Offshore wind power intelligent operation and maintenance management platform based on meteorological data

Abstract

The present application relates to offshore wind power internet platform layer technical field, the present application discloses a kind of offshore wind power wisdom operation and management platform based on meteorological data, including data construction module, abnormality discrimination module, fault prediction and health management module, window analysis module, risk decision module and dispatch feedback module.The platform is by constructing target wind turbine single machine meteorological influence vector, extracts operating abnormality feature, calculates abnormal weather interpretation degree and equipment degradation credibility, realizes the dynamic correction of wind turbine fault prediction weight.Combining current weather window and next weather window task completion probability, calculate the window miss increment risk between the waiting path and the current window repair path, generate optimization operation and maintenance strategy.Through distinguishing and processing meteorological disturbance type exception and equipment degradation type exception, and quantifying incremental risk, the platform can improve the accuracy and reliability of wind power operation and maintenance strategy decision, provide intelligent, efficient and safe operation and maintenance guarantee for offshore wind turbine.

Description

Technical Field

[0001] This invention relates to the field of offshore wind power internet platform technology, and in particular to an intelligent operation and maintenance management platform for offshore wind power based on meteorological data. Background Technology

[0002] Offshore wind turbines are typically deployed in marine environments far from land, and their operation is easily affected by meteorological and sea condition factors such as wind speed, wind direction, gusts, turbulence, wave height, tides, ocean currents, visibility, rainfall, lightning, the outer effects of typhoons, high humidity, and salt spray. Meanwhile, the operation and maintenance of offshore wind turbines are also limited by factors such as ship navigation conditions, personnel boarding safety conditions, safe return time, spare parts availability, maintenance time, and the duration of meteorological windows. Therefore, intelligent operation and maintenance management of offshore wind power typically requires a comprehensive consideration of equipment operating status, fault prediction results, health management results, meteorological window conditions, and the status of operation and maintenance resources.

[0003] Several offshore wind power operation and maintenance management solutions are already capable of accessing wind turbine operating data, fault alarm data, historical maintenance data, meteorological and sea condition data, and operation and maintenance resource data. Based on one or more of these data sources, they can perform fault prediction, health assessment, maintenance plan generation, or operation and maintenance resource scheduling. For example, some existing technologies can determine whether there are future weather windows suitable for departure, boarding, maintenance, and return based on meteorological and sea condition information such as wind speed, wave height, visibility, tides, and lightning warnings. Some solutions can perform fault prediction or health status assessment based on operating data such as wind turbine vibration, temperature, power, speed, yaw, pitch, and alarms. Still other solutions can generate maintenance plans by combining information on vessels, personnel, spare parts, and maintenance duration.

[0004] However, in actual operation of offshore wind turbines, abnormal turbine operation is not necessarily caused by equipment degradation. Short-term gusts, strong turbulence, sudden changes in wave height, typhoon periphery disturbances, and high humidity and salt spray can also cause increased turbine vibration, temperature fluctuations, power fluctuations, increased yaw and pitch frequency, or abnormal alarm frequencies. These operational responses caused by meteorological disturbances may exhibit similarities in data to the early degradation of components such as bearings, gearboxes, generators, converters, yaw systems, or pitch systems.

[0005] If the intelligent operation and maintenance platform directly performs fault prediction and health management based on the original vibration value, original temperature rise value, power fluctuation value, or original fault alarm level, it is easy to misjudge recoverable anomalies caused by weather disturbances as equipment failures, leading to unnecessary overseas deployments, premature maintenance, or occupation of limited weather windows. Conversely, if an operational anomaly is simply attributed to weather fluctuations without further determining whether the anomaly is persistent, synergistic, or related to historical faults, it may also lead to delayed handling of equipment degradation anomalies.

[0006] Furthermore, existing operation and maintenance strategies typically focus more on "whether there is a risk of failure" and "whether there is an available weather window," but lack a mechanism to determine whether the current weather window must be used based on the nature of the anomaly and the difference in path risk. For example, if the current weather window is available but the next weather window is uncertain, the platform needs to determine: if maintenance is not carried out during the current weather window, can the target wind turbine safely wait for the next weather window by not processing and waiting, operating with limited power, monitoring after remote reset, or waiting with shutdown; if waiting for the next weather window, will the target wind turbine experience a escalation of the fault, shutdown, or an increase in the maintenance level; and even if the next weather window is available, is there a sufficient probability of task completion?

[0007] Furthermore, even if existing solutions use meteorological and sea condition data, operational status data, and maintenance resource data simultaneously, they usually only use meteorological and sea condition data as external operational window constraints or as ordinary environmental input variables in fault prediction models. It is difficult to first determine whether "operational abnormalities are caused by meteorological disturbances" and then correct the fault prediction and current meteorological window occupancy decision accordingly.

[0008] Therefore, there is a need for an intelligent operation and maintenance management platform for offshore wind power based on meteorological data, which can distinguish between meteorological disturbance-type anomalies and equipment degradation-type anomalies during fault prediction and health management, and determine whether the current meteorological window should be used based on the path risk difference between the current meteorological window maintenance path and the waiting path. Summary of the Invention

[0009] The technical problem to be solved by this invention is that existing offshore wind power intelligent operation and maintenance platforms have the disadvantage of being unable to distinguish between meteorological disturbances and equipment degradation anomalies, and are unable to quantify the incremental risks of the current window and waiting path, resulting in inaccurate maintenance strategies. To address this, we propose an offshore wind power intelligent operation and maintenance management platform based on meteorological data.

[0010] To achieve the above objectives, this application adopts the following technical solution: a smart operation and maintenance management platform for offshore wind power based on meteorological data, including a data construction module, an anomaly identification module, a fault prediction and health management module, a window analysis module, a risk decision-making module, and a scheduling feedback module.

[0011] The data construction module is used to receive meteorological and sea condition data, wind turbine operation data, fault alarm data, historical maintenance data, and operation and maintenance resource data from offshore wind farms. It also performs time alignment of the meteorological and sea condition data, wind turbine operation data, and fault alarm data according to the wind turbine number and collection time, and generates a single-unit meteorological influence vector corresponding to the target wind turbine based on the turbine location information and meteorological and sea condition data.

[0012] The anomaly detection module is used to determine an anomaly analysis time window based on the occurrence time of the anomaly when the target wind turbine experiences an operational anomaly, and to extract the operational anomaly characteristics of the target wind turbine within the anomaly analysis time window. The operational anomaly characteristics include at least one of the following: abnormal vibration, abnormal temperature rise, abnormal power fluctuation, abnormal yaw action, abnormal pitch action, or abnormal alarm frequency.

[0013] The anomaly detection module is further configured to calculate the anomalous meteorological explanatory power corresponding to the operational anomaly feature based on the single-unit meteorological influence vector and the operational anomaly feature, using at least two of the following: time synchronization information, trend consistency information, duration matching information, and recovery information after the meteorological event ends. The anomalous meteorological explanatory power characterizes the degree to which the operational anomaly feature can be explained by meteorological disturbances.

[0014] The anomaly detection module is also used to generate a device degradation confidence value based on the intensity of the operational anomaly, the duration of the operational anomaly, the synergy between multiple operational anomalies, the interpretability of the abnormal meteorological conditions, and historical fault correlation information. The device degradation confidence value is used to characterize the degree to which the operational anomaly characteristics are caused by the device's own degradation.

[0015] The equipment degradation confidence value does not mechanically decrease simply as the explanatory power of abnormal weather increases. When an operational anomaly recovers after a weather event, lacks synergy among multiple operational anomalies, and lacks historical fault association information, the explanatory power of abnormal weather has a deduction effect on the equipment degradation confidence value. When an operational anomaly has a certain meteorological correlation but continues to exist or intensify after a weather event, or is associated with similar historical faults, the platform maintains or increases the equipment degradation confidence value.

[0016] The fault prediction and health management module is used to perform fault prediction and health management on the target wind turbine based on the equipment degradation confidence value, and generate fault prediction results and health degradation curves for the target wind turbine. Specifically, when generating the fault prediction results, for operational anomaly features where the abnormal meteorological explanatory power meets a preset high explanatory power condition and the equipment degradation confidence value meets a preset low confidence value condition, the fault prediction weight of the operational anomaly feature is reduced; for operational anomaly features where the abnormal meteorological explanatory power meets a preset low explanatory power condition and the equipment degradation confidence value meets a preset high confidence value condition, the fault prediction weight of the operational anomaly feature is increased.

[0017] The window analysis module is used to generate the task completion probability of the current meteorological window and the task completion probability of the next meteorological window based on future meteorological and sea condition data and operation and maintenance resource data.

