A multi-parameter correlation fusion-based offshore wind turbine state evaluation system

Abstract

The present application relates to the technical field of equipment state evaluation, in particular to a kind of offshore wind turbine state evaluation system based on multi-parameter association fusion, system includes disturbance feature identification module, load response determination module, vibration trend induction module, parameter correlation analysis module, state grade determination module.In the present application, by constructing the analysis process with disturbance trigger as starting point, identify wind speed jump and carry out disturbance classification, associate disturbance time window with load torque fluctuation, extract periodic variation characteristics and map as torque level, while combining vibration direction trend change before and after disturbance, comprehensively induce the synchronous association state between each key parameter corresponding to disturbance, establish the logical closed loop among wind speed, load and structure response in dynamic disturbance scene, realize the unified identification of state change trend among multi-parameters, enhance the sensitivity and explanatory power of state evaluation to sudden operating disturbance, improve the support ability of evaluation result to fine operation and maintenance.

Description

Technical Field

[0001] This invention relates to the field of equipment condition assessment technology, and in particular to a condition assessment system for offshore wind turbines based on multi-parameter correlation fusion. Background Technology

[0002] The field of equipment condition assessment technology involves the monitoring, analysis, and judgment of the operating status of industrial equipment. Its core aspects include data acquisition based on multiple operating parameters, condition diagnosis, fault early warning, and health assessment. It is widely used in various industrial sectors such as wind power, thermal power, hydropower, petrochemicals, and transportation. Through multi-source data fusion, condition modeling, and the construction of assessment criteria, it achieves accurate identification and management of equipment operating status, improving the intelligence and precision of equipment operation and maintenance. Among these, the traditional offshore wind turbine condition assessment system refers to a technical system used to monitor and judge the operating status of offshore wind turbines. It primarily addresses issues such as aging, wear, corrosion, and structural fatigue of key components caused by the complex marine environment during long-term operation of offshore wind turbines. By collecting parameters such as wind speed, blade vibration, motor speed, bearing temperature, and power output, it uses empirical formulas, expert judgment, or statistical analysis to assess the health status of the wind turbine, thereby assisting in maintenance decisions and risk management.

[0003] Existing offshore wind turbine condition assessment technologies rely on data acquisition methods for single or isolated parameters such as wind speed, vibration, motor speed, and bearing temperature. They use empirical formulas or expert experience for judgment, lacking in-depth exploration of the logical relationship between the background of operational disturbances and load response. This results in unclear response mechanisms to changes in operating status, and in actual operation, it is easy to overlook the chain reactions caused by disturbances and the dynamic correlation between various parameters. It is impossible to accurately characterize the overall response performance of the equipment during disturbances, and the assessment results are easily affected by experience bias, leading to increased risk of judgment lag or misjudgment. It is difficult to support targeted and reasonable operation and maintenance scheduling strategies, thus limiting the guiding effectiveness of assessment results for actual operation and maintenance work. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a condition assessment system for offshore wind turbines based on multi-parameter correlation fusion.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a condition assessment system for offshore wind turbines based on multi-parameter correlation fusion includes:

[0006] The disturbance feature recognition module collects second-level wind speed data from the wind speed sensor at the front end of the offshore wind turbine blade, calculates the wind speed jump rate, classifies the wind speed disturbance interval, marks the disturbance response window and assigns a corresponding number to the window, and obtains the disturbance window sequence number.

[0007] The load response measurement module acquires synchronous torque sensor data based on the disturbance window sequence number, filters continuous periodic torque waveforms within the disturbance response window, identifies the occurrence time of periodic troughs, compares with the theoretical period to obtain the time difference sequence, selects the maximum time difference value within each disturbance window and maps it to the torque response standard level range to obtain the torque level mapping label.

[0008] The vibration trend summarization module extracts the vertical vibration data of the tower vibration acceleration sensor within the corresponding disturbance response window based on the torque level mapping label, calculates the average acceleration change, constructs a vibration variation identifier using the change direction and disturbance level, and generates a vibration trend identifier label.

[0009] The parameter correlation analysis module statistically analyzes three items in the disturbance response window corresponding to the vibration trend label: wind speed jump rate, torque cycle time difference, and vibration change label. It identifies whether the rising and falling directions of these three items are in the same synchronous direction in the current disturbance response window, records the cumulative number of each direction combination category, extracts the highest combination feature by frequency, and outputs the multi-parameter synchronous combination type.

[0010] Based on the multi-parameter synchronization combination type, the status level determination module maps the current disturbance response window with a level identifier, outputs the status assessment level and assessment label of the current disturbance response window, and obtains the status assessment result of the offshore wind turbine.

[0011] As a further aspect of the present invention, the marked disturbance response window is specifically configured to match a fixed window backward from the start time point corresponding to each disturbance level.

