A method for collecting operation data and health assessment of a box-type substation

Abstract

The application relates to the technical field of power distribution network operation management, and discloses a box-type substation operation data acquisition and health assessment method; the method is used for solving the problems that it is difficult to establish a verifiable unified time reference, it is difficult to quantitatively identify the reliability of cross-source alignment, and there is a lack of evidence access conditions, evidence consistency constraints and conflict resolution rules for the same operation period; the method first freezes a data source list, a data dictionary mapping, a time window strategy, evidence rules and a session identifier; the multi-source message is frozen into an operation entry, a source side timestamp, a serial number, a start count and an arrival timestamp are recorded, and after structure and value range checking, the out-of-order buffer is output; the time reference account book is maintained through heartbeat and anchor points, the unified time window is generated, and the access, verification and resolution are completed according to the clauses, the input package is encapsulated and landed and archived, and the traceable assessment result is output.

Description

Technical Field

[0001] This invention relates to the field of power distribution network operation and management technology, specifically to a method for collecting operation data and assessing the health of prefabricated substations. Background Technology

[0002] As a crucial node in the distribution network, prefabricated substations typically integrate transformers, high-voltage switches, low-voltage distribution, protection and control systems, and environmental control units. Existing operational monitoring solutions generally collect electrical parameters such as voltage, current, power, and power quality, as well as remote signaling data for circuit breaker positions, alarms, and interlocking. They also access fault and event records from protection devices and can be further expanded to include environmental control and status parameters such as temperature, humidity, condensation, smoke detection, access control, and partial discharge, enabling remote monitoring, alarms, and maintenance assistance. To support these needs, existing patents propose online monitoring systems for prefabricated substation environmental parameters. For example, CN103107597B utilizes sensors and communication links to collect and transmit parameters such as temperature, humidity, and partial discharge online, improving the observability of the substation's environmental status. Other patents propose fault analysis systems for substations, such as CN110941918B, which acquires operational data and fault information through a data acquisition module and combines this with data analysis and health assessment modules to achieve status evaluation and early warning.

[0003] However, in distribution network prefabricated substations, multi-source heterogeneous data are typically generated independently by different devices. The sampling frequencies for electrical parameters, remote signaling updates, protection event records, environmental control, and partial discharge monitoring differ, and the timestamp sources are also inconsistent. Simultaneously, field links may experience communication jitter, out-of-order retransmissions, time jumps caused by equipment restarts, and missing data segments. Existing solutions often employ simple timestamp splicing or fixed time window aggregation, making it difficult to establish a verifiable unified time benchmark, quantify the reliability of cross-source alignment, and lack evidence admission conditions, evidence consistency constraints, and conflict resolution rules for the same operating period. This results in unreliable correlation of state evidence from different sources, significant fluctuations in health assessment input data, difficulty in verifying assessment conclusions, and poor consistency in repeated calculations, further leading to false alarms and missed alarms and reducing the credibility of operation and maintenance decisions. Therefore, a method for prefabricated substation operation data acquisition and health assessment is needed. This method should establish a unified time window with quantifiable alignment confidence at the substation or master station side, combined with evidence admission and conflict resolution constraints, to obtain verifiable and repeatable health assessment results and stable alarm outputs. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method for collecting operational data and assessing the health of prefabricated substations. This method solves the problems of traditional methods, such as the difficulty in establishing a verifiable unified time reference, the difficulty in quantifying the reliability of cross-source alignment, and the lack of evidence admission conditions, evidence consistency constraints, and conflict resolution rules for the same operating period.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A method for data acquisition and health assessment of prefabricated substation operation, comprising: S1: Create a running session, freeze the data source list, data dictionary mapping, time window strategy and evidence rule version, generate a session identifier and write it to the session record; S2: Access electrical parameters, remote signaling, events, environmental control and partial discharge data in a unified entry format, register source-side timestamp, sequence number, start count and arrival timestamp, perform field verification and enter the buffer queue; S3: Maintain the time baseline ledger based on heartbeat events and anchor events, perform mapping establishment, downgrading, freezing and reconstruction according to trigger conditions, update the valid interval and mark the alignment confidence level; S4: Generate a unified time window according to the time window strategy, perform evidence admission judgment, evidence consistency constraint verification and conflict adjudication on the items within the window, and record the window timeline and adjudication items; S5: Output health assessment input package and archive session traces, write back mapping snapshots, missing segment summaries, resource summaries and parameter candidate records.

[0006] Further, a running session is created, the data source list, data dictionary mapping, time window strategy, and evidence rule versions are frozen, a session identifier is generated, and written to the session record, including: Sessions are created by timed triggers, event triggers, or manual triggers, and session identifiers and caliber verification identifiers are generated according to site identifier, date, session sequence number, and trigger code; Freeze the list of data sources, data dictionary mappings, time window strategies and evidence rule versions, and fix the version number, mapping number, threshold number, snapshot number and scope of effect; After writing the resource budget and enabling threshold triggering, it enters the access state.

[0007] Furthermore, electrical parameters, remote signaling, events, environmental control and partial discharge data are accessed according to a unified entry format, and source-side timestamps, sequence numbers, start counts and arrival timestamps are registered, including: The access end converts multi-protocol messages into running entry records, registers the source-side timestamp, source-side sequence number, source-side start count and arrival timestamp, and writes the link batch and retransmission flag; When the source-side startup count is unavailable, the restart criterion is that the chain break exceeds the preset threshold and the source-side sequence number wraps around.

[0008] Further, field validation is performed and the data is added to the buffer queue, including: The entries are sequentially subjected to structural and value range verification, and are rejected, downgraded, or retained according to the clauses corresponding to the key fields. Validated running entries are entered into the out-of-order buffer queue according to the data source, and are output according to normal release, timeout release, or out-of-order release, and the release reason is recorded. When the timestamp rollback exceeds the rollback threshold, the system switches to paused alignment and releases the pause once the stability condition is met. Write an investigation flag when the abnormality rate exceeds the abnormality rate threshold.

[0009] Furthermore, based on the maintenance time-baseline ledger for heartbeat events and anchor events, including: Maintain the time-based ledger by creating append-written ledger entries for each data source; The session timeline is constructed using the arrival timestamp of the access point, and the corresponding arrival timestamp is not used as a marker when the session is in a paused alignment state. Anchor point matching is based on the consistency of event type, device location, and event direction, combined with timing tolerance determination, and the matching clause number and alignment confidence level are written into the ledger entry.

[0010] Furthermore, mapping is established, downgraded, frozen, and rebuilt according to triggering conditions, and the valid intervals are updated and aligned confidence levels are marked, including: A state machine is used for mapping establishment, degradation, freezing, and reconstruction. Within the confirmation observation window, mapping is confirmed when consecutive state entries reach the confirmation threshold; When the jitter count, out-of-order count, or missing count exceeds the threshold, it switches to degrade mode. When the restart criterion is met, a freeze is triggered and reconstruction begins. At the same time, the endpoints of the valid interval are updated, and a ledger snapshot indexed by window identifier and aligned confidence label entries are generated.

[0011] Furthermore, a unified time window is generated according to a time window strategy, including: Based on the time window strategy, determine the start and end points of the window on the session timeline, and generate a timed window by timed scrolling or an extended window by event triggering. When the window overlap exceeds the merge threshold, a merge is performed and the source window reference is recorded. Write the window number, credibility flag, and coverage flag for each window; When a missing window number or a rollback is detected, an abnormal number record is written and the generation of the adjudication record for that window is stopped.

[0012] Furthermore, evidence admission determination, evidence consistency constraint verification, and conflict resolution are performed on the items within the window, and the window timeline and resolution items are recorded, including: Within the window, the entries are assessed for field completeness, value range validity, and alignment confidence based on the clause number, and the role of evidence is determined accordingly. Conflict candidates are generated based on timing constraints, state consistency constraints, and association constraints. The decision priority is determined based on the evidence role, alignment confidence and release reason, and decision entries and rollback action codes are generated and written to the window timeline record.

[0013] Furthermore, output the health assessment input package and archive the session history, including: The window record is encapsulated into an input package, which contains the rule version number, clause index identifier, caliber verification identifier, index key and verification code. The index key consists of a session identifier, a window identifier, and a record type code, and is associated with the location mapping snapshot, evidence list, adjudication list, timeline, and payload summary area; When archiving session tracks, write a checksum to each track entry and write a reference to the preceding checksum.

[0014] Furthermore, the write-back mapping snapshot, missing segment summary, resource summary, and parameter candidate records include: Record mapping snapshots, missing segment summaries, and resource summaries during session archiving, and record parameter candidate entries; The missing segment summary includes the judgment criterion code, threshold citation identifier, and evidence role adjustment record; Resource summary write buffer peak value, release count, latency grading identifier, and continuous over-limit judgment criteria; Parameter candidate entries are generated based on event frequency thresholds and statistical criteria, and suggested effective conditions and suggested rollback conditions are written into them.

