A data auditing method based on semantic evidence chain

Abstract

The application discloses a data auditing method based on a semantic evidence chain, comprising the following steps: forming a to-be-audited semantic evidence chain capable of reflecting a fault shutdown process by extracting state, instruction, alarm and disposal evidence nodes from a to-be-audited wind power operation segment; further constructing a wind power auditing evidence chain template through a wind turbine state transition relationship and a component physical association relationship, and comparing the template with the to-be-audited semantic evidence chain, so that the missing position of a node and the broken position of a relationship can be accurately located; on this basis, a supplementary evidence path is selected according to a breakpoint type, and supplementary evidence is connected, so that the workload of manually checking main control records, component records and maintenance records is reduced; finally, the integrity, traceability and result reliability of fault shutdown data storage auditing are improved through joint checking of node sequence, source consistency, state connection relationship, cause-effect explanation relationship and disposal corresponding relationship.

Description

Technical Field

[0001] This invention relates to the technical field of wind power data governance, and in particular to a data auditing method based on semantic evidence chains. Background Technology

[0002] In the data review scenario before wind farm outage data is entered into the database, wind turbines typically generate operating status quantities, control command contents, alarm event contents, and handling results before and after an outage. Among them, operating status quantities reflect changes in the status of the unit such as grid connection, power limitation, shutdown, fault, and recovery; control command contents reflect control actions such as start-up, shutdown, reset, power limitation, and grid connection switching; alarm event contents reflect abnormal triggering situations of the pitch system, yaw system, converter, generator, gearbox, and main control system; and handling results reflect maintenance results such as repair, reset, component replacement, and operation recovery. In the actual data entry process, the same outage segment is often scattered across the SCADA system, main control system, alarm system, and maintenance work order system. If the outage status or alarm code is generated based solely on the outage status or alarm code in a single system, it is difficult to determine whether the outage process has complete control commands, alarm triggers, component abnormalities, and handling and recovery evidence. This can easily lead to data samples with unclear fault causes, inconsistent faulty components, or missing recovery status.

[0003] Existing wind power operation data audits typically focus on checking numerical integrity, temporal continuity, anomaly thresholds, and field consistency. For example, they check whether operating parameters such as wind speed, power, rotational speed, and pitch angle are missing, or whether alarm codes, downtime, and work order numbers have corresponding records. This type of audit can detect null values, duplicate values, out-of-bounds values, and timestamp anomalies, but it is difficult to verify the state transition relationships during wind turbine failure shutdowns. For example, when the operating status changes from generation to shutdown, if there is a lack of shutdown instructions or fault alarms, traditional audits may still determine that the data can be stored because the shutdown status record is complete. Similarly, when the fault status changes to the recovery status, if there is a lack of repair, reset, or component replacement results, traditional audits also find it difficult to determine whether the fault shutdown segment forms a traceable operation and maintenance closed loop.

[0004] While existing multi-source data fusion or knowledge graph methods can link wind power operation data, alarm events, and work order content, their linking processes are mostly based on equipment numbers, time ranges, or event tags. They typically remain at the level of matching fields with the same name, merging events, or linking work order texts, lacking evidence chain verification mechanisms for wind turbine state transition relationships and component physical relationships. For situations such as pitch system alarms but corresponding flowmeter handling results, shutdown status lacking master control shutdown commands, handling records preceding alarm events, and the same component remaining in a fault state after fault handling, existing linking methods struggle to accurately locate missing nodes and broken relationships. They also cannot select appropriate supplementary evidence paths based on the type of breakpoint. This forces data auditors to manually review master control status records, component operation records, maintenance handling records, and adjacent wind power operation segments, resulting in low audit efficiency and unstable audit standards.

[0005] Therefore, existing wind power outage data entry and verification technologies suffer from problems such as insufficient extraction of multi-source evidence nodes, inadequate verification of status continuity relationships, unclear component object associations, unclear paths for supplementing missing evidence, and weak traceability of verification results. The present invention addresses the technical problem of how to locate breakpoints, supplement evidence, and verify consistency based on semantic evidence chains when there are missing nodes or broken relationships among status, instruction, alarm, and handling evidence before wind power outage data is entered into the database. Summary of the Invention

[0006] The purpose of this section is to outline some aspects of the embodiments of the present invention and to briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section, as well as in the abstract and title of the present application, to avoid obscuring the purpose of this section, the abstract and title of the invention. Such simplifications or omissions shall not be used to limit the scope of the present invention.

[0007] In view of the aforementioned existing problems, the present invention is proposed.

[0008] To solve the above-mentioned technical problems, the present invention provides the following technical solution: A data auditing method based on semantic evidence chain includes: extracting status, instruction, alarm and disposal evidence nodes from the wind power operation segment to be audited to form a semantic evidence chain to be audited; The semantic evidence chain to be audited is compared with the wind power audit evidence chain template constructed based on the wind turbine state transition relationship and component physical association relationship to anchor the missing node position and the relationship break position. The breakpoint type is determined based on the missing node location and the broken relationship location. A supplementary evidence path is selected according to the breakpoint type, and supplementary evidence is retrieved. The supplementary evidence is then connected to the semantic evidence chain to be reviewed. Verify the node order, source consistency, state succession relationship, causal explanation relationship, and disposal correspondence of the semantic evidence chain to be reviewed after the supplementary evidence is accessed, and output the data review results.

[0009] The beneficial effects of this invention are as follows: This invention extracts status, instruction, alarm, and handling evidence nodes from the wind power operation segments to be audited, forming a semantic evidence chain that reflects the fault shutdown process; further, it constructs a wind power audit evidence chain template through wind turbine state transition relationships and component physical association relationships, and compares it with the semantic evidence chain to be audited to accurately locate missing nodes and broken relationships; based on this, it selects supplementary evidence paths according to the breakpoint type and connects supplementary evidence, reducing the workload of manually reviewing master control records, component records, and maintenance records; finally, through joint verification of node order, source consistency, status succession relationship, causal explanation relationship, and handling correspondence relationship, it improves the completeness, traceability, and reliability of the fault shutdown data entry audit. Attached Figure Description

[0010] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a flowchart illustrating the data verification method based on semantic evidence chain as shown in this invention. Detailed Implementation

[0011] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0012] Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort should fall within the scope of protection of this invention.

[0013] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0014] In a preferred embodiment, the present invention is applied to a data review scenario before wind farm outage data is entered into the database. Specifically, the wind power operation segment to be reviewed is a data segment obtained by the wind farm data platform from the wind turbine main control system, energy management system, SCADA system, alarm management system, and maintenance management system after the occurrence of an outage. This segment represents the data generated before and after the same outage event for the same wind turbine. The wind power operation segment to be reviewed includes at least operating status variables, control command content, alarm event content, and handling result content. The operating status variables include the wind turbine grid connection status. The system includes the following parameters: wind turbine status, wind turbine shutdown status, power limitation status, fault status, recovery status, active power, speed, pitch angle, yaw angle, DC bus voltage, winding temperature, gearbox bearing temperature, and communication status; control commands include start-up commands, shutdown commands, reset commands, power limitation commands, and grid connection switching commands; alarm events include pitch alarms, yaw alarms, converter alarms, generator alarms, gearbox alarms, main control protection actions, and communication anomalies; and handling results include maintenance work order conclusions, reset results, component replacement results, and operation recovery confirmation results.

[0015] The above data reflects the state changes, control actions, causes of abnormalities, and closed-loop handling before and after the occurrence of a fault shutdown event, and can therefore serve as the basis for auditing fault shutdown data before it is entered into the database.

[0016] According to an embodiment of the present invention, in combination Figure 1 The flowchart shown illustrates a data auditing method based on semantic evidence chains, which specifically includes the following steps: S1. Extract status, instruction, alarm, and handling evidence nodes from the wind power operation segments to be reviewed, forming a semantic evidence chain to be reviewed. Note that the following points should be noted in this step: S1.1. Merge the operating status quantities, control command contents, alarm event contents, and handling results in the wind power operation segments to be reviewed according to the wind turbine number and the time of occurrence to obtain the same machine time sequence evidence segment.

[0017] In a preferred embodiment, the wind power operation segment to be reviewed is obtained as follows: when the wind farm data platform receives a fault shutdown event, the wind turbine number and shutdown time in the fault shutdown event are used as the retrieval basis, and the operating status, control command content, alarm event content and handling result content are obtained within the time range from 30 minutes before the shutdown time to 120 minutes after the shutdown time.

[0018] After acquisition, all content is merged according to turbine number, occurrence time, and source system. When multiple source systems have content with the same turbine number and the occurrence time difference is no greater than 3 seconds, the content of multiple source systems is merged into the same sampling time. When the same sampling time contains more than two of the following categories: operating status quantity, control command content, alarm event content, and handling result content, the evidence categories of operating status quantity, control command content, alarm event content, and handling result content are retained separately and do not overlap with each other. Subsequently, the content with the same turbine number is arranged from earliest to latest according to the occurrence time to obtain the same turbine time sequence evidence segment.