[0018] The risk decision-making module is used to construct, based on the equipment degradation confidence value, the health degradation curve, the task completion probability of the current weather window, and the task completion probability of the next weather window, at least one of the following waiting paths: a maintenance path for the current weather window and a waiting path that abandons the current weather window and waits for the next weather window. The waiting path includes at least one of the following: a no-processing waiting path, a power-limiting waiting path, a remote reset followed by monitoring waiting path, or a shutdown waiting path.

[0019] The risk decision module is also used to calculate the path risk of the waiting path by using the path risk of the current meteorological window maintenance path as the benchmark risk, and to use the difference between the path risk of the waiting path and the benchmark risk as the window missed incremental risk.

[0020] The incremental risk of missing the window is not simply the difference in sea exit costs or maintenance scheduling priority, but is related to at least one or more of the following: risk of increased equipment degradation confidence value, risk of decreased health index, risk of fault expansion, risk of downtime, risk of increased maintenance level, and risk of failure to complete the task in the next meteorological window. It is used to characterize the increased equipment operation risk to the safe operation and maintenance feasibility of the target wind turbine due to abandoning the current meteorological window.

[0021] The scheduling feedback module is used to generate an operation and maintenance strategy for the target wind turbine based on the incremental risk of missed windows. The operation and maintenance strategy includes at least one of the following: maintenance in the current meteorological window, maintenance in the next meteorological window, power limiting and waiting, observation after remote reset, shutdown and waiting for maintenance, or continuous monitoring.

[0022] The operation and maintenance strategy includes not only work orders or scheduling results, but also control strategies oriented towards the target wind turbine operating status. The control strategies include at least one of power-limited operation, monitoring after remote reset, shutdown for maintenance, or immediate maintenance within the current meteorological window. The control strategies are jointly triggered by the abnormal meteorological interpretability, equipment degradation confidence value, and incremental risk of window miss, rather than by the original alarm level or meteorological window availability alone.

[0023] The technical effects and advantages of this invention are as follows: In this invention, to address the issue of easily confusing meteorological disturbance-type anomalies with equipment degradation-type anomalies, the invention first generates a single-unit meteorological influence vector based on the target wind turbine's location information and meteorological and sea state data. Then, it calculates the anomaly meteorological explanatory power based on the single-unit meteorological influence vector and operational anomaly characteristics, and further generates an equipment degradation confidence value. Since the anomaly meteorological explanatory power reflects the degree to which operational anomalies can be explained by meteorological disturbances, and the equipment degradation confidence value reflects the degree of confidence that operational anomalies are caused by the equipment's own degradation, the platform can reduce the misjudgment of meteorological disturbance-type anomalies as equipment malfunctions and also reduce the possibility of incorrectly attributing genuine equipment degradation anomalies to meteorological disturbances.

[0024] In this invention, to address the issue that fault prediction results are easily affected by meteorological disturbances, the invention adjusts the fault prediction weight of operational anomaly features based on the explanatory power of abnormal weather and the confidence value of equipment degradation. When the explanatory power of abnormal weather is high and the confidence value of equipment degradation is low, the fault prediction weight of that operational anomaly feature is reduced; when the explanatory power of abnormal weather is low and the confidence value of equipment degradation is high, the fault prediction weight of that operational anomaly feature is increased; when the explanatory power of abnormal weather is high but the confidence value of equipment degradation has not decreased, the contribution of that operational anomaly feature to the fault prediction results and health degradation curve is retained. Therefore, the fault prediction results and health degradation curve can more effectively reflect the equipment's own degradation risk.

[0025] In this invention, addressing the difficulty existing platforms face in determining whether a current weather window must be occupied, the invention uses the path risk of the maintenance path within the current weather window as the baseline risk. The difference between the path risk of the executable waiting path and the baseline risk is used as the incremental risk of window miss. Based on this incremental risk, the invention outputs maintenance strategies such as continuous monitoring, observation after remote reset, power-limited waiting, shutdown pending maintenance, or maintenance within the current weather window. Thus, the platform can determine whether a target wind turbine can safely wait for the next weather window or should prioritize occupying the current weather window, thereby improving the alignment between maintenance strategies and the actual equipment degradation state and the feasibility of weather windows. Attached Figure Description

[0026] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts: Figure 1 This is a schematic diagram of the system topology provided for an embodiment of the present invention. Figure 2 The overall system flowchart provided for embodiments of the present invention. Figure 3 A schematic diagram illustrating the calculation process for the explanatory power of abnormal weather and the confidence value of equipment degradation provided in this embodiment of the invention. Figure 4 This is a schematic diagram illustrating the path risk comparison process between the current meteorological window maintenance path and the waiting path, provided in an embodiment of the present invention. Figure 5 This is a schematic diagram illustrating the process of generating operation and maintenance strategies and providing feedback updates based on incremental risk of window misses, as provided in an embodiment of the present invention. Figure 6 A schematic diagram of the model training process for the explanatory power of abnormal weather and the confidence value of equipment degradation provided in the embodiments of the present invention. Detailed Implementation

[0027] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0028] The following mathematical formulas illustrate the feasible calculation method of this invention. Besides the explicitly given formulas for calculating the explanatory power of abnormal weather and the incremental risk of missed windows, other scores, probabilities, losses, and risk factors can be normalized to a preset value range, such as 0 to 1, based on wind farm operation and maintenance rules, historical data distribution, or differences in equipment models. Each weight, threshold, and level classification can be determined based on historical samples, expert experience, wind farm safety rules, or model training results. This invention does not limit specific weight values.

[0029] Reference Figure 1 , Figure 1 The diagram illustrates an offshore wind turbine cluster, wind measurement towers, meteorological buoys, sea state monitoring equipment, third-party weather forecasting interfaces, a SCADA system, a condition monitoring system, a central control center server, a cloud server, maintenance vessels, personnel terminals, a spare parts warehouse, and a maintenance work order system. The offshore wind turbine cluster, wind measurement towers, meteorological buoys, and sea state monitoring equipment upload wind turbine operating data, fault alarm data, and meteorological and sea state data to the central control center server or cloud server. The central control center server or cloud server issues maintenance strategies or work order scheduling results to the maintenance vessels, personnel terminals, and the maintenance work order system. This embodiment provides a smart operation and maintenance management platform for offshore wind power based on meteorological data. This platform can be deployed on the offshore wind farm's central control center server, cloud server, edge computing devices, or a cloud-edge collaborative system.

[0030] The platform communicates with offshore wind turbine clusters, wind measurement towers, meteorological buoys, sea state monitoring equipment, third-party weather forecast interfaces, SCADA systems, condition monitoring systems, maintenance work order systems, vessel management systems, personnel terminals, and spare parts management systems. Specifically, the offshore wind turbine clusters provide the platform with wind turbine operating data and fault alarm data; wind measurement towers, meteorological buoys, sea state monitoring equipment, and third-party weather forecast interfaces provide the platform with meteorological and sea state data; the SCADA system and condition monitoring system provide the platform with wind turbine operating data, condition monitoring data, and fault alarm data; and the maintenance work order system, vessel management system, personnel terminals, and spare parts management system provide the platform with maintenance resource data and receive maintenance strategies or work order scheduling results generated by the platform.

[0031] Reference Figure 1 and Figure 2The offshore wind power intelligent operation and maintenance management platform based on meteorological data in this embodiment includes a data construction module, an anomaly identification module, a fault prediction and health management module, a window analysis module, a risk decision-making module, and a scheduling feedback module.

[0032] The platform receives data including meteorological and sea state data, wind turbine operation data, fault alarm data, historical maintenance data, and operation and maintenance resource data. Meteorological and sea state data may include one or more of the following: wind speed, wind direction, gusts, turbulence intensity, wave height, wave period, tides, ocean currents, visibility, rainfall, lightning, typhoon path, humidity, salt spray concentration, weather forecast confidence level, and sea state forecast confidence level. Wind turbine operation data may include one or more of the following: power generation, speed, pitch angle, yaw angle, gearbox temperature, bearing temperature, generator temperature, converter temperature, main shaft vibration, tower vibration, blade load, number of yaw actions, number of pitch actions, start-up / shutdown status, power-limited operation status, and shutdown status. Fault alarm data may include alarm type, alarm level, alarm occurrence time, alarm duration, alarm recovery time, alarm recurrence count, and historical alarm correlations. Historical maintenance data may include historical maintenance work orders, fault handling records, component replacement records, actual maintenance time, post-maintenance recovery status, historical downtime, and historical downtime losses. Operational resource data may include vessel location, vessel speed, vessel wave resistance capability, personnel teams, personnel qualifications, spare parts inventory, tool and equipment status, port location, safe return time, and estimated time for different maintenance tasks.

