not included in the denominator; they are only counted as a separate pending verification count for maintenance reference and are not mixed with healthy or unhealthy directional health tags in the statistics. This independent count is also stored by channel group number for easy individual verification of whether the pending verification ratio is too high. The true connection status is obtained by comparing this consistency ratio with two preset high and low limits. A ratio higher than the high limit indicates a normal connection; a ratio lower than the low limit indicates that the actual connection direction of the sending and receiving ends is reversed; and a ratio between the two limits indicates an unstable connection status requiring manual on-site verification. The specific values ​​of the two limits are set separately according to the line voltage level, and different levels do not share the same pair of limits. The limit values ​​are adjusted annually based on newly added verification cases. If the proportion of "pending verification" tags in the directional health tagging exceeds the set proportion of all tagging times, the actual connection status, regardless of the consistency ratio, will be uniformly judged as "pending verification." Priority will be given to equipment-level maintenance rather than drawing direct conclusions, to avoid giving misleading connection conclusions when the equipment's status is unclear, leading to incorrect replacement or adjustment of wiring. The actual connection status is recorded according to the sending and receiving end channel group number. The latest result is directly available for downstream use, and the change process is recorded in the connection status change record of that channel group. When the connection status of the same group repeatedly flips within several cycles, it will be separately marked as unstable, and priority will be given to investigating whether there are intermittent poor contacts in the sending and receiving end wiring.

[0055] Step S104: Based on the actual connection status, locate the degraded channel and the numerically frozen channel as an abnormal channel group. For the abnormal channel group, perform primary and backup switching with maximum heterogeneity to obtain switching records. Verify connectivity and homogeneity through the switching records to generate channel health records.

[0056] Specifically, based on the actual connection status, degraded channels and numerically frozen channels are identified as abnormal channels. Channels whose actual connection status indicates that the sending and receiving directions are actually opposite or the connection is unstable are directly included in the degraded channel range, without further verification of their numerical fluctuation characteristics. For channels with normal actual connection status, the variance of the recent energy values ​​of the channel is calculated hourly from the raw energy data. When the variance is lower than the square of the minimum resolvable value of the metering chip of that model, it is identified as a numerically frozen channel. The variance calculation window is taken from the most recent consecutive set number of sampling points of the channel. The window length is determined according to the criticality level of the equipment to which the channel belongs. The higher the level, the shorter the window to detect freezing more quickly; the lower the level, the longer the window to avoid misjudging normal low-fluctuation loads. The abnormal channel group merges degraded channels and numerically frozen channels into the same set for management. Within the set, each channel is still labeled with its specific category to avoid confusion between the causes of the two types of channels due to the merged management. Degraded channels often point to problems with the connection direction or the link itself, while frozen channels often point to malfunctions of the metering chip or the sensor itself. The responsible departments for investigating the two types of channels are also different. If a channel's actual connection status is "Pending Verification," it will not be included in this round of abnormal channel group judgment. The channel will be reassessed for signs of degradation or freezing after the sending and receiving end issues are resolved. During the pending verification period, the channel's existing classification will remain unchanged, and historical classification records will not be cleared due to the postponement of assessment. Abnormal channel groups are recorded based on both the channel number and its classification. Theoretically, the same channel should not simultaneously meet both criteria; if it does, it will be treated as a degraded channel.

[0057] In some embodiments, the step of performing primary / backup switching with maximum heterogeneity for the abnormal channel group to obtain switching records includes: identifying the acquisition terminal and transmission path identifier of each channel in the abnormal channel group to obtain a heterogeneity feature set; calculating the path overlap between the backup channel and the primary channel based on the heterogeneity feature set to obtain a heterogeneity score; screening the backup channel with the highest heterogeneity score based on the heterogeneity score to determine a preferred switching target; and performing primary / backup switching for the preferred switching target to obtain switching records.

[0058] The anomaly feature set is obtained by identifying the acquisition terminal and transmission path identifier of each channel in the abnormal channel group. For example... Figure 2As shown, the heterogeneous feature set records two items for each channel: the acquisition terminal connected to the channel (e.g., main channel acquisition terminal 1), and the segment identifiers of the transmission path through which the data is uploaded—main channel transmission path 2, backup channel A transmission path 3, and backup channel B transmission path 6 correspond to the actual wiring of different channels. Both items are only written after being verified against the equipment ledger and network topology records. The path segment identifiers are arranged in the order of the relay nodes that the channel data actually passes through, without omitting intermediate nodes due to the detour of physical wiring. The number of nodes varies depending on the actual network structure on site and there is no uniform upper limit. A multi-level relay line may contain four to five segment identifiers. The terminal segment identifier always corresponds to the last level relay device before the data reaches the dispatch center 7. Each channel in the abnormal channel group establishes a corresponding record in the above manner. The two identifiers together characterize the complete link features of the channel from physical acquisition to the arrival of data at the receiving end, and are used to compare the degree of link overlap between different channels. For channels belonging to the same data acquisition terminal that have undergone equipment model changes due to capacity expansion, the terminal identifier in the heterogeneous feature set is recorded according to the actual model after the most recent change, without retaining the old model information before the change. The change history is separately included in the equipment ledger change record and does not affect the judgment and use of the current feature set. If such channels also involve path segmentation adjustments, the segmentation identifiers are also re-registered according to the actual network topology after the change. If a channel's transmission path is adjusted only due to relay node expansion or line relocation, and the data acquisition terminal itself remains unchanged (this channel still belongs to the abnormal channel group), the heterogeneous feature set only updates the path segmentation identifier, while the terminal identifier remains unchanged. The two items are updated independently and do not refresh each other due to a change in one, avoiding accidental overwriting of terminal information that has not yet changed during the update process.