[0019] For example, if wind turbine number WTG-014 experiences a fault shutdown at 10:24:16 on 2026-05-12, then the operating status variables, control command contents, alarm event contents, and handling results between 09:54:16 and 12:24:16 on 2026-05-12 are obtained. Among them, at 10:18:05 on 2026-05-12, there is an operating status variable showing a gearbox high-speed shaft bearing temperature of 92℃; at 10:18:08 on 2026-05-12, there is a high gearbox high-speed shaft bearing temperature alarm; at 10:18:12 on 2026-05-12, there is a main control protection shutdown command; and at 10:18:20 on 2026-05-12, the wind turbine is in a shutdown state. The above contents are arranged according to WTG-014 and the time of occurrence to form a time sequence evidence segment for the same machine.

[0020] S1.2. Mark the operating status quantities representing grid connection, shutdown, power limitation, fault and recovery in the same machine time sequence evidence segment as status evidence nodes.

[0021] In a preferred embodiment, the method for calibrating the state evidence node is as follows: check the operating state quantities in the same machine time sequence evidence segment one by one, and when the operating state quantity indicates that the wind turbine is in any operating state of grid connection, shutdown, power limitation, fault or recovery, the operating state quantity is calibrated as the state evidence node.

[0022] Taking a doubly-fed wind turbine generator set with a rated power of 2500kW and a grid-connected operating speed range of 9rpm to 18rpm as an example, the status evidence node corresponding to grid connection is determined by the grid connection status being grid-connected, the active power being greater than 125kW, and the main shaft speed being between 9rpm and 18rpm. Among them, the active power being greater than 125kW is 5% of the rated power of 2500kW, which is used to exclude idling, standby, and low-speed drive states. The main shaft speed being between 9rpm and 18rpm is used to limit the wind turbine to have entered the normal grid-connected power generation speed range.

[0023] The status evidence node corresponding to the shutdown is determined by the wind turbine status being shutdown, active power being less than 25kW, and main shaft speed being less than 2rpm. Among them, active power less than 25kW is 1% of the rated power of 2500kW, which is used to indicate that the wind turbine has exited the effective power generation state, and main shaft speed less than 2rpm is used to indicate that the wind turbine is in a stopped or near-stopped state. If the shutdown status flag, active power less than 25kW, and main shaft speed less than 2rpm exist simultaneously in the same time sequence evidence segment, then the corresponding operating status quantity is marked as the status evidence node corresponding to the shutdown.

[0024] The status evidence node corresponding to the power limitation is determined by the power limitation flag being in effect, the power setpoint being lower than 2000kW, and the active power decreasing to within 50kW above or below the power setpoint within 60 seconds after the power limitation command is issued. Among them, the power setpoint being lower than 2000kW indicates that the wind turbine has switched from the normal power generation capacity corresponding to the rated power of 2500kW to the limited power generation state, and the range of 50kW above or below is used to eliminate the impact of wind speed fluctuations and power regulation lag on the identification of the power limitation status. If the power setpoint is 1500kW, and the active power decreases from 2380kW to between 1460kW and 1550kW within 60 seconds, then the corresponding operating status quantity is marked as the status evidence node corresponding to the power limitation.

[0025] The status evidence node corresponding to the fault is determined by the fault status flag being valid, the fault code existing, and the fault duration being greater than 10 seconds. Among them, the fault code should be a fault code in the main control system or alarm management system that can point to the component object. The fault duration being greater than 10 seconds is used to exclude instantaneous communication jitter, short-term sampling anomalies, and non-continuous alarms. If there is a GBX-TEMP-HIGH gearbox high temperature fault code in the same machine timing evidence segment, and this fault code lasts from 2026-05-12 10:18:08 to 2026-05-12 10:18:25, then the fault duration is 17 seconds, and the corresponding operating status quantity is marked as the status evidence node corresponding to the fault.

[0026] The recovery of the corresponding state evidence node is jointly determined by the clearing of the fault status flag, the successful reset result, the spindle speed recovering to above 3 rpm, and the operating status changing from shutdown or fault to pending grid connection. Among them, the clearing of the fault status flag indicates that the fault lockout condition has been released, the successful reset result indicates that the main control system has accepted the reset operation, the spindle speed recovering to above 3 rpm indicates that the fan has left the shutdown state and entered the start-up preparation process, and the operating status changing from shutdown or fault to pending grid connection indicates that there is a state succession relationship between this recovery state and the previous fault shutdown event. If the same machine timing evidence segment shows that the fault status flag was cleared, the reset result was successful, the spindle speed was 4.2 rpm, and the operating status was pending grid connection at 2026-05-12 11:55:20, then this operating status quantity is marked as the state evidence node corresponding to the recovery.

[0027] Each state evidence node includes at least the wind turbine number, the time of occurrence, the type of operating state, the component object, and the system from which the state originates. The component object is determined according to the system to which the operating state quantity belongs. For example, the gearbox bearing temperature corresponds to the gearbox, the converter module temperature corresponds to the converter, and the main control operating state corresponds to the main control system.

[0028] S1.3. Mark the control command content representing start-up, shutdown, reset, power limiting and grid connection switching in the same machine timing evidence segment as command evidence nodes.

[0029] In a preferred embodiment, the method for identifying instruction evidence nodes is as follows: check the control instruction content in the same time sequence evidence segment item by item. When the control instruction content indicates that the main control system, the station energy management system or the remote monitoring system has issued any control instruction among start-up, shutdown, reset, power limiting or grid connection switching, the control instruction content is identified as an instruction evidence node.

[0030] Taking a doubly-fed wind turbine generator set with a rated power of 2500kW and a grid-connected operating speed range of 9rpm to 18rpm as an example, the command evidence node corresponding to the start-up is determined by the start-up command status being issued, the start-up command reception result being received successfully, and the main control system forming a grid-connection pending state within 30 seconds after the start-up command is issued. Among them, 30 seconds is the time range for the main control system to complete the start-up command reception, yaw correction, pitch initial angle adjustment, and the start-up process entering the grid-connection pending state. If the same machine timing evidence segment shows that a start-up command exists at 11:58:30 on 2026-05-12, the start-up command reception result is received successfully at 11:58:35 on 2026-05-12, and the operating state changes to grid-connection pending at 11:58:52 on 2026-05-12, then the content of this control command is marked as the command evidence node corresponding to the start-up.

[0031] The evidence node for the shutdown command is determined by the following: the shutdown command status is "issued," the shutdown reason exists, the shutdown command reception result is "successfully received," and the turbine status changes to shutdown within 60 seconds after the shutdown command is issued. The shutdown reason includes any one of manual shutdown, protection shutdown, fault shutdown, or scheduled shutdown. 60 seconds is the time range for determining whether the turbine completes power reduction, pitch adjustment, and grid connection disconnection. If the same machine timing evidence segment shows a protection shutdown command at 2026-05-12 10:18:12, the shutdown reason being high gearbox high-speed shaft bearing temperature, the shutdown command reception result at 2026-05-12 10:18:16 being "successfully received," and the turbine status changing to shutdown at 2026-05-12 10:18:20, then this control command content is marked as the evidence node for the shutdown command.

[0032] The evidence node for the reset command is determined by the following: the reset command status is "issued," the reset target is clearly defined, the reset command reception result is "successfully received," and the fault status flag is cleared within 120 seconds after the reset command is issued. The reset target includes any one of the following: main control system, converter, pitch system, yaw system, generator, and gearbox. 120 seconds is the time range for determining whether the reset command completes fault lockout release, status refresh, and status feedback. If the same machine timing evidence segment shows a gearbox fault reset command at 11:54:00 on 2026-05-12, the reset command reception result is "successfully received," and the fault status flag is cleared at 11:55:20 on 2026-05-12, then this control command content is designated as the evidence node for the reset command.

[0033] The evidence node for the power limiting command is determined by the following conditions: the power limiting command status is issued, the power setpoint is below 2000kW, the power limiting command reception result is successful, and the active power decreases to within 50kW above or below the power setpoint within 60 seconds after the power limiting command is issued. Among these conditions, a power setpoint below 2000kW indicates that the wind turbine has switched from normal power generation capacity to limited power generation, and the range of 50kW above or below is the allowable deviation range corresponding to wind speed fluctuations and power regulation lag. If the power setpoint is 1500kW, and the active power decreases from 2380kW to between 1460kW and 1550kW within 60 seconds, then the content of this control command is marked as the evidence node for the power limiting command.

[0034] The evidence nodes for grid connection switching are determined by the following: the grid connection switching command status is "issued"; the grid connection switch action feedback is "successfully closed"; the grid connection status changes to grid connection within 45 seconds after the grid connection switching command is issued; and the active power is greater than 125kW within 90 seconds after the grid connection switching command is issued. Here, 45 seconds is the judgment time range for grid connection switch action and main control status feedback; 90 seconds is the judgment time range for power ramp-up after the wind turbine completes grid connection; and 125kW is 5% of the rated power of 2500kW. For example, if the same machine timing evidence segment shows that a grid connection switching command existed at 12:02:30 on 2026-05-12, the grid connection switch action feedback was "successfully closed" at 12:02:48 on 2026-05-12, and the grid connection status was "grid connected" at 12:03:00 on 2026-05-12, then... If the active power is 180kW at 12:03:45, then the content of this control command is marked as the command evidence node corresponding to the grid connection switching.