[0033] The platform's main data processing logic includes: a data construction module that generates a single-unit meteorological impact vector; an anomaly detection module that calculates the explanatory power of abnormal weather based on the single-unit meteorological impact vector and operational anomaly characteristics, and generates a reliable value for equipment degradation; a fault prediction and health management module that performs fault prediction and health management based on the reliable value for equipment degradation; a window analysis module that generates the task completion probability for the current meteorological window and the task completion probability for the next meteorological window; a risk decision module that calculates the incremental risk of window misses based on the path risk difference between the maintenance path and the waiting path for the current meteorological window; and a scheduling feedback module that generates an operation and maintenance strategy based on the incremental risk of window misses.

[0034] In this implementation, the platform does not directly use meteorological and sea state data as ordinary input variables for the fault prediction model, nor does it determine the maintenance plan solely based on the availability of a meteorological window. The platform first determines the degree to which the operational anomaly can be explained by meteorological disturbances based on the individual unit meteorological influence vector and operational anomaly characteristics corresponding to the target wind turbine. Then, it generates a reliable equipment degradation value based on this explanatory power, anomaly persistence, anomaly synergy, and historical fault association information. Next, it corrects the fault prediction results and health degradation curve based on the reliable equipment degradation value. Finally, it inputs the corrected health degradation curve, the task completion probability of the current meteorological window, and the task completion probability of the next meteorological window into the path risk comparison process.

[0035] Thus, this implementation method forms the following continuous processing chain: from single-machine meteorological influence vector to abnormal meteorological explanatory power to equipment degradation confidence value to fault prediction weight correction to health degradation curve to path risk comparison between current meteorological window maintenance path and waiting path to window missed incremental risk to operation and maintenance strategy.

[0036] The aforementioned continuous processing chain enables meteorological and sea state data to no longer serve merely as external environmental constraints, but to participate in the judgment of operational anomalies; at the same time, it enables equipment degradation reliability values ​​to no longer serve merely as health assessment results, but to further participate in the path risk judgment of whether the current meteorological window should be occupied.

[0037] The core of this implementation method lies not in the platform comprising the aforementioned multiple modules, but in the inseparable and continuous data processing relationship formed among these modules: the single-machine meteorological impact vector is first used to determine the meteorological interpretability of operational anomalies; the interpretability of the anomalies further participates in the generation of equipment degradation confidence values; and the equipment degradation confidence values ​​then simultaneously affect the fault prediction weight correction and the current meteorological window occupancy decision. Thus, the platform can simultaneously solve the problems of meteorological disturbance misjudgment and current meteorological window miss risk assessment within the same processing chain.

[0038] In one implementation, the data construction module performs data preprocessing on meteorological and sea state data, wind turbine operation data, and fault alarm data before constructing the meteorological impact vector for a single unit.

[0039] Data preprocessing can include data cleaning, outlier removal, missing value completion, sampling frequency standardization, and time alignment. The platform can establish a unified index based on wind turbine number, acquisition time, data source, and data type, so that the target wind turbine corresponds to meteorological and sea condition data, wind turbine operation data, and fault alarm data on the same timeline.

[0040] For example, wind turbine operating data can be sampled at the second, minute, or ten-minute intervals; meteorological and sea condition data can be sampled at the minute, hour, or forecast time step levels; and fault alarm data can be event-based. The platform can aggregate or interpolate the above data according to preset time windows, enabling comparisons within the same anomaly analysis time window.

[0041] When some meteorological and sea state data is missing, the platform can supplement it using data from adjacent wind measurement points, adjacent meteorological grids, target wind turbine nacelle wind speed data, or historical statistical data. When there are obvious acquisition errors in the wind turbine operating data, the platform can remove or correct them based on upstream and downstream time points, adjacent sensor data, or equipment status indicators.

[0042] The above data preprocessing is used to ensure that all types of data have temporal consistency and source traceability when calculating the explanatory power of abnormal weather and the confidence value of equipment degradation.

[0043] Reference Figure 2 The data construction module converts regional or station-level meteorological and sea state data into a single-unit meteorological influence vector corresponding to the target wind turbine based on at least two of the following: the turbine location coordinates, wind measurement point location, meteorological grid location, prevailing wind direction, wind field layout, upstream wind turbine operating status, wake influence, sea surface humidity, salt spray exposure level, and data acquisition time.

[0044] A single-unit meteorological impact vector may include at least two of the following: effective wind speed, effective turbulence intensity, gust impact value, wake disturbance value, wave height impact value, humidity impact value, salt spray exposure value, meteorological forecast confidence level, and meteorological event label.

[0045] Effective wind speed can be obtained by adjusting for wind speed at the wind farm, nacelle wind speed, wind direction, and the target turbine location. Effective turbulence intensity can be determined based on the wind speed variation within a preset time window. Gust impact value can be determined based on the sudden increase in wind speed over a short period. Wake disturbance value can be determined based on wind direction, upstream turbine operating status, and the relative position of the target turbine. Wave height impact value can be determined based on wave height, wave period, and the location of the turbine in the sea area. Humidity impact value can be determined based on relative humidity, sea fog, or rainfall information. Salt spray exposure value can be determined based on salt spray sensor data, or estimated based on the wind farm's sea area location, humidity, wind direction, and historical corrosion environment level.

[0046] Meteorological event tags can include short-term gust events, strong turbulence events, sudden wave height events, high humidity and high salt spray events, lightning warning events, typhoon outer periphery impact events, low visibility events, or continuous severe weather events.

[0047] For example, when the wind speed suddenly increases beyond a preset threshold within a preset time window, the platform generates a short-term gust event label; when the effective turbulence intensity exceeds a preset threshold, the platform generates a strong turbulence event label; when the wave height increases rapidly within a preset time window, the platform generates a wave height change event label; when the meteorological interface returns lightning warnings or typhoon path impact information, the platform generates corresponding meteorological event labels.

[0048] In this embodiment, the single-unit meteorological influence vector is not only used to determine whether the target wind turbine has a maintenance window, but also serves as the basic input for subsequent calculation of the explanatory power of abnormal meteorology, and is used to explain the correlation between the abnormal operation of the target wind turbine and meteorological disturbances.

[0049] In other words, the role of the single-unit meteorological impact vector is not to judge whether the weather is suitable for operation, but to transform the meteorological and sea state data at the station level or regional level into interpretive inputs corresponding to the abnormal operation of the target wind turbine, so that the subsequent abnormal meteorological interpretation degree can be calculated for the specific target wind turbine, the specific abnormal analysis time window, and the specific meteorological event.

[0050] Reference Figure 3 When a target wind turbine experiences an operational anomaly, the anomaly detection module determines an anomaly analysis time window based on the occurrence time of the anomaly and extracts the anomaly characteristics of the target wind turbine within this time window. These anomaly characteristics can include one or more of the following: vibration anomaly, temperature rise anomaly, power fluctuation anomaly, yaw anomaly, pitch anomaly, and alarm frequency anomaly. Each type of anomaly characteristic can include the anomaly occurrence time, anomaly duration, anomaly intensity, anomaly trend, anomaly recovery time, and anomaly repetition count. Vibration anomalies can be determined based on changes in main shaft vibration, tower vibration, or blade load; temperature rise anomalies can be determined based on changes in gearbox temperature, bearing temperature, generator temperature, or converter temperature; power fluctuation anomalies can be determined based on the deviation between power generation and wind speed; and alarm frequency anomalies can be determined based on the number of alarms or repeated alarms per unit time. The anomaly analysis time window can be preset and extended forward and backward based on the anomaly occurrence time to match the corresponding single-unit meteorological influence vector and meteorological event label. The preset duration can be set according to the data sampling period, fault type, and wind farm operation and maintenance rules. The anomaly meteorological interpretability is used to characterize the degree to which the operational anomaly characteristics can be explained by meteorological disturbances. The higher the explanatory power of anomalous weather, the more likely the operational anomalies are caused by meteorological disturbances; the lower the explanatory power of anomalous weather, the more difficult it is for operational anomalies to be explained by meteorological disturbances.

[0051] In one implementation, the anomaly detection module determines at least two of the following: time synchronization score, trend consistency score, duration matching score, recovery score, historical similarity score, and similar operating condition response score.

[0052] The time synchronization score characterizes the degree of synchronization between the occurrence time of the operational anomaly and the occurrence time of the meteorological event. The trend consistency score characterizes the degree of consistency between the trend of the operational anomaly and the trend of the meteorological disturbance. The duration matching score characterizes the degree of matching between the duration of the operational anomaly and the duration of the meteorological event. The recovery score characterizes whether the operational anomaly has recovered after the meteorological event. The historical similarity score characterizes whether the target wind turbine has experienced similar operational anomalies under similar historical meteorological events. The similar operating condition response score characterizes whether other wind turbines in the same wind farm with similar locations, similar wind directions, or similar operating conditions have shown similar operational responses under the same meteorological event.

