A high-reliability redundant communication method for a wind farm industrial Ethernet switch

Abstract

The application discloses a high-reliability redundant communication method of a wind farm industrial Ethernet switch and particularly relates to the technical field of wind farm industrial communication, and is used for solving the problems of inaccurate communication state judgment and chaotic message identification in the existing redundant link switching process. Through collection of link states, error code changes, adjacency changes, time delay drift and queue backlog information of each redundant port of a target switch, a switching influence event fingerprint is generated, and under the switching influence state, a service segment identifier, a path identifier, a sequence identifier and a window identifier are established for the arriving messages, path screening, message collection and verification, isolation and merging and recovery consistency verification are further performed, so that the abnormal communication segment under the switching disturbance can be accurately identified, the interference of abnormal messages on the service transmission judgment can be inhibited, and the effects of keeping the consistency of the active path, the service sequence, the processing queue and the output result before and after are achieved.

Description

Technical Field

[0001] This invention relates to the field of industrial communication technology in wind farms, specifically a highly reliable redundant communication method for industrial Ethernet switches in wind farms. Background Technology

[0002] Industrial Ethernet switches in wind farm substations typically handle the aggregation and forwarding of wind turbine control commands, operational status data, and fault information. To ensure communication continuity within the substation, existing systems generally employ dual-link, ring network, or other redundant communication methods. In the event of link anomalies, port fluctuations, equipment switching, or short-term disturbances, communication switching is completed through redundant paths. While existing technologies can meet basic communication requirements under normal steady-state conditions, in the actual operating environment of wind farms, link switching is not an ideal instantaneous process. It is often accompanied by port jitter, bit error fluctuations, short-term changes in adjacency relationships, transmission delay offsets, and queue backlogs. In such cases, the same service message may arrive along the old and new paths sequentially, or experience delays, duplication, or backflow, leading to complex message timing and path relationships received by the switch during the switching phase.

[0003] In existing technologies, the handling of redundant communication handover mostly focuses on link reachability assessment, port primary / backup switching, or network-level path recovery, typically using port connectivity status, link interruption events, or simple timeout rules as the basis for handover. For service message processing after handover, it is often assumed that arriving messages can directly participate in subsequent control, status, or fault information transmission, or only partial corrections are made through single message deduplication, buffer delay, and sequence number checks. While this approach works to some extent in general network environments, in wind farm booster station scenarios, due to the parallel transmission of control, status, and fault messages, the large number of devices within the station, the complex types of link disturbances, and the tendency for message arrival order to interleave during handover, it is difficult to accurately distinguish which messages belong to the current valid path transmission results and which belong to old path remnants, cross-flow, duplicate arrivals, or windowed delays, based solely on a single link event or single message characteristic. This can easily lead to inaccurate judgment of the current communication status, confused service segment associations, and inconsistent processing results.

[0004] Therefore, existing technologies lack a mechanism for continuously identifying, screening, verifying, and converging abnormal communication segments during redundant link switching of industrial Ethernet switches in wind farms, based on the relationship between port status changes and message timing during switching disturbances. This makes it difficult to reliably distinguish abnormal messages under the influence of switching from valid messages, thereby affecting communication status determination, service processing continuity, and recovery consistency after switching exit. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a highly reliable redundant communication method for industrial Ethernet switches in wind farms, thereby solving the problems mentioned in the background section.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a highly reliable redundant communication method for industrial Ethernet switches in wind farms, comprising:

[0007] S1. Collect link status, bit error rate changes, adjacency changes, delay drift and queue backlog information of each redundant port of the target switch, generate handover impact event fingerprint, and determine the current communication status accordingly.

[0008] S2. When the current communication state enters the handover impact state, establish service segment identifier, path identifier, sequence identifier and window identifier for the arriving message to form a service communication segment table.

[0009] S3. Based on the business communication segment table, perform path screening on the arriving messages, identify active path messages, old path residual messages, cross-flow messages, duplicate messages, and windowed late messages, and assign them to the corresponding processing queues.

[0010] S4. Based on the business communication segment table, perform sequence continuity verification, path uniqueness verification, and segment correspondence verification on the messages in each processing queue to determine the message acceptance result;

[0011] S5. Isolate and merge unaccepted messages, and output control results, status results, and fault results based on the message acceptance results.

[0012] S6. When the event fingerprint of the switching event meets the exit conditions, perform a recovery consistency check on the activity path, business sequence, processing queue, and output results. After the check passes, remove the window flag and restore normal communication.

[0013] Furthermore, S1 includes:

[0014] The event record is formed by combining the number of link state mutations, cumulative bit error changes, number of adjacency changes, number of delay drift out-of-bounds, and duration of queue backlog for each redundant port within the same observation window in a fixed field order.

[0015] When at least two disturbances are met within the same observation window, a switching impact event fingerprint is generated based on the event log.

[0016] Furthermore, S2 includes:

[0017] When establishing a service segment identifier, it is merged with the preset service mapping table according to the service field summary, source address record, and destination address record of the arriving message;

[0018] Under the same service mapping result, a service segment identifier is formed based on consecutively arriving messages and the time interval between adjacent arriving messages not exceeding the segment retention duration;

[0019] The business segment identifier is formed by combining the observation window number, the business mapping result identifier, and the segment starting number.

[0020] Furthermore, S3 includes:

[0021] Under the same window identifier, arriving messages whose path identifier is consistent with the active path record locked at the current observation window starting point are identified as active path messages;

[0022] Arrival messages whose path identifiers within the same business segment are inconsistent with the active path record locked at the starting point of the current observation window, and whose arrival timestamps are later than the arrival time of the first active path message, are identified as residual messages of the old path.

[0023] Furthermore, S3 also includes:

[0024] Arrival messages that form a path state switch and switch back to an inactive path within the same service segment identifier are identified as cross-flow back packets.

[0025] Arriving messages with identical business field summaries under the same window identifier, the same business segment identifier, and the same path identifier, and whose time interval between adjacent arriving messages does not exceed the duplicate judgment duration, are identified as duplicate messages.

[0026] The cross-flow messages and duplicate messages are then assigned to the corresponding processing queues.

[0027] Furthermore, S4 includes:

[0028] Form a verification group based on the same window identifier, the same business segment identifier, and the same processing queue;

[0029] Perform sequence continuity verification on messages within the same check group;

[0030] The sequence continuity verification results are then matched one by one with the message registration numbers within the group.

[0031] Furthermore, S4 also includes:

[0032] Perform path uniqueness verification and fragment correspondence verification on messages within the same verification group;

[0033] When the sequence continuity check result, path uniqueness check result, and fragment correspondence check result corresponding to the same message registration number are all true, the message is determined to be accepted.

[0034] Furthermore, S5 includes:

[0035] Unaccepted messages are logically removed and written to the local isolation record area;

[0036] Unaccepted messages that have been written to the local isolation record area are merged into a result merge record based on the identification result and path identifier;

[0037] Control results, status results, and fault results are generated based on the fixed correspondence between the result merging records and the result type correspondence table.

[0038] Furthermore, S6 includes:

[0039] When the event fingerprint of the switching event meets the exit conditions, the activity path, business sequence, processing queue and output result corresponding to the event fingerprint number that has formed the exit establishment record are sequentially checked for recovery consistency.

[0040] When the activity path, business sequence, processing queue, and output result are all consistent under the same event fingerprint number, the window identifier is removed first, and then normal communication is restored.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] 1. By collecting link status, bit error rate changes, adjacency changes, delay drift, and queue backlog information of each redundant port of the target switch, a handover impact event fingerprint is generated. After the current communication state enters the handover impact state, service segment identifiers, path identifiers, sequence identifiers, and window identifiers are established for the arriving packets. Further, path screening, packet acceptance verification, isolation and merging, and consistency verification are performed on the arriving packets. This achieves the effect of accurately identifying abnormal communication segments affected by handover in redundant link handover disturbance scenarios, suppressing the interference of old path residual packets, cross-flow packets, duplicate packets, and windowed late packets on service transmission judgment, and maintaining the consistency of active paths, service sequences, processing queues, and output results.

[0043] 2. By isolating and merging unaccepted messages, and outputting control, status, and fault results based on the message acceptance results, the window flag is removed and normal communication is restored when the fingerprint of the switching impact event meets the exit conditions. This achieves the effect of orderly convergence of result output, traceability of abnormal records, clear state switching boundaries, and continuous and stable communication recovery process during the recovery process of redundant communication from a disturbed state to a stable state. Attached Figure Description

[0044] Figure 1 A block diagram of the overall structure of a high-reliability redundant communication system for industrial Ethernet switches in a wind farm.

[0045] Figure 2A flowchart of a high-reliability redundant communication method for industrial Ethernet switches in wind farms;

[0046] Figure 3 Flowchart for switching impact event fingerprint generation and current communication status determination;

[0047] Figure 4 A schematic diagram illustrating the construction and path filtering of the business communication segment table;

[0048] Figure 5 Flowchart for restoring consistency verification and normal communication. Detailed Implementation

[0049] 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.