[0035] Each instruction evidence node includes at least the wind turbine number, the time of occurrence, the instruction type, the instruction source system, the instruction reception result, the component object, and the instruction feedback result. The component object is determined according to the direct target of the control instruction content. When the reset instruction is applied to the converter, the component object is the converter. When the power limiting instruction is executed by the main control system to regulate the overall power, the component object is the main control system. When the protection shutdown instruction is triggered by the gearbox alarm, the component object retains the association identifier between the main control system and the gearbox.

[0036] S1.4 Mark the alarm event content representing component alarms, protection actions, and communication anomalies in the same machine time sequence evidence segment as alarm evidence nodes.

[0037] In a preferred embodiment, the method for marking alarm evidence nodes is as follows: check the alarm event content in the same machine time sequence evidence segment item by item, and when the alarm event content indicates that there is any event among component alarm, protection action or communication abnormality, mark the alarm event content as an alarm evidence node.

[0038] Taking a doubly-fed wind turbine generator set with a rated power of 2500kW and a grid-connected operating speed range of 9rpm to 18rpm as an example, the alarm evidence node corresponding to the component alarm is determined by the alarm status being valid, the alarm code existing, the alarm level not lower than level 2, and the alarm duration being greater than 10s. Among them, the alarm level is divided into level 1 alert alarm, level 2 general alarm, level 3 important alarm, and level 4 shutdown alarm according to the wind farm alarm management system. An alarm level not lower than level 2 indicates that the alarm event has reached the abnormality level required for entry into the database. The alarm duration is greater than 10s to eliminate instantaneous sampling disturbances. If there is an alarm code GBX-TEMP-HIGH in the same machine time sequence evidence segment, the alarm level is level 3, and the alarm status lasts from 2026-05-12 10:18:08 to 2026-05-12 10:18:25, then the content of this alarm event is marked as the alarm evidence node corresponding to the gearbox.

[0039] The alarm evidence node corresponding to the protection action is determined by the following factors: the protection action status is triggered, the protection action name exists, the protection action triggering object is clear, and the fan status changes to shutdown or power limitation within 60 seconds after the protection action occurs. Among them, the protection action name includes any one of over-temperature protection, over-speed protection, over-voltage protection, over-current protection, vibration protection, and communication interruption protection. 60 seconds is the judgment time range for the main control system to complete power reduction, disconnection from grid, or power limitation after the protection action is triggered. If the same machine timing evidence segment shows that there is a gearbox over-temperature protection action at 10:18:10 on 2026-05-12 and the fan status changes to shutdown at 10:18:20 on 2026-05-12, then the alarm event content is marked as the alarm evidence node corresponding to the protection action.

[0040] The alarm evidence node corresponding to the communication anomaly is determined by the following: the communication object is clearly identified, the communication status is interrupted or packet loss is abnormal, the duration of the communication anomaly is greater than 30 seconds, and the device object corresponding to the communication anomaly exists. Among them, the communication object includes any one of the following: the main control system and the pitch system, the main control system and the converter, the main control system and the yaw system, and the main control system and the SCADA system. The duration of the communication anomaly is greater than 30 seconds to exclude instantaneous network jitter. If the same time sequence evidence segment shows that the communication between the main control system and the converter was interrupted at 10:10:00 on 2026-05-12 and communication was restored at 10:10:42 on 2026-05-12, then the duration of the communication anomaly is 42 seconds, and the content of this alarm event is marked as the alarm evidence node corresponding to the communication anomaly.

[0041] The alarm evidence nodes corresponding to converter alarms are determined by the presence of a converter alarm code, a DC bus voltage greater than 1150V or a module temperature greater than 85℃, and an alarm duration greater than 10s. The alarm evidence nodes corresponding to generator alarms are determined by the presence of a generator alarm code, a winding temperature greater than 130℃ or a speed exceeding 18rpm, and an alarm duration greater than 10s. The alarm evidence nodes corresponding to gearbox alarms are determined by the presence of a gearbox alarm code, a bearing temperature greater than 85℃ or a gearbox vibration characteristic exceeding the level 3 alarm limit in the alarm management system, and an alarm duration greater than 10s. The alarm evidence nodes corresponding to pitch alarms are determined by the presence of a pitch alarm code, a pitch angle feedback deviation greater than 3° or a pitch motor current greater than 120% of the rated current, and an alarm duration greater than 10s. The alarm evidence nodes corresponding to yaw alarms are determined by the presence of a yaw alarm code, a wind direction deviation greater than 15° or a yaw motor current greater than 120% of the rated current, and an alarm duration greater than 10s.

[0042] Each alarm evidence node includes at least the turbine number, occurrence time, alarm type, alarm code, alarm level, alarm source system, component object, and alarm duration. The component object is determined according to the component to which the alarm belongs. For example, if the alarm name is "High temperature of gearbox high-speed shaft bearing," the component object is the gearbox; if the alarm name is "Overvoltage of converter DC bus," the component object is the converter; if the alarm name is "Communication interruption between main control system and pitch system," the component object is the communication object between the main control system and pitch system. For the same alarm code generated consecutively by the same component object within 30 seconds, the first alarm is marked as alarm occurrence evidence, and subsequent alarms are merged into alarm continuity evidence and assigned to the same alarm evidence node.

[0043] S1.5 Mark the handling results content representing maintenance, reset, component replacement and operation recovery in the same machine time sequence evidence segment as handling evidence nodes.

[0044] In a preferred embodiment, the method for marking the disposal evidence node is as follows: check the disposal result content in the same time sequence evidence segment item by item, and when the disposal result content indicates that there is any disposal result in maintenance, reset, component replacement or operation recovery, mark the disposal result content as the disposal evidence node.

[0045] Taking a doubly-fed wind turbine generator set with a rated power of 2500kW and a grid-connected operating speed range of 9rpm to 18rpm as an example, the corresponding maintenance handling evidence node is determined by the following: the maintenance object is clearly identified, the maintenance start time exists, the maintenance end time exists, the maintenance conclusion is "completed", and the maintenance end time is later than the corresponding alarm occurrence time. Among them, the maintenance object should be consistent with the component object of the previous alarm evidence node or fault status evidence node. The maintenance end time being later than the corresponding alarm occurrence time is used to exclude the possibility that historical maintenance records are incorrectly connected to the current fault shutdown event. If the same machine time sequence evidence segment shows that there is a gearbox alarm evidence node at 10:18:08 on 2026-05-12, and there is a gearbox maintenance record from 11:05:00 to 11:35:00 on 2026-05-12, with a maintenance conclusion of "completed", then the content of this handling result is marked as the maintenance corresponding handling evidence node.

[0046] The evidence node for resetting is determined by the following: the reset object is clearly identified, the reset result is successful, the reset time is later than the corresponding alarm time, and the fault status flag is cleared within 120 seconds after the reset. The 120-second period is the time range within which the main control system completes reset feedback, fault lockout release, and status refresh. If a gearbox fault reset record exists at 11:54:00 on 2026-05-12, and the fault status flag is cleared at 11:55:20 on 2026-05-12, then this result is marked as the evidence node for resetting. If the reset result is unsuccessful, or the reset time is earlier than the corresponding alarm time, then this result is not marked as the evidence node for resetting.

[0047] The evidence node for component replacement is determined by the existence of the replacement component name, the replacement component number, the component status before replacement, the component status after replacement, and the replacement completion time being later than the corresponding alarm occurrence time. Among them, the replacement component name and replacement component number are used to identify the unique component of the object to be replaced, and the component status before replacement and the component status after replacement are used to distinguish between actual replacement and maintenance only. If the same machine time sequence evidence segment shows "Replaced gearbox high-speed shaft bearing temperature sensor, old part number GBX-SEN-014-03, new part number GBX-SEN-014-09, replacement completion time is 2026-05-12 11:36:00", and the preceding alarm evidence node is high gearbox high-speed shaft bearing temperature, then this disposal result content is marked as the disposal evidence node corresponding to the component replacement.

[0048] The evidence nodes for operation recovery are determined by the following: the recovery confirmation result is normal, the fault status flag is cleared, the grid connection status is switched to grid connection within 30 minutes after recovery confirmation, and the active power is greater than 125kW within 30 minutes after recovery confirmation. Among them, 125kW is 5% of the rated power of 2500kW, and 30 minutes is the audit time range for completing reset, start-up, grid connection, and initial power ramp-up after fault handling. If the same machine timing evidence segment shows that the recovery confirmation result is normal at 11:58:00 on 2026-05-12, the grid connection status is grid connection at 12:03:00 on 2026-05-12, and the active power is 680kW at 12:10:00 on 2026-05-12, then the content of the handling result is marked as the evidence node for operation recovery.

[0049] Each handling evidence node shall include at least the wind turbine number, the time of occurrence, the handling type, the handling object, the handling source system, the handling result, the handling completion time, and the corresponding alarm object; wherein the handling object shall be consistent with the component object in the preceding alarm evidence node or fault status evidence node; when the preceding alarm evidence node is a gearbox alarm evidence node, the handling object of the handling evidence node shall be the gearbox, gearbox sensor, gearbox lubrication unit, or the main control protection reset object that is directly related to the gearbox.

[0050] S1.6 Arrange the status evidence nodes, instruction evidence nodes, alarm evidence nodes, and handling evidence nodes in the order of occurrence. Connect adjacent evidence nodes with the same turbine number and the same component object, and connect adjacent evidence nodes with the same turbine number and consecutive statuses to obtain the semantic evidence chain to be reviewed.