[0053] In one implementation, the explanatory power of anomalous weather conditions is calculated using the following mathematical formula: Where E represents the explanatory power of abnormal weather; This represents the i-th explanatory score, which includes at least one of the following: time synchronization score, trend consistency score, duration matching score, recovery score, historical similarity score, and similar operating condition response score. represents the weight corresponding to the i-th explanation score; n represents the number of explanation scores involved in the calculation.

[0054] Each explanatory score can be normalized to a score between 0 and 1. A higher score indicates that the corresponding explanatory factor strongly supports the claim that the operational anomaly was caused by a meteorological disturbance. When data corresponding to any explanatory score is missing, the anomaly detection module can reset the weight of that score to zero and normalize the weights of the remaining explanatory scores. If all explanatory scores are unavailable, the anomaly detection module can set the abnormal meteorological explanatory power to a preset default value or mark the operational anomaly as an anomaly to be confirmed, to avoid directly judging it as a meteorological disturbance-type anomaly when data is insufficient.

[0055] In one implementation, the time synchronization score can be determined based on the time difference between the peak occurrence time of the operational anomaly and the peak occurrence time of the meteorological event; the closer the two times are, the higher the time synchronization score. The trend consistency score can be determined based on the consistency or correlation of the changing directions of the operational anomaly sequence and the meteorological disturbance sequence within the anomaly analysis time window; the more consistent their changing trends, the higher the trend consistency score. The duration matching score can be determined based on the proximity between the duration of the operational anomaly and the duration of the meteorological event; the more similar their durations, the higher the duration matching score. The recovery score can be determined based on the degree of recovery of the operational anomaly relative to the anomaly peak and the pre-anomaly baseline value after the meteorological event ends; the more significant the recovery, the higher the recovery score.

[0056] In this embodiment, the value of the explanatory power of abnormal weather conditions lies not in the weighted calculation method itself, but in using explanatory scores such as time synchronization score, trend consistency score, duration matching score, and recovery score to determine the cause of operational anomalies. This cause determination result is then used to generate equipment degradation confidence values ​​and correct fault prediction weights. Therefore, the explanatory power of abnormal weather conditions is not a typical environmental score, but rather an intermediate technical variable connecting weather disturbance identification, equipment degradation assessment, and subsequent maintenance window decisions.

[0057] In another implementation, the explanatory power of anomalous weather can be generated by a trained meteorological interpretation model. The platform can extract historical anomalous segments containing anomalies from historical operational data and match each historical anomalous segment with a single-unit meteorological impact vector, meteorological event label, operational anomaly characteristics, anomaly duration, anomaly recovery status, historical alarm information, and historical maintenance results.

[0058] If a historical anomaly segment recovers after a meteorological event ends, and no similar fault alarms, corresponding component maintenance records, or confirmed equipment failures after maintenance occur within the preset observation period, then this historical anomaly segment can be considered a meteorological disturbance-type anomaly sample. If a historical anomaly segment persists under stable meteorological conditions, or if similar fault alarms, corresponding component maintenance records, post-maintenance recurrence records, or confirmed equipment failures occur within the preset observation period, then this historical anomaly segment can be considered an equipment degradation-type anomaly sample.

[0059] The meteorological interpretation model can take single-aircraft meteorological impact vectors, meteorological event labels, and operational anomaly features as inputs, and output the explanatory power of abnormal meteorological conditions. The meteorological interpretation model can be a logistic regression model, decision tree model, random forest model, gradient boosting tree model, support vector machine model, or neural network model. This invention does not limit the specific type of meteorological interpretation model. When historical samples are insufficient, the platform can first use a rule-based calculation method to generate initial labels or weak labels, and update the training samples after subsequent maintenance results are fed back.

[0060] Equipment degradation confidence values ​​characterize the degree to which abnormal operational features are caused by equipment degradation itself. Equipment degradation confidence values ​​differ from health indices and failure probabilities. Health indices characterize the overall health status of equipment, failure probabilities characterize the likelihood of future failures, while equipment degradation confidence values ​​characterize whether current anomalies are more likely to originate from equipment degradation itself rather than a response to weather disturbances.

[0061] In one implementation, the equipment degradation confidence value is determined based on a basic anomaly score, a meteorological deduction factor, and a historical correlation correction factor. The basic anomaly score represents the fundamental contribution of the intensity and duration of operational anomalies, as well as the synergy among multiple operational anomalies, to the equipment degradation confidence value. The meteorological deduction factor is negatively correlated with the explanatory power of anomalous weather conditions and is used to reduce the equipment degradation confidence value when operational anomalies are more likely to be explained by meteorological disturbances. The historical correlation correction factor is positively correlated with historical fault association information and is used to increase the equipment degradation confidence value when the correlation between similar historical anomalies and actual faults is high.

[0062] The basic anomaly score can be determined based on the anomaly intensity score, the anomaly duration score, and the synergy score among multiple anomalies. The anomaly intensity score is determined based on the deviation of the current anomaly value from the normal baseline value and the preset anomaly upper limit value. The anomaly duration score is determined based on the ratio of the anomaly duration to the preset reference duration. The synergy score is determined based on the number of anomalies simultaneously pointing to the same component or the same fault chain.

[0063] The meteorological deduction factor can be determined based on the explanatory power of abnormal weather. A higher explanatory power indicates that the operational anomaly is more likely to be caused by meteorological disturbances, resulting in a lower meteorological deduction factor; conversely, a lower explanatory power indicates that the operational anomaly is less likely to be explained by meteorological disturbances, resulting in a higher meteorological deduction factor. The historical correlation correction factor can be determined based on historical fault correlation information. A higher degree of correlation between similar historical anomalies and actual faults results in a higher historical correlation correction factor.

[0064] In a more preferred embodiment, the equipment degradation confidence value is generated using a hierarchical judgment logic. First, the anomaly discrimination module determines whether the operational anomaly has a meteorological disturbance explanation basis based on the anomaly meteorological interpretability. Second, the anomaly discrimination module further determines whether the operational anomaly recovers after the meteorological event ends, whether it continues to intensify, whether it points to the same component or the same fault chain as other operational anomalies, and whether there is a correlation with similar historical faults. Finally, the anomaly discrimination module integrates the above results to generate the equipment degradation confidence value.

[0065] When an operational anomaly has a high degree of explanatory power for abnormal weather, and the operational anomaly recovers after the weather event ends, there is a lack of coordination among multiple operational anomalies, and the historical fault correlation information is weak, the anomaly identification module reduces the equipment degradation confidence value; when an operational anomaly has a high degree of explanatory power for abnormal weather, but continues to exist or continue to intensify after the weather event ends, or there is coordination among multiple operational anomalies, or the historical fault correlation information is strong, the anomaly identification module maintains or increases the equipment degradation confidence value.

[0066] Through the above processing, the equipment degradation confidence value is affected not only by the intensity and duration of operational anomalies and the synergy among multiple operational anomalies, but also by the reduction in the explanatory power of abnormal weather and the enhancement of historical fault correlation information. Therefore, the equipment degradation confidence value is not a general health index, nor a direct failure probability, but rather represents the degree of confidence that the current operational anomaly is caused by the equipment's own degradation.

[0067] In one implementation, the anomaly detection module generates at least two of the following: vibration degradation confidence value, temperature rise degradation confidence value, power fluctuation degradation confidence value, yaw degradation confidence value, pitch degradation confidence value, and alarm degradation confidence value. Based on the fault correlation between different operational anomaly characteristics and wind turbine components, the module fuses these values ​​to obtain the equipment degradation confidence value of the target wind turbine.

[0068] Vibration degradation confidence values ​​can be associated with at least one of bearing, gearbox, or spindle failures; temperature rise degradation confidence values ​​can be associated with at least one of gearbox, generator, or converter failures; yaw degradation confidence values ​​can be associated with yaw system failures; pitch degradation confidence values ​​can be associated with pitch system failures; and power fluctuation degradation confidence values ​​can be associated with at least one of pitch system, converter, or weather disturbances. Through sub-confidence value fusion, the platform can avoid directly equating a single operational anomaly with equipment degradation and can combine the synergy of multiple operational anomalies in the component failure chain to determine the degree of equipment degradation confidence.

[0069] In another implementation, the equipment degradation confidence value can be generated by a trained degradation confidence model. The degradation confidence model can take as input operational anomaly characteristics, anomaly duration, anomaly synergy, anomaly weather interpretability, historical fault correlation information, recurring alarm information, and post-maintenance recurrence information, and output the equipment degradation confidence value.