[0015] Compared with existing technologies, the present invention provides a method for data acquisition and health assessment of prefabricated substations, which has the following advantages: 1. This invention manages the data source list, data dictionary mapping, time window strategy, and evidence rule version of the collection starting point through session-based management. Session identifiers connect running entries, unify time windows, and assessment inputs, reducing caliber drift caused by changes in caliber. At the access end, electrical parameters, remote signaling, events, environmental control, and partial discharge data are collected and unified as running entries, registering source timestamps, sequence numbers, start counts, and arrival timestamps. Combined with structure verification, value range verification, out-of-order buffering, retransmission, and restart criteria, the impact of communication jitter, out-of-order retransmission, equipment restarts, and data loss on cross-source correlation is reduced. Based on heartbeat events and cross-source anchor events, a time baseline ledger is maintained. A state machine is used to complete mapping establishment, degradation, freezing, and reconstruction, marking alignment confidence levels and valid intervals. Within a unified time window, evidence access, timing and state consistency constraint verification, and conflict adjudication are performed according to clause numbers. An indexable and verifiable health assessment input package is output, and session trajectories and snapshot summaries are archived. Finally, verifiable, traceable, and reproducible consistent health assessment inputs and stable alarm outputs are obtained.

[0016] 2. This invention freezes parameters such as resource budget, anomaly ratio threshold, missing segment judgment threshold, window merging threshold, and continuous over-limit judgment rules, and combines them with threshold triggers. When the station-side processing latency, memory usage, or buffer size exceeds the budget limit, the verification scope and output path are subject to controlled degradation based on evidence role priority. Structured summaries such as degradation source, missing segment information, retransmission batch markers, and unresolved conflicts are written into the health assessment input package for trajectory chain verification. Parameter candidate records are generated based on frequently triggered statistical criteria, and effective and rollback conditions are suggested. This prevents parameter adjustments from replacing effective configurations within the session and entering a controlled evolution process, thereby reducing the risks of instantaneous accumulation, repeated adjudication, and output rhythm loss under link fluctuations and resource constraints. Ultimately, this invention achieves stable window operation under edge resource constraints and enables controlled iterative parameter operation, improving the operational capability to maintain stable window processing and support controlled iterative parameter operation under edge resource constraints. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the process for a prefabricated substation operation data acquisition and health assessment method according to the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1: Figure 1A method for data acquisition and health assessment of prefabricated substations is presented, including: S1: Create a runtime session, freeze the data source list, data dictionary mapping, time window strategy and evidence rule versions, generate a session identifier and write it to the session record. The specific implementation is as follows: In a data collection and health assessment task for a distribution network prefabricated substation, an operational session is established and the scope is frozen. The session identifier and resource constraints are added to the session header. The session identifier is used to uniformly reference and associate operational items, unified time windows, adjudication records, and health assessment input packages. Various records within the session can be continuously traced and support consistent verification. The version number, mapping number, threshold number, snapshot number, and effective scope related to the scope are permanently stored with the session header or session record. It includes at least the following fields: version number field, mapping table number field, threshold number field, snapshot number field, and effective scope field. The version number comes from the version repository record frozen when the session is created. The mapping number and threshold number come from the corresponding mapping table and policy table numbering system. The snapshot number comes from the ledger or configuration snapshot record collected when the session is created. The effective scope comes from the session start and end boundaries and version freezing rules. This avoids inconsistencies in scope and uncertainties in verification caused by relying solely on timestamps or external configuration changes. The session is initiated by a trigger source, which can be one of three types: timed trigger, event trigger, or manual trigger. The trigger source type and trigger parameters are written to the session record when the session is created, serving as the basis for subsequent time window generation path selection and resource budget strategy matching. Timed triggers are controlled by session period parameters, with the session period ranging from 15 minutes to 24 hours, preferably 30 minutes to 4 hours. The parameter range is set as a compromise between the frequency of operation and maintenance inspections and the density of station reporting, and is constrained by the resource budget to avoid exceeding the buffer budget within the session period in the shortest reporting cycle scenario, while ensuring the continuous traceability of the window sequence; event triggers... Event triggering is controlled by event thresholds. The event threshold value is triggered when any of the following conditions are met: alarm level reaches level 2 or above, switch status changes, protection action record appears, partial discharge status changes from normal to alert or alarm, and temperature and humidity exceed the duration threshold. The alarm level field is taken from the main station alarm classification table or the station alarm policy table. Level 2 is the level marked as 2 in the level field of the table or its equivalent mapping value. When there is a level inconsistency between the main station alarm classification table and the station alarm policy table, the alarm caliber version frozen at the time of session creation shall prevail, and the level mapping table number shall be written into the session record to ensure consistent resolution across systems. The partial discharge status is determined using mutually exclusive enumeration items mapped from a discrete code table. The enumeration items include at least fields such as normal, attention, and alarm, and lock mutually exclusive relationships. The upper limit thresholds for temperature and humidity are taken from the station-side environmental control strategy table or equipment operation procedure clauses. When the thresholds are default, the default thresholds in the equipment ledger are used and the default reason code is recorded. At the same time, the threshold number and unit caliber are written into the session record to support review and traceability. The duration threshold is set from 5 seconds to 300 seconds, preferably from 10 seconds to 60 seconds. The parameter setting is based on filtering single refresh jitter and taking into account the response speed. The threshold is not less than twice the remote signaling refresh cycle or environmental control sampling cycle to reduce the risk of false triggering by short-term disturbances. Manual triggering is used for forced collection in operation and maintenance review or maintenance window. Manual triggering adds a manual reason code to the session record. The manual reason code is a numeric alphanumeric code of no more than 16 characters, which comes from the mapping table of work order type code and maintenance plan type code. When the mapping is missing, a default value is written and the default reason is marked. A session is created and written to the session header. The session identifier consists of a site identifier, date, session sequence number, and trigger source type code. The site identifier is a set of numbers and letters not exceeding 32 characters, derived from the distribution network asset ledger or the dispatch master station number system. Its length is appropriate for the field length and message payload of the station-side database. The date is year, month, and day. The session sequence number is read from 1 within the same date and persisted locally. If the sequence number rolls back, the session is closed and the sequence number exception event is recorded. The session header synchronously writes the caliber verification identifier field, which consists of the data source list version number, data dictionary mapping version number, time window strategy version number, and evidence rule. The system consists of fields such as version number, key parameter table version number, number of data source entries and clause number range. The maximum length of the caliber verification identifier is 64 characters, which can be used for recalculation. The key parameter table should include at least the following fields: session period, duration threshold, temperature upper limit threshold, humidity upper limit threshold, window duration, step size, event extension before and after, splicing upper limit, minimum coverage level rule, minimum confidence level rule, arrival delay threshold, and missing proportion threshold. The corresponding version number and effective range should be written in the session header. The thresholds and limits are derived from the site-side policy table, equipment operation procedures or asset ledger default configuration and snapshot number for traceability. During session creation, the data source list version, data dictionary mapping version, time window strategy version, and evidence rule version are frozen. These versions are frozen upon session creation and are prohibited from being directly changed or effective within the session. Subsequent changes are entered into the database as candidate records and processed in subsequent write-back processes. Candidate records do not replace effective parameters within the current session to avoid caliber drift leading to irreproducible verification. The data source list version is used to solidify the set of data sources participating in this session. The data source set is stored in an entry format. Each data source entry includes at least the following fields: data source identifier, data source type, electrical location identifier, channel set identifier, reporting cycle, heartbeat interval, source-side sequence number monotonicity flag, source-side start-up count availability flag, clock type, and link type. The data source type can be selected from electrical parameters, remote signaling, events, environmental control, and partial discharge. The identifier value is a combination of the interval number and the cabinet number. The channel set identifier is used to lock the measurement point set and the range of the point table. The reporting period is used as the input for access buffer and out-of-order waiting parameters. For example, electrical parameters are 1 to 10 seconds, remote signaling is triggered by changes or polled every 2 seconds, events are reported immediately, environmental control is 5 to 60 seconds, and partial discharge is 1 to 30 seconds. The above ranges are derived from the common configuration of field devices and the constraints of the station link carrying capacity. The heartbeat interval is 2 to 5 seconds, used to keep the link observable and limit the bandwidth occupied by the heartbeat when there are no service entries. Its range is derived from the field link jitter level and bandwidth occupancy constraints. The clock type should at least distinguish between the device local clock, network time synchronization and master station time synchronization. The link type is used to identify the link type such as private network, wireless, Ethernet or serial port forwarding, and provides condition input for subsequent mapping confirmation threshold settings and abnormal handling strategy selection. The data dictionary mapping version is used to solidify field meanings, unit definitions, discrete value semantics, and legal ranges. The mapping table establishes mapping records based on the data source identifier and maintains consistency within the session. Each mapping record includes at least the source field name, target field name, target unit, lower limit of legal range, upper limit of legal range, discrete code table, and default value handling rules. The legal range is jointly determined by the equipment rated values, allowable deviation clauses, and short-term overload clauses. The rated values ​​are taken from the rated parameter table in the equipment ledger or the rated parameter area in the device configuration file. After reading, a snapshot of the rated values ​​is written to the session header. Numbering and reading time are used for verification and traceability; for example, the legal range of the low-voltage side voltage of the transformer substation is 0.7 to 1.3 times the rated voltage, and the legal range of the current is 0 to 1.5 times the rated current. The range boundaries are derived from the constraints of allowable deviation and short-time overload clauses; the default value handling rules can be selected as three types: deny access, retain with missing mark, and replace with safety default code, and are configured according to the data source type; the discrete code table is used to map the source-side code values ​​of remote signaling, events and partial discharge status to unified semantic values ​​and solidify mutual exclusion relationships to ensure semantic consistency of the same status across different devices; The time window strategy version is used to freeze the parameters of a unified time window, including at least the window duration, sliding step, event pre- and post-event expansion boundaries, upper limit for missing segment splicing, minimum coverage rule, minimum confidence rule, and overlap / merging threshold. The window duration ranges from 120 to 600 seconds, and the sliding step ranges from 10 to 30 seconds. Its selection is based on the coverage requirements of protecting the event propagation chain and environmental control response delays, and is constrained by the maximum processing delay budget. The event expansion boundary uses pre- and post-event expansion parameters, such as 10 seconds before the event to 120 seconds after the event, to cover the cause and effect of the event. The upper limit for missing segment splicing... The limit value is 3 to 10 seconds, which is used to tolerate short-term link drops and retransmissions and is not used for inference across missing segments. Missing segments exceeding the splicing limit are written into the window header as the basis for coverage limitation. The minimum coverage and minimum confidence are fixed by level rules. The coverage can be selected as satisfied, limited, or not satisfied, and the confidence can be selected as reliable, available, conservative, or unavailable. The level definition is frozen in the session to maintain comparability between windows. The overlap merging threshold value is set to perform merging when the overlap ratio exceeds 30%. Its value is based on the engineering requirements of reducing duplicate decisions and resource consumption and avoiding duplicate decisions of the same conflict in different windows. The evidence rule version freezes the evidence admission conditions, evidence consistency constraint clauses, and conflict resolution clauses, and retains the clause number within the same session. The evidence admission clause must include at least the fields of field completeness, value range legality, alignment confidence condition, duplicate handling rule, and missing handling rule. The consistency constraint clause must include at least the fields of time sequence tolerance, state consistency condition, and association condition. The resolution clause must include at least the fields of conflict type enumeration, priority chain, and fallback action. These fields are the result of the frozen evidence rule version and clause index. The clause index is marked in the session header, and the clause number is used in subsequent resolution records and input packets to facilitate version traceability of the basis for admission, verification, and resolution. After session creation, a resource budget is written and a resource threshold trigger is enabled. The resource budget includes at least the following fields: maximum number of buffer entries, maximum memory usage, maximum processing latency, and maximum retention period. The maximum number of buffer entries is between 5000 and 50000; the maximum memory usage is between 64 MB and 1024 MB; the maximum processing latency is between 300 milliseconds and 2000 milliseconds; and the maximum retention period is between 30 days and 180 days. These ranges are determined by constraints imposed by the site's resource configuration capabilities, alarm handling timeliness requirements, and maintenance traceability cycles. The site's resource configuration snapshot number or configuration version number is also written synchronously to ensure traceable review. The system includes fields such as the sum of the current lengths of the buffer queues of each data source, the statistical period, and the peak value. The trigger judgment value is the peak value within the most recent second. When the peak value reaches 80% of the upper limit, the system enters an early warning state to reserve space for peak retransmission. When the peak value reaches 100% of the upper limit, the system enters a degradation state and triggers degradation path selection. The above threshold settings are matched with the instantaneous queue surge characteristics caused by sudden retransmission and the memory limit protection requirements. The statistical period, trigger time, and trigger reason associated with the trigger event are written to the session record for subsequent review. The session state switches from the creation state to the access state, enabling the session to enter the unified entry access and buffer management operation phase.