[0051] In this embodiment, "state continuity" refers to the relationship between adjacent state evidence nodes that conforms to the state change relationship in the wind turbine fault shutdown process. This state change relationship includes: entering the power limiting state from the grid-connected state via a power limiting command; entering the shutdown state from the grid-connected state via a shutdown command or fault alarm; entering the fault state from the shutdown state via fault confirmation; entering the recovery state from the fault state via fault handling; and entering the grid-connected state from the recovery state via a grid-connected switching command. Specifically, the state evidence nodes, command evidence nodes, alarm evidence nodes, and handling evidence nodes are first arranged in the order of occurrence. Then, the wind turbine number and component object of adjacent evidence nodes are checked. When adjacent evidence nodes have the same wind turbine number and the same component object, a component object connection is established between the adjacent evidence nodes. When adjacent state evidence nodes have the same wind turbine number and the state change belongs to the above-mentioned state change relationship, a state continuity connection is established between the adjacent state evidence nodes.

[0052] Taking WTG-014 as an example, the semantic evidence chain to be reviewed can include, in sequence: the grid connection status evidence node at 10:17:40 on 2026-05-12, the gearbox bearing temperature rise status evidence node at 10:18:05 on 2026-05-12, the gearbox alarm evidence node at 10:18:08 on 2026-05-12, the protection shutdown command evidence node at 10:18:12 on 2026-05-12, the shutdown status evidence node at 10:18:20 on 2026-05-12, the fault confirmation status evidence node at 10:42:00 on 2026-05-12, the gearbox handling evidence node at 11:36:00 on 2026-05-12, the recovery status evidence node at 11:58:00 on 2026-05-12, and the [unclear - possibly related to a date or timeline]. The grid connection status evidence node at 12:03:00; the above evidence nodes are connected according to the time of occurrence, and component object connections are formed between the gearbox alarm evidence node, the protection shutdown command evidence node, the shutdown status evidence node and the gearbox disposal evidence node, thereby obtaining the semantic evidence chain to be reviewed.

[0053] It should be noted that step S1 marks the status evidence nodes, instruction evidence nodes, alarm evidence nodes, and handling evidence nodes from the wind power operation segments to be audited, and forms a semantic evidence chain to be audited according to the occurrence time, wind turbine number, component object, and the sequential relationship of status. This transforms the fault shutdown data, which was originally scattered in the main control, SCADA, alarm, and maintenance systems, into audit objects with time sequence and causal orientation. This solves the technical problems of scattered evidence sources, difficulty in matching status with handling, and difficulty in tracing the cause of faults before the wind farm fault shutdown data is entered into the database. It improves the accuracy of subsequent template comparison and anomaly location, and reduces the probability of fault shutdown data being incorrectly entered into the database.

[0054] S2. Compare the semantic evidence chain to be reviewed with the wind power review evidence chain template constructed based on the wind turbine state transition relationship and component physical association relationship, and anchor the missing node positions and relationship break positions. Note that the following should be noted in this step: S2.1. Based on the state evidence nodes in the semantic evidence chain to be reviewed, select a wind power review evidence chain template that contains the same wind turbine operation stage.

[0055] In a preferred embodiment, the method for constructing a wind power audit evidence chain template includes: The wind turbine operation is divided into power generation stage, power limitation stage, shutdown stage, fault stage and recovery stage, and a corresponding template status node is configured for each wind turbine operation stage. Based on the wind turbine state transition relationship, template instruction nodes, template alarm nodes, and template handling nodes are configured between adjacent template state nodes to form a state transition template chain; Based on the physical relationships between components, the pitch system, yaw system, converter, generator, gearbox and main control system are respectively associated with the corresponding template status node, template command node, template alarm node and template handling node; Generate a wind power audit evidence chain template based on the order of nodes in the state transition template chain and the component association results.

[0056] Specifically, the wind turbine state transition relationship includes the state transition relationship between power generation state, power limitation state, shutdown state, fault state, and recovery state. Among them, the power generation state is converted to the power limitation state after receiving the power limitation command; the power generation state is converted to the shutdown state after receiving the shutdown command or generating a fault alarm; the shutdown state is converted to the fault state after completing the fault confirmation; the fault state is converted to the recovery state after completing the fault handling; and the recovery state is converted back to the power generation state after receiving the grid connection command.

[0057] Furthermore, the physical relationships between components include the relationships between the pitch system, yaw system, converter, generator, gearbox, and main control system. Specifically, the pitch system is associated with pitch angle, pitch motor current, and pitch alarms; the yaw system is associated with yaw angle, wind direction deviation, and yaw alarms; the converter is associated with DC bus voltage, AC output current, module temperature, and converter alarms; the generator is associated with speed, active power, winding temperature, and generator alarms; the gearbox is associated with bearing temperature, gearbox vibration characteristics, lubrication status, and gearbox alarms; and the main control system is associated with control commands, operating status, and status switching records.

[0058] In a preferred embodiment, when selecting a wind power audit evidence chain template based on the state evidence nodes in the semantic evidence chain to be audited, the types of operating states appearing in the semantic evidence chain to be audited are first counted to form an operating phase set; then, a wind power audit evidence chain template containing the operating phase set is selected from the template library; when the operating phase set includes the power generation phase, shutdown phase, fault phase, recovery phase, and power generation phase, the fault shutdown recovery template is selected; when the operating phase set includes the power generation phase, power limitation phase, and power generation phase, the power limitation recovery template is selected; when the operating phase set includes the power generation phase, fault phase, and shutdown phase but lacks a recovery phase, the fault non-closed-loop template is selected.

[0059] For data review scenarios before wind farm fault outage data is entered into the database, the fault outage recovery template should be selected first. The fault outage recovery template should include at least the following nodes: generation status template node, alarm template node, outage command template node, outage status template node, fault confirmation template node, handling template node, recovery status template node, grid connection command template node, and generation status template node.

[0060] In a preferred embodiment, the division of the power generation stage, power limitation stage, shutdown stage, fault stage, and recovery stage is based on the wind turbine operating status indicators, active power, main shaft speed, alarm status, and reset status. Taking a doubly-fed induction generator (DFIG) with a rated power of 2500kW and a main shaft grid-connected operating speed range of 9rpm to 18rpm as an example, the power generation stage is the stage where the wind turbine is effectively grid-connected, the active power is greater than 125kW, the main shaft speed is between 9rpm and 18rpm, and there is no shutdown protection action. The active power greater than 125kW is 5% of the rated power of 2500kW, and the main shaft speed range of 9rpm to 18rpm is used for limitation. The wind turbine has entered a stable grid-connected power generation state, excluding standby, idling, and the initial startup phase. The power limiting phase is characterized by the power limiting command being in effect, the power setpoint being below 2000kW, the active power decreasing to within 50kW above or below the power setpoint within 60 seconds after the power limiting command is issued, the wind turbine's grid connection status remaining valid, and the main shaft speed remaining between 9rpm and 18rpm. The power setpoint below 2000kW indicates that the wind turbine has transitioned from its normal power generation capacity (corresponding to a rated power of 2500kW) to a limited power generation state, and the main shaft speed remaining between 9rpm and 18rpm distinguishes between power limiting operation and the shutdown speed reduction process. The shutdown phase is characterized by the wind turbine being in grid-connected state. The following are the stages for the wind turbine's operation: Invalid, active power less than 25kW, main shaft speed less than 2rpm, and a shutdown command or protection action has been issued. Specifically, active power less than 25kW is 1% of the rated power of 2500kW, and a main shaft speed less than 2rpm is defined as the shutdown range, indicating that the turbine has entered a shutdown or near-shutdown state. If the turbine's grid connection is invalid but the main shaft speed is still above 2rpm, this period is considered a shutdown transition process and is not directly classified as a shutdown stage until the main shaft speed drops below 2rpm. The fault stage is defined as a stage where the fault code is valid, the fault confirmation result exists, the fault duration is greater than 10 seconds, and the fault cause has not yet been resolved. The fault code should be able to point to at least one component in the pitch system, yaw system, converter, generator, gearbox, or main control system. The fault duration is greater than 10 seconds to eliminate instantaneous sampling anomalies and short-term communication jitter. The recovery phase is the stage where the fault code is cleared, the reset result is successful, the handling result exists, the main shaft speed is restored to more than 3 rpm, and the wind turbine is in the process of waiting to be connected to the grid or the grid connection recovery process. Among them, the main shaft speed is restored to more than 3 rpm to indicate that the wind turbine has left the shutdown state and entered the start-up preparation process. If the grid connection status becomes effective within 30 minutes after the recovery confirmation, and the active power is greater than 125kW, then the recovery phase is followed by the power generation phase.

[0061] It should be noted that a power generation status template node is configured during the power generation phase, a power limitation status template node is configured during the power limitation phase, a shutdown status template node is configured during the shutdown phase, a fault status template node is configured during the fault phase, and a recovery status template node is configured during the recovery phase.