[0070] The training samples for the degradation credibility model can be derived from historical anomaly segments and maintenance confirmation results. If maintenance confirms no equipment fault, the corresponding anomaly segment can be labeled as a meteorological disturbance anomaly or a low-degradation credibility sample; if maintenance confirms an equipment fault, the corresponding anomaly segment can be labeled as an equipment degradation anomaly or a high-degradation credibility sample. The degradation credibility model can be a logistic regression model, a decision tree model, a random forest model, a gradient boosting tree model, a support vector machine model, or a neural network model. This invention does not limit the specific type of degradation credibility model.

[0071] The fault prediction and health management module performs fault prediction and health management on the target wind turbine based on the equipment degradation confidence value, and generates fault prediction results and health degradation curves.

[0072] The fault prediction results can include the suspected fault type, the current fault probability, the probability of fault expansion in the first future time period, the probability of fault expansion in the second future time period, the time interval in which the fault may occur, and the fault prediction confidence level.

[0073] When generating fault prediction results, the fault prediction and health management module adjusts the fault prediction weights of operational anomaly features based on the interpretability of abnormal weather and the confidence value of equipment degradation. The fault prediction weights represent the degree of contribution of a particular operational anomaly feature to fault probability calculation, fault level determination, or health degradation curve generation.

[0074] For operational anomalies where the explanatory power of abnormal weather meets a preset high explanatory power condition and the confidence value of equipment degradation meets a preset low confidence value condition, the platform determines that the operational anomaly is more likely caused by weather disturbances, and therefore reduces the fault prediction weight of the operational anomaly. For operational anomalies where the explanatory power of abnormal weather meets a preset low explanatory power condition and the confidence value of equipment degradation meets a preset high confidence value condition, the platform determines that the operational anomaly is more likely caused by equipment degradation itself, and therefore increases the fault prediction weight of the operational anomaly.

[0075] When the interpretability of abnormal weather is high but the confidence value of equipment degradation has not decreased, the fault prediction and health management module does not directly filter this operational anomaly as a weather disturbance anomaly. Instead, it continues to retain the contribution of this operational anomaly to the fault prediction results and health degradation curve. Thus, the platform can avoid incorrectly excluding genuine equipment degradation anomalies that are "weather-related but not recoverable."

[0076] The fault prediction and health management module can also generate at least two of the following based on the changing trends of the equipment degradation confidence value under different operating strategies: full-load operation health degradation curve, power-limited operation health degradation curve, operation health degradation curve after remote reset, shutdown waiting for maintenance health degradation curve, and continuous monitoring without intervention health degradation curve.

[0077] The fault prediction and health management module can also determine the waitability level of the target wind turbine based on the change in the equipment degradation confidence value under at least one of the following states: limited power operation, remote reset, or shutdown waiting.

[0078] When the equipment degradation confidence value decreases or its growth slows down after power-limited operation, the waitability level of the target wind turbine is increased; when the equipment degradation confidence value decreases and the alarm is cleared after remote reset, the waitability level of the target wind turbine is increased; when the equipment degradation confidence value continues to rise after power-limited operation, remote reset, or shutdown waiting, the waitability level of the target wind turbine is decreased. The waitability level characterizes whether the target wind turbine can wait for the next meteorological window under controllable risk conditions. The waitability level can include at least one of the following: not waitable, short-term waitable, power-limited waitable, waitable after remote reset, or controllable shutdown waiting.

[0079] The window analysis module is used to generate the task completion probability for the current weather window and the task completion probability for the next weather window.

[0080] In one implementation, the window analysis module generates the task completion probability for the current weather window and the task completion probability for the next weather window based on at least two of the following: the available working time of the weather window, the maintenance time distribution corresponding to the target fault type, the ship's sailing time, the personnel boarding preparation time, the personnel qualification matching status, the availability of spare parts, the availability of tools, the safe return time, and the risk of window shortening.

[0081] The longer the weather window is available for operation, the shorter the maintenance time corresponding to the target fault type, the shorter the ship's sailing time and personnel's boarding preparation time, and the better the personnel's qualifications, spare parts, and tools are matched, the higher the probability of completing the task in the corresponding weather window. Conversely, if the weather window is close to the time required for the target maintenance task, or if the risk of shortening the window is high, the probability of completing the task in the corresponding weather window will decrease.

[0082] The matching degree of operation and maintenance resources can be determined based on the matching degree of personnel qualifications, the availability of spare parts, and the availability of tools. Personnel qualification matching degree indicates whether the qualifications of available personnel meet the requirements of the target maintenance task; spare parts availability degree indicates whether the spare parts required for the target maintenance task are complete; and tool availability degree indicates whether the tools and equipment required for the target maintenance task are complete. The higher the matching degree of personnel qualifications, spare parts availability, and tool availability, the higher the matching degree of operation and maintenance resources.

[0083] In this embodiment, the task completion probability of the current meteorological window and the task completion probability of the next meteorological window are not used alone to determine the maintenance plan, but are used as path risk calculation variables for the maintenance path and waiting path of the current meteorological window to calculate the incremental risk of window miss.

[0084] Reference Figure 4 The risk decision-making module is used to construct the maintenance path and waiting path for the current meteorological window, and to calculate the incremental risk of missing the window.

[0085] The current meteorological window maintenance path indicates the target wind turbine's scheduled offshore maintenance within the current meteorological window. The platform uses the path risk of the current meteorological window maintenance path as the baseline risk. The waiting path indicates the path for the target wind turbine to wait for the next meteorological window after abandoning the current meteorological window. Waiting paths include at least one of the following: no-processing waiting path, power-limited waiting path, remote reset and monitoring waiting path, or shutdown waiting path.

[0086] The "No Processing Waiting Path" means that after abandoning the current meteorological window, no measures such as power limiting, remote reset, or shutdown will be taken against the target wind turbine. The target wind turbine will remain in its current operating state and wait for the next meteorological window. The "Power Limiting Waiting Path" means that after abandoning the current meteorological window, power limiting will be implemented on the target wind turbine, allowing it to wait for the next meteorological window under reduced load conditions. The "Remote Reset and Monitoring Waiting Path" means that after abandoning the current meteorological window, the target wind turbine will be remotely reset or its status restored, and its status will be continuously monitored after the reset, waiting for the next meteorological window. The "Shutdown Waiting Path" means that after abandoning the current meteorological window, the target wind turbine will be shut down or kept in a shut-down state, waiting for maintenance in the next meteorological window.

[0087] Before calculating the path risk of multiple waiting paths, the risk decision module can assess the executability of each waiting path. If a waiting path does not meet safety constraints, equipment control conditions, or operation and maintenance rules, then that waiting path will not participate in the selection of the optimal waiting path risk.

[0088] For example, when the target fault type does not support remote reset, the monitoring waiting path after remote reset is not included in the window miss incremental risk calculation; when the continued operation of the target wind turbine would trigger a safety interlock or pose a significant risk of expanded damage, the waiting path is not processed and is not included in the window miss incremental risk calculation; when the target wind turbine does not meet the conditions for power-limited operation, the power-limited waiting path is not included in the window miss incremental risk calculation. The path risk of the maintenance path in the current meteorological window can include at least two of the following: risk of incomplete task in the current meteorological window, risk of shortened current meteorological window, risk of residual fault after maintenance, risk of resource occupation, and power generation loss during maintenance. The path risk of the unprocessed waiting path can include at least two of the following: risk of increased equipment degradation confidence value during the waiting period, risk of decreased health index, risk of fault expansion, risk of shutdown, risk of unreachable next meteorological window, and risk of incomplete task in the next meteorological window. The path risk of the power-limited waiting path can include at least two of the following: risk of residual increase in equipment degradation confidence value after power limiting, risk of power-limited power generation loss, risk of unreachable next meteorological window, and risk of incomplete task in the next meteorological window.

[0089] The path risks of monitoring the waiting path after remote reset can include at least two of the following: remote reset failure risk, abnormal recurrence risk after reset, continued increase in the equipment degradation confidence value after reset, unreachable next meteorological window risk, and uncompleted task risk in the next meteorological window.

[0090] The path risks of a shutdown waiting path can include at least two of the following: shutdown power generation loss risk, shutdown waiting residual degradation risk, next weather window unreachable risk, and next weather window uncompleted task risk.

[0091] While the risks of power generation loss and resource occupation in the aforementioned path risks can serve as supplementary factors, the main basis for assessing the incremental risk of missed weather windows includes changes in equipment degradation confidence values, changes in health indices, the probability of fault expansion, the probability of outages, the risk of increased maintenance requirements, and the risk of not completing tasks in the next weather window. Therefore, the incremental risk of missed weather windows reflects the risk of changes in equipment condition resulting from the continued operation of the target wind turbine or waiting for the next weather window, rather than simply economic or scheduling costs.