[0050] Example: Combined with Appendix Figure 1-5 This embodiment provides a highly reliable redundant communication method for industrial Ethernet switches in wind farms, including:

[0051] S1. Collect link status, bit error rate changes, adjacency changes, latency drift, and queue backlog information for each redundant port of the target switch, generate a handover impact event fingerprint, and determine the current communication status accordingly. The specific implementation is as follows:

[0052] The target switch management side, located in the communication cabinet of the wind farm substation, continuously collects link status, bit error rate changes, adjacency changes, latency drift, and queue backlog information for each redundant port. The execution entity is the acquisition control program running on the onboard management unit of the target switch. The acquisition period covers normal power generation, wind turbine start-up and shutdown, collection line switching, in-station maintenance and connection switching, and short-term disturbance recovery. The acquisition interval is bounded by all physical ports of the target switch participating in the current redundant communication group. The redundant communication group is locked by the group identifier in the effective configuration file and the running port table. Ports that have been included in the group at the beginning of the observation window are included in the acquisition range of this observation window. Ports added in the middle of the observation window are included from the next observation window. Ports removed in the middle of the observation window are retained until the end of this observation window. Maintenance ports, mirror ports, disabled ports, ports with failed rate negotiation, and isolated ports with only one-way transmission and reception anomalies that are not included in the current redundant communication range are not included.

[0053] The link status is the port's connectivity record within the current sampling period, directly taken from the port's physical layer status register record and port event log. The record caliber is uniformly classified into three categories: connected, disconnected, and jitter-locked, with the unit being status codes, dimensionless. The sampling rhythm can be set from 100ms to 500ms, preferably 200ms. The single sampling timestamp accuracy is preferably not less than 1ms. The local time synchronization accuracy is the minimum resolution capability of the target switch's onboard time synchronization unit to uniformly allocate timestamps to each recording object, and serves as the uniform execution caliber for time tolerance in this embodiment. A connected record is a record in which the physical layer remains connected and a valid frame is received within the current sampling period. Valid frame records are those where the port's received count increases and the frame verification passes. Damaged frames, mirrored frames, and maintenance test frames are not included. A disconnected record is a record in which the physical layer is disconnected and lasts for more than one sampling period. A jitter-locked record is a record in which connection and disconnection alternate twice within the same observation window and do not remain stable for two consecutive sampling periods.

[0054] Bit error rate changes are recorded as incremental changes in the bit error count of the same port within a continuous observation window. These changes originate from the port transmit / receive count register and the optical port diagnostic record. The optical port diagnostic record is the port physical layer operation record written periodically by the target switch's optical interface diagnostic function and is associated with the port transmit / receive count register in the same sampling period, measured in times. It is aggregated based on the difference between adjacent sampling points, prioritizing the use of the port cumulative count register. When a register record is missing, the optical port diagnostic record is used to fill the gap. The optical port diagnostic record filling is only used to maintain the continuity of bit error rate changes and does not change the original record source identifier. The filling record can participate in event fingerprint generation within the current observation window, but it must not overwrite existing register records. When the device restarts, the port resets, or the count wraps around, the identification condition is that the count at the previous sampling point is greater than the count at the current sampling point and a reset log exists simultaneously. The bit error rate change in that sampling period is recorded as unreliable and transferred to pending verification. No automatic filling is performed for missing counts. If two consecutive sampling periods are missing, the data processing stops.

[0055] Adjacency changes are the results of changes in the neighboring device identifier, neighboring port identifier, and adjacency duration obtained by the target switch from the port adjacency record. The record sources are the switch's port adjacency table and link establishment log. The port adjacency table is a local record table of the target switch that records the current port's neighboring devices and neighboring ports, and is continuously updated according to the port link establishment, disconnection, and reconstruction time, with units of identifier value and ms, respectively. They are only merged within the same redundant communication group. Changes in neighboring device identifiers, changes in neighboring port identifiers, and re-establishment of the same adjacency record after interruption with an interruption duration exceeding 1000ms are all considered adjacency changes. Instantaneous adjacencies formed during maintenance access with a duration of less than 3000ms are not included.

[0056] Delay drift is the offset of the round-trip time of a message on the same port relative to the reference value within a continuous observation window. The reference value is the stable operation reference record formed when the port has no sudden changes in link status, no adjacency changes, bit error rate changes below the locking threshold, and queue backlog retention time below the retention threshold after the current rule version takes effect. The unit is μs. The sampling rhythm is consistent with the link status. The stable operation reference record is formed by merging continuous valid samples of the same port and locking the version. Records during wind turbine shutdown maintenance, station communication self-test, and rule version switching transition periods are not used. If more than 3 consecutive null values ​​are collected, they will not be automatically filled and will be transferred to the waiting record. The reference band is the allowable offset range pre-written around the reference value in the rule version. It is executed according to the same standard for the same port and remains unchanged within the validity period of the current rule version. The queue backlog information is the queue length and retention time of the port's transmit queue and receive queue at the sampling time. It comes from the total queue statistics and register records of the switch port. The units are frames and ms, respectively. The retention time is recorded as the duration from the first frame entering the queue to leaving the queue. Broadcast traffic and multicast traffic are included in the statistics, while mirror traffic is not included.

[0057] After collection, the above information is first aligned according to timestamp, port identifier, and record source order. Alignment is based on the local time synchronization record of the target switch. When timestamps are the same but sources conflict, registered records take priority over log records. Then, noise reduction is performed. During noise reduction, only manual plug-in / plug test points corresponding to the maintenance work order period and operation account, as well as initialization gaps after device restart, are removed. Then, completion is performed. Completion is limited to single-point missing tests during periods of continuous and stable link status. The completion boundary is that the record sources of two consecutive sampling points are consistent, the time interval does not exceed one sampling period, and the status values ​​are consistent. Error changes, adjacency changes, jitter lockout status, and adjacency are also considered. The duration must not be automatically filled in; the processed records form a continuous observation segment of the port, which is written in chronological order to the local non-volatile recording area of ​​the target switch, and at the same time generates the rule version number, the acquisition program version number, the device serial number, the writing timestamp, and the operation account trace. Old version records must not overwrite new version records. Only appending is allowed within the same rule version. The lock threshold, hold threshold, and baseband are all executed with the threshold table corresponding to the current rule version as the sole execution caliber, and are permanently stored in the local non-volatile recording area of ​​the target switch and the synchronous recording area of ​​the station communication management host along with the rule version. The threshold table must not be switched within the same observation window.

[0058] When generating handover impact event fingerprints, the number of link state mutations, cumulative bit error changes, number of adjacency changes, number of delay drift out-of-bounds, and duration of queue backlog for each redundant port within the same observation window are combined in a fixed field order to form an event record. Only one event record is formed for the same device and the same observation window. When consecutive observation windows are established, the fingerprint number of the previous event is continued and the observation window sequence number is appended. Overwriting of the fixed record is not performed. Only the addition of supplementary related records is allowed. The port set identifier is a fixed identifier formed by concatenating all port identifiers in the current redundant communication group of the target switch in ascending order of port number when the observation window starts. It remains unchanged within the observation window.

[0059] The observation window can be set from 2s to 10s, preferably 5s. The threshold is determined based on the stable operating condition samples and disturbance operating condition samples collected during the on-site commissioning phase. The stable operating condition samples should have no less than 500 observation windows covering day and night periods, and the disturbance operating condition samples should have no less than 100 observation windows covering periods of link interruption, manual connection switching, port congestion, and adjacency reconstruction. The threshold takes effect along with the rule version after being confirmed by the on-site commissioning supervisor. A sudden change is defined as two alternating connections and disconnections in the link status within the observation window. A bit error rate disturbance is defined as the cumulative value of the bit error rate change reaching the lockout threshold. An adjacency rate disturbance is defined as two consecutive rewrites of the adjacency change within the same observation window. A time delay disturbance is defined as three consecutive sampling points exceeding the reference band. The three consecutive sampling points must be three sampling points that are adjacent in time sequence, have the same sampling rhythm, and have not switched recording sources. A queue backlog that lasts for more than the preset holding time is defined as a queuing disturbance. When at least two disturbance conditions are met within the same observation window, a handover impact event fingerprint is generated and written to the local event record area of ​​the target switch and the synchronization record area of ​​the station's communication management host. The writing format adopts a fixed field order message, with fields written in a fixed length or fixed separator order. Missing fields are reserved with a preset null value. Field positions cannot be deleted. The minimum field set includes device identifier, port set identifier, observation window start and end time, five types of original record summaries, rule version number, event fingerprint number, and writing result code. The writing result code includes at least six types: success, record missing, version inconsistency, timestamp conflict, duplicate writing, and storage area full. When a record is missing, a maximum of three retries can be made, with a retry interval of 200ms. If there are consecutive failures, the acquisition process of other redundant ports within the same observation window will not be interrupted, and the record of that port will be transferred to the verification queue.

[0060] When determining the current communication status based on the handover impact event fingerprint, the determination is only performed under the condition that all necessary fields in the current observation window are complete, the versions are consistent, and the time order is not reversed. The determination result is limited to two categories: normal communication status and handover impact status. The normal communication status corresponds to continuous record segments in the observation window that do not meet the conditions for the establishment of the event fingerprint. The handover impact status corresponds to continuous record segments that meet the conditions for the establishment of the event fingerprint and are maintained for at least 2 observation windows. The release condition is that in the subsequent two consecutive observation windows, the link status of each port involved in the triggering does not change abruptly, the bit error rate change is lower than the locking threshold, there is no effective adjacency change, the delay drift does not exceed the reference band, and the queue backlog maintenance time is lower than the maintenance threshold.