[0062] When configuring template instruction nodes, template alarm nodes, and template handling nodes between adjacent template status nodes, the corresponding template nodes are configured according to the triggering evidence required for the status change; for example, alarm template nodes and shutdown instruction template nodes are configured between the power generation status template node and the shutdown status template node, fault confirmation template nodes are configured between the shutdown status template node and the fault status template node, handling template nodes and reset template nodes are configured between the fault status template node and the recovery status template node, and grid connection instruction template nodes and grid connection feedback template nodes are configured between the recovery status template node and the power generation status template node.

[0063] In a preferred embodiment, when associating the pitch system, yaw system, converter, generator, gearbox, and main control system to their respective template status nodes, template command nodes, template alarm nodes, and template handling nodes, the association is performed according to the engineering attribution relationship between the component objects and the operating status quantities, control command content, alarm event content, and handling result content. Specifically, the pitch system is associated with pitch angle, pitch motor current, pitch limit status, pitch alarm, and pitch reset result; the yaw system is associated with yaw angle, wind direction deviation, yaw motor status, yaw alarm, and yaw reset result; the converter is associated with DC bus voltage, AC output current, module temperature, converter alarm, and converter reset result; the generator is associated with speed, active power, winding temperature, generator alarm, and generator maintenance result; the gearbox is associated with bearing temperature, gearbox vibration characteristics, lubrication status, gearbox alarm, and gearbox maintenance result; and the main control system is associated with control commands, operating status, status switching records, protection actions, and grid connection switching result.

[0064] As an example, for the pre-entry audit event WTG-014, where high gearbox high-speed shaft bearing temperature led to protection shutdown, the wind power audit evidence chain template sequentially includes the following nodes: power generation status template node, gearbox abnormal temperature status template node, gearbox alarm template node, main control protection shutdown command template node, shutdown status template node, gearbox fault confirmation template node, gearbox repair or component replacement template node, reset template node, recovery status template node, grid connection switching template node, and power generation status template node. In this template, the gearbox alarm template node is associated with the gearbox abnormal temperature status template node, the gearbox fault confirmation template node, and the gearbox repair or component replacement template node; the main control protection shutdown command template node is associated with the shutdown status template node; and the grid connection switching template node is associated with the recovery status template node and the power generation status template node.

[0065] In a preferred embodiment, the transition methods in the wind turbine state transition relationship are as follows: When the power generation state transitions to the power limiting state, the semantic evidence chain to be audited should first show the power generation state evidence node, then the power limiting command evidence node, and finally the power limiting state evidence node; when the power generation state transitions to the shutdown state, the semantic evidence chain to be audited should first show the power generation state evidence node, then the shutdown command evidence node or fault alarm evidence node, and finally the shutdown state evidence node; when the shutdown state transitions to the fault state, the semantic evidence chain to be audited should first show the shutdown state evidence node, then the fault confirmation result or fault cause determination result, and finally the fault state evidence node; when the fault state transitions to the recovery state, the semantic evidence chain to be audited should first show the fault state evidence node, then the handling evidence node and the reset evidence node, and finally the recovery state evidence node; when the recovery state transitions to the power generation state, the semantic evidence chain to be audited should first show the recovery state evidence node, then the grid connection switching command evidence node, and finally the grid connection state evidence node; if any of the above transitions lacks intermediate triggering evidence, or the subsequent state precedes the triggering evidence, then the transition does not satisfy the wind turbine state transition relationship.

[0066] In a preferred embodiment, the alarm association method includes: matching the alarm name, alarm code, alarm level, alarm source system, and alarm object in the alarm event content with the physical association relationship of the components; when the alarm name or alarm code belongs to the alarm range of a certain component object, and there is an abnormal operating status quantity of the same component object near the time of alarm occurrence, then the alarm evidence node is associated with the component object; for example, when a high temperature alarm of the gearbox high-speed shaft bearing and an operating status quantity of gearbox bearing temperature greater than 85°C occur together, the alarm evidence node is associated with the gearbox.

[0067] The alarm scope includes the valid alarm code, alarm name, alarm level, and duration range corresponding to each component object in the wind farm alarm management system; among them, the alarm level shall not be lower than level 2, the alarm duration shall be greater than 10s, and the alarm code or alarm name shall be able to uniquely point to one component object in the pitch system, yaw system, converter, generator, gearbox, or main control system.

[0068] In a preferred embodiment, the recording association method includes: mapping the control commands, operating states, state switching records, protection action records, and grid connection switching results generated by the main control system according to the wind turbine number and the time of occurrence; when a state switching record appears after a control command, and the state change in the state switching record is consistent with the control command, the control command is associated with the state switching record; for example, if an invalid grid connection state and a valid shutdown state appear 8 seconds after a shutdown command, the shutdown command is associated with the shutdown state switching record.

[0069] S2.2 Align the status evidence nodes, instruction evidence nodes, alarm evidence nodes, and disposal evidence nodes in the semantic evidence chain to be reviewed with the template nodes in the wind power review evidence chain template to obtain the node alignment results.

[0070] In a preferred embodiment, the rules for type alignment include consistency in evidence category, operational phase, component object, and occurrence order. Specifically, the status evidence nodes in the semantic evidence chain to be audited are aligned with the template status nodes in the wind power audit evidence chain template; the instruction evidence nodes are aligned with the template instruction nodes; the alarm evidence nodes are aligned with the template alarm nodes; and the disposal evidence nodes are aligned with the template disposal nodes. Based on the consistency of evidence categories, the operational phase and component objects are further compared. If the gearbox alarm evidence node in the semantic evidence chain to be audited is located between the power generation status evidence node and the shutdown status evidence node, and there is a gearbox alarm template node in the wind power audit evidence chain template located between the power generation status template node and the shutdown status template node, then the gearbox alarm evidence node is aligned with the gearbox alarm template node. If the evidence categories are consistent but the component objects are inconsistent, then it is only marked as a category candidate and is not considered a node that has completed type alignment.

[0071] The node alignment results include aligned template nodes, aligned evidence nodes, category candidate nodes, and unaligned template nodes.

[0072] S2.3 Mark the template nodes in the node alignment results that are not aligned with the completion type of the semantic evidence chain to be reviewed as missing nodes, and record the missing node positions in the wind power review evidence chain template.

[0073] In a preferred embodiment, when identifying missing nodes and recording their missing positions, the template nodes are first checked to see if they have aligned evidence nodes according to the order of the wind power audit evidence chain templates. If a template node does not have aligned evidence nodes, then the template node is identified as a missing node. The missing position of the node is determined by the previous template node, the next template node, the operating stage, and the component object of the missing node in the wind power audit evidence chain template.

[0074] For example, in the wind power audit evidence chain template, there should be a main control protection shutdown command template node between the gearbox alarm template node and the shutdown status template node. However, if the semantic evidence chain to be audited only contains the gearbox alarm evidence node and the shutdown status evidence node, then the main control protection shutdown command template node will be marked as a missing node, and its missing node position will be recorded as: after the gearbox alarm template node, before the shutdown status template node, the power generation to shutdown transition stage, and the main control system.

[0075] S2.4. According to the wind turbine state transition relationship in the wind power audit evidence chain template, check the state continuity relationship of adjacent state evidence nodes in the node alignment result, and mark the connection position that does not conform to the wind turbine state transition relationship as the state break position.

[0076] In a preferred embodiment, when verifying the state succession relationship, the adjacent state evidence nodes in the semantic evidence chain to be reviewed are checked one by one in sequence. If the previous state evidence node is a power generation state evidence node and the next state evidence node is a shutdown state evidence node, then it is checked whether there is a shutdown command evidence node or a fault alarm evidence node between them. If not, the connection position between them is marked as a state break position. If the previous state evidence node is a fault state evidence node and the next state evidence node is a power generation state evidence node, and there are no disposal evidence node, reset evidence node, or recovery state evidence node between them, then the connection position between the fault state evidence node and the power generation state evidence node is marked as a state break position. If the next state evidence node does not belong to the transferable state of the previous state evidence node in the wind turbine state transition relationship, for example, the shutdown state evidence node is directly converted into a power limitation state evidence node, then the connection position between the adjacent state evidence nodes is marked as a state break position.

[0077] S2.5. Based on the physical relationship of components in the wind power audit evidence chain template, verify the consistency of component objects among status evidence nodes, instruction evidence nodes, alarm evidence nodes and disposal evidence nodes in the node alignment results, and mark the connection position of inconsistent component objects as the component break position.

[0078] Specifically, if the status evidence node, command evidence node, alarm evidence node, and handling evidence node all point to the same component object, then the connection position satisfies the component object consistency relationship. If the component objects of adjacent evidence nodes are different, but they are directly related objects in the component physical association relationship, then the connection position satisfies the component object consistency relationship. If the component objects of adjacent evidence nodes are different and there is no direct association relationship, then the connection position is marked as the component breakage position. For example, if a gearbox handling evidence node appears after a gearbox alarm evidence node, the connection position satisfies the component object consistency relationship. If a main control protection shutdown command evidence node appears after a gearbox alarm evidence node, the connection position is a reasonable association because the main control system is associated with protection actions, operating status, and status switching records. If a yaw system handling evidence node appears directly after a gearbox alarm evidence node, and there is no yaw alarm or abnormal yaw status, then the connection position is marked as the component breakage position.

[0079] S2.6. Merge the state fracture location and component fracture location into the relation fracture location, and output the node missing location and relation fracture location.