[0092] In one implementation, the incremental risk of window misses is calculated using the following mathematical formula: .in, This indicates the risk of missing incremental growth during the window period; This represents the path risk of the current meteorological window maintenance path, i.e., the baseline risk; Ω represents the set of executable waiting paths. This represents the path risk of the j-th waiting path in the set of executable waiting paths; This represents the minimum path risk in the set of executable waiting paths. When there is only one executable waiting path... The path risk of the executable waiting path is determined. When multiple executable waiting paths exist, the risk decision module selects the waiting path with the lowest path risk from among the executable waiting paths that are not processed, power limiting waiting paths, remote reset monitoring waiting paths, and shutdown waiting paths, and compares its path risk with the path risk of the maintenance path in the current meteorological window.

[0093] When the calculated incremental risk of missed windows is less than 0, it indicates that under the current data conditions and executable waiting path constraints, the path risk of the waiting path is no higher than the path risk of the current meteorological window maintenance path. In this case, the platform can classify the incremental risk of missed windows as a low-risk level, or correct the incremental risk of missed windows to 0 before using it for risk level classification, in order to avoid misjudgment of the strategy output due to negative risk values.

[0094] The incremental risk of missed window in this embodiment differs from the risk of ordinary delayed failure. Ordinary delayed failure risk typically evaluates whether postponing maintenance itself is dangerous, while this embodiment evaluates the difference in path risk between abandoning the current weather window and using the current weather window for maintenance. Furthermore, this path risk difference is combined with the equipment degradation confidence value, health degradation curve, the probability of task completion in the current weather window, and the probability of task completion in the next weather window.

[0095] Reference Figure 5 The scheduling feedback module generates an operation and maintenance strategy for the target wind turbine based on the incremental risk of missed window periods. In one implementation, the incremental risk of missed window periods can be classified into low-risk, medium-risk, high-risk, and extremely high-risk levels.

[0096] When the incremental risk of missing the window is at a low risk level, and the explanatory power of abnormal weather meets the preset high explanatory power condition and the confidence value of equipment degradation meets the preset low confidence value, the scheduling feedback module generates a continuous monitoring strategy or a remote reset observation strategy.

[0097] When the incremental risk of missing the window is at a medium risk level, and the equipment degradation confidence value under the power-limited waiting path decreases or slows down, and the task completion probability of the next meteorological window meets the preset requirements, the scheduling feedback module generates a power-limited waiting strategy or a maintenance strategy for the next meteorological window.

[0098] When the risk of missing incremental updates in the current weather window is at a high risk level, and the equipment degradation confidence value meets the preset high confidence condition and the task completion probability of the current weather window meets the preset requirements, the scheduling feedback module generates the maintenance strategy for the current weather window.

[0099] When the risk of missing incremental updates in the current weather window is at an extremely high risk level, and continued operation would increase the probability of the fault escalating, the scheduling feedback module generates an immediate maintenance strategy for the current weather window, or generates a shutdown and maintenance waiting strategy when the probability of task completion in the current weather window does not meet the requirements.

[0100] The aforementioned continuous monitoring strategy, remote reset observation strategy, power limiting waiting strategy, shutdown waiting for maintenance strategy, and current meteorological window maintenance strategy are all related to the target wind turbine's operation control or maintenance execution status. Specifically, the power limiting waiting strategy is used to reduce the target wind turbine's operating load; the remote reset observation strategy is used to continue monitoring changes in the equipment's degradation confidence value after the remote status is restored; the shutdown waiting for maintenance strategy is used to prevent continued operation from causing the fault to escalate; and the current meteorological window maintenance strategy is used to schedule the target wind turbine for offshore maintenance within the current meteorological window.

[0101] Therefore, the operation and maintenance strategy output by the scheduling feedback module is not simply a management schedule, but rather a technical control action that performs on the target wind turbine based on the identification of the cause of the anomaly and the incremental risk of missing the window, such as continuous monitoring, remote reset, power limiting, shutdown waiting, or maintenance within the current meteorological window.

[0102] The scheduling feedback module can also convert operation and maintenance strategies into work orders and scheduling plans. Work order content can include target wind turbine number, suspected fault type, health level, health index, abnormal weather interpretation degree, equipment degradation confidence value, window miss incremental risk level, recommended maintenance items, estimated maintenance time, departure time, return time, required vessels, required personnel, spare parts list, tool list, and emergency evacuation plan.

[0103] Once the actual maintenance results are returned, the scheduling feedback module reverse-labels the operational anomalies based on these results. If the maintenance confirms no equipment fault, the corresponding anomaly is marked as a meteorological disturbance anomaly, and the anomaly meteorological interpretation parameters are updated. If the maintenance confirms an equipment fault, the corresponding anomaly is marked as an equipment degradation anomaly, and the equipment degradation confidence value calculation parameters are updated.

[0104] The platform can also update the calculation parameters for the task completion probability of the current meteorological window and the task completion probability of the next meteorological window based on the difference between the actual window duration and the predicted window duration; update the health degradation curve based on the changes in the health index after actual power limitation, remote reset or shutdown waiting; and update the calculation parameters for incremental risk of missed window based on whether a fault upgrade or maintenance level increase occurs after a missed window.

[0105] Example 1: This example illustrates how the platform identifies meteorological disturbance anomalies and avoids unnecessarily occupying the current meteorological window.

[0106] In this embodiment, the target wind turbine experiences a short-term increase in main shaft vibration and fluctuations in power generation over a certain period of time. The platform synchronously receives the individual meteorological influence vector corresponding to the target wind turbine, which indicates that the target wind turbine is under the influence of short-term gust events and strong turbulence events.

[0107] The platform extracts vibration anomalies and power fluctuation anomalies from the wind turbine's operating data. The anomaly identification module calculates the meteorological explanatory power corresponding to the vibration anomalies and power fluctuation anomalies. The platform determines that: the vibration anomaly is synchronous with the gust impact in time; the power fluctuation is consistent with the trend of effective turbulence intensity change; the duration of the anomaly matches the duration of the meteorological event; after the meteorological event weakens, the vibration anomaly and power fluctuation have a recovery trend; the target wind turbine has experienced similar short-term anomalies under similar historical meteorological events, and no equipment failure has been confirmed.

[0108] Based on the above judgments, the platform generates a high explanatory power for the abnormal weather. Subsequently, the anomaly discrimination module generates a reliable equipment degradation value based on the anomaly intensity, duration, synergy, explanatory power of the abnormal weather, and historical fault correlation information. Since the anomaly in this embodiment is synchronized with the meteorological event and has a recovery trend, the platform reduces the contribution of vibration anomalies and power fluctuation anomalies to the reliable equipment degradation value, resulting in a lower reliable equipment degradation value.

[0109] The fault prediction and health management module performs fault prediction and health management based on the equipment's degradation confidence value. For vibration anomalies and power fluctuation anomalies where the abnormal meteorological interpretability meets the preset high interpretability condition and the equipment degradation confidence value meets the preset low confidence condition, the platform reduces their fault prediction weight, so that the fault level will not be significantly increased simply because of the increase in the original vibration value or power fluctuation. The health degradation curve generated by the platform shows that the target wind turbine does not have a significant and continuous degradation trend.

[0110] Subsequently, the risk decision-making module constructs the maintenance path for the current meteorological window, the no-processing waiting path, and the monitoring waiting path after remote reset. The platform uses the path risk of the maintenance path for the current meteorological window as the baseline risk and calculates the path risks of the no-processing waiting path and the monitoring waiting path after remote reset, respectively.

[0111] Because the target wind turbine equipment has a low confidence value for degradation, a high explanatory power for abnormal weather, and an anomaly with a recovery trend, and because the health index is not expected to decline significantly under the waiting path, the path risk difference between the waiting path and the baseline risk is small. The platform calculates a low incremental risk of window misses.

[0112] Based on this, the scheduling feedback module generates a remote reset observation strategy, a continuous monitoring strategy, or a maintenance strategy for the next meteorological window, without prioritizing the use of the current meteorological window for offshore maintenance.

[0113] In this embodiment, the platform does not refrain from going to sea because "the current weather window is unavailable", nor does it simply postpone processing because "the alarm level is low". Instead, it determines that abandoning the current weather window will not significantly increase the risk of equipment operation based on the high interpretability of abnormal weather, the low confidence value of equipment degradation, and the low incremental risk of window loss. Thus, it generates a remote reset observation strategy or a continuous monitoring strategy.