[0059] For example, the step of calculating the path overlap between the backup channel and the main channel based on the heterogeneous feature set to obtain the heterogeneity score includes: identifying the acquisition terminals and line segments shared by the backup channel and the main channel as a shared element set based on the heterogeneous feature set; calculating the proportion of the number of shared elements to the total number of elements based on the shared element set to obtain the path overlap; converting the path overlap according to a preset weight coefficient to generate a basic heterogeneous score; and verifying whether historical fault correlation correction is added to the basic heterogeneous score to determine the heterogeneity score.

[0060] Based on heterogeneous feature sets, the acquisition terminals and line segments shared by the backup channel and the main channel are identified as shared element sets. The heterogeneous feature set records of the main channel and the candidate backup channel A are compared item by item. Acquisition terminals with the same identifier are recorded as a shared element (i.e., backup channel A and the main channel share the same main channel acquisition terminal 1). The same segment number in the transmission path segment identifier is also recorded as a shared element, i.e., the overlapping segment 4 between transmission path 3 of backup channel A and transmission path 2 of the main channel. The comparison is performed item by item, without skipping the comparison of path segments because the terminals are the same, and without skipping the comparison of terminal identifiers because all path segments in the heterogeneous feature set are different. The two types of comparisons are completed independently without interference. The comparison order is fixed as terminal first and then path to avoid missing any item due to different execution order. The shared element set includes all shared elements obtained from the comparison. Each element is labeled with its specific type (terminal shared or path shared) and corresponding identifier number. If multiple instances of the same type of shared element appear, they are recorded separately and not merged. Duplicate counting is crucial for reflecting differences in the degree of sharing when calculating path overlap. Both the type and the number are indispensable; recording only the type without the number will make it impossible to trace the specific shared location. When there are no shared elements between the candidate backup channel and the main channel, the shared element set is recorded as an empty set. The empty set itself is a direct basis for judging that the physical links between the two are completely independent, rather than a signal of an abnormality in the comparison process. It is not necessary to re-perform the comparison due to an empty result. The empty set record still occupies a complete storage location and is not omitted due to empty content. The shared element set is established separately according to the number combination of the main channel and the candidate backup channel. When the same main channel corresponds to multiple candidate backup channels, each combination generates its own independent shared element set. The combination number follows the channel number already established in the heterogeneous feature set and is not newly coded.

[0061] The path overlap is obtained by calculating the proportion of shared elements to the total number of elements based on the shared element set. The denominator of the total number of elements is the sum of the number of terminal items and path segment items in the heterogeneous feature sets of the primary and backup channels. The number of elements in the shared element set is included in the numerator. The numerator is not distinguished by type, and terminal sharing and path sharing are included together. The weakening effect of the two on physical independence is temporarily regarded as equivalent and no weight is distinguished. This equivalent treatment takes effect uniformly before the weight coefficient conversion. Path overlap is calculated using the formula R = 2S / (P + Q), where S is the number of elements in the shared element set, P is the total number of elements in the primary channel heterogeneous feature set, and Q is the total number of elements in the candidate backup channel heterogeneous feature set. The path overlap is R. The numerator is multiplied by two to ensure that R falls between zero and one. A zero R indicates that the two links are completely independent. The closer R is to one, the higher the proportion of shared elements and the weaker the physical independence of the two links. The intermediate values ​​linearly reflect the degree of overlap without any non-linear weighting. The formula itself does not distinguish between P and Q, as they have completely equal status in the denominator. When the shared element set is empty, the path overlap is directly recorded as zero and is not included in the formula calculation to avoid ambiguity when the denominator is a normal value and the numerator is zero. This treatment is consistent with the formula calculation result itself, only saving an unnecessary division operation. When either P or Q is zero (the channel's heterogeneous feature set itself lacks terminal and path records, rather than the shared features being empty), the denominator of the formula becomes abnormal. The path overlap is then marked as missing data, and the value calculated by forcibly omitting it from the formula may be distorted. This is clearly distinguished from the case of zero overlap when the shared feature set is empty but P and Q are both normal.

[0062] The path overlap is converted into a basic heterogeneous source score using a preset weighting coefficient. The basic heterogeneous source score is calculated using the formula B=w s Calculated using ×(1-R), where R is the path overlap, w s As preset weighting coefficients, B is the basic heterogeneous score, w s The values ​​are taken from the constants obtained by referencing previous link independence assessment cases of this type of power dispatch data network. Different voltage levels of the network have their own values ​​for w. s Values ​​are not shared by a unified constant. The calibration process is organized by the network planning department for on-site verification. The verification method involves selecting several groups of known physically independent or known shared-cable channel pairs and comparing them with the calibrated values. s The calculated score must be verified to ensure it matches the actual degree of independence. The results are reviewed every two years. The larger the path overlap R, the smaller the (1-R) ​​term, and the lower the basic heterogeneous score B. The conversion between the two is entirely determined by the subtraction and multiplication structure of the formula itself, without any additional piecewise functions or correction terms. For candidate backup channels with a path overlap of zero, the basic heterogeneous score is directly equal to w. sThe score itself represents the highest possible value under that weighting coefficient. At this point, multiplication is no longer performed to avoid introducing deviations due to floating-point errors. s The value itself is also recorded in the metadata of this calculation for traceability. The metadata also indicates whether this calculation triggered the simplified processing branch of zero overlap. When the path overlap itself is marked as missing data (i.e., the R value cannot be calculated because P or Q is zero), the basic heterogeneous score is also not substituted into the formula calculation, but is directly marked as pending. It does not take extreme assumptions such as R equal to zero or equal to one because R is missing, so as to avoid treating "cannot be determined" as two completely different deterministic conclusions of "completely independent" or "completely overlapping".