[0061] After the judgment result is formed, a timestamped status record is written to the local status area of ​​the target switch. This record is then sent to the next stage via a wired management link within the station, using a message with a fixed field order. The notification only includes the device identifier, judgment time, current communication status, event fingerprint number, and rule version number, without carrying the message classification results required by subsequent stages. The maximum delay for this step can be set to complete record solidification and status writing within 500ms after the end of a single observation window. Concurrency is determined by simultaneously collecting data from all redundant ports of the target switch. The idempotent strategy uses the same device identifier, the same observation window start and end time, and the same... The rule version number is written only once. An incomplete record arriving earlier must not overwrite a complete record arriving later, and a duplicate complete record arriving later must not overwrite a valid record arriving earlier. The queue for verification and the queue for supplementary records are retained for no less than 30 days after the current rule version expires. It is prohibited to participate in event fingerprint generation and current communication status determination before manual review is completed. The rule version expires when the new rule version is written to take effect and confirmed by the target switch to replace the old rule version. The security boundary is limited to collecting only communication operation records and device status records, without rewriting business message content, issuing control commands, or extracting records from other switches across stations.

[0062] In a preferred embodiment, a target switch in a wind farm booster station connects to 6 redundant ports. The sampling rhythm is set to 200ms, the observation window is set to 5s, the link state change threshold is 2 times, the adjacency change threshold is 2 times, the delay drift consecutive over-limit threshold is 3 sampling points, and the queue backlog holding threshold is 800ms. During a short-term disturbance of the collector line, the second port experiences 2 connection disconnections within 5s, the third port accumulates 128 bit error changes, the second and fourth ports each experience 2 adjacency rewrites, the delay drift of the third port exceeds the reference band for 4 consecutive sampling points, and the fourth port transmits queue backlog is held for 920ms. Within 320ms after the observation window ends, the system forms a handover impact event fingerprint and writes the current communication state as a handover impact state into the local state area. Alternatively, queue backlog information can be recorded using the percentage of frames not dequeued within a fixed period instead of queuing length and dwell time being recorded in parallel. As long as the recording source, time caliber, version caliber, and event fingerprint generation caliber are consistent with the other technical features of this step, this step can still be implemented.

[0063] S2. When the current communication state enters the handover impact state, service segment identifiers, path identifiers, sequence identifiers, and window identifiers are established for the arriving messages to form a service communication segment table. The specific implementation is as follows:

[0064] After the current communication state written in the previous step enters the handover impact state, the management side of the target switch located in the communication cabinet of the wind farm booster station establishes service segment identifier, path identifier, sequence identifier, and window identifier for the arriving messages. The execution entity is the message registration program running in the onboard management unit of the target switch. The effective period is limited to the continuous period during which the current communication state remains in the handover impact state. The applicable scope is limited to arriving messages that have completed full frame reception, passed verification, and been written to the local receive buffer area via the redundant communication group port of the target switch. The receive buffer area is the local buffer area of ​​the target switch used to temporarily store messages that have completed full frame reception. The message registration number is written according to the port reception order. It does not include maintenance test messages, mirror messages, equipment self-test messages, verification failure messages, truncated messages, and messages that have been marked as unreliable sources in the previous step. Broadcast messages, multicast messages, and unicast messages are all included when the aforementioned conditions are met.

[0065] The arriving messages are taken from the target switch port reception record, the redundant communication group record to which the port belongs, the current communication status record written in the previous step, and the handover impact event fingerprint record. The collection criteria are uniformly defined as message registration number, arrival timestamp, receiving port identifier, source address record, destination address record, message length, service field summary, and receiving sequence number, with units of count value, μs, identifier value, identifier value, identifier value, Byte, summary value, and count value, respectively. The message registration number is a unique number generated by the target switch in a local global incrementing order when the whole frame is written into the receiving buffer. The receiving sequence number is the queuing order number of the same receiving buffer in the current observation window. The collection rhythm is continuously registered according to the actual arrival rhythm of the messages, and the time tolerance is not higher than the local time synchronization accuracy of the previous step. The version criteria are consistent with the current rule version.

[0066] When establishing service segment identifiers, the service field summary, source address record, destination address record, and preset service mapping table of the arriving message are first merged. The service field summary is taken from the fixed field position defined by the field extraction table pre-written in the current rule version. The field extraction table and the service mapping table are fixed in the local non-volatile record area of ​​the target switch and take effect synchronously with the rule version. They cannot be switched within the same observation window. The field extraction table includes at least the field sequence number, field start position, field length, field value method, null value judgment rule, and effective rule version. If the service field is missing, the field length is insufficient, the field value is illegal, or multiple mappings of the same message are hit at the same time and have the same priority, the message should not be automatically filled in. The processing of the message should be stopped and the message should be transferred to the pending verification stage.

[0067] The business mapping table includes at least the mapping sequence number, the business field summary corresponding rule, the source address constraint, the destination address constraint, the business mapping result, and the effective rule version. During mapping, each item is compared according to the mapping sequence number. The matching stops when the first valid mapping is matched. The business mapping result is represented by a unique fixed identifier value under the current rule version. The name cannot be duplicated or reused within the same rule version.

[0068] The service segment identifier is formed by a set of consecutively arriving messages under the same service mapping result, where the time interval between two adjacent messages does not exceed the segment retention duration. The segment retention duration can be set from 1ms to 1000ms, preferably determined by a threshold table formed by the interval samples of consecutive messages of the same link during the field commissioning phase and fixed into the current rule version. When a message with a different service mapping result is inserted in the middle, the current service segment identifier ends, and subsequent messages with the same service mapping result re-form the starting number. Existing service segment identifiers cannot be continued across observation windows. The service segment identifier is formed by combining the observation window number, the service mapping result identifier, and the segment starting number in a fixed order. The segment starting number is the incrementing number when the same service mapping result in the same observation window first forms the service segment identifier, and a new starting number is formed when entering the next observation window. The fixed value is formed by concatenating fixed-length fields in order, and the field position remains unchanged within the validity period of the current rule version.

[0069] When establishing a path identifier, it is associated with the receiving port identifier of the arriving message, the port set identifier locked at the starting point of the observation window, and the fingerprint number of the current handover impact event. The path identifier is a fixed value formed by combining the receiving port identifier of the arriving message that actually enters the target switch with the position number of the port set to which it belongs in a fixed order. The position number of the port set is determined in ascending order of port number and remains unchanged within the same observation window. The fixed value is formed by concatenating fixed-length fields in order. The position of the fields remains unchanged within the validity period of the current rule version. If the receiving port identifier is missing, the port is not within the range of the current port set identifier, or the port set identifier is inconsistent with the current rule version, a path identifier shall not be generated and the process shall be transferred to the pending record.

[0070] When establishing sequence identifiers, consecutive sequence identifiers are formed by arranging the arrival timestamps and receiving sequence numbers of arriving messages under the same service segment identifier in ascending order. The sequence identifier records the order position within the same service segment, with the unit being a count value. The initial value is fixed at 1 and increases continuously only within the current service segment identifier. Arriving messages with the same timestamp are ranked first by the one with the smaller receiving sequence number. If a timestamp is missing, it cannot be filled in by estimated time. If the receiving sequence number is repeated, the first valid record is retained and the later record is appended as a duplicate associated record, but it does not occupy a new sequence identifier.

[0071] When establishing a window identifier, the handover impact event fingerprint number, observation window start and end time and current communication status record already written to the local status area in the previous step are directly called. Arrival messages whose actual arrival time falls into the corresponding observation window during the continuous period of the same handover impact state are assigned the corresponding window identifier. The window identifier is a fixed value formed by combining the event fingerprint number and the observation window sequence number under the event fingerprint. Window identifiers cannot be established when the current communication state has not entered the handover impact state. Messages arriving across the observation window boundary are assigned to the corresponding window identifier when they are written into the receive buffer as a whole frame.

[0072] Before establishing the above identifiers, the arriving message records are first aligned according to the arrival timestamp, receiving port identifier, and receiving sequence number. Then, noise reduction is performed according to the maintenance work order period, mirror port list, and device self-test record. After that, only single missing test points in the message length and destination address records are allowed to be filled in. The filling boundary is that the sources of adjacent records are consistent and the time interval does not exceed one message registration cycle. The message registration cycle is the maximum continuous duration allowed by the target switch to complete the registration of one arriving message, and is based on the upper limit of the registration duration written in the current rule version. It remains unchanged within the same observation window. The service field summary, receiving port identifier, arrival timestamp, receiving sequence number, service segment identifier, path identifier, sequence identifier, window identifier, and current communication status record must not be automatically filled in.

[0073] Upon completion, a business communication segment table is formed. The business communication segment table is a set of structured records that are continuously appended in the order of message registration. The minimum set of fields includes message registration number, business segment identifier, path identifier, sequence identifier, window identifier, arrival timestamp, receiving port identifier, rule version number, and write result code. Fields are written in a fixed length or fixed delimiter order. Missing fields are reserved in the position of a preset null value. Field positions cannot be deleted. The write result code includes at least 7 categories: success, field missing, mapping conflict, version inconsistency, duplicate registration, synchronization failure, and storage area full. Duplicate registration is limited to duplicate writing of the same message registration number.