[0080] In a preferred embodiment, when merging state fracture locations and component fracture locations into relational fracture locations, the merging is performed according to the wind turbine number, the time of occurrence, the previous evidence node, the next evidence node, and the fracture type.

[0081] If both state breakage and component breakage exist at the same connection location, then both state breakage type and component breakage type will be retained at the same relationship breakage location. If multiple breakage locations occur in the same operating phase and belong to the same component object, they will be arranged in the order of occurrence and output as continuous relationship breakage locations. The final output of node missing locations and relationship breakage locations will include at least the wind turbine number, the operating phase, the previous evidence node, the next evidence node, the missing or breakage type, the component object, and the corresponding supplementary evidence direction.

[0082] For example, if WTG-014 lacks a main control protection shutdown command template node during the power generation to shutdown transition phase, then the output node is missing; if the gearbox alarm evidence node is directly connected to the yaw system maintenance evidence node, then the output relationship is broken, and the break type is component breakage.

[0083] It should be noted that step S2, by aligning the semantic evidence chain to be audited with the wind power audit evidence chain template in terms of type, identifying missing nodes, verifying status continuity, and verifying the consistency of component objects, locates the missing certificate issues and relationship errors in the fault shutdown event to specific node and connection positions. This solves the technical problem in the fault shutdown data audit that it is difficult to judge whether the status change is reasonable and whether the alarm and the handling correspond, even if only missing data items can be found. It provides a clear direction for the subsequent retrieval of supplementary evidence and improves the traceability and consistency of the pre-entry audit.

[0084] S3. Determine the breakpoint type based on the location of missing nodes and broken relationships, select the supplementary evidence path according to the breakpoint type, retrieve the supplementary evidence, and integrate the supplementary evidence into the semantic evidence chain to be reviewed. Note that the following should be noted in this step: S3.1 Mark the locations between adjacent sampling times in the missing node locations as time breakpoints; mark the locations lacking data sources in the missing node locations as source breakpoints; mark the locations with inconsistent state continuity in the broken relationship locations as state breakpoints; mark the locations with inconsistent component objects in the broken relationship locations as causal breakpoints; mark the locations lacking evidence of maintenance, reset, component replacement, or operation recovery in the missing node locations as disposal breakpoints.

[0085] In a preferred embodiment, the method for calibrating time breakpoints is as follows: when the missing node location is located between adjacent sampling times of the same wind turbine number, and the time interval between the previous evidence node and the next evidence node is greater than the normal interval between adjacent sampling times, the missing node location is calibrated as a time breakpoint; for example, if the sampling interval of the SCADA system is 10s, and WTG-014 is missing the operating status quantities of 2026-05-12 10:18:10 and 2026-05-12 10:18:20 between 2026-05-12 10:18:00 and 2026-05-12 10:18:30, then this location is calibrated as a time breakpoint.

[0086] In a preferred embodiment, the method for identifying source breakpoints is as follows: when the template node corresponding to the missing node position is required to come from a source system among the main control system, alarm management system, SCADA system, or maintenance management system, but the corresponding evidence content of that source system is not present in the semantic evidence chain to be reviewed, the missing node position is marked as a source breakpoint; for example, if the shutdown status evidence node already exists, but the main control system lacks shutdown instruction evidence content, then that position is marked as a source breakpoint.

[0087] In a preferred embodiment, the method for calibrating state breakpoints is as follows: when the breakpoint is caused by the adjacent state evidence nodes not satisfying the wind turbine state transition relationship, the breakpoint is calibrated as a state breakpoint.

[0088] In a preferred embodiment, the method for identifying causal breakpoints is as follows: when the breakpoint in a relationship is caused by inconsistencies between component objects among alarm evidence nodes, instruction evidence nodes, status evidence nodes, and disposal evidence nodes, the breakpoint in the relationship is identified as a causal breakpoint.

[0089] In a preferred embodiment, the method for marking the handling breakpoint is as follows: when the missing node location is located after the alarm evidence node or fault status evidence node, and there is a lack of evidence for repair, reset, component replacement, or operation recovery, the missing node location is marked as the handling breakpoint.

[0090] S3.2. Retrieve adjacent wind power operation segments by time breakpoint, retrieve corresponding source logs by source breakpoint, retrieve main control status records by status breakpoint, retrieve component operation records by cause-effect breakpoint, and retrieve maintenance and handling records by handling breakpoint to obtain supplementary evidence.

[0091] In a preferred embodiment, when retrieving adjacent wind power operation segments by time breakpoint, adjacent wind power operation segments are obtained by extending forward and backward by 5 minutes based on the occurrence times of the preceding and following evidence nodes of the time breakpoint; adjacent wind power operation segments include at least the operation status quantities and alarm event contents of the same wind turbine number; when retrieving the corresponding source log by source breakpoint, control command contents, state switching records, or alarm event contents are obtained from the source system according to the source system name, wind turbine number, and occurrence time range corresponding to the source breakpoint; when retrieving the main control status record by state breakpoint, the main control status log is retrieved. During recording, based on the previous and next state evidence nodes of the state breakpoint, the state switching record, protection action record, and grid connection switching result of the main control system are obtained; when retrieving component operation records by causal breakpoint, based on the component object corresponding to the causal breakpoint, the operating status quantity, alarm event content, and associated control instruction content of the component object within 30 minutes before and after the breakpoint are obtained; when retrieving maintenance and handling records by handling breakpoint, based on the wind turbine number, component object, and alarm occurrence time corresponding to the handling breakpoint, the maintenance work order, reset result, component replacement result, and trial operation result are obtained, thus obtaining supplementary evidence.

[0092] For example, the semantic evidence chain to be audited in WTG-014 contains evidence nodes for a high gearbox high-speed shaft bearing temperature alarm at 10:18:08 on 2026-05-12 and a shutdown status evidence node at 10:18:20 on 2026-05-12, but lacks a main control protection shutdown command evidence node. This missing node is marked as a source breakpoint. For this source breakpoint, the control command content from 10:18:00 to 10:18:30 on 2026-05-12 is obtained from the main control system, and the protection shutdown command at 10:18:12 on 2026-05-12 is obtained. The shutdown reason is high gearbox high-speed shaft bearing temperature, and the command feedback result is successful reception. This protection shutdown command is the supplementary evidence. If the same event also lacks evidence for 2026-05-12... At the 11:36:00 gearbox disposal evidence node, the maintenance disposal record is obtained from the maintenance management system for the disposal breakpoint. The disposal result is obtained that the gearbox high-speed shaft bearing temperature sensor was replaced and the normal operation was achieved after 30 minutes of trial run. This disposal result is used as supplementary evidence and is connected to the semantic evidence chain to be reviewed.

[0093] S3.3. Connect the supplementary evidence to the missing node or broken relationship position in the semantic evidence chain to be reviewed, according to the time of occurrence, wind turbine number and component object.

[0094] Specifically, when integrating supplementary evidence into the semantic evidence chain to be reviewed, the integration order is first determined based on the occurrence time of the supplementary evidence, then the integration object is determined based on the wind turbine number of the supplementary evidence, and finally the integration position is determined based on the component object of the supplementary evidence. If the corresponding node of the supplementary evidence is missing, the supplementary evidence is converted into an evidence node of the corresponding type and placed between the previous and next evidence nodes. If the correspondence of the supplementary evidence is broken, the supplementary evidence is placed between the two ends of the broken connection, and a new state connection or component object connection is established with the previous and next evidence nodes. After the supplementary evidence is integrated, its source system, occurrence time, and source are retained so that the original evidence and supplementary evidence can be distinguished in the subsequent review results.

[0095] For example, after the protection shutdown command at 10:18:12 on 2026-05-12 is accessed as supplementary evidence, it is located between the gearbox alarm evidence node at 10:18:08 on 2026-05-12 and the shutdown status evidence node at 10:18:20 on 2026-05-12, forming a causal connection with the gearbox alarm evidence node and a status succession connection with the shutdown status evidence node.

[0096] It should be noted that step S3 transforms missing node locations and broken relationship locations into traceable supplementary evidence operation objects by classifying breakpoint types, selecting supplementary evidence paths, and accessing supplementary evidence. This solves the problem of lacking targeted supplementary evidence paths after missing evidence is discovered before the data is entered into the database during a system downtime due to a fault. It also avoids the indiscriminate retrieval of large amounts of historical data, which would reduce audit efficiency. Furthermore, by differentiating between time breakpoints, source breakpoints, status breakpoints, causal breakpoints, and disposal breakpoints, it improves the accuracy of the correspondence between supplementary evidence and audit anomaly locations.

[0097] S4. Verify the node order, source consistency, state continuity, causal explanation, and corresponding handling of the semantic evidence chain to be reviewed after supplementing the evidence, and output the data review results. Note that the following should be noted in this step: S4.1. Verify the node order of the semantic evidence chain to be reviewed according to the time of occurrence. When the occurrence time of the later evidence node is earlier than the occurrence time of the previous evidence node, or the occurrence time of the instruction evidence node is later than the occurrence time of the state evidence node triggered by the instruction, mark the corresponding evidence node as an abnormal node in the order.