[0063] The heterogeneity score is determined by verifying whether historical fault correlation corrections are added to the basic heterogeneity score. The heterogeneity score is calculated by subtracting the historical fault correlation correction, converted based on the correlation between the historical fault times of the main channel and the candidate backup channel, from the basic heterogeneity score. The stronger the correlation, the larger the correction, and the greater the decrease in the final score. The correlation conversion relationship is derived from statistical fitting of similar events for this equipment model, without using subjective judgment of the correlation level. The correlation verification first retrieves the historical fault records of the main and backup channels from the equipment ledger fault archive, comparing whether the fault times of the two channels fall within the same power grid disturbance event window. The proportion of abnormal records appearing in both channels before and after the same disturbance event to the total number of historical faults for both channels is used as the quantitative value of the correlation, directly used for correction conversion. The basic heterogeneity score itself only measures the static overlap of physical paths, while the correlation verification supplements the dynamic evidence of whether the two channels experienced the same disturbance and synchronized anomalies. These two types of evidence have different sources and are indispensable. If there are no historical fault correlation records between the primary channel and the candidate backup channel (e.g., both have a short operational history and have not yet experienced a common disturbance event), the heterogeneity score will not be corrected and will be directly equal to the basic heterogeneity score itself, without artificially setting a discount due to the lack of historical data. For candidate backup channels whose basic heterogeneity score is marked as pending (i.e., the path overlap cannot be calculated to a definite value due to missing data), the heterogeneity score will also be marked as pending. The calculation of subtracting the correction amount will not be performed, and the correlation verification process will also skip this combination. The verification will be carried out again after the path overlap data is supplemented and the basic heterogeneity score is recalculated to a definite value.

[0064] The preferred switching target is determined based on the highest heterogeneity score of the backup channel. All candidate backup channels corresponding to the same main channel are summed and ranked according to their heterogeneity scores. A higher score indicates a stronger physical independence between the candidate backup channel and the main channel. For example, backup channel B, which uses terminal 5 and transmission path 6 and has no shared sections with the main channel, typically has a higher score than backup channel A, which shares terminals and cable corridors with the main channel. The ranking process covers all currently available candidate backup channels for the main channel, ensuring no eligible candidate is overlooked. The ranking itself is unidirectional, sorted from highest to lowest heterogeneity score, without grouping or grading. The ranking result retains the original score value for each candidate, not just the ranking. The preferred switching target is the candidate backup channel with the highest score in the ranking, marked with a "preferred switching target" symbol (8). In cases of a tie for the highest score, the path overlap is further compared, with the lower overlap rate prioritized. If the overlap rates are also the same, the candidate backup channel with the smaller numerical value is prioritized. These three criteria are compared sequentially without weighted summation. Once a higher criterion is determined, the comparison stops, and the remaining levels are not verified. If all candidate backup channels are unavailable due to other main channels being in service, the preferred switching target is temporarily suspended and marked as "no available target." The main channel retains its original primary / backup configuration without adjustment, waiting for a candidate channel to become available before recalculating the heterogeneity score ranking and determining the target. During recalculation, the old scores are retained for comparison, including whether the candidate pool itself has changed. The preferred switching target is recorded according to the main channel number. If the preferred target is frequently changed for the same main channel, it is separately marked as having scarce backup resources, prompting the expansion of independent backup links for that main channel.

[0065] For the preferred switching target, a primary / backup switch is performed to obtain a switchover record. Once the preferred switching target is determined, a switchover command is simultaneously issued to the corresponding primary channel and the switching target. The command includes two parameters: the target role (primary or backup) and an effective time limit. The switching target is required to complete the role switch from backup to primary within a fixed time after receiving the command, and the original primary channel simultaneously switches to backup status. The roles of the two are swapped, but no simultaneous parallel data collection is performed. The swap action is transparent to the other system modules corresponding to the primary channel and requires no additional configuration changes. The specific value of the fixed time limit is set according to the line voltage level. The switchover record records the execution time of this switchover action, the original primary channel number, and the number of the new primary channel after the switchover (i.e., the preferred switching target number). These three items are written all at once, without batch appending. Immediately after writing, the connectivity verification process is triggered. The triggering action and the switchover record writing action are completed in the same operation without any time interval. If no status confirmation is received from the target channel within a fixed time after the switchover command is issued, the switchover is deemed a failure. A failure flag is marked on the switchover record, the original primary channel reverts to primary status, and the transitional state during the switchover is not maintained. An alarm is simultaneously generated to remind maintenance personnel to verify the actual status of the preferred target channel on-site. If the same preferred target channel fails to receive two consecutive switchover commands, the command will not be reissued. Instead, the candidate backup channel with the second-highest heterogeneity score will be selected as the new preferred target for the switchover. The number of consecutive failures of the original target channel is recorded in its own record for future reference when determining whether the channel is still suitable as a switchover target.