[0074] The business communication segment table is stored in the local non-volatile record area of ​​the target switch and synchronously written to the segment registration area of ​​the station's communication management host. Successful local writing is considered a successful write in this step. Remote synchronization failure does not affect the effectiveness of the local segment table, but synchronization should be retried according to the original field order. Older version records must not overwrite newer version records. The same message registration number is only allowed to be written once. First-arriving complete records take priority; later-arriving duplicate complete records must not be overwritten. A maximum of three retries are allowed for writing failures, with a retry interval of 200ms. In case of consecutive failures, other arriving message registrations within the same observation window will not be interrupted, and failed records will be transferred to the verification queue. The verification queue and records to be supplemented only allow... Supplementing related records is permitted, but overwriting the already fixed business communication segment table is not allowed. Furthermore, supplementing related records must not alter the already written business segment identifier, path identifier, sequence identifier, and window identifier, and must be retained for at least 30 days after the current rule version expires. The maximum delay from when the arriving packet enters the receive buffer to when the business communication segment table is locally written can be set to 500ms. Concurrency capability is determined based on the simultaneous arrival of packets on all redundant ports of the target switch. The security boundary is limited to registering only the packet arrival record and its identifier result, without rewriting the original packet content, generating subsequent screening results, or extracting packet records from other switches across stations.

[0075] During on-site inspection, the consistency rate of service segment identifiers, the correspondence rate of path identifiers, the continuity rate of sequence identifiers, the correctness rate of window identifier attachment, and the success rate of segment table writing are used as inspection criteria. Preferably, no less than 1,000 arriving messages are sampled for inspection, covering different ports, different service mapping results, and no less than two observation windows. The samples are checked item by item using on-site packet capture records, target switch local registration records, and station synchronization records.

[0076] In a preferred embodiment, a target switch in a wind farm substation receives a total of 248 incoming messages from 6 redundant ports during a one-window observation period of the switching impact state. Among them, 36 consecutive messages from the same wind turbine control link with the same service field digest are formed when the fragment holding time is set to 20ms. The 12 messages received by the second port establish the same path identifier, and the 24 messages received by the fourth port establish another path identifier. Sequence identifiers 1 to 36 are formed according to the arrival timestamp and the receiving sequence number, and are uniformly attached to the window identifier corresponding to the current observation window. The service communication fragment table is locally written within 280ms after the observation window ends.

[0077] Alternatively, when the source address record and destination address record constraints are merged and written into the service mapping table in the current rule version, the service segment identifier can also be merged with the preset service mapping table based solely on the service field summary. As long as its version caliber, time caliber, and segment duration caliber remain consistent with the other technical features of this step, this step can still be implemented.

[0078] S3. Based on the business communication segment table, perform path screening on arriving messages, identify active path messages, old path residual messages, cross-flow messages, duplicate messages, and windowed late messages, and allocate them to the corresponding processing queues. The specific implementation is as follows:

[0079] The target switch management side, located in the communication cabinet of the wind farm booster station, performs path screening on arriving packets based on the service communication segment table already written to the local non-volatile record area in the previous step. The execution entity is the path screening program running in the onboard management unit of the target switch. Path screening is performed one by one on arriving packets, and is only performed on the service communication segment table records that have been written locally in the observation window corresponding to the current window identifier. It does not merge and compare across window identifiers. The effective period is limited to the continuous period during which the current communication status remains in the switching impact state. The applicable scope is limited to arriving packets that have completed the registration of service segment identifier, path identifier, sequence identifier, and window identifier and have been written to the service communication segment table. It does not include packet records in the pending verification queue, records to be supplemented, or packet records that have not been written locally to the service communication segment table.

[0080] During path screening, the message registration number, service segment identifier, path identifier, sequence identifier, window identifier, arrival timestamp, receiving port identifier, and rule version number are first read from the service communication segment table. Then, the current communication status, observation window start and end time, port set identifier, and handover impact event fingerprint number are read from the status area of ​​the previous stage. The reading rhythm is executed synchronously according to the rhythm appended to the service communication segment table, with the time tolerance not exceeding the local time synchronization accuracy. The version caliber is consistent with the current rule version. The alignment order is window identifier, service segment identifier, arrival timestamp, and sequence identifier in sequence. After alignment, only incomplete records with duplicate writing and manual test records that have been clearly marked by maintenance work orders are removed. The path identifier, window identifier, arrival timestamp, and sequence identifier are not automatically filled in. If the above fields are missing, the processing of the record is stopped and it is transferred to the pending verification stage.

[0081] When performing path filtering on arriving messages based on the business communication segment table, active path messages are first determined. Active path messages are those arriving within the same window identifier and the same business segment identifier, whose path identifier matches the active path record locked at the start of the current observation window, and whose arrival timestamp is greater than or equal to the start time of the current observation window and less than the end time of the current observation window. Active path records are taken from the pre-fixed active path determination table under the current rule version and the association result of the port set identifier at the start of the current observation window. An active path record is a single-value record formed by matching the active path determination table with the port set identifier under the current rule version, and is within the same observation window. The window serves as the sole basis for determining the active path. The active path determination table includes at least the port set identifier, active path record, scope of application for the service segment, and version of the effective rule. The scope of application for the service segment is taken from the set of service mapping result identifiers under the current rule version. The service mapping result identifiers are matched one-to-one with the service mapping result identifiers in the service segment identifiers. At the starting point of the same observation window, one active path record is obtained by matching the port set identifier. The record remains unchanged within the same observation window. Fluctuations in the port status during the process do not write back the already formed active path message identification result. If no active path record is matched, the service segment will not enter the active path message identification.

[0082] Continue to identify residual messages of old paths. Residual messages of old paths are those whose window identifier is consistent with the current observation window and whose service segment identifier is consistent with the active path message, but whose path identifier is not equal to the active path record in the current observation window and whose arrival timestamp is later than the arrival time of the first active path message in the same service segment. The physical meaning of residual messages of old paths is the message record that enters the target switch's receive buffer lagging behind the previous valid transmission path under the handover impact state. The previous valid transmission path is the path identifier that was registered before the first active path message in the same service segment, is different from the current active path record, and is the last one to form a valid arrival record. It is only identified when the current communication state remains under the handover impact state and the inactive path only forms a single late arrival. No new identification results are added after the observation window ends.

[0083] Continue to identify cross-path backflow messages. Cross-path backflow messages are arrival messages that, when checking the path identifier changes in the same service segment identifier according to the arrival time order, show continuous alternating backflow of inactive path, active path, and inactive path, or after continuous alternating backflow of active path, inactive path, and active path, re-enter inactive path. In addition, the same inactive path forms no less than 2 backflow registrations in the current observation window. Cross-path backflow messages are established when there are 3 path state switches in the same service segment, and the arrival message after the 3rd switch back to inactive path is used as the starting record. The aforementioned no less than 2 backflow registrations are judged by the 3 path state switches. If there is only a single inactive path arriving late and no further backflow is formed, it is not identified as a cross-path backflow message and the original identification result is retained.

[0084] Continue to identify duplicate messages. Duplicate messages are multiple arriving messages under the same window identifier, the same service segment identifier, and the same path identifier, with the first valid registration record as the comparison benchmark. These messages have the same service field summary, the same message length, and the time interval between the arrival of two adjacent messages does not exceed the duplicate judgment duration under the current rule version. The duplicate judgment duration is determined by the threshold table formed by duplicate arrival samples of the same link during the field commissioning phase and is fixed into the current rule version. It cannot be switched within the same observation window. Continuous service messages with consecutively increasing sequence identifiers and consecutive message registration numbers within the same service segment are not directly registered as duplicate messages due to the consistency of service field summaries. Under the same window identifier, the same service segment identifier, and the same path identifier, once the first valid registration record is determined, it cannot be replaced before the end of the service segment. Only the first valid registration record is retained as the benchmark for duplicate messages. Other arriving messages that meet the above conditions are registered as duplicate messages. The identification of duplicate messages must not change the registration order of existing active path messages, old path residual messages, and cross-loop messages.

[0085] Continue to identify late-arriving messages. Late-arriving messages are those whose entire frame is written to the receive buffer later than the end time of the observation window corresponding to the window identifier. The time when the entire frame is written to the receive buffer is used as the sole time reference, and the arrival timestamp is only used as an auxiliary record and is not used as a priority determination basis. The identification of the above 5 types of messages is performed in a fixed order: active path messages, old path residual messages, cross-flow messages, duplicate messages, and late-arriving messages. Once the previous identification is successful, the subsequent category comparison is stopped immediately. Only the first successful identification result is retained for the same arriving message, and no auxiliary identification results shall be registered for subsequent categories.