[0114] Example 2: This example illustrates how the platform identifies equipment degradation anomalies and prioritizes occupying the current meteorological window when the risk of missing incremental window opportunities is high.

[0115] In this embodiment, the target wind turbine experienced a continuous increase in gearbox temperature, a slight increase in main shaft vibration, and repeated occurrences of similar alarms during a period of relatively stable meteorological conditions. The single-unit meteorological influence vector received by the platform indicates that there were no significant abrupt changes in the current wind speed, turbulence intensity, gust impact value, and wave height influence value, and the target wind turbine was not in the midst of a short-term gust, strong turbulence, or wave height change event.

[0116] The anomaly detection module extracts abnormal temperature rise, vibration, and alarm frequency, and calculates the corresponding meteorological explanatory power. The platform determines that: the abnormal temperature rise is not synchronized with short-term gusts, strong turbulence, or sudden changes in wave height; the abnormal temperature rise persists even after meteorological conditions stabilize; there is no clear temporal synchronization between alarm frequency and meteorological events; there is a certain degree of synergy between abnormal temperature rise, vibration, and recurring alarms; and historical similar abnormal temperature rises or recurring alarms are associated with faults in components such as gearboxes, bearings, or generators. Based on these judgments, the platform generates a low meteorological explanatory power. Subsequently, the anomaly detection module generates a reliable equipment degradation value based on the anomaly intensity, duration, synergy, meteorological explanatory power, and historical fault association information. Due to the low meteorological explanatory power and the persistence and synergy of the anomalies, the platform generates a high reliable equipment degradation value.

[0117] The fault prediction and health management module performs fault prediction and health management based on the equipment degradation confidence value. For abnormal weather conditions that meet the preset low interpretation criteria and the equipment degradation confidence value meets the preset high confidence criteria, such as abnormal temperature rise, abnormal vibration, and abnormal alarm frequency, the platform increases their fault prediction weight, thereby generating suspected fault types, fault prediction results, and health degradation curves.

[0118] If the health degradation curve shows that the health status of the target wind turbine is in a continuous downward trend, the platform will further generate a full-load operation health degradation curve, a power-limited operation health degradation curve, and a shutdown waiting for maintenance health degradation curve.

[0119] When the platform determines that the equipment degradation confidence value does not decrease significantly after power limiting operation, it indicates that the target wind turbine is not suitable for safe waiting for a long time by power limiting alone.

[0120] The window analysis module generates the task completion probability for the current weather window and the task completion probability for the next weather window. If the current weather window meets the safety requirements for ship navigation, personnel boarding, maintenance operations, and return to port, but the next weather window has a low reachability probability, a high risk of window shortening, or a low task completion probability, then the platform enters the risk decision-making process.

[0121] The risk decision-making module constructs maintenance paths, no-processing waiting paths, power-limited waiting paths, and shutdown waiting paths for the current weather window. The platform uses the path risk of the maintenance path within the current weather window as the baseline risk and calculates the changes in health index, changes in equipment degradation confidence value, probability of fault escalation, probability of task completion, and risk of increased maintenance level for each waiting path.

[0122] Because the target wind turbine equipment has a high reliability value for degradation, the health degradation curve shows a continuous downward trend, the equipment degradation reliability value under the power-limited waiting path has not decreased significantly, and the probability of completing the task in the next meteorological window is low, the platform determines that the path risk of the waiting path is significantly increased compared to the path risk of the maintenance path in the current meteorological window, and the incremental risk of missing the window is high.

[0123] Based on this, the scheduling feedback module generates a maintenance strategy for the current meteorological window and converts this strategy into work orders and scheduling plans. If all maintenance tasks cannot be completed within the current meteorological window, the platform prioritizes scheduling wind turbines with high equipment degradation confidence values ​​and high risk of missing incremental maintenance windows.

[0124] In this embodiment, the platform does not schedule sea trips solely based on the availability of the current weather window. Instead, it prioritizes occupying the current weather window when the equipment degradation confidence value is high, the health degradation curve continues to decline, and the path risk difference between the waiting path and the maintenance path in the current weather window is large.

[0125] Example 3: This example illustrates how the risk decision-making module calculates the incremental risk of window misses when multiple waiting paths exist.

[0126] In this embodiment, the target wind turbine exhibits abnormal temperature rise and recurring alarms with a high confidence value for equipment degradation. The fault prediction and health management module generates a health degradation curve and determines that the target wind turbine's health index declines rapidly under full-load operation. The window analysis module generates the task completion probability for the current meteorological window and the task completion probability for the next meteorological window. The task completion probability for the current meteorological window meets the maintenance requirements, while the task completion probability for the next meteorological window is reduced due to the risk of window shortening.

[0127] The risk decision-making module constructs maintenance paths, no-processing waiting paths, power-limited waiting paths, and shutdown waiting paths for the current weather window. For the maintenance path within the current weather window, the module calculates the risks of unfinished tasks, shortened weather windows, residual fault risks after maintenance, and resource occupancy risks, and obtains the baseline risk. For the no-processing waiting path, the module calculates the risks of increased equipment degradation confidence values, fault expansion, unreachability in the next weather window, and unfinished tasks in the next weather window. For the power-limited waiting path, the module calculates the residual growth risk of equipment degradation confidence values ​​after power limiting, power-limited generation loss risk, and unfinished tasks in the next weather window. For the shutdown waiting path, the module calculates the shutdown generation loss risk, residual degradation risk during shutdown waiting, and unfinished tasks in the next weather window.

[0128] In this embodiment, if a certain waiting path does not meet the safety interlocking conditions, equipment control conditions, or wind farm operation and maintenance rules, the risk decision module marks the waiting path as an unexecutable waiting path and does not include the unexecutable waiting path in the selection range of optimal waiting path risk.

[0129] The risk decision module takes the lower path risk among the executable waiting paths as the optimal waiting path risk, and takes the path risk difference between the optimal waiting path risk and the baseline risk of the current meteorological window maintenance path as the window missed incremental risk.

[0130] When the incremental risk of missing the window is at a high-risk or extremely high-risk level, the scheduling feedback module generates the maintenance strategy for the current meteorological window; when the incremental risk of missing the window is at a low-risk or medium-risk level, the scheduling feedback module can generate a power-limiting waiting strategy, a shutdown waiting maintenance strategy, or a maintenance strategy for the next meteorological window.

[0131] This embodiment demonstrates that the platform does not simply determine whether there is a risk in postponing maintenance, but rather uses the current meteorological window maintenance path as a benchmark to compare the increased path risk of multiple executable waiting paths relative to the benchmark path, thereby determining whether abandoning the current meteorological window will cause additional risks.

[0132] Furthermore, the optimal waiting path in this embodiment is not the lowest-cost path in a management sense, but rather a waiting path with a relatively low overall risk of increased equipment degradation confidence value, decreased health index, expanded fault, downtime, increased maintenance level, and failure to complete the next meteorological window task, provided that safety interlocking conditions, equipment control conditions, and wind farm operation and maintenance rules are met. Therefore, the incremental risk of missed window can be used to support the selection of technical control strategies for the target wind turbine.

[0133] The technical scope of this invention is not limited to the content described above. Those skilled in the art can make various modifications and variations to the above embodiments without departing from the technical concept of this invention, and all such modifications and variations should fall within the protection scope of this invention.

Claims

1. A smart operation and maintenance management platform for offshore wind power based on meteorological data, characterized in that, include: The data construction module is used to time-align meteorological and sea condition data, wind turbine operation data, and fault alarm data according to the wind turbine number and collection time, and generate a single-unit meteorological impact vector for the target wind turbine. abnormal The discrimination module is used to extract operational anomaly features within the anomaly analysis time window, calculate the anomaly meteorological explanatory power based on the single-machine meteorological influence vector, and generate equipment degradation confidence value by combining the operational anomaly intensity, operational anomaly duration, synergy between multiple operational anomalies and historical fault association information. The explanatory power of abnormal weather is used to characterize the extent to which abnormal operational characteristics can be explained by meteorological disturbances; The fault prediction and health management module is used to correct the fault prediction weights of abnormal operating characteristics based on the equipment degradation confidence value and generate a health degradation curve. The window analysis module is used to generate the task completion probability for the current weather window and the task completion probability for the next weather window. The risk decision module is used to construct a set of maintenance paths and executable waiting paths for the current meteorological window based on the equipment degradation confidence value, health degradation curve and task completion probability. The path risk of the maintenance path for the current meteorological window is used as the benchmark risk. The difference between the minimum path risk in the set of executable waiting paths and the benchmark risk is determined as the window miss incremental risk. The incremental risk of missing a window reflects the risk of equipment status changes caused by the continued operation of the target wind turbine or waiting for the next meteorological window, rather than simply economic or scheduling costs; the set of executable waiting paths includes at least one of the following: no-processing waiting path, power-limited waiting path, remote reset monitoring waiting path, or shutdown waiting path; the scheduling feedback module is used to output operation and maintenance strategies based on the incremental risk of missing a window.