[0066] A channel health profile is generated by verifying connectivity and homogeneity through the switchover record. After the switchover, the new primary channel, designated as the preferred switchover target in the switchover record, first undergoes a connectivity verification. This verification is performed using the same method as in the previous steps used to determine the true connection status, confirming that the new primary channel is indeed transmitting data normally and in the correct direction. Subsequently, a homogeneity verification is performed to check whether its recently collected data shows abnormal convergence with the historical data of the original primary channel before the switchover. Convergence indicates that the actual links are not truly independent, contradicting the judgment made before the switchover based on heterogeneity scoring. These two verifications are performed sequentially in a fixed order and are not conducted in parallel. The channel health profile integrates information from four aspects: the channel's degradation or freeze status in the abnormal channel group, the connectivity verification result, the homogeneity verification result, and the current switchover record. This forms a comprehensive record of the channel's current status. These four aspects are separately labeled and not combined into a single health / non-health conclusion, but share the timestamp of the same switchover event as the correlation basis. When the homogeneity verification reveals signs of actual convergence, the channel health record will trigger a heterogeneity score review reminder, indicating that the heterogeneity feature set or path overlap calculation for the primary and backup channel pairs may have been missed, requiring re-verification of the actual physical link routing. The review reminder will also list the data window used for the convergence verification for the verification personnel to locate. If the connectivity verification itself is determined to be unhealthy or pending review, the channel health record will mark the new primary channel of this switch as having abnormal connectivity, and trigger a rollback: the original primary channel will be reset to primary status, the switch target will be marked as unavailable, and will not be included in the candidate pool of preferred switch targets until the connection problem of the target itself is verified and resolved on-site and the restriction is manually lifted. The rollback process itself is also recorded in the switch record as an independent event.

[0067] Step S105: Based on the channel health record, the criticality classification of abnormal channels is corrected to form a channel risk classification. Based on the channel risk classification, a window with abnormally stable margin and fluctuating load is determined to obtain a dynamic dead zone threshold. The channel is frozen by the dynamic dead zone threshold and bandwidth weights are allocated according to the inverse variance to generate a fused data stream.

[0068] Specifically, channel risk grading is formed by correcting the criticality grading of abnormal channels based on channel health records. If the number of records in the channel health record marked as deteriorated, frozen, or failed to switch over within a preset review period exceeds the set upper limit for the proportion of all verification records accumulated in the health record for that period, the preset equipment criticality grading for that channel will be increased by one level. If the proportion exceeds the upper limit of the next higher level, it will be increased by two levels. Channels already at the highest level will not be further increased. The denominator of the proportion is based on the actual accumulated number of records in the health record for that period, and the judgment threshold will not be lowered due to insufficient records. The channel risk grading records the final grade of the channel after correction, while retaining both the original criticality grading and the magnitude of the increase, making it easy to distinguish whether the high importance of the equipment itself is inherent or the increase is passive due to operational performance. If no abnormalities are found for three consecutive review periods after the increase, the corresponding channel can submit a manual review application to be downgraded by one level. The downgrade only takes effect after the responsible team confirms the equipment status on-site; it will not be automatically downgraded due to good performance in a single period. In cases where a channel has never had any abnormal channel group records (i.e., all channel health records are normal), the preset equipment criticality level is directly used without any modification. The original level and final level of this type of channel remain consistent. The abnormal records used for proportional calculations are only traced back within a preset number of periods. Historical records exceeding the traceability range are not included in the denominator of the current proportional calculation. An isolated anomaly from many years ago will not indefinitely affect the current channel risk level. The traceability period is set separately according to the equipment criticality level itself; the higher the level, the longer the traceability period, and the less likely long-standing hidden dangers will be overlooked.

[0069] In some embodiments, determining the dynamic dead zone threshold by identifying a window with abnormally stable margin and load fluctuation based on the channel risk classification includes: determining a load fluctuation window of a preset duration based on the channel risk classification; calculating the variance of the operating margin change from the load fluctuation window to obtain a margin variance value; verifying whether the margin variance value is lower than a preset variance threshold and whether the load fluctuation amplitude within the window reaches the lower limit to generate an abnormally stable marker; and determining the corresponding dead zone adjustment amplitude for the abnormally stable marker to obtain the dynamic dead zone threshold.

[0070] The load fluctuation window is determined based on the channel risk classification. For channels with a risk classification of Level 1, the load fluctuation window duration is 15 minutes; for Level 2, it's 30 minutes; and for Level 3, it's 60 minutes. The higher the risk classification, the shorter the window. A shorter window means more batches can be verified per unit of time. The specific duration is calibrated according to the typical cyclical characteristics of the load curve for that type of equipment. Different types of equipment maintain their own duration lookup tables, which are updated every six months based on newly added equipment types. The start time of the load fluctuation window is the moment when the channel last completed its risk classification correction. The window continuously covers all load data collected by the channel within the corresponding duration from that moment. After the window ends, a new window immediately begins, with no gaps or overlaps between windows. When a channel's risk classification changes during the window's operation (e.g., triggering a new round of upward adjustment), the current load fluctuation window continues to operate for its original duration until its end. It is not prematurely truncated or extended due to mid-term changes in classification. The next window is then recalculated based on the duration corresponding to the new classification. The switch between old and new durations only occurs at the window boundary, and the boundary time itself is taken from the end time of the previous window, not the actual time when the classification change occurred. If a channel has not yet completed any channel risk classification determination (e.g., it has only recently been put into monitoring), the first load fluctuation window is uniformly opened with a duration of 60 minutes corresponding to Level 3. Only after the channel completes its first risk classification will the next window switch to a duration matching its classification, avoiding insufficient sample size for margin variance calculation due to the application of a shorter window caused by missing classification information.