[0098] In a preferred embodiment, when verifying the node order, the occurrence times of adjacent evidence nodes are compared one by one according to the connection order in the semantic evidence chain to be reviewed; if the occurrence time of the later evidence node is earlier than the occurrence time of the previous evidence node, the later evidence node is marked as an order abnormal node; if an instruction evidence node is connected to a state evidence node triggered by the instruction, but the occurrence time of the instruction evidence node is later than the occurrence time of the state evidence node, both the instruction evidence node and the state evidence node are marked as order abnormal nodes.

[0099] For example, if the shutdown status evidence node occurs at 10:18:20 on 2026-05-12, while the protection shutdown command evidence node occurs at 10:18:35 on 2026-05-12, then the protection shutdown command evidence node is later than the shutdown status evidence node it triggered, and both are marked as nodes with abnormal order. If the alarm evidence node occurs at 10:18:08 on 2026-05-12, the protection shutdown command evidence node occurs at 10:18:12 on 2026-05-12, and the shutdown status evidence node occurs at 10:18:20 on 2026-05-12, then these three evidence nodes meet the node order verification requirements.

[0100] S4.2 Verify the consistency of the source of the semantic evidence chain to be audited by wind turbine number and component object. When the wind turbine numbers of adjacent evidence nodes are inconsistent, or the component objects of adjacent evidence nodes are inconsistent and do not belong to the physical association relationship of components, mark the corresponding evidence node as an abnormal source node.

[0101] In a preferred embodiment, when verifying source consistency, first check whether the fan numbers of adjacent evidence nodes are consistent; if the fan numbers of adjacent evidence nodes are inconsistent, then mark the adjacent evidence node as a source abnormal node; then check whether the component objects of adjacent evidence nodes are consistent; if the component objects are consistent, then pass the source consistency verification; if the component objects are inconsistent, then continue to check whether the component object belongs to the component physical association relationship; adjacent evidence nodes that belong to the component physical association relationship can pass the source consistency verification, and adjacent evidence nodes that do not belong to the component physical association relationship are marked as source abnormal nodes.

[0102] For example, if the gearbox alarm evidence node of WTG-014 is connected to the main control protection shutdown command evidence node of WTG-014, the connection is verified by source consistency because the main control system is associated with the protection action and status switching records; if the gearbox alarm evidence node of WTG-014 is connected to the gearbox handling evidence node of WTG-021, the turbine numbers are inconsistent, and the two are marked as abnormal source nodes; if the gearbox alarm evidence node of WTG-014 is connected to the yaw maintenance handling evidence node of WTG-014, and there is no yaw alarm or abnormal yaw status, then the yaw maintenance handling evidence node is marked as an abnormal source node.

[0103] S4.3. Verify the state succession relationship of the semantic evidence chain to be audited according to the wind turbine state transfer relationship. When there is a lack of corresponding instruction evidence node or alarm evidence node between adjacent state evidence nodes, or when the later state evidence node does not belong to the transferable state of the previous state evidence node, mark the corresponding state evidence node as a state abnormal node.

[0104] In a preferred embodiment, when verifying the state succession relationship, the state changes between adjacent state evidence nodes are checked according to the wind turbine state transition relationship to see if there is corresponding triggering evidence; when a power generation state evidence node changes to a power limiting state evidence node, there should be a power limiting command evidence node between them; when a power generation state evidence node changes to a shutdown state evidence node, there should be a shutdown command evidence node or alarm evidence node between them; when a shutdown state evidence node changes to a fault state evidence node, there should be a fault confirmation result between them; when a fault state evidence node changes to a recovery state evidence node, there should be a handling evidence node and a reset evidence node between them; when a recovery state evidence node changes to a power generation state evidence node, there should be a grid connection switching command evidence node between them; if the adjacent state evidence nodes lack the above-mentioned corresponding triggering evidence, or the latter state evidence node does not belong to the transferable state of the former state evidence node, then the latter state evidence node is marked as a state abnormal node.

[0105] For example, if a power generation status evidence node appears directly after a fault status evidence node, and there is no recovery status evidence node or grid connection switching instruction evidence node between them, then the power generation status evidence node is marked as an abnormal status node.

[0106] S4.4 Verify the causal interpretation relationship of the semantic evidence chain to be audited according to the physical relationship of the components. When the component objects between the instruction evidence node, alarm evidence node and status evidence node are inconsistent, or when the status change evidence node of the same component object is missing after the instruction evidence node, the corresponding evidence node will be marked as a causal abnormal node.

[0107] In a preferred embodiment, when verifying the causal explanation relationship, firstly, check whether the component objects among the instruction evidence node, alarm evidence node, and status evidence node are consistent or directly related according to the physical correlation of components; then check whether there are status change evidence nodes for the same component object or directly related component objects after the instruction evidence node; if the component object of the alarm evidence node is a gearbox, its subsequent status evidence nodes should reflect gearbox-related status changes, main control protection actions, or overall machine shutdown status; if the instruction evidence node is a power limiting instruction, its subsequent status evidence nodes should reflect power limiting status or active power reduction; if the instruction evidence node is a reset instruction, its subsequent status evidence nodes should reflect fault status clearing or recovery status; if there is no corresponding status change evidence node after the instruction evidence node, or if the component objects of the alarm evidence node and the handling evidence node are inconsistent and not directly related, then the corresponding evidence node is marked as a causal abnormal node.

[0108] For example, if a yaw reset evidence node appears after a high temperature alarm of the high-speed shaft bearing in the gearbox, but no gearbox repair or reset evidence node appears, then the yaw reset evidence node is marked as a causal abnormal node.

[0109] S4.5. Verify the processing correspondence of the semantic evidence chain to be audited according to the component object and the time of occurrence. When there is no processing evidence node for the same component object after the alarm evidence node, or when the processing evidence node is earlier than the corresponding alarm evidence node, mark the corresponding evidence node as a processing abnormal node.

[0110] In a preferred embodiment, when verifying the correspondence between alarm and response, the system checks whether a corresponding response evidence node exists after each alarm evidence node according to the component object and the time of occurrence. If the alarm evidence node is a component alarm, there should be evidence of repair, reset, component replacement, or operation recovery for the same component object after the alarm evidence node. If the alarm evidence node is a communication anomaly, there should be a communication recovery result, master control reset result, or maintenance confirmation result after the alarm evidence node. If the response evidence node is earlier than the corresponding alarm evidence node, the response evidence node is marked as a response anomaly node. If there is still no corresponding response evidence node after the alarm evidence node before the end time of the wind power operation segment to be reviewed, the alarm evidence node is marked as a response anomaly node.

[0111] For example, there is a gearbox alarm evidence node at 10:18:08 on 2026-05-12, and a gearbox component replacement evidence node at 11:36:00 on 2026-05-12. The two are verified through the corresponding relationship of the two. If the evidence node for the handling at 11:36:00 on 2026-05-12 is yaw system maintenance, and there is no gearbox handling evidence node, then the gearbox alarm evidence node at 10:18:08 on 2026-05-12 is marked as an abnormal handling node.

[0112] S4.6. Based on the sequence anomaly nodes, source anomaly nodes, status anomaly nodes, causal anomaly nodes, and handling anomaly nodes, output the data audit results; including: If no abnormal nodes are found, the approved data review result will be output. When there are abnormal nodes and the number of abnormal nodes is less than the preset threshold, the audit result of the data with doubts will be output. When the number of abnormal nodes reaches or exceeds the preset threshold, or when there is any one of the key abnormal nodes among the source abnormal node, state abnormal node, and causal abnormal node, the data audit result of failing the audit will be output. The preset threshold is 15% of the total number of evidence nodes in the semantic evidence chain to be reviewed, and 15% is the statistical median of the proportion of abnormal nodes in the historically approved wind turbine operation and maintenance evidence chain samples.

[0113] In a preferred embodiment, when outputting the data review results, the total number of evidence nodes in the semantic evidence chain to be reviewed is first counted, and then the number of abnormal nodes in sequence, abnormal nodes in source, abnormal nodes in state, abnormal nodes in causality, and abnormal nodes in handling are counted. The results are then determined based on the ratio of the number of abnormal nodes to the total number of evidence nodes.

[0114] In a specific example, the semantic evidence chain to be reviewed for WTG-014 has 20 evidence nodes after supplementary evidence is added. The judgment limit corresponding to 15% is 3 abnormal nodes. If no abnormal nodes are found after verification by S4.1 to S4.5, the review is approved and the fault shutdown data can enter the fault shutdown data entry process. If there are 2 abnormal handling nodes and no abnormal source nodes, abnormal status nodes, or abnormal causal nodes, the review is questionable and the alarm evidence nodes and handling evidence nodes corresponding to the 2 abnormal handling nodes are submitted for manual review. If there is 1 abnormal source node, such as the gearbox alarm evidence node of WTG-014 being incorrectly connected to the handling evidence node of WTG-021, the review is rejected even if the number of abnormal nodes is less than 3. If there are 4 abnormal nodes, the review is rejected even if all 4 abnormal nodes are either sequential abnormal nodes or handling abnormal nodes.