[0012] As a further embodiment of the present invention, the disturbance window sequence number includes the disturbance level number, the disturbance start time, and the disturbance response window number; the torque level mapping label includes the maximum period time difference value, the torque response level, and the period matching label; the vibration trend identification label includes the acceleration change direction, the disturbance level corresponding category, and the vibration change combination identifier; the multi-parameter synchronization combination type includes the wind speed jump rate change direction, the torque period time difference difference change direction, and the vibration change direction combination category; and the offshore wind turbine status assessment result includes the status level label, the assessment label, and the mapping type matching status.

[0013] As a further aspect of the present invention, the disturbance feature recognition module includes:

[0014] The data acquisition submodule collects second-level wind speed data from the wind speed sensor nodes of the front blades of the offshore wind turbine, obtains the raw wind speed value sequence at each moment, performs time series alignment processing on the data, unifies the timestamp format and data frequency, and generates wind speed time series data.

[0015] The jump rate calculation submodule calculates the wind speed difference between adjacent moments based on the wind speed time series data, and performs division operation at unit time intervals to obtain the wind speed jump rate value sequence for the corresponding time period. Based on the absolute value of the wind speed jump rate and the preset disturbance classification threshold, it makes hourly judgments and classifies the levels to generate wind speed disturbance level intervals.

[0016] The disturbance window marking submodule locates a fixed time window backward based on the start time point corresponding to each disturbance level in the wind speed disturbance level range, and assigns a number to each window, associates it with the corresponding disturbance level, establishes a window number sequence, and generates a disturbance window sequence number.

[0017] As a further aspect of the present invention, the load response measurement module includes:

[0018] The torque data extraction submodule locates the start and end time points corresponding to each disturbance response window based on the disturbance window sequence number, extracts the yaw side electronic control system torque sensor data sequence within the time period from the synchronization timestamp, performs time domain integrity filtering on the data, removes sampling anomalies and missing points, and obtains the disturbance torque data array.

[0019] The periodic trough identification submodule extracts each group of data window by window based on the disturbance torque data array, marks local minimum points using a differential comparison method, judges the trough position within a continuous period, extracts the timestamp of each periodic trough, compares the periodic time position with the theoretical period template, calculates the difference between each trough time and the theoretical periodic trough, and generates a periodic response time difference sequence.

[0020] The response level mapping submodule extracts the maximum value of the time difference in each segment of the periodic response time difference sequence according to the disturbance response window. Combined with the preset torque response standard level range, the maximum time difference value is compared with each level interval. The level label is assigned according to the corresponding interval, and a one-to-one mapping relationship between the disturbance response window and the response level is established to generate torque level mapping labels.

[0021] As a further aspect of the present invention, the vibration trend summarization module includes:

[0022] The vibration data extraction submodule, based on the torque level mapping label, sequentially locates the start and end time periods of each disturbance response window, selects the vibration acceleration sensor node corresponding to the wind turbine tower in the vertical direction, collects the acceleration sampling data sequence before and after each window, performs unified calibration on the sampling frequency, and generates acceleration data pairs before and after the disturbance.

[0023] The acceleration change calculation submodule performs mean calculation operation on each set of data based on the acceleration data before and after the disturbance, extracts the mean acceleration of the front window and the mean acceleration of the back window, calculates the direction of change by subtracting the front mean from the back mean, and marks the difference as rising, falling, or no change state, respectively, and obtains the acceleration change direction label.

[0024] The trend label construction submodule constructs a mapping relationship table structure based on the acceleration change direction label and the torque level mapping label in the corresponding disturbance response window. It assigns a unique identifier to each pair of disturbance level and change direction combinations, arranges the combination pairs in chronological order, and writes them into the label sequence to generate vibration trend label.

[0025] As a further aspect of the present invention, the parameter correlation analysis module includes:

[0026] The directional trend extraction submodule, based on the vibration trend identification label, analyzes the changing direction of three indicators corresponding to each disturbance response window: wind speed jump rate, torque period time difference, and vibration acceleration change. It determines the rising and falling states and converts them into three types of symbol labels: rising, falling, and no change, generating a multi-parameter change direction sequence.

[0027] The direction combination statistics submodule combines the wind speed jump direction, torque response direction, and vibration trend direction within each disturbance response window based on the multi-parameter change direction sequence. It also filters and judges based on direction consistency, calculates the frequency of combinations where the three directions are completely consistent, calculates the cumulative contribution value of direction consistency, and counts the frequency of occurrence of various direction combinations to obtain a direction combination frequency distribution table.

[0028] The combined feature recognition submodule sorts the first and second items according to the frequency distribution table of the direction combination from high to low, analyzes the direction state combination pattern of the three indicators corresponding to the combination, assigns a unique code to establish a mapping relationship, and generates a multi-parameter synchronous combination type.