[0071] The variance of the operating margin is obtained by calculating the variance of the operating margin change from the load fluctuation window. The operating margin is calculated hourly as the ratio of the difference between the rated capacity and the measured load to the rated capacity. Each sampling time within the load fluctuation window corresponds to an operating margin value. The length of the margin sequence is consistent with the actual number of sampling points within the window. The sampling interval follows the original conventional acquisition cycle of the channel without additional interpolation. The variance of the margin is the variance of all operating margin values ​​within the window. The variance is calculated according to the standard statistical formula: first, the mean of the margin sequence within the window is calculated, then the square of the difference from the mean is calculated point by point, and the average is taken. The larger the variance value, the more drastic the margin fluctuation within the window; the smaller the value, the more stable the margin. The mean value itself is also included in the calculation result, together with the variance value, to form a complete statistical summary of the window. If any missing data points appear in the operating margin sequence within the load fluctuation window (e.g., the channel happens to fail to transmit data at that moment), these missing data points are not included in the numerator and denominator of the margin variance calculation. The variance is recalculated based on the actual number of valid sampling points, without filling the missing data points with zero values ​​or interpolation. If the proportion of valid sampling points to the theoretically required number of points in the window is lower than the set lower limit, the current margin variance value is considered to have insufficient labeled samples. The labeled results are passed to the abnormal stability marking judgment link as additional reference and do not trigger any alarms independently. If the rated capacity of the equipment has changed within the corresponding period of the load fluctuation window (e.g., capacity expansion), the calculation basis for the two segments of the operating margin sequence is not consistent. The margin variance value is then divided into two segments according to the time of the change, and the variance is calculated separately for each segment. The segment with more sample points is taken as the basis for this judgment to avoid the sudden change in the rated capacity itself being mistakenly included in the variance as a drastic fluctuation in the margin.

[0072] An abnormal stability marker is generated based on whether the margin variance value is lower than a preset variance threshold and whether the load fluctuation amplitude within the window reaches the lower limit. An abnormal stability marker is generated only when both conditions are met: the margin variance value is lower than the preset variance threshold, and the fluctuation amplitude of the system-side reference load curve within the corresponding time period of the window reaches or exceeds the set lower limit. Situations where only the margin variance value is low while the system-side load itself is in a stable period are not considered abnormal stability, as margin stability in this case is simply a normal result of load stability, not an anomaly. The reference load curve is taken from the historical typical load curve of the power supply area to which the channel belongs. The fluctuation amplitude is measured by the difference between the maximum and minimum values ​​of the curve within the corresponding time period of the window. Using historical curves rather than real-time curves avoids using potentially distorted data from the same channel itself for verification. The abnormal stability marker records the judgment result along with the original values ​​of the margin variance value and the reference load fluctuation amplitude used for judgment. This allows for direct verification of the reasonableness of the judgment basis during manual review, without needing to retrieve the original load fluctuation window data again. The preset variance threshold is set separately for each type of equipment to which the channel belongs. Different types of equipment have different magnitudes of margin fluctuations during normal operation. A uniform threshold would cause some equipment types to be systematically misjudged. The specific value of the threshold is taken from the lower limit of the distribution of margin variance during the historical normal operation of this type of equipment. Each type of equipment is maintained separately and does not share the same value. When multiple consecutive windows are marked as abnormally stable, the consecutive counts are accumulated in the abnormally stable marking record. The accumulated value is directly referenced in the dead zone adjustment amplitude measurement stage. When a window is not marked as abnormally stable (i.e., the margin returns to normal fluctuation), the consecutive counts are reset to zero and the old value is not used after the interruption due to the previously accumulated counts.

[0073] The dynamic dead zone threshold is obtained by measuring the corresponding dead zone adjustment amplitude for abnormally stable markers. Each time the number of consecutive accumulations of abnormally stable markers reaches a set threshold, the dead zone adjustment amplitude increases by a fixed step size. A higher number of consecutive abnormally stable marker accumulations indicates a longer duration of suspected freezing in the channel, necessitating a tightening of the judgment criteria for early detection. The specific step size is one-tenth of the original dead zone setting for this model of metering equipment. The threshold value itself is also set separately for each equipment type, without a shared set number of values. The dynamic dead zone threshold is calculated by subtracting the dead zone adjustment amplitude from the preset fixed dead zone reference value for the channel. This subtraction gradually tightens the dead zone as the number of consecutive abnormally stable accumulations increases. After tightening, even small numerical changes will be judged as exceeding the dead zone and thus considered valid changes, no longer treated as noise. This tightening direction aligns with the target of identifying frozen channels. The dynamic dead zone threshold has a tightening lower limit. The adjustment range accumulates until the dead zone value reaches this lower limit and then stops tightening. This prevents the dead zone from tightening to zero or even a negative value, which could cause normal measurement noise to be mistakenly judged as a valid change. The specific value of the lower limit is half of the measured upper limit of the noise level of this type of metering equipment. Even if the number of consecutive occurrences of the abnormal stability marker continues to increase after reaching the lower limit, the dynamic dead zone threshold will remain at the lower limit and will not tighten further. When the abnormal stability marker of a certain channel turns to unmarked in the next window (i.e., the margin returns to normal fluctuation), the number of consecutive occurrences returns to zero according to the rules described in the previous step, the dead zone adjustment range returns to zero simultaneously, and the dynamic dead zone threshold immediately returns to the fixed dead zone reference value. There is no longer a separate window-by-window gradual retreat path, which avoids the adjustment range's downward pace lagging behind the determination result of the consecutive occurrences returning to zero.

[0074] In some embodiments, the step of allocating bandwidth weights to the numerically frozen channels in an inverse-variance ratio using the dynamic dead-zone threshold to generate a fused data stream includes: identifying the baseline variance value corresponding to the numerically frozen channel using the dynamic dead-zone threshold; determining the weight coefficient based on the inverse coefficient between the current variance and the baseline variance value; allocating the detection bandwidth of the numerically frozen channels using the weight coefficient to obtain a bandwidth allocation value; and summarizing the allocation results of each channel based on the bandwidth allocation value to generate a fused data stream.