[0086] After identification, arriving packets are assigned to their corresponding processing queues: active path packets to the active path queue, old path remnant packets to the old path remnant queue, cross-loop packets to the cross-loop queue, duplicate packets to the duplicate packet queue, and late-arriving packets to the late-arriving queue. Each processing queue is an independent logical queue in the target switch's local queue record area. Packets are appended in the order of packet registration number, arrival timestamp, and sequence identifier. The same packet registration number cannot enter multiple processing queues simultaneously. Each processing queue uses the packet registration number, identification result, service segment identifier, path identifier, sequence identifier, window identifier, arrival timestamp, and rule version number as the minimum set of fields, written in a fixed-length or fixed-separator order to the target switch's local queue record area. Simultaneously, the queues are written to the queue registration area of ​​the station's communication management host. The writing result code must contain at least one packet. The system includes six categories: success, missing fields, sequence conflicts, version inconsistencies, duplicate queuing, and full storage area. Sequence conflicts are limited to message registration numbers in reverse order, arrival timestamps in reverse order, or sequence identifiers backdated within the same business segment. Successful local writing is considered successful queuing, and the next step only reads the local queue record area. The remote queue registration area is only used for synchronization and is not used as the basis for subsequent steps. Remote synchronization failure does not affect the effectiveness of the local processing queue, but synchronization should be retried according to the original field order. The number of retries can be set to 3, and the retry interval can be set to 200ms. If there are consecutive failures, the queuing of other arriving messages in the same observation window will not be interrupted, and the failure record will be appended to the associated record of the local queue to be checked with the message registration number, identification result, failure time, and failure reason. After the verification is completed, only the appended associated record is allowed, and the existing processing queue record will not be overwritten.

[0087] During on-site inspection, the accuracy rates of active path message identification, old path residual message identification, cross-flow message identification, duplicate message identification, windowed late message identification, and local enqueue success rate of the processing queue are used as the inspection criteria. Preferably, no less than 1,000 arriving messages are sampled for inspection, covering day and night time periods, different ports, different service segment lengths, and no less than two observation windows. The results are checked item by item using on-site packet capture records, service communication segment table records, and processing queue registration records.

[0088] In a preferred embodiment, a target switch in a wind farm substation performs path screening on 248 registered arriving messages within one observation window under handover impact conditions. Of these, 182 are identified as active path messages, 21 as residual messages from old paths, 14 as cross-loop messages, 19 as duplicate messages, and 12 as late-arriving messages exceeding the window limit. Each processing queue completes local enqueue writing within 340ms after the observation window ends. Alternatively, duplicate messages can be identified not by identical message length, but solely by the same window identifier, the same service segment identifier, the same path identifier, identical service field digests, and the arrival time interval between two adjacent messages not exceeding the duplicate detection duration. As long as the rule version, time boundary, and enqueue record criteria remain consistent with the other technical features of this step, this step can still be implemented.

[0089] S4. Based on the business communication segment table, perform sequence continuity verification, path uniqueness verification, and segment correspondence verification on the messages in each processing queue to determine the message acceptance result. The specific implementation is as follows:

[0090] The target switch management side, located in the communication cabinet of the wind farm booster station, performs sequence continuity verification, path uniqueness verification, and fragment correspondence verification on the messages in each processing queue based on the processing queues already written to the local queue record area and the service communication segment table already written to the local non-volatile record area in the previous step. Based on this, it determines the message acceptance result. The execution entity is the message acceptance verification program running in the onboard management unit of the target switch. The effective period is limited to the continuous period during which the current communication state remains in the handover impact state. The applicable scope is limited to active path messages, old path residual messages, cross-flow messages, duplicate messages, and windowed late messages that have completed local queuing and have registered identification results. It does not include message records in the queue to be verified, records to be supplemented, or message records that have not completed local queuing.

[0091] During execution, the process first reads the message registration number, identification result, service segment identifier, path identifier, sequence identifier, window identifier, arrival timestamp, and rule version number from each processing queue. Then, it reads the service segment identifier, path identifier, sequence identifier, window identifier, receiving port identifier, and service mapping result identifier corresponding to the same message registration number from the service communication segment table. The reading rhythm is executed synchronously according to the append rhythm in the local queue record area, with a time tolerance not exceeding the local time synchronization accuracy. The version caliber is consistent with the current rule version. The alignment order is window identifier, service segment identifier, identification result, and sequence identifier in that order. After alignment, only records with missing fields, incomplete writing, and those explicitly marked in the maintenance work order are removed. The test record does not automatically complete the business segment identifier, path identifier, sequence identifier, window identifier, and recognition result. If any of the above fields are missing, the record will stop processing and be transferred to pending verification. The above three verifications are performed group by group according to the same window identifier, the same business segment identifier, and the same processing queue. The single message registration number in the verification group is the smallest unit of determination. The verification group result is assigned to each message registration number in the group. If there is only one message registration number in the verification group, that message registration number will receive the corresponding verification result alone. If there are multiple message registration numbers in the verification group, each message registration number will be written according to the corresponding verification result of its respective verification group. Verification results cannot be borrowed from other verification groups.

[0092] When performing sequence continuity verification, the sequence identifiers within the same verification group are compared one by one in ascending order. Sequence continuity verification checks whether the difference between the sequence identifiers of two adjacent messages is equal to 1. The starting verification point is the message record with the smallest sequence identifier in the verification group. If there is only one message record in the verification group, the sequence continuity verification is directly recorded as successful. If the difference between the sequence identifiers of two adjacent messages is greater than 1, the sequence identifier is reversed, or the same sequence identifier appears repeatedly and the corresponding message registration numbers are different, then the sequence continuity verification is recorded as unsuccessful. Among them, duplicate registration records in the duplicate message queue do not participate in the starting calculation of sequence continuity verification. Messages in the late message queue are only independently verified within the same window identifier and do not supplement the previous sequence identifier across window identifiers.

[0093] Continue performing path uniqueness verification. Path uniqueness verification checks whether each message participating in the verification within the same verification group corresponds to only one valid path identifier. The valid path identifier is the set of path identifiers allowed to enter the corresponding processing queue in the path permission table pre-fixed according to the identification results under the current rule version. During path uniqueness verification, the path identifiers actually appearing in the verification group should be checked one by one against the path permission table to determine whether they belong to valid path identifiers. Path identifiers not included in the path permission table are directly recorded as path uniqueness verification failure. The path permission table includes at least the identification results, the set of allowed path identifiers, and the effective rule version. Different processing queues correspond to different sets of allowed path identifiers. During path uniqueness verification, each path identifier is merged and counted. When the merging result is 1, the path uniqueness verification is recorded as successful. When the merging result is greater than 1, the path uniqueness verification is recorded as unsuccessful. Among them, the duplicate message queue uses the path identifier corresponding to the first valid registration record as the unique valid path identifier. The old path residual message queue and the cross-flow message queue both participate in the verification with the original value of their respective registered path identifiers and are not rewritten according to the active path record.

[0094] Continue performing fragment correspondence verification. Fragment correspondence verification checks whether the service fragment identifier of each message in the same verification group is consistent with the service fragment identifier corresponding to the same message registration number in the service communication fragment table, and whether the service mapping result identifier in the service fragment identifier belongs to the allowed registration range given by the fragment allow table pre-fixed according to the window identifier and identification result under the current rule version. The fragment allow table includes at least the window identifier, identification result, allowed registration range of service mapping result identifiers, and effective rule version. Fragment correspondence verification compares each message registration number with the records in the service communication fragment table one by one. It does not require that the entire processing queue contains only one service fragment identifier. Consistency comparison is only performed on messages within the same verification group. If the service fragment identifier, service mapping result identifier, or window identifier of the same message registration number are inconsistent in the processing queue and the service communication fragment table, or if the service mapping result identifier of the message does not belong to the allowed registration range of the current verification group, then the fragment correspondence verification is recorded as invalid.

[0095] The above three checks are performed in a fixed order: sequence continuity check, path uniqueness check, and fragment correspondence check. If the first check fails, the subsequent checks will still be performed and all check records will be retained. The failure of the previous check will not replace the result of the subsequent check. When determining the message acceptance result, the registration number of a single message is used as the smallest unit of determination. The results of the three checks corresponding to the check group to which the message belongs are summarized. If all three checks of the same message registration number are true in its check group, it is determined to be an accepted message. If any one check fails, it is determined to be an unacceptable message. Messages that have not completed the check due to missing fields will not directly form a message acceptance result but will be transferred to pending verification. The message acceptance result is limited to two categories: accepted message and unacceptable message, without further subdivision.

[0096] Upon completion, a message acceptance result record is generated. This record uses the message registration number, identification result, service segment identifier, path identifier, sequence identifier, window identifier, three verification results, message acceptance result, and rule version number as the minimum set of fields. These fields are written to the target switch's local acceptance result area in a fixed-length or fixed-separator order, and simultaneously written to the acceptance result registration area of ​​the station's communication management host. The write result code must include at least six categories: success, missing fields, verification conflict, version inconsistency, duplicate write, and storage area full. Verification conflict is limited to inconsistent combinations of three verification results or different message acceptance results for the same message registration number under the same rule version. Successful local writing signifies completion of this step. Remote synchronization failure does not affect the effectiveness of the local acceptance result, but synchronization should be retried according to the original field order. The number of retries is specified. It can be set to 3 times, and the retry interval can be set to 200ms. If there are consecutive failures, the writing of the other message acceptance results in the same observation window will not be interrupted, and the failure record will be appended to the associated record in the local pending verification queue with the message registration number, failure time and failure reason. The old version record shall not overwrite the new version record. The same message registration number is only allowed to be written once under the same rule version. The first complete record is given priority, and the later duplicate complete records shall not be overwritten. The pending verification queue and its attached supplementary records shall not change the message acceptance results already written to the local acceptance result area. The next stage only reads the local acceptance result area. The remote acceptance result registration area is only used for synchronous tracekeeping and is not used as the basis for reading in subsequent stages. The security boundary is limited to only performing verification on the registered messages and forming message acceptance results. The original message content is not rewritten and the existing processing queue identification results are not changed.