2. The offshore wind power intelligent operation and maintenance management platform based on meteorological data as described in claim 1, characterized in that, The data construction module converts regional or station-level meteorological and sea state data into a single-unit meteorological influence vector corresponding to the target wind turbine based on at least two of the following: turbine location coordinates, wind measurement point location, meteorological grid location, prevailing wind direction, wind field layout, upstream wind turbine operating status, wake influence, sea surface humidity, salt spray exposure level, and data acquisition time. The single-unit meteorological influence vector includes at least two of the following: effective wind speed, effective turbulence intensity, gust impact value, wake disturbance value, wave height influence value, humidity influence value, salt spray exposure value, meteorological prediction confidence level, and meteorological event label. The meteorological event label includes at least one of the following: short-term gust event, strong turbulence event, wave height change event, high humidity and high salt spray event, lightning warning event, typhoon outer periphery influence event, low visibility event, or continuous severe weather event.

3. The offshore wind power intelligent operation and maintenance management platform based on meteorological data as described in claim 1, characterized in that, When the target wind turbine experiences an operational anomaly, the anomaly detection module determines an anomaly analysis time window based on the occurrence time of the anomaly, and extracts the operational anomaly features within the anomaly analysis time window. The operational anomaly features include at least one of the following: vibration anomaly, temperature rise anomaly, power fluctuation anomaly, yaw action anomaly, pitch action anomaly, or alarm frequency anomaly. Each type of operational anomaly feature includes at least two of the following: anomaly occurrence time, anomaly duration, anomaly intensity, anomaly change trend, anomaly recovery time, and anomaly repetition number.

4. The offshore wind power intelligent operation and maintenance management platform based on meteorological data according to claim 1, characterized in that, When calculating the explanatory power of the abnormal meteorological conditions, the anomaly detection module determines at least two of the following: time synchronization score, trend consistency score, duration matching score, recovery score, historical similarity score, and similar operating condition response score. Specifically, the time synchronization score characterizes the degree of synchronization between the time of the operational anomaly and the time of the meteorological event; the trend consistency score characterizes the degree of consistency between the trend of the operational anomaly and the trend of the meteorological disturbance; the duration matching score characterizes the degree of matching between the duration of the operational anomaly and the duration of the meteorological event; the recovery score characterizes whether the operational anomaly has recovered after the meteorological event; the historical similarity score characterizes whether the target wind turbine has experienced similar operational anomalies under similar historical meteorological events; and the similar operating condition response score characterizes whether other wind turbines in the same wind farm with similar locations, wind directions, or operating conditions exhibit similar operational responses under the same meteorological event.

5. The offshore wind power intelligent operation and maintenance management platform based on meteorological data according to claim 4, characterized in that, The anomaly detection module performs a weighted fusion of at least two of the time synchronization score, trend consistency score, duration matching score, recovery score, historical similarity score, and similar working condition response score to obtain the explanatory power of the abnormal weather. When data corresponding to any explanation score is missing, the anomaly detection module resets the weight corresponding to that explanation score to zero, and normalizes the weights corresponding to the remaining explanation scores before calculating the explanatory power of the abnormal weather.

6. The offshore wind power intelligent operation and maintenance management platform based on meteorological data according to claim 1, characterized in that, The anomaly detection module generates the equipment degradation confidence value based on hierarchical judgment logic. Specifically, the module first determines whether the operational anomaly corresponding to the operational anomaly feature has a meteorological disturbance explanation basis based on the anomaly meteorological explanatory power. Then, it determines whether the operational anomaly meets preset recovery conditions, preset continuous enhancement conditions, points to the same component or fault chain as other operational anomalies, and whether there is a historical correlation with similar faults after the meteorological event corresponding to the single-unit meteorological influence vector ends. Finally, it comprehensively generates the equipment degradation confidence value. When the operational anomaly meets the preset high explanatory power condition, and meets the preset recovery condition after the meteorological event ends, multiple operational anomalies meet the preset low coordination condition, and the historical fault correlation information meets the preset low correlation condition, the equipment degradation confidence value is reduced. When the operational anomaly meets the preset high explanatory power condition but continues to exist or continuously enhance after the meteorological event ends, or multiple operational anomalies meet the preset high coordination condition, or the historical fault correlation information meets the preset high correlation condition, the equipment degradation confidence value is maintained or increased.

7. The offshore wind power intelligent operation and maintenance management platform based on meteorological data according to claim 6, characterized in that, The anomaly detection module generates at least two of the following: vibration degradation confidence value, temperature rise degradation confidence value, power fluctuation degradation confidence value, yaw degradation confidence value, pitch degradation confidence value, and alarm degradation confidence value. Based on the fault correlation between different operational anomaly characteristics and wind turbine components, it fuses to obtain the equipment degradation confidence value of the target wind turbine. Among them, the vibration degradation confidence value is associated with at least one of bearing, gearbox, or main shaft failure; the temperature rise degradation confidence value is associated with at least one of gearbox, generator, or converter failure; the yaw degradation confidence value is associated with yaw system failure; the pitch degradation confidence value is associated with pitch system failure; and the power fluctuation degradation confidence value is associated with at least one of pitch system, converter, or weather disturbance.

8. The offshore wind power intelligent operation and maintenance management platform based on meteorological data according to claim 1, characterized in that, When generating fault prediction results, the fault prediction and health management module reduces the fault prediction weight of an abnormal operation feature if the explanatory power of abnormal weather meets the preset high explanatory power condition and the confidence value of equipment degradation meets the preset low confidence value condition. For operational anomalies where the explanatory power of abnormal weather meets the preset low explanatory power condition and the confidence value of equipment degradation meets the preset high confidence value condition, the fault prediction weight of the operational anomaly feature is increased. When the explanatory power of abnormal weather meets the preset high explanatory power condition but the confidence value of equipment degradation does not meet the preset low confidence value condition, the contribution of this abnormal operation feature to the fault prediction results and health degradation curve is retained.

9. The offshore wind power intelligent operation and maintenance management platform based on meteorological data according to claim 8, characterized in that, The fault prediction and health management module generates at least two of the following based on the changing trend of the equipment degradation confidence value under different operating strategies: full-load operation health degradation curve, power-limited operation health degradation curve, operation health degradation curve after remote reset, shutdown waiting for maintenance health degradation curve, and continuous monitoring without intervention health degradation curve. It then determines the waitability level of the target wind turbine based on the change in the equipment degradation confidence value under at least one of the following states: power-limited operation, remote reset, or shutdown waiting. Specifically, when the equipment degradation confidence value decreases or its growth slows down after power-limited operation, the waitability level of the target wind turbine is increased; when the equipment degradation confidence value decreases and the alarm is cleared after remote reset, the waitability level of the target wind turbine is increased; and when the equipment degradation confidence value continues to rise after power-limited operation, remote reset, or shutdown waiting, the waitability level of the target wind turbine is decreased.

10. The offshore wind power intelligent operation and maintenance management platform based on meteorological data according to claim 1, characterized in that, The window analysis module generates the task completion probability for the current weather window and the task completion probability for the next weather window based on at least two of the following: the operable duration of the weather window, the maintenance time distribution corresponding to the target fault type, the ship's sailing time, the personnel boarding preparation time, the personnel qualification matching status, the availability of spare parts, the availability of tools, the return safety time, and the risk of window shortening. The risk decision module constructs the maintenance path for the current weather window and a set of executable waiting paths. The set of executable waiting paths includes at least one of the following: no-processing waiting path, power-limited waiting path, remote reset monitoring waiting path, or shutdown waiting path. Before determining the set of executable waiting paths, the risk decision module performs an executability judgment on each waiting path. If a waiting path does not meet safety constraints, equipment control conditions, or operation and maintenance rules, then that waiting path will not participate in the selection of the optimal waiting path risk; the path risk of the current meteorological window maintenance path and the path risk of each waiting path in the set of executable waiting paths are determined based on at least two of the following: changes in equipment degradation confidence value, changes in health index, risk of fault expansion, downtime risk, risk of improved maintenance level, and the probability of task completion for the corresponding meteorological window; the scheduling feedback module outputs at least one of the following strategies based on the incremental risk of window miss: continuous monitoring strategy, remote reset observation strategy, power limiting waiting strategy, next meteorological window maintenance strategy, current meteorological window maintenance strategy, current meteorological window immediate maintenance strategy, or downtime waiting maintenance strategy.

Images