[0115] It should be noted that step S4 verifies the node order, source consistency, state succession relationship, causal explanation relationship, and handling correspondence of the semantic evidence chain to be audited after supplementary verification. This ensures that the audit judgment before the fault shutdown data is entered into the database is implemented to specific evidence nodes and specific anomaly types. This solves the technical problem that relying solely on the integrity of a single time or a single alarm for the audit of fault shutdown data makes it impossible to discover cross-system errors, cross-component errors, and causal errors. This improves the quality of fault shutdown data entering the database and reduces the impact of erroneous fault causes, erroneous handling objects, and erroneous state transitions on subsequent fault statistics, operation and maintenance analysis, and wind turbine reliability evaluation.

[0116] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A data verification method based on semantic evidence chain, characterized in that, include: Extract status, instruction, alarm, and handling evidence nodes from the wind power operation segments to be reviewed, and form a semantic evidence chain to be reviewed; The semantic evidence chain to be audited is compared with the wind power audit evidence chain template constructed based on the wind turbine state transition relationship and component physical association relationship to anchor the missing node position and the relationship break position. The breakpoint type is determined based on the missing node location and the broken relationship location. A supplementary evidence path is selected according to the breakpoint type, and supplementary evidence is retrieved. The supplementary evidence is then connected to the semantic evidence chain to be reviewed. Verify the node order, source consistency, state succession relationship, causal explanation relationship, and disposal correspondence of the semantic evidence chain to be reviewed after the supplementary evidence is accessed, and output the data review results.

2. The data verification method based on semantic evidence chain according to claim 1, characterized in that, Extract status, instruction, alarm, and handling evidence nodes from the wind power operation segments to be reviewed, including: The operating status quantities, control command contents, alarm event contents, and handling result contents in the wind power operation segment to be audited are merged according to the wind turbine number and the time of occurrence to obtain the same machine time sequence evidence segment; The operating state quantities representing grid connection, shutdown, power limitation, fault and recovery in the same machine time sequence evidence segment are labeled as state evidence nodes; The control command content representing start-up, shutdown, reset, power limiting, and grid connection switching in the same machine timing evidence segment is marked as the command evidence node; The alarm event content representing component alarms, protection actions, and communication anomalies in the same machine time sequence evidence segment is marked as alarm evidence nodes; The content representing the handling results of maintenance, reset, component replacement and operation recovery in the same time sequence evidence segment is marked as the handling evidence node.

3. The data verification method based on semantic evidence chain according to claim 2, characterized in that, The process of forming a semantic evidence chain to be reviewed includes: arranging the status evidence nodes, instruction evidence nodes, alarm evidence nodes, and disposal evidence nodes in the order of their occurrence; connecting adjacent evidence nodes with the same wind turbine number and the same component object; and connecting adjacent evidence nodes with the same wind turbine number and consecutive states to obtain the semantic evidence chain to be reviewed.

4. The data verification method based on semantic evidence chain according to claim 3, characterized in that, The locations of missing anchor nodes and broken relationships include: Based on the state evidence nodes in the semantic evidence chain to be reviewed, select a wind power review evidence chain template that contains the same wind turbine operation stage. The status evidence nodes, instruction evidence nodes, alarm evidence nodes, and disposal evidence nodes in the semantic evidence chain to be reviewed are respectively aligned with the template nodes in the wind power review evidence chain template to obtain the node alignment result. Template nodes in the node alignment results that are not aligned with the completion type of the semantic evidence chain to be reviewed are marked as missing nodes, and the missing node positions in the wind power review evidence chain template are recorded. According to the wind turbine state transition relationship in the wind power audit evidence chain template, the state succession relationship of adjacent state evidence nodes in the node alignment result is checked, and the connection position that does not conform to the wind turbine state transition relationship is marked as the state break position. Based on the physical association of components in the wind power audit evidence chain template, verify the consistency of component objects among the status evidence node, instruction evidence node, alarm evidence node and disposal evidence node in the node alignment result, and mark the connection position of inconsistent component objects as the component break position. The state fracture location and the component fracture location are merged into the relationship fracture location, and the node missing location and the relationship fracture location are output.

5. The data verification method based on semantic evidence chain according to claim 4, characterized in that, The wind turbine state transition relationship includes the state transition relationship between power generation state, power limitation state, shutdown state, fault state and recovery state; Specifically, the power generation state is converted to the power limiting state after receiving a power limiting command; the power generation state is converted to the shutdown state after receiving a shutdown command or generating a fault alarm; the shutdown state is converted to the fault state after completing fault confirmation; the fault state is converted to the recovery state after completing fault handling; and the recovery state is converted back to the power generation state after receiving a grid connection command.

6. The data verification method based on semantic evidence chain according to claim 4, characterized in that, The physical relationships between the components include the relationships between the pitch system, yaw system, converter, generator, gearbox, and main control system; Specifically, the pitch system is associated with pitch angle, pitch motor current, and pitch alarm; the yaw system is associated with yaw angle, wind direction deviation, and yaw alarm; the converter is associated with DC bus voltage, AC output current, module temperature, and converter alarm; the generator is associated with speed, active power, winding temperature, and generator alarm; the gearbox is associated with bearing temperature, gearbox vibration characteristics, lubrication status, and gearbox alarm; and the main control system is associated with control commands, operating status, and status switching records.

7. The data verification method based on semantic evidence chain according to claim 1, characterized in that, The method for constructing the wind power audit evidence chain template includes: The wind turbine operation is divided into power generation stage, power limitation stage, shutdown stage, fault stage and recovery stage, and a corresponding template status node is configured for each wind turbine operation stage. Based on the wind turbine state transition relationship, template instruction nodes, template alarm nodes, and template handling nodes are configured between adjacent template state nodes to form a state transition template chain; Based on the physical relationships between the components, the pitch system, yaw system, converter, generator, gearbox, and main control system are respectively associated with the corresponding template status node, template command node, template alarm node, and template handling node; The wind power audit evidence chain template is generated according to the order of nodes in the state transition template chain and the component association results.

8. The data verification method based on semantic evidence chain according to claim 1, characterized in that, Integrating the supplementary evidence into the semantic evidence chain to be reviewed includes: The missing node locations located between adjacent sampling times are marked as time breakpoints; the missing node locations lacking data sources are marked as source breakpoints; the broken relationship locations with inconsistent state continuity are marked as state breakpoints; the broken relationship locations with inconsistent component objects are marked as causal breakpoints; and the missing node locations lacking evidence of maintenance, reset, component replacement, or operation recovery are marked as disposal breakpoints. Retrieve adjacent wind power operation segments according to the time breakpoints, retrieve corresponding source logs according to the source breakpoints, retrieve main control status records according to the status breakpoints, retrieve component operation records according to the causal breakpoints, and retrieve maintenance and handling records according to the handling breakpoints to obtain supplementary evidence. The supplementary evidence is inserted into the missing node or broken relationship position in the semantic evidence chain to be reviewed, according to the time of occurrence, wind turbine number, and component object.

9. The data verification method based on semantic evidence chain according to claim 8, characterized in that, The verification of the node order, source consistency, state succession relationship, causal explanation relationship, and disposal correspondence of the semantic evidence chain to be reviewed after the supplementary evidence is incorporated includes: The order of nodes in the semantic evidence chain to be audited is checked according to the time of occurrence. When the occurrence time of the later evidence node is earlier than the occurrence time of the previous evidence node, or the occurrence time of the instruction evidence node is later than the occurrence time of the state evidence node triggered by the instruction, the corresponding evidence node is marked as an abnormal node in the order. The consistency of the source of the semantic evidence chain to be audited is verified by wind turbine number and component object. When the wind turbine number of adjacent evidence nodes is inconsistent, or the component object of adjacent evidence nodes is inconsistent and does not belong to the physical association relationship of the component, the corresponding evidence node is marked as an abnormal source node. The state succession relationship of the semantic evidence chain to be audited is verified according to the wind turbine state transition relationship. When there is a lack of corresponding instruction evidence node or alarm evidence node between adjacent state evidence nodes, or when the subsequent state evidence node does not belong to the transitional state of the previous state evidence node, the corresponding state evidence node is marked as a state abnormal node. The causal interpretation relationship of the semantic evidence chain to be audited is verified according to the physical relationship of the components. When the component objects between the instruction evidence node, alarm evidence node and status evidence node are inconsistent, or when the status change evidence node of the same component object is missing after the instruction evidence node, the corresponding evidence node is marked as a causal abnormal node. The processing correspondence of the semantic evidence chain to be audited is verified according to the component object and the time of occurrence. When there is no processing evidence node for the same component object after the alarm evidence node, or when the processing evidence node is earlier than the corresponding alarm evidence node, the corresponding evidence node is marked as a processing abnormal node.

10. The data verification method based on semantic evidence chain according to claim 9, characterized in that, Based on the sequence anomaly node, the source anomaly node, the state anomaly node, the causal anomaly node, and the handling anomaly node, output the data audit results, including: If no abnormal nodes are found, the approved data review result will be output. When there are abnormal nodes and the number of abnormal nodes is less than the preset threshold, the audit result of the data with doubts will be output. When the number of abnormal nodes reaches or exceeds the preset threshold, or when there is any one of the key abnormal nodes among the source abnormal node, state abnormal node, and causal abnormal node, the data audit result of failing the audit will be output. The preset threshold is 15% of the total number of evidence nodes in the semantic evidence chain to be reviewed, and the 15% is the statistical median of the proportion of abnormal nodes in the historically approved wind turbine operation and maintenance evidence chain samples.

Images