[0097] During on-site inspection, the correctness of sequence continuity verification, path uniqueness verification, fragment correspondence verification, message acceptance result consistency rate, and local write success rate are used as the inspection criteria. Preferably, no less than 1,000 queued messages are sampled for inspection, covering different processing queues, different business fragment lengths, different identification results, and no less than two observation windows. The results of on-site manual review are used as the comparison benchmark, and the on-site packet capture records, business communication fragment table records, processing queue registration records, and local acceptance result records are checked item by item.

[0098] In a preferred embodiment, within one observation window of a target switch in a wind farm substation under handover influence, this step of verification is performed on 246 locally enqueued packets. Of these, 221 packets pass the sequence continuity check, path uniqueness check, and fragment correspondence check, and are thus identified as accepted packets. The remaining 25 packets fail at least one check and are thus identified as unacceptable packets. The local acceptance result area is written within 360ms after the observation window ends. Alternatively, the path uniqueness check can be performed not by merging all participating packets in the current check group, but by comparing the path identifiers of adjacent sequence identifier packets under the same service fragment identifier. As long as the rule version, window boundary, and acceptance result recording criteria remain consistent with the other technical features of this step, this step can still be implemented.

[0099] S5. Isolate and merge unaccepted messages, and output control results, status results, and fault results based on the message acceptance results. Specifically, the implementation is as follows:

[0100] The target switch management side, located in the communication cabinet of the wind farm booster station, performs isolation and merging processing on the unaccepted messages that have been written into the local acceptance result area in the previous stage. Based on the message acceptance results, it outputs control results, status results, and fault results. The execution entity is the result compilation program running in the onboard management unit of the target switch. The effective period is limited to the continuous period during which the current communication state remains in the switching impact state. The applicable scope is limited to active path messages, old path residual messages, cross-loop messages, duplicate messages, and windowed late messages that have formed message acceptance results and have been determined to be unaccepted messages. It does not include message records in the pending verification queue, records to be supplemented, or message records for which local acceptance results have not been written.

[0101] During execution, the system first reads the message registration number, identification result, service segment identifier, path identifier, sequence identifier, window identifier, message acceptance result, and rule version number from the local acceptance result area. Then, it reads the service mapping result identifier, arrival timestamp, and receiving port identifier corresponding to the same message registration number from the service communication segment table. The reading rhythm is executed synchronously according to the appended rhythm in the local acceptance result area, with a time tolerance not exceeding the local time synchronization accuracy. The version caliber is consistent with the current rule version. The alignment order is window identifier, service segment identifier, message acceptance result, and sequence identifier in sequence. After alignment, only records with missing or incomplete fields and manually tested records explicitly marked in the maintenance work order are removed. The system does not automatically complete the service segment identifier, path identifier, sequence identifier, window identifier, and message acceptance result. If any of the above fields are missing, the system stops processing the record and moves it to pending verification.

[0102] During isolation processing, a single message registration number is used as the smallest execution unit. Unaccepted messages are logically removed. Logical removal excludes the message from the reading range of subsequent steps after the local acceptance result area without physically deleting the original record. The readable range refers only to the scope of subsequent steps calling the local acceptance result area, excluding the scope of existing local acceptance result record retention and traceability query. The message registration number, identification result, business segment identifier, path identifier, sequence identifier, window identifier, isolation flag, and rule version number are written to the local isolation record area. The isolation flag is limited to "isolated," indicating that the message will no longer be called in subsequent steps in the current rule version and current result formation chain, but it does not change its existing message acceptance result record. The same message registration number is only allowed to be written once under the same rule version. The first complete record is given priority, and duplicate complete records cannot be overwritten. Only unaccepted messages that have completed isolation processing and have been written to the local isolation record area are included in the merging process.

[0103] Continue with the merging process. This process involves merging isolated but unaccepted messages under the same window identifier, service segment identifier, and service mapping result identifier, based on their identification results and path identifiers, to form a result merging record. Records with identical identification results and path identifiers, and consecutive message registration numbers, arrival timestamps, and sequence identifiers, are merged into one result merging record. If the identification results are different, the path identifiers are different, or any of the aforementioned consecutiveness conditions are not met, the previous result merging record ends and the next result merging record begins. Multiple result merging records are allowed under the same service segment identifier. During the merging process, the original message registration number and its existing identification results are not changed. Merging does not cross window identifiers or business segment identifiers. If the merging cannot be completed due to missing verification fields, the merging process is stopped and the process is transferred to pending verification. The result merging record includes at least the merging record number, the first message registration number, the last message registration number, the identification result, the business segment identifier, the business mapping result identifier, the path identifier, the window identifier, and the rule version number. The merging record number is a unique number within the same window identifier that is generated sequentially. The result merging record is stored in the local merging record area. Only appending is allowed; writing back for correction is not permitted.

[0104] When outputting control results, status results, and fault results based on message acceptance results, the results are merged and fixed according to the pre-fixed result type correspondence table under the current rule version. The result type correspondence table includes at least the business mapping result identifier, the result type identifier, and the effective rule version. The result type identifier is a fixed identifier value used to distinguish control results, status results, and fault results in the result type correspondence table, and maintains a unique correspondence within the same rule version. The same business mapping result identifier is only allowed to correspond to one result type identifier. Result merge records that do not match the result type correspondence table are transferred to pending verification and are not directly output as results.

[0105] A result merging record that matches the control result type identifier forms one control result; a result merging record that matches the status result type identifier forms one status result; and a result merging record that matches the fault result type identifier forms one fault result. Each result merging record forms only one control result, status result, or fault result; multiple results are not allowed to be formed simultaneously. Control results, status results, and fault results are all written as three record types in the local result output area. Control results, status results, and fault results are all result output record objects in the local result output area and correspond to the result record types of the control link, status link, and fault link, respectively. The minimum set of fields includes the merged record number, the first message registration number, the last message registration number, the business segment identifier, the window identifier, the result type identifier, and the rule version number. The first message registration number and the last message registration number are taken from the result merging record. The minimum and maximum values ​​of the message registration number are the same when the merged record contains only one message. They are written to the local result output area of ​​the target switch in a fixed length or fixed delimiter order, and simultaneously written to the result registration area of ​​the station's communication management host. The written result code includes at least six categories: success, missing field, merge conflict, version inconsistency, duplicate write, and storage area full. Merge conflict is limited to multiple result type identifiers formed within the same merge range or the same message registration number being merged into multiple result merge records. Successful local writing is considered a completion of this step. Remote synchronization failure does not affect the effectiveness of the local result output area, but synchronization should be retried in the original field order. The number of retries can be set to 3, and the retry interval can be set to 200ms. If there are consecutive failures, the writing of other result records in the same observation window will not be interrupted, and the failed record will be appended to the associated record in the local pending verification queue with the merge record number, failure time, and failure reason.

[0106] Old version records must not overwrite new version records. The next step only reads the local result output area. The local isolation record area, merge record area, and remote result registration area are only used for tracekeeping and synchronization and are not used as the basis for reading in subsequent steps. The security boundary is limited to performing isolation and merging processing on unaccepted messages to form control results, status results, and fault results. The original message content is not rewritten, and the existing message acceptance results are not changed.

[0107] During on-site inspection, the inspection criteria are: correctness of isolation execution, consistency of merging, correspondence of control results, correspondence of status results, correspondence of fault results, and success rate of local write. Preferably, no less than 1,000 messages with accepted message results are sampled for inspection. The sampling covers different business mapping result identifiers, different processing queues, and no less than two observation windows. The results of on-site manual review are used as the comparison benchmark. The records in the local accepted result area, local isolation record area, local merging record area, and local result output area are checked item by item.

[0108] In a preferred embodiment, within one observation window of a target switch in a wind farm substation under switching influence status, this step is performed on 25 unaccepted messages. Of these, 9 messages generate control results, 11 generate status results, and 5 generate fault results. The local result output area is written within 380ms after the observation window ends. Alternatively, only when the current rule version merges the identification results and path identifier constraints into the result type correspondence table, the merging process can also be performed using only the same window identifier, the same service segment identifier, and the same service mapping result identifier. As long as the rule version caliber, window boundary caliber, and result output caliber remain consistent with the other technical features of this step, this step can still be implemented.