[0029] As a further aspect of the present invention, the formula for calculating the cumulative contribution value of directional consistency is as follows:

[0030] ;

[0031] in, Indicates the first The direction of wind speed jump rate change within each disturbance response window. This indicates the direction of the torque period difference change within the corresponding disturbance response window. This indicates the direction of vibration trend change within the disturbance response window. The three parameter value ranges are set as follows: increase = +1, decrease = -1, and no change = 0. The cumulative contribution to synchronization consistency across all directions.

[0032] As a further aspect of the present invention, the state level determination module includes:

[0033] Based on the multi-parameter synchronization combination type, the template matching submodule performs item-by-item matching between the combination code of each disturbance response window and the built-in offshore wind power operation and maintenance experience template, judges according to structural consistency, extracts the successfully matched combination records and the risk level segment to which they belong in the template, and generates a combination matching level index table.

[0034] The level mapping submodule classifies the disturbance response window number according to the level index table of the combination matching level, according to the definition rules of the risk segment, sets the level identifier code, and maps the disturbance response window to the corresponding level label to generate the disturbance window status level label.

[0035] The output module combines the disturbance window status level labels, packages and encapsulates each disturbance response window number and its level label in sequence, and pushes it to the data receiving channel of the operation and maintenance scheduling platform to generate the offshore wind turbine status assessment results.

[0036] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0037] In this invention, an analysis process starting from disturbance triggering is constructed to identify wind speed jumps and classify disturbances. The disturbance time window is associated with load torque fluctuations, periodic change features are extracted and mapped to torque levels, and the synchronous correlation between key parameters corresponding to the disturbance is comprehensively summarized by combining the trend changes in vibration direction before and after the disturbance. By combining the combined features of the change directions of the three, the mapping and label output of the state level are completed. In the dynamic disturbance scenario, a logical closed loop is established between wind speed, load, and structural response, realizing the unified identification of state change trends among multiple parameters. This enhances the sensitivity and interpretability of state assessment to sudden operational disturbances and improves the ability of assessment results to support refined operation and maintenance. Attached Figure Description

[0038] Figure 1 This is a system flowchart of the present invention;

[0039] Figure 2 This is a flowchart of the disturbance feature recognition module of the present invention;

[0040] Figure 3 This is a flowchart of the load response measurement module of the present invention;

[0041] Figure 4 This is a flowchart of the vibration trend summarization module of the present invention;

[0042] Figure 5 This is a flowchart of the parameter correlation analysis module of the present invention;

[0043] Figure 6 This is a flowchart of the status level determination module of the present invention. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0045] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0046] Please see Figure 1 A condition assessment system for offshore wind turbines based on multi-parameter correlation fusion includes:

[0047] The disturbance feature recognition module collects second-level wind speed data from the wind speed sensor nodes of the front blades of the offshore wind turbine, calculates the wind speed difference between adjacent moments and divides it by the time interval to output the wind speed jump rate, classifies the wind speed disturbance interval according to the absolute value of the wind speed jump rate, matches the starting time point corresponding to each disturbance level with a fixed window, marks the corresponding interval as the disturbance response window, assigns the window a corresponding number, and obtains the disturbance window sequence number.

[0048] The load response measurement module obtains torque sensor data from the yaw side electronic control system based on the disturbance window sequence number, filters continuous periodic torque waveforms within the disturbance response window, identifies the occurrence time of periodic troughs, compares with the theoretical period to obtain the time difference sequence, selects the maximum time difference value within each disturbance window and maps it to the torque response standard level range to obtain the torque level mapping label;

[0049] The vibration trend summarization module extracts the vertical vibration data of the tower vibration acceleration sensor monitoring node within the corresponding disturbance response window based on the torque level mapping label, calculates the average acceleration change within a 0.5-second window before and after the disturbance, classifies the change direction, and uses the change direction and disturbance level to construct a vibration change label to generate a vibration trend label.

[0050] The parameter correlation analysis module is based on three items in the disturbance response window corresponding to the vibration trend label: wind speed jump rate, torque period time difference, and vibration change label. The direction of rise and fall of these three items in the current disturbance response window is set as the judgment criterion. It identifies whether there is a situation where the synchronization direction is consistent, records the cumulative number of each direction combination category, extracts the highest combination feature by frequency sorting, and outputs the multi-parameter synchronization combination type.

[0051] The status level determination module is based on the multi-parameter synchronization combination type, sets the synchronization direction and matches it with the known offshore wind power operation and maintenance experience template. According to the interval category to which the matching type belongs, the current disturbance response window is mapped with a level identifier. The status assessment level and assessment label of the current disturbance response window are output to the operation and maintenance scheduling platform to obtain the status assessment result of the offshore wind turbine.