[0020] S2: Access electrical parameters, remote signaling, events, environmental control and partial discharge data according to a unified entry format, register the source-side timestamp, sequence number, start count and arrival timestamp, perform field verification and enter the buffer queue. The specific implementation is as follows: After the session enters the access state, the reported data of electrical parameters, remote signaling, events, environmental control and partial discharge are uniformly converted into running entry records and the fields are verified and buffered and shaped. The protocol difference only affects the unpacking method and does not change the entry field caliber and threshold caliber. The running entry record adopts a fixed structure, including at least the following fields: entry identifier, session identifier, data source identifier, data source type, device location identifier, channel identifier, source-side timestamp, source-side sequence number, source-side start count, arrival timestamp, payload category, payload field area, quality flag area, integrity flag, and link flag. The entry identifier monotonically increments within the session, and is generated by the access end by assigning an incrementing number to records that have been successfully parsed and passed the structure verification, and combining it with the session identifier. When the access end restarts, causing the entry identifier count to wrap around, an access end restart event is written to the access end restart event record, and the session state is switched to pause alignment. The source-side timestamp is taken from the time the data source was generated and retains the original value, with the time unit labeled as milliseconds or seconds according to the data dictionary mapping version. The arrival timestamp is taken from the time the access end's local clock receives the data, for out-of-order delivery. Buffering, arrival delay, and backoff determination; source-side sequence number is used to identify out-of-order and dropped frames, and source-side startup count is used to identify time jumps caused by device restarts; when the startup count is unavailable in the data source list version, the access end maintains the original meaning of the source-side sequence number and source-side timestamp, and uses alternative criteria to write the quality flag. The alternative criteria include at least the heartbeat break exceeding the restart threshold and source-side sequence number wrapping; the heartbeat entry value is the status entry or protocol heartbeat frame reported by the data source at heartbeat intervals, and the heartbeat identification field name is taken from the heartbeat type identifier field in the data dictionary mapping version; the restart threshold is 20 to 60 times the heartbeat interval, preferably 30 to 40 times. This multiple range matches the difference in short-term link fluctuation duration and device-level restart duration to ensure consistent determination criteria under different heartbeat configurations; The load category distinguishes between instantaneous values, change events, alarm events, statistical segments, and waveform segments. The load field area binds the target field name and unit caliber according to the data dictionary mapping version. For fields that cannot be mapped, the original field name and original unit are written into the extended field area. An unknown field or unknown code table mark is added to the quality mark area. The link mark marks information such as link source and transmission batch, including at least the link source, batch number, and retransmission mark. The link source is the link type enumeration value of the data source list version. The retransmission mark takes the retransmission indication field built into the protocol, or is set when the access end observes no less than 20 entries from the same data source with source timestamps earlier than the current access time and continuously increasing within a 5-second observation window. The observation window and the number of entries threshold match the short-term dense characteristics of on-site retransmission playback. It is written into the link mark along with the batch number. The access point performs two-level verification on the running entry record and generates a branch path; the structure verification determines whether the entry fields are interpretable. Required fields must include at least the session identifier, data source identifier, data source type, source-side timestamp, arrival timestamp, and payload category. When a required field is empty, parsing fails, or the format violates the format specification, the integrity is marked as "field not full," the failure field name and failure reason code are written, the entry enters the exception retention queue, and the structure verification failure event record is written to the session track. It does not enter the buffer queue or the alignment processing link. Value range verification is used to determine the range of the load field and the validity of the discrete code. The verification method adopts the data dictionary mapping version, and verifies the valid range and discrete code table segment by segment in the load field area. When a continuous quantity goes out of bounds, a field out-of-bounds flag is written to the quality flag area, recording the field out-of-bounds name, out-of-bounds direction, and out-of-bounds magnitude. The out-of-bounds magnitude is the difference between the actual value and the nearest valid boundary, and the target unit is also attached. When the discrete quantity code value is not in the code table, an unknown code table flag is written to the quality flag area, recording the original code value and the reason for the mapping failure. Three paths can be selected: rejection, downgrading, and retention, depending on the project availability and review consistency. When a critical field goes out of bounds or the critical discrete code is unknown, a rejection path is taken. The critical field value is the set of critical fields frozen in the evidence rule version and referenced by the clause number. The types of critical fields include switch open / close position, protection action type, fault trip mark, etc. When a non-critical field goes out of bounds, a de-weighted path is taken. The entry can enter the buffer and be preset as circumstantial evidence, and the out-of-bounds information is retained. When the information is incomplete but can be used for link diagnosis, a retention path is taken. The entry can enter the abnormal retention queue and be included in the resource summary. The hierarchical handling rules and the set of critical fields are frozen with the evidence rule version to ensure consistency in the interpretation during review and replay. Entries that pass structure and range validation and are not rejected enter the buffer queue for reordering and retransmission, and output a data stream that is as ordered as possible. Buffer queues are established separately according to data source identifiers, with each data source corresponding to a buffer instance. The buffer instance includes at least the fields of buffer wait time, single-source queue limit, release policy, and overflow policy. The buffer wait time is 5 to 15 seconds, preferably 8 to 12 seconds. This range matches the jitter of field communication, retransmission and playback latency, and the buffer carrying capacity of the access end, and is coordinated with the window duration and step size rolling frequency. It is also constrained by the maximum number of buffer entries and the maximum memory usage in the resource budget. The single-source queue limit is 200 to 20,000 entries, preferably 1,000 to 10,000 entries. Its range is constrained by the data source reporting cycle, entry density, and resource budget. The upper limit is larger when the reporting cycle is shorter and smaller when the reporting cycle is longer. The sum of the lengths of each single-source queue is constrained by the maximum number of buffer entries in the session header. Release methods include: normal release, timeout release, and out-of-order release. Normal release is triggered and outputs in order when the source-side sequence numbers in the buffer queue are consecutive and the waiting time does not exceed the buffer waiting time. Timeout release is triggered and outputs the earliest entry when the buffer waiting time exceeds the buffer waiting time. Out-of-order release is triggered and outputs the earliest entry when the source-side sequence number jumps within the buffer waiting time and the waiting time exceeds half of the buffer waiting time. An out-of-order count is added to the access statistics and a release reason flag is written. For cases where duplicate entries affect subsequent alignment, the access end uses first-arrival retention to determine duplicate entries. The conditions for duplicate determination are that the same data source identifier, the same channel identifier, and the same source-side sequence number appear twice or more within the buffer waiting time. The first-arrival entry is retained in the buffer, and subsequent entries are written to the abnormal retention queue and recorded as duplicates. The number of duplicates is written to the duplicate count and included in the resource digest. To handle the issue of reversed arrival timestamps caused by local clock anomalies at the access end, a backoff threshold is set at the access end. The backoff threshold is set between 100 milliseconds and 1000 milliseconds, preferably between 200 milliseconds and 500 milliseconds. This range matches the requirements for distinguishing the operating system's time fine-tuning range and the hardware clock jump range. When the arrival timestamp of an adjacent entry backs away from the backoff threshold, an access end time anomaly event record is generated, and the session state is switched to pause alignment. The time anomaly event record includes at least the event occurrence time, backoff range, associated entry identifier, associated data source identifier, and link marker. In the pause alignment state, access and buffer shaping are maintained, and output of buffer-ready entries is stopped. The release condition is that no entry with an arrival timestamp backs away from the backoff threshold appears within a 30-second observation window, and the arrival timestamp remains monotonically increasing. The stable duration of 30 seconds is consistent with the characteristics of the heartbeat interval covering multiple cycles and the continuous characteristics of the time synchronization fine-tuning window. Buffer overflow employs an explicit overflow strategy and is linked to resource budget. When the length of a single-source queue exceeds its upper limit, a single-source overflow is executed and written to the overflow event log. When the total number of buffered entries reaches 80% of the maximum number of buffered entries in the session header, overflow preprocessing begins. When the total number of buffered entries reaches 100% of the maximum number of buffered entries in the session header, a forced overflow is executed. The selection of overflow objects follows the priority of evidence roles, with the option to prioritize discarding entries pre-set as circumstantial evidence, followed by discarding entries with reduced weight. The overflow event log includes at least the trigger threshold level, the number of discarded entries, the distribution of discarded evidence roles, the link marker, and the resource usage snapshot, providing a traceable record basis for overflow handling and enabling review and replay. The access point generates two types of structured outputs and associates them with subsequent processing stages. The buffered ready entry stream is an output stream of entries ordered as much as possible according to the internal sequence number of the data source. Each entry includes at least fields such as release reason flag, quality flag, integrity flag, and link flag, which are used for time base ledger maintenance. The abnormal retention entry stream is used for session trajectory archiving and resource summary statistics. The abnormal ratio threshold is set at 1% to 20%, with a preference for 5% to 10%. Its range is determined by referring to the normal packet loss level on site, the probability of errors in point table configuration, and the acceptable boundary of short-term interference. The abnormal ratio is the ratio of the number of abnormal retention entries to the total number of entries successfully parsed by the access point, and is written to the session record according to the session rolling statistics cycle. The statistics cycle is 10 seconds to 60 seconds, or 1 to 3 times the window step size and is fixed in the session record. When the abnormal ratio exceeds the abnormal ratio threshold, an investigation flag is generated and written to the session record. At the same time, when the subsequent window is generated, the starting value of the alignment confidence level of the data source is set to conservative.