[0109] S6. When the event fingerprint of the switching event meets the exit conditions, a recovery consistency check is performed on the activity path, service sequence, processing queue, and output results. After the check passes, the window flag is removed and normal communication is restored. The specific implementation is as follows:

[0110] When the fingerprint of a handover impact event meets the exit conditions, the management side of the target switch, located in the communication cabinet of the wind farm's booster station, sequentially performs a recovery consistency check on the activity path, service sequence, processing queue, and output results. The execution entity is the recovery consistency check program running in the onboard management unit of the target switch. The effective period is limited to after the current communication state remains in the handover impact state and the previous step has completed writing the control results, status results, and fault results to the local result output area. The applicable scope is limited to the activity path records, service sequence records, processing queue records, and output result records written to the local status area, local queue record area, local accepted result area, and local result output area within the observation window corresponding to the current window identifier and subsequent consecutive observation windows. It does not include records in the pending verification queue, records to be supplemented, or records that have not been completed in local writing. Records; where the handover impact event fingerprint meets the exit conditions, refers to reading the handover impact event fingerprint number, observation window start and end time, current communication status and exit condition record written into the local status area in the previous step, comparing them continuously in the order of the observation windows under the current rule version, and confirming that each port participating in triggering the handover impact event fingerprint meets the following conditions in the exit judgment observation window: no sudden change in link status, bit error rate change below the locking threshold, no effective adjacency change, delay drift not exceeding the reference band, queue backlog retention time below the retention threshold, and the exit is established after meeting the preset number of times in the natural continuous observation window. The exit judgment observation window is counted continuously from the first observation window that meets all the above conditions. If any intermediate observation window does not meet the conditions, the count is restarted. If any port participating in the trigger has no record in the exit judgment observation window, it is not considered to meet the conditions and is transferred to the pending verification.

[0111] The exit establishment record is written to the local exit judgment area using the event fingerprint number, observation window sequence number, exit establishment flag, and rule version number as the minimum field set. The event fingerprint number and observation window sequence number form a unique identifier. Only one exit establishment record is allowed to be formed for the same event fingerprint number under the same rule version. Subsequent repeated formation will only add duplicate associated records and will not write back existing records. If fields are missing, versions are inconsistent, or the observation window order is reversed, the processing will stop and the record will be transferred to pending verification. The event fingerprint number, observation window sequence number, and exit establishment flag will not be automatically filled in. Recovery consistency verification will only be initiated for event fingerprint numbers that have formed exit establishment records. Event fingerprint numbers that have not formed exit establishment records will not be allowed to enter the recovery consistency verification process of activity path, business sequence, processing queue, and output results. The results of the recovery consistency verification of the four items of activity path, business sequence, processing queue, and output results will be finally summarized to the same event fingerprint number.

[0112] When performing a recovery consistency check on an active path, the current active path record, port set identifier, and current observation window start-up lock record are read from the local status area, and the path identifier corresponding to the active path message is read from the local queue record area. Each business segment under the same event fingerprint number is compared one by one to confirm that the path identifier of the current active path record is consistent with the path identifier of the last valid active path message in the observation window corresponding to the exit establishment record. The last valid active path message is the active path message with the latest arrival timestamp and valid message registration number in the active path queue within the coverage of the exit establishment record, and no new inactive path switching record is formed in the continuous exit judgment observation window. If the active path recovery consistency check is successful, it is recorded as active path consistent; otherwise, it is recorded as active path inconsistent.

[0113] When performing a consistency check on the service sequence, the sequence identifier, message registration number, and message acceptance result under the same service segment identifier are read from the service communication segment table and the local acceptance result area. They are aligned in ascending order by window identifier, service segment identifier, and sequence identifier. It is confirmed that there is no backtracking, skipping, or duplicate occupation after the sequence identifier of the last accepted message. Duplicate occupation is limited to the same sequence identifier being occupied by multiple different message registration numbers within the same service segment. Furthermore, the last sequence identifier of each service segment within the coverage of the exit establishment record corresponds one-to-one with the last message registration number of the corresponding result record in the local result output area according to the same service segment identifier. When the service sequence consistency check is successful, it is recorded as a consistent service sequence. When a single service segment contains only one accepted message, it is directly recorded as successful.

[0114] When performing a recovery consistency check on the processing queue, the check scope is all processing queue records under the same event fingerprint number. All processing queue records include all local enqueue records formed within the coverage area of ​​the event fingerprint number for the active path queue, old path residual queue, cross-flow queue, duplicate message queue, and windowed late message queue. The enqueue records of the active path queue, old path residual queue, cross-flow queue, duplicate message queue, and windowed late message queue are read from the local queue record area and merged one by one according to the message registration number, identification result, and window identifier. It is confirmed that there are no duplicate enqueues with the same message registration number, cross-queue coexistence, or incomplete local enqueue writing in each processing queue. It is also confirmed that all isolated messages before the end of the observation window corresponding to the exit establishment record remain in the isolated state and all enqueueed messages have formed message acceptance results. When the processing queue recovery consistency check is successful, it is recorded as processing queue consistency.

[0115] When performing a consistency check on the output results, control results, status results, and fault results records are read from the local output results area, and their corresponding merge record number, first message registration number, last message registration number, and message acceptance result are read from the local merge record area and the local accepted result area. Each record is compared according to the result type identifier to confirm that each output result record can be associated with one and only one merge result record, the first message registration number is equal, and the last message registration number is equal. The association between the output result record and the merge result record is based on the same merge record number as the primary association basis, and the first message registration number and the last message registration number are used as the boundary verification basis. When the merge result record contains only one message, the first message registration number and the last message registration number are the same, and the result type identifier is consistent with the result type correspondence table under the current rule version. When the output result consistency check is successful, it is recorded as consistent output result; when the merge result record cannot be associated, it is recorded as inconsistent output result.

[0116] The above four consistency checks are executed in a fixed order: activity path, business sequence, processing queue, and output result. If the first check fails, subsequent checks will still be executed and all check records will be retained. The results of previous checks will not replace the results of subsequent checks. A successful check means that the activity path, business sequence, processing queue, and output result are all consistent under the same event fingerprint number. A failed check is limited to any one of the four consistency checks under the same event fingerprint number failing. After a successful check, the window identifier is removed and normal communication is restored. Removing the window identifier involves writing all records corresponding to the associated window identifiers under the same event fingerprint number into the local window removal area and assigning a removed mark. The local window removal area is the removal record partition in the local status area. Removed marks cannot be reversed. If the window is removed, subsequent duplicate removal requests will only add duplicate association records, without overwriting existing status records. Furthermore, no new switch-affected status records associated with the window identifier are allowed, and subsequent steps are not permitted to use the window identifier as the basis for determining switch-affected status or for message registration. To restore normal communication, the window removal write must be completed first, and then the current communication status in the local status area must be changed from the switch-affected status to the normal communication status and an effective timestamp must be written. The latter cannot take effect if the former fails. Only after the window removal write is successful can the current communication status rewriting proceed. If the window removal write is incomplete, fails, or is determined to be a duplicate removal, the current communication status remains unchanged as the switch-affected status, and the effective timestamp is the moment when the current communication status rewriting in the local status area is completed.

[0117] Both window release records and communication status recovery records are written to the local status area in a fixed length or fixed separator order and simultaneously written to the recovery registration area of ​​the station's communication management host. The writing result code includes at least six categories: success, missing field, verification failure, version mismatch, duplicate write, and storage area full. Successful local writing indicates that this step is complete. Remote synchronization failure does not affect the effectiveness of local window release and normal communication recovery, but synchronization should be retried in the original field order. The number of retries can be set to 3, and the retry interval can be set to 200ms. If there are consecutive failures, the writing of other recovery records in the same observation window will not be interrupted. The failure record is appended to the associated record in the local queue to be verified, with the event fingerprint number, failure time and failure reason. The old version record shall not overwrite the new version record. The same event fingerprint number is only allowed to be deactivated once under the same rule version. The next step only reads the window deactivation record and the current communication status record in the local status area. The remote recovery registration area is only used for synchronization and is not used as the basis for reading in subsequent steps. The security boundary is limited to performing only the recovery consistency check, deactivating the window identifier and restoring normal communication. The original message content is not rewritten, and the existing message acceptance result and existing result output record are not changed.

[0118] During on-site inspection, the following criteria are used for inspection: activity path recovery consistency rate, business sequence recovery consistency rate, processing queue recovery consistency rate, output result recovery consistency rate, window release accuracy rate, and normal communication recovery write success rate. Preferably, no less than 1,000 records with completed output results are sampled for inspection, covering different result types, different observation windows, and no less than two event fingerprint numbers. The results of on-site manual review are used as the comparison benchmark, and the records in the local status area, local queue record area, local result output area, and local window release area are checked item by item.

[0119] In a preferred embodiment, after two consecutive observation windows of a switching event fingerprint in a wind farm substation meet the exit conditions, a target switch performs a recovery consistency check on the activity path, service sequence number, processing queue, and output result corresponding to an event fingerprint number. If all four checks are successful, the system completes the window release write within 320ms after the exit is successful, and restores the current communication state to the normal communication state within 80ms thereafter. Alternatively, the service sequence number recovery consistency check may not use the one-to-one correspondence between the last valid sequence number identifier and the last message registration number as the identification criterion, but rather use the consistency between the range of consecutive valid sequence numbers under the same service segment identifier and the coverage area of ​​the result output record as the identification criterion. As long as its rule version criterion, observation window boundary criterion, and state recovery write criterion are consistent with the other technical features of this step, this step can still be implemented.

[0120] In the operational scenario shown in this embodiment: a target switch is installed in the communication cabinet of a wind farm booster station. The target switch is connected to 6 physical ports participating in the current redundant communication group, which are respectively connected to the wind turbine side control device, status acquisition device, fault transmission device and station communication management host. The target switch's onboard management unit has pre-fixed rule versions, threshold tables, service mapping tables, field extraction tables, activity path determination tables, path permission tables, fragment permission tables and result type correspondence tables.