[0075] The baseline variance value corresponding to the numerically frozen channel is identified by using a dynamic dead zone threshold. The dynamic dead zone threshold for each numerically frozen channel is taken from its most recent window value. This threshold itself implies the severity of the channel's current suspected freezing. The tighter the threshold, the more serious the freezing suspicion. Based on this, the identification process filters out a list of all channels currently marked as numerically frozen from the dynamic dead zone threshold record. The list also carries the equipment type identifier of each channel for matching when determining the alternative benchmark for the baseline variance value. This list itself is updated synchronously with each round of dynamic dead zone threshold recalculation. The baseline variance value is taken as the mean variance of the original energy value during the most recent period in the channel's history when it was confirmed to be in a normal fluctuation state (i.e., never judged as anomalously stable or numerically frozen). The mean is obtained by directly calculating the arithmetic mean of the variance values ​​of all windows within that period, without weighting. The period length is taken as the typical span of the normal operating cycle of the equipment type to which the channel belongs. If a numerical freeze channel has never had a confirmed normal period in its history (e.g., freezes appear shortly after equipment is put into operation), the baseline variance value is replaced by the mean variance of similar equipment under similar operating conditions. The specific source of the replacement benchmark is simultaneously recorded in the baseline variance value record of that channel for traceability. The baseline variance value is not fixed after a single calculation. Every time a set review cycle is completed, if the channel recently has a confirmed normal period, the baseline is recalculated using the latest data to replace the old value. This avoids the variance changes caused by natural equipment aging being misjudged as abnormal by the long-used old baseline. The baseline in the case of a replacement benchmark also follows this refresh rule. Once the channel itself accumulates confirmed normal periods, individual data is used.

[0076] The weighting coefficients are determined by calculating the inverse ratio between the current variance and the baseline variance based on the baseline variance. The weighting coefficients are calculated using the formula w = σ0. 2 / σ_c 2 Calculate, where σ0 2 σ_c represents the baseline variance. 2 σ_c represents the actual variance of the raw values ​​measured in the current statistical window for this channel. w is the weighting coefficient; this ratio increases as the current variance decreases, and is the inverse coefficient of the current variance relative to the baseline variance. A larger inverse coefficient indicates that the current fluctuation deviates more from historical normal levels. 2 The closer w is to zero, the larger the value of σ_c 2 The variance σ0 of the baseline 2 When w approaches one, it indicates that the current fluctuation is consistent with historical normal levels and does not constitute an anomaly. σ_c 2When an extremely small value causes w to be abnormally large, the weighting coefficient has an upper limit. Once this limit is reached, it will not increase further. The specific upper limit is taken as the high-order cutoff point of the observed w value distribution for that equipment model. The cutoff point is calculated statistically from the w value samples actually calculated for that equipment type over the past periods. This is re-verified every six months in conjunction with new samples to avoid individual channels consuming excessive bandwidth due to extremely small variance calculations, thus compressing the detection resources that should be allocated to other channels. σ_c 2 In the special case of zero (channel values ​​remain completely unchanged within the statistical window), the weighting coefficient is taken directly as the upper limit, without division to avoid division by zero errors. This special case is also recorded next to the baseline variance record for that channel as an additional reminder. σ_c 2 When the number of valid sampling points in the current window is insufficient to meet the minimum set value, the weighting coefficients will use the values ​​already calculated in the previous window for this channel, instead of using the σ_c value due to insufficient samples. 2 Recalculate to avoid insufficient sample size leading to σ_c 2 The distortion itself renders the w value meaningless; the continued use of this value is marked in the weight coefficient record so that the bandwidth allocation process can identify that the current value is not the actual measurement result of the current round.

[0077] The bandwidth allocation value is obtained by allocating the detection bandwidth of the numerically frozen channels using weighted coefficients. The bandwidth allocation value is calculated according to the formula W_i = W_total × w_i / Σw_j, where W_total is the total bandwidth available for detection of numerically frozen channels in this round, w_i is the weight coefficient w of the i-th numerically frozen channel calculated by the inverse proportional coefficient in the previous step, and Σw_j is the sum of the weight coefficients of all numerically frozen channels. W_i is the bandwidth allocation value allocated to that channel. The formula distributes the total bandwidth according to the relative proportion of the weight coefficients of each channel, and the sum of the allocation results is exactly equal to W_total. The higher the weight coefficient of the channel, the larger the bandwidth allocation value is allocated. That is, the channel with a more severe suspected freezing degree receives more detection resources, which is used to increase its resampling frequency and speed up the confirmation of whether the freezing is real or has been lifted. W_total is determined by the remaining bandwidth after deducting the bandwidth required for regular testing of other non-frozen channels from the current total network bandwidth. This remaining portion fluctuates round by round based on the real-time occupancy of the remaining channels, without a fixed value. In rounds with large fluctuations, W_total is directly adopted based on the measured value of that round without cross-round smoothing. When the number of frozen channels in a round is zero (i.e., no channels are currently frozen), the bandwidth allocation calculation is skipped entirely, and the reserved portion of W_total is returned to the regular bandwidth pool for unified scheduling, without being reserved for idle channels. If the bandwidth allocation value calculated by the formula is lower than the minimum bandwidth required for basic testing of this device model, the channel is allocated the minimum bandwidth as a safety net. The difference is deducted proportionally from other channels with lower weighting coefficients that would have originally received more bandwidth. After deduction, the sum of the bandwidth allocation values ​​for all channels still equals W_total, without exceeding the total bandwidth limit.