[0021] S3: Based on the heartbeat event and anchor point event maintenance time baseline ledger, perform mapping establishment, downgrading, freezing, and reconstruction according to trigger conditions, update the valid interval and mark the alignment confidence level. The specific implementation is as follows: During the continuous output of the buffered ready entry stream, a time base ledger is maintained. Verifiable session timeline mappings are established between different data sources under absolute time synchronization conditions. When the mapping is unstable or broken, it is downgraded, frozen, and rebuilt according to a state machine. The session timeline is a monotonic timeline formed by the arrival timestamps of the access points. Sessions are written starting from the arrival timestamp corresponding to the session creation time, including the time unit and the access point's time source type. The session timeline marker for each entry is the valid arrival timestamp after filtering by the pause alignment rules, and buffered metadata such as the release reason flag is retained. The time base ledger establishes ledger entries according to the data source identifier. Ledger entries are appended and include at least the following fields: mapping status, valid interval start point, valid interval end point, alignment confidence level, jitter counter, out-of-order counter, missing counter, jump counter, anchor list, heartbeat continuity count, recent release reason statistics, and recent arrival delay statistics. The mapping status can be selected as not established, pending confirmation, established, downgraded, frozen, or rebuilt. The alignment confidence level can be selected as reliable, available, conservative, or unavailable, with the meaning solidified in the evidence rule version and frozen in the form of rule clauses. The time-based ledger uses heartbeat events and anchor events as the mapping and verification basis. Heartbeat events are status entries or protocol heartbeat frames reported by the data source at the heartbeat interval frozen according to the data source list version; the common configuration range is 2 to 5 seconds. The heartbeat entry payload includes at least the heartbeat sequence number, source-side timestamp, source-side start count, and link status code. The access end performs structural verification on the heartbeat entries and registers the sequence number continuity, without absolutely correcting the source-side timestamp. The continuous heartbeat count is used to characterize the online stability of the data source. The continuous heartbeat count threshold is recorded as the acknowledgment threshold, ranging from 5 to 1. 0 times, used to cover multiple consecutive heartbeat cycles within the confirmation observation window and suppress false confirmations caused by single heartbeat jitter; to enhance reproducibility, the confirmation threshold is fixed in the evidence rule version using segmented rules and written into the session record. For example, the confirmation threshold is 6 times when the heartbeat interval is not less than 4 seconds, and the confirmation threshold is 8 times when the heartbeat interval is less than 4 seconds. It is still allowed to make slight adjustments within the range of 5 to 10 times in combination with the site resource budget and write the rule version number and parameter version number of this session. The relevant settings refer to the matching relationship between heartbeat interval configuration, buffer waiting time and site processing latency constraints. Anchor events are business edge events that can be observed simultaneously from multiple data sources. The anchor event type set can be selected from six categories: switch on / off position changes, protection action record occurrence, fault alarm transition, fan start / stop, dehumidifier start / stop, and access control opening / closing. When a site does not have a certain type of equipment configured, that type of anchor point is marked as unavailable in the data source list version and does not participate in matching. Anchor matching is expressed using consistency rules and timing tolerance rules. The consistency conditions include at least consistent event type, consistent equipment location identifier, and consistent event direction. The timing tolerance is denoted as the anchor tolerance, ranging from 50 milliseconds to 500 milliseconds, with 100 milliseconds being preferred. The millisecond to 300 millisecond range is set to maintain compatibility with the secondary record resolution and typical latency of the station link. Each anchor matching adds a record to the anchor list in the ledger entry. The anchor record includes fields such as anchor type, participating data source identifier set, timestamps of each source side, marker points of the session timeline, matching clause number, and link marker summary. When the participating data source set is frozen or the alignment confidence level is unavailable, the anchor is only used for diagnostic retention and is not used as a reference factor to improve the alignment confidence level, so as to avoid the impact of time jump entries on the mapping reliability after restart. Mapping establishment, downgrading, freezing, and rebuilding are executed using a state machine. State transition conditions are set as auditable trigger conditions and linked to the ledger counter. Mapping establishment transitions from "not established" to "pending confirmation," triggered by the first heartbeat or business entry received from the data source. During the pending confirmation period, a confirmation observation window is enabled, and a sequence number monotonic check is performed. The confirmation observation window is recorded as the observation duration, ranging from 10 to 120 seconds, preferably 20 to 60 seconds, to cover multiple heartbeat cycles while also considering online latency. Within the observation duration, continuous heartbeat counting must be satisfied simultaneously. When the number reaches the confirmation threshold, the source-side sequence number does not wrap around, and the jump counter does not reach the jump threshold, the mapping state changes to established. The jump counter is used to record events such as changes in the source-side start count, the source-side timestamp rollback exceeding the source-side rollback threshold, and the source-side sequence number wrapback. The source-side rollback threshold is between 1 second and 30 seconds, preferably between 3 seconds and 10 seconds, to distinguish between device-side fine-tuning and obvious jumps. The source-side rollback threshold applies to the source-side timestamp field, and the arrival timestamp rollback threshold applies to the arrival timestamp field. The two objects are different and are frozen in their respective stages. When mapping enters an established state, the start point of the effective interval is set to the session timeline marker corresponding to the time of entry into the established state, and the end point of the effective interval remains open until a freeze is triggered or the session ends. The alignment confidence level is conservative in the initial stage of establishment, and subsequent upgrades are executed according to the rules and can be reviewed and referenced: when there is at least one anchor match within the window and no unresolved conflict flag is received, it is upgraded to available; when there are at least two consecutive anchor matches and no unresolved conflict flags are received in the last five windows, it is upgraded to trusted. Unresolved conflict flags are written to the window adjudication entry list with the window identifier. In this stage, the participating data source list is associated with the window identifier, the flag summary is appended to the status change record of the corresponding ledger entry, and the trigger clause number is referenced so that the confidence level adjustment can be traced back to the window and clause. Degradation triggering conditions are executed based on three types of counter thresholds: jitter, out-of-order delivery, and missing data. The jitter counter records arrival delay and the number of times the arrival delay threshold is exceeded. The arrival delay is the time difference between the entry's arrival timestamp and the arrival timestamp of the most recent heartbeat entry within the current observation window, and is grouped and registered according to link tags. The arrival delay threshold is set between 200 milliseconds and 3000 milliseconds, preferably between 500 milliseconds and 1500 milliseconds. This range is determined by combining the normal latency and retransmission latency tolerance boundaries of power distribution communication links, and different preferred intervals can be adopted according to link type. For example, 500 milliseconds to 800 milliseconds is preferred for private network links, and 800 milliseconds to 1500 milliseconds is preferred for wireless links. The link type value follows the data source list version and is... The ledger statistics are grouped; the out-of-order counter is a cumulative summary of the number of out-of-order releases and timeout releases in the buffer output, and the most recent release reason is written into the ledger entry for verification; the out-of-order threshold is 20 to 50 times per window, or 100 to 500 times cumulatively per session segment. The session segment is an observation interval consisting of 3 to 10 consecutive uniform time windows, and its length is written into the session record and fixed within the session. The session segment is updated by sliding according to the window number to distinguish between short-term fluctuations and long-term instability; the missing counter records the number of times heartbeats are missing to form missing segments. The missing segment judgment threshold is 5 to 10 times the heartbeat interval, preferably 6 to 8 times. This range is used to distinguish between short-term packet loss and continuous loss of connection and to avoid misjudging short-term packet loss as mapping breakage; When any counter exceeds the threshold within the current observation window, the mapping status switches from established to degraded, the alignment confidence level is lowered by one level, and the degraded reason code, trigger window identifier reference, and counter snapshot summary fields are written into the ledger entry. The degraded reason code can be selected from three types: jitter degraded, out-of-order degraded, and missing degraded, which facilitates locating the trigger source during review. In the degraded state, the anchor matching and counter fallback are continuously observed, and the recovery observation window takes 3 to 8 unified time windows, preferably 4 to 6 windows. When an anchor matching occurs in the recovery observation window and all three types of counters fall back below the threshold, the mapping status is restored to established and the alignment confidence level is raised by one level. When no anchor matching occurs in the recovery observation window but the heartbeat continuous count is stable, the mapping status remains degraded and the alignment confidence level is maintained at a conservative level to adapt to site scenarios lacking anchors and avoid misjudging cross-source alignment as trustworthy when there are no verifiable anchors. Freezing is used to handle device restarts and significant time jumps. Freezing trigger conditions can be selected from any of the following: changes in the source-side startup count, source-side sequence number wrapping, or heartbeat break exceeding the restart threshold. The restart threshold adopts the restart threshold parameter fixed at the access end and is expressed as a multiple of the heartbeat interval to adapt to different heartbeat configurations and distinguish between short-term link loss and device-level restart. After freezing is triggered, the effective interval endpoint is set to the session timeline marker point corresponding to the last entry with an alignment confidence level not lower than available before the freeze trigger, and a freeze reason code is written. The reason code can be either restart freeze or jump freeze. When there are no entries with an alignment confidence level not lower than available before the freeze trigger, the effective interval endpoint is taken as the session timeline marker point corresponding to the last entry that passed the structure verification and whose release reason is marked as normal release before the freeze trigger, thus ensuring that the effective interval endpoint can be determined and is easy to verify. During the freeze, the data source entry continues to be connected and buffered, and the alignment confidence level is uniformly marked as unavailable and does not participate in anchor matching to avoid the source-side timestamp after the restart from entering the established mapping. After freezing, reconstruction enters the reconstruction state. The reconstruction trigger condition is the reappearance of continuous heartbeats reaching the confirmation threshold and the source-side start count stabilizing and no longer changing. Once these conditions are met, the process enters the pending confirmation and reuse mapping establishment process. The alignment confidence level after reconstruction is initially conservative and can only be upgraded to usable after at least one anchor point match, to avoid judging cross-source alignment solely based on heartbeat stability. If a data source repeatedly triggers freezing during reconstruction and the number of freezing exceeds the freezing number threshold, the data source is marked as an unstable source, and fields such as the unstable flag and freezing number statistics summary are written into the ledger entry. The freezing number threshold is 2 to 5 times, preferably 3 times. This range needs to match the frequency of occasional restarts and the distinction between continuous instability on site, and allows for adjustment of its evidence role priority in subsequent window phases. This phase generates two types of structured outputs: a time-based ledger snapshot and an alignment confidence annotation entry stream. The time-based ledger snapshot is generated at a uniform time window granularity and includes at least the following fields: mapping status of each data source, valid interval, alignment confidence level, jitter counter summary, out-of-order counter summary, missing counter summary, jump counter summary, recent anchor list summary, and state change event reference. It is written into the session track using the window identifier as the index key. The alignment confidence annotation entry stream adds fields such as alignment confidence level, mapping status, valid interval marker, recent release reason summary, and arrival delay grading marker to each buffered ready entry. The arrival delay grading marker can be selected as three levels: normal, slightly late, and severely slightly late. The grading boundary is taken as 1 times and 2 times the arrival delay threshold. This boundary setting is consistent with the arrival delay threshold caliber and is fixed in the rule version. When the session enters the paused alignment state, the ledger side marks the alignment confidence level as unavailable and records the pause reason code. After the pause is lifted, the mapping status switches to pending confirmation and re-enters the confirmation observation window.