[0121] When the wind farm is in normal power generation mode, the data acquisition and control program continuously collects the link status, bit error rate changes, adjacency changes, delay drift, and queue backlog information of 6 redundant ports at a sampling rhythm of 200ms. It also aligns, denoises, and completes the records of each port according to a 5s observation window. During a short-term disturbance superimposed on a power collection line switch, port 2 experienced two alternating connections and disconnections within the current observation window. The bit error rate change of port 3 reached the lockout threshold. Adjacency changes occurred between port 2 and port 4. The delay drift of port 3 continuously exceeded the reference band. The queue backlog of port 4 continuously exceeded the hold threshold. Based on this, the data acquisition and control program generates a switch impact event fingerprint and determines the current communication state from the normal communication state to the switch impact state, writing it into the local state area.

[0122] Subsequently, the message registration procedure registers each arriving message that was received in full frame by the six redundant ports and written to the receive buffer during the handover impact period. It extracts the service field summary and combines it with the service mapping table to form a service segment identifier, combines the receiving port identifier and the port set identifier to form a path identifier, combines the arrival timestamp and the receiving sequence number to form a sequence identifier, and combines the event fingerprint number and the observation window sequence number to form a window identifier. The relevant records are then continuously written to the service communication segment table.

[0123] Within one observation window of the switching impact state, continuous control messages from the same wind turbine control link form a service segment identifier. Some messages enter through port 2 to form a path identifier, while other messages enter through port 4 to form another path identifier. The sequence identifiers are numbered consecutively according to the order of arrival.

[0124] The path screening program then reads the service communication segment table and, in conjunction with the active path record locked at the current observation window start point, performs path screening on each arriving message. Messages whose path identifier matches the active path record and whose arrival time falls within the current observation window are identified as active path messages. Messages whose path identifier is not equal to the active path record and arrive later than the first active path message are identified as old path remnant messages. Messages that form three path state switches within the same service segment and then return to an inactive path are identified as cross-flow messages. Subsequent messages with consistent service field summaries and adjacent arrival time intervals not exceeding the repetition determination duration are identified as duplicate messages. Messages whose entire frame is written to the receive buffer later than the end time of the observation window corresponding to their respective window identifier are identified as late-arriving messages exceeding the window limit. These messages are then written to the active path queue, old path remnant queue, cross-flow queue, duplicate message queue, and late-arriving message queue, respectively.

[0125] The message acceptance verification procedure then forms a verification group based on the same window identifier, the same business segment identifier, and the same processing queue. It performs sequence continuity verification, path uniqueness verification, and segment correspondence verification on the messages in each processing queue. Messages that pass all three verifications are identified as accepted messages, and messages that fail any one of the verifications are identified as unaccepted messages. The message acceptance result is then written to the local acceptance result area.

[0126] The compilation program then performs logical elimination on each unaccepted message and writes it to the local isolation record area. Then, it merges the isolated unaccepted messages according to the same window identifier, the same service segment identifier, and the same service mapping result identifier to form a result merging record. For the result merging record that hits the control result type identifier, a control result is formed; for the result merging record that hits the status result type identifier, a status result is formed; and for the result merging record that hits the fault result type identifier, a fault result is formed. The corresponding record is then written to the local result output area.

[0127] As the disturbance subsides, within the subsequent two consecutive observation windows, all ports involved in triggering the handover impact event fingerprint meet the following criteria: no sudden changes in link state, bit error rate changes below the locking threshold, no effective adjacency changes, delay drift not exceeding the reference band, and queue backlog retention time below the retention threshold. Based on this, the recovery consistency verification program generates an exit success record and sequentially performs recovery consistency verification on the active path, service sequence, processing queue, and output results around the same event fingerprint number. After confirming that the current active path record is consistent with the path identifier of the last valid active path message, that the service sequence has not regressed or duplicated, that there is no duplicate enqueueing or cross-queue coexistence in each processing queue, and that the control results, status results, and fault results can all be uniquely associated with the corresponding results and merged records, the recovery consistency verification program determines that the recovery consistency verification has passed. First, it writes all records corresponding to the associated window identifiers under the event fingerprint number into the local window release area and assigns a release mark. Then, it rewrites the current communication state in the local status area from the handover impact state to the normal communication state, thereby completing the entire process of the target switch transitioning from the handover impact state to the normal communication state in the short-term link disturbance scenario of the wind farm.

[0128] All calculations involved in the embodiments are dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to the actual situation.

[0129] 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.

[0130] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as 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, all or part of the processes or functions described in the embodiments of this application are generated. 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 from one website, computer, server, or data center to another website, computer, server, or data center via wireless or wired transmission; wired transmission methods include optical fiber, twisted pair, coaxial cable, etc.; wireless transmission includes infrared, microwave, etc. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center containing one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0131] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0132] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0133] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0134] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0135] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0136] 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.

[0137] 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 highly reliable redundant communication method for industrial Ethernet switches in wind farms, characterized in that, include: S1. Collect link status, bit error rate changes, adjacency changes, delay drift and queue backlog information of each redundant port of the target switch, generate handover impact event fingerprint, and determine the current communication status accordingly. S2. When the current communication state enters the handover impact state, establish service segment identifier, path identifier, sequence identifier and window identifier for the arriving message to form a service communication segment table. S3. Based on the business communication segment table, perform path screening on the arriving messages, identify active path messages, old path residual messages, cross-flow messages, duplicate messages, and windowed late messages, and assign them to the corresponding processing queues. S4. Based on the business communication segment table, perform sequence continuity verification, path uniqueness verification, and segment correspondence verification on the messages in each processing queue to determine the message acceptance result; S5. Isolate and merge unaccepted messages, and output control results, status results, and fault results based on the message acceptance results. S6. When the event fingerprint of the switching event meets the exit conditions, perform a recovery consistency check on the activity path, business sequence, processing queue, and output results. After the check passes, remove the window flag and restore normal communication.

2. The high-reliability redundant communication method for an industrial Ethernet switch in a wind farm according to claim 1, characterized in that, S1 includes: The event record is formed by combining the number of link state mutations, cumulative bit error changes, number of adjacency changes, number of delay drift out-of-bounds, and duration of queue backlog for each redundant port within the same observation window in a fixed field order. When at least two disturbances are met within the same observation window, a switching impact event fingerprint is generated based on the event log.

3. A highly reliable redundant communication method for an industrial Ethernet switch in a wind farm according to claim 1, characterized in that, S2 include: When establishing a service segment identifier, it is merged with the preset service mapping table according to the service field summary, source address record, and destination address record of the arriving message; Under the same service mapping result, a service segment identifier is formed based on consecutively arriving messages and the time interval between adjacent arriving messages not exceeding the segment retention duration; The business segment identifier is formed by combining the observation window number, the business mapping result identifier, and the segment starting number.

4. A highly reliable redundant communication method for an industrial Ethernet switch in a wind farm according to claim 1, characterized in that, S3 includes: Under the same window identifier, arriving messages whose path identifier is consistent with the active path record locked at the current observation window starting point are identified as active path messages; Arrival messages whose path identifiers within the same business segment are inconsistent with the active path record locked at the starting point of the current observation window, and whose arrival timestamps are later than the arrival time of the first active path message, are identified as residual messages of the old path.

5. A highly reliable redundant communication method for an industrial Ethernet switch in a wind farm according to claim 4, characterized in that, S3 also includes: Arrival messages that form a path state switch and switch back to an inactive path within the same service segment identifier are identified as cross-flow back packets. Arriving messages with identical business field summaries under the same window identifier, the same business segment identifier, and the same path identifier, and whose time interval between adjacent arriving messages does not exceed the duplicate judgment duration, are identified as duplicate messages. The cross-flow messages and duplicate messages are then assigned to the corresponding processing queues.

6. A highly reliable redundant communication method for an industrial Ethernet switch in a wind farm according to claim 1, characterized in that, S4 include: Form a verification group based on the same window identifier, the same business segment identifier, and the same processing queue; Perform sequence continuity verification on messages within the same check group; The sequence continuity verification results are then matched one by one with the message registration numbers within the group.

7. A highly reliable redundant communication method for an industrial Ethernet switch in a wind farm according to claim 6, characterized in that, S4 also includes: Perform path uniqueness verification and fragment correspondence verification on messages within the same verification group; When the sequence continuity check result, path uniqueness check result, and fragment correspondence check result corresponding to the same message registration number are all true, the message is determined to be accepted.

8. A highly reliable redundant communication method for an industrial Ethernet switch in a wind farm according to claim 1, characterized in that, S5 include: Unaccepted messages are logically removed and written to the local isolation record area; Unaccepted messages that have been written to the local isolation record area are merged into a result merge record based on the identification result and path identifier; Control results, status results, and fault results are generated based on the fixed correspondence between the result merging records and the result type correspondence table.

9. A highly reliable redundant communication method for an industrial Ethernet switch in a wind farm according to claim 1, characterized in that, S6 include: When the event fingerprint of the switching event meets the exit conditions, the activity path, business sequence, processing queue and output result corresponding to the event fingerprint number that has formed the exit establishment record are sequentially checked for recovery consistency. When the activity path, business sequence, processing queue, and output result are all consistent under the same event fingerprint number, the window identifier is removed first, and then normal communication is restored.

Images