[0078] A fused data stream is generated by aggregating the bandwidth allocation results of each channel based on the bandwidth allocation values. The bandwidth allocation values ​​of all frozen channels, along with the original regular detection bandwidths of the remaining unfrozen channels, are aggregated and written into a unified bandwidth scheduling table, one by one according to the channel number. The scheduling table is refreshed entirely after each round of bandwidth allocation value recalculation, without incremental modifications. The old scheduling table before the refresh is not retained separately and is directly overwritten by the new table. The fused data stream sequentially merges the data actually collected by each channel in this round into a single data stream output according to the bandwidth share given in the scheduling table. Frozen channels occupy a correspondingly higher sampling density in the merged result due to their increased bandwidth allocation, while unfrozen channels maintain their original regular density unaffected. If a channel fails to generate data according to its allocated bandwidth share in this round due to a link failure, the corresponding position for that channel is left blank in the fused data stream and is not filled with data from other channels. The blank itself is marked with the specific reason (such as switchover failure or link interruption) in the data stream's metadata for downstream system identification. When aggregating the fused data stream, three source identifiers are simultaneously carried: channel risk classification, dynamic dead zone threshold, and current weighting coefficient. These three identifiers are taken from the latest record of their respective stages, enabling downstream applications to trace the current monitoring status of each data stream while reading the data, rather than simply obtaining raw values ​​without source information. Degraded channels, whose problems lie in the connection direction or the link itself and can be resolved through master / slave switching, do not require encrypted sampling confirmation. Therefore, bandwidth allocation in this stage is only for channels with frozen values ​​and no additional bandwidth is allocated to degraded channels. The fused data stream is generated and output round by round. Complete historical status tracing relies on the original records retained by previous stages, such as channel health profiles and channel risk classifications, and only the latest fusion results of the current round are presented.

[0079] To implement the above-described method embodiment, a multi-channel energy data acquisition method for a power dispatching data network is provided to achieve the corresponding functions and technical effects. See also... Figure 4 , Figure 4 This diagram illustrates a structural block diagram of a multi-channel energy data acquisition device 400 for a power dispatching data network according to an embodiment of this application. For ease of explanation, only the parts relevant to this embodiment are shown. The multi-channel energy data acquisition device 400 for a power dispatching data network provided in this embodiment includes:

[0080] The data acquisition unit 401 is used to acquire the channel configuration parameters of the acquisition terminal, and to perform interleaved acquisition according to the topological correlation phase difference based on the channel configuration parameters to generate raw energy data;

[0081] Hazard screening unit 402 is used to locate the margin deviation interval of the original energy data to determine the criticality level of the equipment, filter low load records below a preset proportion for the criticality level of the equipment to obtain a hazard channel set, and establish a channel priority sequence based on the hazard channel set by frequency and duration weighting.

[0082] The connection diagnostic unit 403 is used to track the channel uploading records according to the channel priority sequence to obtain the abnormal uploading interval duration, verify the abnormal uploading interval duration with the same window change to determine the associated response status, and verify the consistency of the sending and receiving end channel direction from the associated response status with deviation tolerance to output the true connection status.

[0083] The redundancy switching unit 404 is used to locate the degraded channel and the numerically frozen channel as an abnormal channel group based on the actual connection status, and to perform primary and backup switching for the abnormal channel group with maximum heterogeneity to obtain switching records. The switching records are used to verify connectivity and homogeneity to generate channel health records.

[0084] The risk allocation unit 405 is used to correct the criticality of abnormal channels based on the channel health record to form a channel risk classification, determine a window with abnormally stable margin and fluctuating load based on the channel risk classification to obtain a dynamic dead zone threshold, and allocate bandwidth weights to the numerically frozen channels according to the inverse variance ratio through the dynamic dead zone threshold to generate a fused data stream.

[0085] The aforementioned multi-channel energy data acquisition device 400 for a power dispatch data network can implement a multi-channel energy data acquisition method for a power dispatch data network as described in the above method embodiments. The options in the above method embodiments are also applicable to this embodiment and will not be detailed here. The remaining contents of this application embodiment can be referred to the contents of the above method embodiments, and will not be repeated in this embodiment.

[0086] The above embodiments are not an exhaustive list based on the present invention, and there may be many other embodiments not listed. Any substitutions and improvements made without departing from the concept of the present invention are within the protection scope of the present invention.

Claims

1. A method for acquiring multi-channel energy data in a power dispatching data network, characterized in that, include: Obtain the channel configuration parameters of the acquisition terminal, and perform interleaved acquisition based on the topological correlation phase difference according to the channel configuration parameters to generate raw energy data; The original energy data is used to locate the margin deviation interval to determine the equipment criticality level. Low load records below a preset proportion are selected for the equipment criticality level to obtain a set of hidden danger channels. Based on the set of hidden danger channels, a channel priority sequence is established by weighting frequency and duration. Based on the channel priority sequence, track the channel uplink records to obtain the abnormal uplink interval duration. Verify the abnormal uplink interval duration against the window change to determine the associated response status. From the associated response status, verify the consistency of the sending and receiving end channel directions with deviation tolerance and output the true connection status. Based on the actual connection status, the degraded channels and numerically frozen channels are identified as abnormal channel groups. For the abnormal channel groups, the primary and backup switching is performed with maximum heterogeneity to obtain switching records. The connectivity and homogeneity are verified through the switching records to generate channel health records. Based on the channel health records, the criticality of abnormal channels is graded and corrected to form a channel risk classification. Based on the channel risk classification, a window with abnormally stable margin and fluctuating load is determined to obtain a dynamic dead zone threshold. The bandwidth weights of the numerically frozen channels are allocated inversely proportional to the variance using the dynamic dead zone threshold to generate a fused data stream.

2. The method according to claim 1, characterized in that, The step of generating raw energy data by interleaving acquisition according to the channel configuration parameters and the topologically correlated phase difference includes: Based on the channel configuration parameters, the phase offset is determined by evenly distributing the preset period according to the number of channels within the topology association group; Based on the phase offset, channel interleaved sampling times are allocated to generate a comb-shaped sampling sequence; The effectiveness of the interleaving is determined by calculating the overlap ratio of sampling times of adjacent channels based on the comb sampling sequence. For channels where the interleaving effectiveness is below a preset threshold, the phase offset is reallocated and the original energy data is output.