[0022] S4: Generate a unified time window according to the time window strategy, perform evidence admission determination, evidence consistency constraint verification, and conflict adjudication on the items within the window, and record the window timeline and adjudication items. The specific implementation is as follows: After the unified time-based ledger is stably output at a unified time window granularity, a unified time window is formed and evidence governance within the window is performed. This enables multi-source entries to form a window-level evidence set with clear boundaries, consistent standards, and traceability on the same session timeline. This provides a verifiable window header, evidence list, constraint verification record, and adjudication record for subsequent health assessment input encapsulation. The unified time window is based on the session timeline. The window record must include at least the following fields: window identifier, session identifier, window start point, window end point, trigger type, window credibility level, window coverage level, list of participating data sources, and list of original window references. The window identifier is composed of a session identifier and a window sequence number. The session trajectory is read in every time the window sequence number increases by 1 within the session. When the window sequence number is missing or rolled back, the adjudication output for that window is stopped and an abnormal event of the window number is written. The abnormal event must include at least the following fields: abnormal window identifier, most recently valid window identifier, and abnormal reason code, to avoid the adjudication record from breaking the reference chain of the input package. The start and end points of the window are determined by the time window policy version and fixed when the session is created. The trigger path of the time window policy can be a timed window or an event window. The timed window is generated by rolling according to the window duration and step size. The window duration is 120 seconds to 600 seconds, and the step size is 10 seconds to 30 seconds. The above range is determined with reference to the minute-level change characteristics of the power distribution operation status, the combination of station reporting cycles, and the processing delay budget, so that the window covers multiple reporting cycles of electrical parameters and environmental control and controls the processing frequency. The event window is generated when an anchor event marked as triggerable by the evidence rule version is detected. The range of the anchor event is consistent with the anchor availability mark in the data source list. The trigger detection takes the entry in the admission evidence list with the evidence role as the main evidence and the event direction as the trigger source. The trigger time is the session time axis mark point of that entry. The event window is generated by moving forward from the trigger time. The event window is then extended, with the extension period ranging from 10 seconds before the event to 120 seconds after the event. This range is determined by referencing the resolution of secondary equipment event records, the timing of station-end records of switching and protection actions, and the response lag interval of the environmental control actuator. This range is used as a time window strategy version parameter for verification. To reduce duplicate decisions in the same time period and take into account the station-end processing delay budget, an event window and time window overlap and merge rule is set. The overlap and merge threshold is set when the ratio of the intersection duration to the shorter window duration exceeds 30%. Both the intersection duration and the window duration are calculated on the session timeline and written into the window record. After merging, the window start point is the earlier of the two, and the window end point is the later of the two. The trigger type record is a composite trigger. The window record includes at least fields such as the original window reference list. The original window reference list is written into the merged window identifier set to verify the merging source and merging scope. The window credibility level and window coverage level are determined according to rules and written into the window header when the window is generated. The rules are written into the evidence rule version with the clause number and remain consistent within the session. The window credibility level is jointly determined by the alignment confidence level snapshot of the participating data sources and the coverage of key evidence. The key evidence value is the main evidence type entry, which includes at least the switch opening and closing event and protection action event entries. The key evidence completeness value is the presence of a main evidence entry that matches the trigger type within the window, or the presence of a most recent state entry of the main evidence type within the timed window with an alignment confidence level not lower than available. The most recent state entry is a similar main evidence entry with a duration not exceeding one window before the window start point, and its effective interval covers the window start point. The credibility rule can be set as follows: when the key evidence is complete and the alignment confidence level of the data sources participating in the key evidence is not lower than available, the window credibility is recorded as available or credible; when the key evidence is missing or the data source of the key evidence has an alignment confidence level of unavailable, the window credibility is recorded as conservative or unavailable. The window coverage level is determined according to the coverage rules of valid entries from various data sources within the window. Valid entries must include at least those that have passed the admission criteria and whose alignment confidence level is not unusable. Entries rejected due to duplication, missing fields, out-of-bounds fields, unknown code tables, or unusable alignment are not counted as valid entries. The allowable missing percentage for coverage is fixed in the evidence rule version. For example, electrical parameters are allowed to have no more than 10% missing, environmental control is allowed to have no more than 20% missing, and remote signaling and events are allowed to have no more than 1% missing. The above ranges are set in combination with the sensitivity of different data to causal determination and the stability of common station-level reporting, and are consistent with the station-level parameters. The missing percentage is the ratio of the difference between the expected number of entries and the actual number of valid entries within the window to the expected number of entries. For example, the expected number of entries is calculated from the reporting cycle or heartbeat interval in the data source list, and the expected number of entries range, the actual number of valid entries, and the list of data source identifiers affecting the data source are recorded in the header of the window. When remote information and event changes trigger reporting, the expected number of entries is not calculated from the cycle, but is determined according to the minimum required number of entries threshold of the main evidence and key remote information. The threshold is fixed in the evidence rule version and given by the missing and exceeding clause. When the coverage is not satisfied, a coverage limited or unsatisfactory mark is written in the header of the window, and a trigger reason field is written. The trigger reason includes at least the missing data source identifier list, missing segment reference, etc. The missing segment reference is taken from the missing segment summary reference identifier in the time base ledger snapshot, so that the coverage determination can be recalculated under the same snapshot caliber. Evidence governance within the window is based on the version of the evidence rules, including a list of admissible evidence, a list of rejected evidence, constraint verification records, and adjudication items. Evidence admission is based on clause numbering, and clauses must include at least field integrity conditions, value range legality conditions, alignment confidence conditions, duplicate processing conditions, missing processing conditions, and release reason conditions. Alignment confidence conditions remove items with an alignment confidence level of unusable from the adjudication set. Release reason conditions downgrade the role of evidence in cases of incomplete or timed release. Evidence roles can be selected as primary evidence, secondary evidence, or circumstantial evidence. The release reason marker and quality mark carried by the item... The record area can serve as an admission criterion, allowing auxiliary evidence to be downgraded to circumstantial evidence, or circumstantial evidence to be retained without entering the adjudication. The reason code for retention is recorded in the list of rejected evidence. The list of admitted evidence should include at least the following fields: item identifier, data source identifier, evidence role, admission clause number, evidence confidence mark, and admission time mark. The list of rejected evidence should include at least the following fields: item identifier, rejection reason code, and rejection clause number. The rejection reason code can include fields such as missing field, field out of bounds, unknown code table, alignment unavailable, duplicate discard, and missing / exceeding limits. Reviewers can reproduce the admission boundary based on the clause number and reason code. The verification of evidence consistency constraints is conducted on the set of admission evidence. Constraint clauses can be categorized into three types: timing constraints, state consistency constraints, and correlation constraints, and are fixed in the evidence rule version by clause number. Timing constraints regulate the sequential relationship between switch opening / closing events and protection action events, with action tolerances ranging from 50 milliseconds to 500 milliseconds, and commonly used configurations from 100 milliseconds to 300 milliseconds. Action tolerances are set in tiers according to link type and assigned tier numbers. Tier boundaries are referenced to the arrival delay statistical distribution of the link and the device event record resolution. Verification records are written with the adopted tolerance tier number. State consistency constraints regulate that after a switch is closed, the loop current is not zero or not lower than the minimum current threshold during the stable holding time. The minimum current threshold is 1% to 5% of the equipment's rated current. The equipment's rated current is derived from... Originating from asset ledgers or device setting area records and solidified in the data dictionary mapping version; the stable holding time is set to 500 milliseconds to 2 seconds, with this range referencing the electrical parameter sampling period, load transient attenuation characteristics, and sampling jitter range, and is executed as a rule parameter; the associated constraints are used to regulate the reasonable delay relationship between environmental control start / stop and temperature and humidity changes, with a delay range of 5 seconds to 60 seconds, which refers to the environmental control execution response time and the thermal and humidity inertial characteristics of the enclosure, and is executed as a rule parameter; when any constraint clause is violated, a conflict candidate entry is generated. The conflict candidate entry includes at least the following fields: conflict type, related evidence list, trigger clause number, window identifier, key evidence marker, and aligned confidence snapshot reference. The key evidence marker is used to distinguish whether the conflict involves the main evidence and is used for subsequent rollback actions. Conflict identification and adjudication are executed according to the adjudication rule version. The adjudication entry uses conflict type, which can be state conflict, timing conflict, amplitude conflict, or missing conflict. Adjudication employs a priority chain and a rollback strategy. The priority chain prioritizes primary evidence over secondary evidence, secondary evidence over circumstantial evidence, and within the same evidence role, those with higher confidence levels take priority. Those marked as having complete payload integrity take priority, and those marked as having normal release reason take priority. If still unable to distinguish, the order is sorted by the retransmission marker on the link, with non-retransmission taking priority over retransmission to reduce timing deviations introduced by out-of-order retransmission. The adjudication output forms an adjudication entry, which includes at least the following fields: adjudication identifier, window identifier, conflict type, list of accepted evidence, list of rejected evidence, adjudication level, adjudication clause number, and rollback action code. The coding options are: extended window observation, triggered supplementary data collection, marked for verification, and marked as unavailable. The extended window observation duration ranges from 30 to 600 seconds, with a common configuration of 60 to 300 seconds. This range matches the requirements for anomaly persistence confirmation and session resource budget, and is written to the window record. The maximum number of supplementary data collections triggered is 1 to 3. This range is used to limit cyclic supplementary windowing and retain necessary secondary observation opportunities. The maximum number is executed according to the time window strategy version parameters and accumulated in the window record, while the original window reference list is saved. After the rollback action is triggered, it is written to the window record, and the unresolved conflict marker is written back to the unresolved conflict marker area of ​​the time base ledger according to the window identifier and the list of participating data sources, so that the alignment confidence level rule can reference the marker and trace back to the corresponding window. After the ruling is completed, a window timeline is generated and appended to the session track. Timeline entries must include at least the following fields: event type, event direction, event occurrence point marker, source data source identifier, evidence role marker, ruling association identifier, and trigger clause number. During window processing, processing latency is checked against the session resource budget, with a maximum processing latency of 300 to 2000 milliseconds, determined by referring to the station's real-time alarm response requirements and processing resource limits. When processing latency exceeds the maximum processing latency, the verification scope is reduced according to evidence role priority, first reducing the verification of circumstantial evidence, then reducing the verification of supplementary evidence, and the data is recorded. Source degradation flag; when the window credibility is unavailable and exceeds the continuous window threshold for consecutive times, a session-level pause output is triggered. The continuous window threshold is 3 to 10 windows, and the commonly used configuration is 4 to 6 windows. This range is determined with reference to the duration of short-term fluctuations in the link and the requirements for suppressing misjudgments. During the pause, fields such as window header, coverage, credibility, and trigger reason are still written. The adjudication entry list is set to empty and the pause reason code is written. The recovery conditions are executed according to the recovery observation window parameters in the evidence rule version, and the stable output flag of the time base ledger snapshot is used as the standard. After the condition is met, the adjudication output is restored and the window timeline continues to be generated.