3. The method according to claim 1, characterized in that, The process of obtaining a potential hazard channel set by filtering low-load records below a preset proportion based on the criticality level of the equipment includes: Based on the aforementioned equipment criticality classification, a low-load judgment benchmark is established according to the rated capacity ratio range; Candidate potential hazard periods are obtained by statistically analyzing the duration of continuous satisfaction of the low load judgment criteria. The confidence level of a potential hazard is formed by calculating the margin increase of adjacent normal periods from the candidate hazard periods. Based on the aforementioned hazard confidence level screening, channels with confidence levels higher than a preset threshold are used to obtain a hazard channel set.

4. The method according to claim 1, characterized in that, The step of verifying the consistency of the sending and receiving end channel directions from the associated response state with deviation tolerance and outputting the true connection state includes: The calibration direction relationship is obtained by using the preset power flow direction labels between the sending and receiving channels to identify the associated response status. The deviation angle value of the real-time direction is determined based on the aforementioned calibration direction relationship; The deviation angle value is checked to see if it falls within the preset deviation tolerance range to obtain a directional health mark; The true connection status is output by statistically analyzing the consistency ratio of the sending and receiving channel groups based on the directional health markers.

5. The method according to claim 1, characterized in that, The step of obtaining handover records by performing primary / backup handover with maximum heterogeneity for the abnormal channel group includes: Based on the abnormal channel group, the acquisition terminal and transmission path identifier of each channel are identified to obtain the heterogeneous feature set; The heterogeneity score is obtained by calculating the path overlap between the backup channel and the main channel based on the heterogeneous feature set. Based on the heterogeneity score, the backup channel with the highest screening score is selected as the preferred switching target; Perform a primary / backup switchover for the preferred switchover target and obtain the switchover record.

6. The method according to claim 1, characterized in that, The process of determining the dynamic dead zone threshold based on the channel risk classification, which involves identifying a window with abnormally stable margins and fluctuating loads, includes: A load fluctuation window of preset duration is determined based on the channel risk classification; The margin variance value is obtained by calculating the variance of the operating margin change from the load fluctuation window; Based on the margin variance value, verify whether it is lower than the preset variance threshold and whether the load fluctuation amplitude within the window reaches the lower limit to generate an abnormal stability mark; The dynamic dead zone threshold is obtained by measuring the corresponding dead zone adjustment amplitude for the anomalous stationary marker.

7. The method according to claim 1, characterized in that, The step of generating a fused data stream by allocating bandwidth weights to the numerically frozen channels in inverse variance proportions using the dynamic dead zone threshold includes: The baseline variance value corresponding to the numerical freeze channel is identified by the dynamic dead zone threshold. The weighting coefficient is determined by measuring the inverse ratio of the current variance to the baseline variance based on the baseline variance value. The bandwidth allocation value is obtained by allocating the detection bandwidth of the numerical freezing channel using the weighting coefficients. Based on the bandwidth allocation values, the allocation results of each channel are summarized to generate a fused data stream.

8. The method according to claim 4, characterized in that, The step of verifying whether the deviation angle value falls within the preset deviation tolerance range to obtain a directional health mark includes: Numerical aggregation is performed on the aforementioned deviation angle values ​​to establish an angle distribution range; The samples whose angles deviate from the center by more than a preset multiple of the standard deviation are identified as abnormal angle points from the angular distribution interval. Based on the percentage of occurrence of the abnormal angle points within a preset period, an interval correction coefficient is obtained. The directional health marker is obtained by adjusting the preset deviation tolerance range using the interval correction coefficient.

9. The method according to claim 5, characterized in that, The process of calculating the path overlap between the backup channel and the main channel based on the heterogeneous feature set to obtain a heterogeneity score includes: Based on the heterogeneous feature set, the acquisition terminals and line segments shared by the backup channel and the main channel are identified as a shared element set. The path overlap is obtained by calculating the proportion of shared elements to the total number of elements in the shared element set. The path overlap is converted into a basic heterogeneous score according to a preset weighting coefficient; The heterogeneity score is determined by verifying whether the historical fault correlation correction amount is superimposed on the basic heterogeneity score.

10. A multi-channel energy data acquisition device for a power dispatching data network, characterized in that, include: The data acquisition unit is used to acquire the channel configuration parameters of the acquisition terminal, and to perform interleaved acquisition based on the topological correlation phase difference according to the channel configuration parameters to generate raw energy data; The hidden danger screening unit is used to determine the criticality level of equipment by locating the margin deviation range of the original energy data, to filter low load records below a preset proportion for the criticality level of the equipment to obtain a set of hidden danger channels, and to establish a channel priority sequence based on the set of hidden danger channels by frequency and duration. The connection diagnostic unit is used to track the channel uplink records according to the channel priority sequence to obtain the abnormal uplink interval duration, verify the abnormal uplink interval duration with the same window change to determine the associated response status, and verify the consistency of the sending and receiving end channel direction from the associated response status with deviation tolerance to output the true connection status. The redundant switching unit is used to locate the degraded channels and numerically frozen channels as abnormal channel groups based on the actual connection status, and to perform primary and backup switching for the abnormal channel groups with maximum heterogeneity to obtain switching records. The switching records are used to verify connectivity and homogeneity to generate channel health records. The risk allocation unit is used to correct the criticality of abnormal channels based on the channel health records to form a channel risk classification, determine a window with abnormally stable margin and fluctuating load based on the channel risk classification to obtain a dynamic dead zone threshold, and allocate bandwidth weights to the numerically frozen channels according to the inverse variance ratio using the dynamic dead zone threshold to generate a fused data stream.

Images