[0023] S5: Output the health assessment input package and archive the session trace, write back the mapping snapshot, missing segment summary, resource summary and parameter candidate record, specifically as follows: After a window record is generated, the window-level output is encapsulated and written to storage in the form of a health assessment input package, while the session trajectory and parameter candidate records are archived in an append-only manner. The header of the health assessment input package includes at least the following fields: session identifier, window identifier, window start and end boundaries, trigger type, window trust level, window coverage level, rule version reference, clause index area identifier, index key, caliber verification identifier, and checksum. The rule version reference is represented by a combination of version number and clause index area identifier. The clause index area identifier includes at least the following numbering information: admission clause number, constraint clause number, and adjudication clause number. During review, the number corresponds to the rule caliber fixed within the session. The index key set is used to uniformly locate various referenced objects within the health assessment input package. The index key includes at least the following fields: session identifier, window identifier, and record type code. The record type code is expressed as a 1-byte to 2-byte unsigned integer and is written to the enumeration table during session creation to solidify the parsing criteria. The value range of the record type code is determined by the number of referenced object categories and subsequent expansion reservations. Using 1 byte covers common categories and reduces storage overhead, while using 2 bytes ensures compatibility when adding referenced object categories. The enumeration table is set, for example, as follows: 01 Mapping Snapshot, 02 Missing Segment Summary, 03 Resource Summary, 04 Admission Evidence List, 05 Rejection Evidence List, 06 Decision Entries List, 07 Window Timeline, and 08 Load Summary Area. Each enumeration value corresponds one-to-one with the corresponding object type and remains unchanged throughout the session's lifecycle. The index key is written to both the input package and the session trace, enabling the referenced object to be located and traced using the same index key even after migration across storage media, database sharding, or offline archiving, thus maintaining consistency in the review process. The data source mapping snapshot references a static snapshot of the time-based ledger within this window. The snapshot includes at least the following fields: data source mapping status, valid interval, alignment confidence level, counter summary, recent anchor list summary, and state change event reference. The counter summary records the count values ​​within the window and marks them with grading labels for jitter counters, out-of-order counters, missing counters, and jump counters. The grading labels are divided into three levels: low, medium, and high. The grading boundaries are determined according to the threshold ratio rules fixed in the evidence rule version: low grade corresponds to a count value not exceeding one-third of the counter threshold; medium grade corresponds to a count value exceeding one-third but not exceeding two-thirds; and high grade corresponds to a count value exceeding two-thirds or triggering a mapping state switch event. The recent anchor list summary includes at least the following fields: anchor type, participating data source set, matching clause number, and anchor entry index, used for adjudicating the reference and review of anchor evidence by the adjudicating entries. The missing segment summary reference is used to associate the set of missing segment records. The missing segment record should include at least the following fields: missing segment start point, missing segment end point, judgment basis code, judgment threshold reference, list of affected data source identifiers, and evidence role adjustment record. The judgment basis code can be any of the following types: missing heartbeat, discontinuous sequence number, or arrival interval exceeding the threshold. The judgment threshold reference is associated with the missing processing clause number and its corresponding threshold caliber. The evidence role adjustment record should include at least the following fields: the type of evidence role being downgraded, the downgrade trigger clause number, and the set of indexes of the downgraded items. The resource summary references the window resource record, which includes at least the following fields: peak number of buffer entries, number of timeout releases, number of out-of-order releases, number of abnormal retention entries, processing latency flag, resource degradation flag, and resource over-limit flag. The processing latency flag is fixed into three levels according to the proportion of the maximum processing latency budget: the first level is a processing latency not exceeding one-quarter of the maximum processing latency, the second level is exceeding one-quarter but not exceeding three-quarters, and the third level is exceeding three-quarters. For example, when the maximum processing latency is 2000 milliseconds, the corresponding level boundaries are 500 milliseconds and 1500 milliseconds. The resource over-limit flag is used to give the result of continuous over-limit judgment. The continuous over-limit value is determined if one of the following conditions is met: the third level of processing latency occurs in M ​​consecutive windows, or the resource degradation flag is triggered in M ​​consecutive windows. M is 3 to 10 and is fixed in the session record. The commonly used configuration is 4 to 6, and its range is determined according to the duration of communication jitter and the station's real-time handling tolerance. The access evidence list, rejection evidence list, and adjudication item list are structured data that reference each other through index keys and item indexes. Items in the access evidence list include fields such as item identifier, data source identifier, evidence role, access clause number, release reason marker, alignment confidence level marker, and load summary index. The load summary has fixed minimum necessary fields according to load category, with the same unit caliber. The load category can be one of instantaneous value, change event, alarm event, statistical segment, or waveform segment. The instantaneous value summary includes fields such as measurement point identifier, value, unit, sampling quality marker, and source-side timestamp. The change event and alarm event summary includes fields such as event code, event direction, status code, alarm level, and source-side timestamp. The protection or fault event summary includes fields such as action type, action phase or loop identifier, action direction, action result code, and source-side timestamp. The environmental control summary includes fields such as equipment status code, temperature, humidity, over-limit marker, and source-side timestamp. The partial discharge summary includes fields such as level code, amplitude, channel identifier, status code, and source-side timestamp. When a site does not have a certain type of field, a missing marker is used as a placeholder, and the field name is no longer retained. Each item in the rejected evidence list must include at least the item identifier, data source identifier, rejection reason code, rejection clause number, rejection time marker, ruling identifier, conflict type, trigger clause number, index of accepted evidence list, index of rejected evidence list, ruling level, rollback action code, and downgrade source marker. The ruling level is divided into three types: ruled, conservative ruling, and unresolved. The rollback action code is divided into four types: extended window observation, triggered supplementary collection, marked for review, and marked as unavailable. The window timeline corresponds to the ruling item index through the ruling association identifier. Timeline items must include at least the event type, event direction, event occurrence point marker, source data source identifier, evidence role marker, ruling association identifier, and trigger clause number. The session trajectory uses session-level archiving. After the input packet is written, the session header record, window sequence record, time base ledger status change record, abnormal event record, adjudication entry sequence, resource degradation record, parameter candidate record, and other trajectory entries are concatenated in chronological order and appended to the trajectory storage. The trajectory storage has a retention period parameter, ranging from 30 to 180 days, with a commonly used configuration of 60 to 120 days. This range matches the operation and maintenance work order closed-loop cycle and the monthly to quarterly review cycle, and is constrained by the site storage budget. The append write strategy does not overwrite existing trajectories, but only adds new entries, and writes a checksum field for each trajectory entry for integrity verification. The checksum value is a checksum value of no more than 64 bits, with a checksum type code attached. The checksum type code can be selected as cyclic redundancy check or digest truncation. The checksum is generated according to the fixed field sequence of the trajectory entry. The fixed field sequence is fixed and written to the session header when the session is created. Each append write also writes a reference to the checksum of the previous entry to form a chained checksum relationship. The write-back mechanism and trajectory archiving are completed in the same batch. The write-back object is the parameter candidate record rather than directly modifying the effective parameter, in order to maintain consistency within the session and support review and recalculation. The generation of candidate parameters is driven by trigger conditions, which are any one of the following: frequent mapping downgrades, frequent freeze-rebuilds, frequent missing segments, frequent unresolved conflicts, and frequent resource downgrades. The frequency threshold is fixed according to the window counting caliber. If the same type of event occurs 3 or more times within 10 consecutive windows, it is judged as frequent. The observation span is consistent with 10 times the window duration and is written into the session record. The candidate parameter record includes at least the parameter name, parameter meaning, original value, candidate value, trigger window list, trigger clause number list, statistical caliber identifier, suggested effective conditions, and suggested rollback conditions. The statistical caliber identifier is used to fix the statistical denominator and statistical period to avoid inconsistencies in the caliber during review. The candidate value for buffer wait time is increased by 2 to 5 seconds from the original value. This candidate is generated when the timeout release ratio exceeds 10% of the single-source entries and the peak number of buffered entries does not exceed 80% of the single-source queue limit for three consecutive windows. The timeout release ratio is the ratio of the number of timeout releases to the number of ready buffered entries for that data source in that window. The above threshold is configured to balance the jitter absorption capacity of retransmission and the safety margin of queue occupancy. The candidate value for missing segment judgment threshold is adjusted from 5 times the heartbeat interval to 8 times. This candidate is generated when the heartbeat missing segment judgment is triggered for three consecutive windows and the proportion of retransmission marked entries exceeds 5%. The proportion of retransmission marked entries is the ratio of the number of retransmission marked entries to the total number of successfully parsed entries for that data source in that window. The above threshold is set based on the principle that missing segment judgment should be more conservative in retransmission scenarios to reduce false positives. The candidate value for action tolerance is to adjust the action tolerance from 200 milliseconds to 500 milliseconds for specific link types. A candidate parameter is generated when timing conflicts account for more than 50% of similar conflicts within 10 consecutive windows and the conflicts mainly originate from the same link type data source. The number of candidate entries for the same type of conflict is taken as the threshold. The above threshold is selected based on the impact of link latency differences on timing consistency. It is recommended that the effective condition be that the candidate parameter is triggered in the next 3 consecutive sessions and the confidence level of the triggering window is not lower than the corresponding level of the baseline session before the candidate is generated. The baseline session is the most recent session or the most recent 3 sessions before the candidate is generated and is fixed in the candidate parameter record. It is recommended that the rollback condition be that the number of unresolved conflicts increases or the window unavailability ratio increases beyond the threshold after the candidate parameter is effective. The window unavailability ratio is the ratio of the number of unavailable windows to the total number of windows within N consecutive windows, where N is 10 to 50 and is written to the session record. The rollback threshold is 5% to 20%, and the commonly used configuration is 10%. The configuration is based on the operational stability constraints and the tolerance boundaries for false alarms and false negatives. When a review request occurs during the retention period, the session trajectory, health assessment input package, and the referenced object pointed to by its index key are retrieved according to the session identifier for reproduction verification. When the rule version reference is missing or the clause index area is incomplete, an unreproducible flag is written and a version missing alarm entry is generated and archived. When a resource limit exceeding flag occurs, a resource budget candidate adjustment record is generated and used as a candidate version input in the next session creation stage. The candidate version is still subject to the suggested effective conditions and suggested rollback conditions to avoid caliber drift. The session end condition can be one of three categories: the timed session reaches the maximum session duration, the event session observation period ends and no new key events are added in the continuous window, or the manual session receives an end instruction. The maximum session duration is 10 minutes to 24 hours, and the commonly used configuration is 30 minutes to 6 hours. This range is determined in combination with the coverage period required for anomaly tracking and the site storage budget. When the session ends, an end record is written. The end record includes at least the end time, end reason code, number of windows, key anomaly summary code, number of candidate parameters, and check code. The end record check code is included in the trajectory chain.

[0024] The technical solution of this embodiment takes a distribution network box-type substation as an application scenario. The station end simultaneously receives multi-source data such as electrical parameter sampling, remote signaling of switch quantities, protection events, environmental control temperature and humidity, and partial discharge status. When the session is created, the data source list, data dictionary mapping, time window strategy, and evidence rule version are frozen, and a session identifier is generated and placed in the session header, so that the records in the session can be traced by a unified standard. The access end converts different protocol messages into running entries, registers the source side timestamp, sequence number, start count, and arrival timestamp, and verifies the legality of required fields and value ranges. After the verification is passed, the entries enter the out-of-order buffer of the data source queue. According to the sequence number continuity and waiting time, the normal release, timeout release, or out-of-order release is selected for output, and duplicate, out-of-bounds, and other data are written. Quality markers such as unknown code tables are retained; a session timeline mapping is established based on the heartbeat entries and cross-source jointly observable anchor events, maintaining a timeline ledger; when jitter, out-of-order, missing, or restart conditions occur, the mapping is downgraded, frozen, rebuilt, and the alignment confidence level and valid interval are marked; after the ledger stabilizes, a unified time window is generated according to the strategy, and evidence admission judgment, time sequence and state consistency constraint verification, and conflict adjudication are completed according to the clause number, with adjudication entries and window timelines; after the window ends, the health assessment input package is encapsulated to archive the session trajectory, and mapping snapshots, missing segment summaries, resource summaries, and parameter candidate records are written, and the reproduction is verified with the help of version references and checksum chains; when frequent downgrades or resource limits are exceeded, only candidate parameters are generated that conditionally take effect for subsequent sessions.

[0025] It should be noted that this invention can be deployed on the device itself to realize embedded applications, or it can run on a PC or other terminal with a user interface, thereby meeting various hardware environments and usage requirements.

[0026] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented in software, the above embodiments can be implemented in whole or in part by a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions of the embodiments of this application are implemented in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted wirelessly or wiredly from one website, computer, server, or data center to another website, computer, server, or data center. Wired methods include optical fiber, twisted pair, coaxial cable, etc. Wireless methods include infrared, microwave, etc. Available media include any available media that can be accessed by a computer or data storage devices such as servers and data centers that contain one or more sets of available media. Available media can be magnetic media (floppy disks, hard disks, magnetic tapes), optical media (DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0027] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0028] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for data acquisition and health assessment of prefabricated substation operation, characterized in that, include: S1: Create a running session, freeze the data source list, data dictionary mapping, time window strategy and evidence rule version, generate a session identifier and write it to the session record; S2: Access electrical parameters, remote signaling, events, environmental control and partial discharge data in a unified entry format, register source-side timestamp, sequence number, start count and arrival timestamp, perform field verification and enter the buffer queue; S3: Maintain the time baseline ledger based on heartbeat events and anchor events, perform mapping establishment, downgrading, freezing and reconstruction according to trigger conditions, update the valid interval and mark the alignment confidence level; S4: Generate a unified time window according to the time window strategy, perform evidence admission judgment, evidence consistency constraint verification and conflict adjudication on the items within the window, and record the window timeline and adjudication items; S5: Output health assessment input package and archive session traces, write back mapping snapshots, missing segment summaries, resource summaries and parameter candidate records.

2. The method for data acquisition and health assessment of a prefabricated substation according to claim 1, characterized in that, Create a running session, freeze the data source list, data dictionary mapping, time window policy and evidence rule versions, generate a session identifier and write it to the session record, including: Sessions are created by timed triggers, event triggers, or manual triggers, and session identifiers and caliber verification identifiers are generated according to site identifier, date, session sequence number, and trigger code; Freeze the list of data sources, data dictionary mappings, time window strategies and evidence rule versions, and fix the version number, mapping number, threshold number, snapshot number and scope of effect; After writing the resource budget and enabling threshold triggering, it enters the access state.

3. The method for data acquisition and health assessment of a prefabricated substation according to claim 1, characterized in that, Access electrical parameters, remote signaling, events, environmental control and partial discharge data according to a unified entry format, and register the source-side timestamp, sequence number, start count and arrival timestamp, including: The access end converts multi-protocol messages into running entry records, registers the source-side timestamp, source-side sequence number, source-side start count and arrival timestamp, and writes the link batch and retransmission flag; When the source-side startup count is unavailable, the restart criterion is that the chain break exceeds the preset threshold and the source-side sequence number wraps around.

4. The method for data acquisition and health assessment of a prefabricated substation according to claim 1, characterized in that, Perform field validation and add the data to the buffer queue, including: The entries are sequentially subjected to structural and value range verification, and are rejected, downgraded, or retained according to the clauses corresponding to the key fields. Validated running entries are entered into the out-of-order buffer queue according to the data source, and are output according to normal release, timeout release, or out-of-order release, and the release reason is recorded. When the timestamp rollback exceeds the rollback threshold, the system switches to paused alignment and releases the pause once the stability condition is met. Write an investigation flag when the abnormality rate exceeds the abnormality rate threshold.

5. The method for data acquisition and health assessment of a prefabricated substation according to claim 1, characterized in that, The time-based ledger is maintained based on heartbeat events and anchor events, including: Maintain the time-based ledger by creating append-written ledger entries for each data source; The session timeline is constructed using the arrival timestamp of the access point, and the corresponding arrival timestamp is not used as a marker when the session is in a paused alignment state. Anchor point matching is based on the consistency of event type, device location, and event direction, combined with timing tolerance determination, and the matching clause number and alignment confidence level are written into the ledger entry.

6. The method for data acquisition and health assessment of a prefabricated substation according to claim 1, characterized in that, Mapping is established, downgraded, frozen, and rebuilt according to trigger conditions; valid intervals are updated and aligned confidence levels are marked, including: A state machine is used for mapping establishment, degradation, freezing, and reconstruction. Within the confirmation observation window, mapping is confirmed when consecutive state entries reach the confirmation threshold; When the jitter count, out-of-order count, or missing count exceeds the threshold, it switches to degrade mode. When the restart criterion is met, a freeze is triggered and reconstruction begins. At the same time, the endpoints of the valid interval are updated, and a ledger snapshot indexed by window identifier and aligned confidence label entries are generated.

7. The method for data acquisition and health assessment of a prefabricated substation according to claim 1, characterized in that, Generate a unified time window according to the time window strategy, including: Based on the time window strategy, determine the start and end points of the window on the session timeline, and generate a timed window by timed scrolling or an extended window by event triggering. When the window overlap exceeds the merge threshold, a merge is performed and the source window reference is recorded. Write the window number, credibility flag, and coverage flag for each window; When a missing window number or a rollback is detected, an abnormal number record is written and the generation of the adjudication record for that window is stopped.

8. The method for data acquisition and health assessment of a prefabricated substation according to claim 1, characterized in that, Perform evidence admission determination, evidence consistency constraint verification, and conflict resolution on the items within the window, and record the window timeline and resolution items, including: Within the window, the entries are assessed for field completeness, value range validity, and alignment confidence based on the clause number, and the role of evidence is determined accordingly. Conflict candidates are generated based on timing constraints, state consistency constraints, and association constraints. The decision priority is determined based on the evidence role, alignment confidence and release reason, and decision entries and rollback action codes are generated and written to the window timeline record.

9. The method for data acquisition and health assessment of a prefabricated substation according to claim 1, characterized in that, Output health assessment input packets and archive session traces, including: The window record is encapsulated into an input package, which contains the rule version number, clause index identifier, caliber verification identifier, index key and verification code. The index key consists of a session identifier, a window identifier, and a record type code, and is associated with the location mapping snapshot, evidence list, adjudication list, timeline, and payload summary area; When archiving session tracks, write a checksum to each track entry and write a reference to the preceding checksum.

10. The method for data acquisition and health assessment of a prefabricated substation according to claim 1, characterized in that, Write-back mapping snapshot, missing segment summary, resource summary, and parameter candidate records, including: Record mapping snapshots, missing segment summaries, and resource summaries during session archiving, and record parameter candidate entries; The missing segment summary includes the judgment criterion code, threshold citation identifier, and evidence role adjustment record; Resource summary write buffer peak value, release count, latency grading identifier, and continuous over-limit judgment criteria; Parameter candidate entries are generated based on event frequency thresholds and statistical criteria, and suggested effective conditions and suggested rollback conditions are written into them.

Images