[0052] The disturbance window sequence number includes the disturbance level number, disturbance start time, and disturbance response window number. The torque level mapping label includes the maximum period time difference value, torque response level, and period matching label. The vibration trend label includes the acceleration change direction, disturbance level corresponding category, and vibration change combination label. The multi-parameter synchronization combination type includes the wind speed jump rate change direction, torque period time difference value change direction, and vibration change direction combination category. The offshore wind turbine status assessment result includes the status level label, assessment label, and mapping type matching status.

[0053] Please see Figure 2 The disturbance feature recognition module includes:

[0054] The data acquisition submodule collects second-level wind speed data from the wind speed sensor nodes of the front blades of the offshore wind turbine, obtains the raw wind speed value sequence at each moment, performs time series alignment processing on the data, unifies the timestamp format and data frequency, and generates wind speed time series data.

[0055] First, it is necessary to collect second-level wind speed data corresponding to each node of the wind speed sensor on the front end of the offshore wind turbine blade. The specific deployment location and sampling frequency of the sensors must first be determined. Assuming that wind speed sensors numbered A01 to A10 are installed on the front end of the blade, and the sampling frequency is set to 1Hz (i.e., wind speed value is collected once per second), after the system starts at time T, it sequentially collects the wind speed values ​​of each node within seconds T+1, T+2…T+60 and generates time stamps. For example, if the sampling value of node A01 is 12.6 m / s at second T+1, 13.1 m / s at second T+2, and 13.8 m / s at second T+3, then these values ​​will be recorded sequentially as { (T+1, 12.6), (T+2, 13.1), (T+3, 13.8)}, all sensor node data must be aggregated into the wind speed data center after being formatted uniformly. Then, the collected data undergoes time alignment processing, i.e., determining whether there are inconsistencies in time labels among the sensors. If the sampling time of node A02 is T+1.05, T+2.03, or T+3.00 seconds, its corresponding time labels need to be aligned to whole seconds, and missing or biased data is corrected using interpolation. For example, if A02 has no valid value at T+2 seconds, it can be processed into 13.4 m / s using linear interpolation of preceding and following values. The formula used here is:

[0056] ;

[0057] in , Substituting into the calculation, we get:

[0058] ;

[0059] All corrected data are organized into a standard wind speed sequence array in seconds. The sequence of each node should meet the length consistency requirement. For example, the sampling result within 60 seconds should be 60 wind speed values. After the data is unified, it is stored in a two-dimensional array structure. Each row represents a different sensor number, and each column represents the data at the same time, as shown in Table 1.

[0060] Table 1. Sensor wind speed sampling table:

[0061] ;

[0062] As shown in Table 1, the wind speed data is aligned to the second level and stored uniformly to facilitate the subsequent calculation of the jump rate and the execution of the disturbance identification process, thereby obtaining wind speed time series data.

[0063] The jump rate calculation submodule calculates the wind speed difference between adjacent moments based on wind speed time series data, and performs division operations at unit time intervals to obtain the wind speed jump rate value sequence for the corresponding time period. Based on the absolute value of the wind speed jump rate and the preset disturbance classification threshold, it makes hourly judgments and classifies the levels to generate wind speed disturbance level intervals.

[0064] Based on wind speed time series data, the first step is to calculate the wind speed difference between adjacent times for any node. Let the wind speed values ​​at node A01 from T+1 to T+4 be 12.6, 13.1, 13.8, and 14.2 m / s, respectively. Then the adjacent differences are 0.5, 0.7, and 0.4 m / s. Using a unit time interval Δt = 1 second for division, the jump rate sequence is obtained as 0.5, 0.7, and 0.4 m / s². A general formula is used here. ,in Indicates the wind speed at the current moment. The formula calculates the slope of the wind speed change per second per unit time. After the calculation, an array of abrupt change rates will be obtained. For the absolute value of the wind speed abrupt change rate, the following formula needs to be called. After converting all jump rate values ​​to absolute values, a preset disturbance level classification threshold sequence is called. The disturbance level is divided into four categories according to the absolute value, with the following classification criteria: 0–0.3 m / s² is level 1 disturbance, 0.3–0.6 m / s² is level 2 disturbance, 0.6–0.9 m / s² is level 3 disturbance, and greater than 0.9 m / s² is level 4 disturbance. Each level is assigned a label. Taking node A01 as an example, its jump rate value sequence is 0.5, 0.7, 0.4 m / s², corresponding to levels 2, 3, and 2 respectively, forming a wind speed disturbance level array. There is a level value corresponding to each second. The wind speed disturbance level sequence of each node is finally saved in the form of a two-dimensional matrix, with the horizontal axis representing time and the vertical axis representing the sensor number, which is used for subsequent disturbance response window marking, and finally the wind speed disturbance level range is obtained.

[0065] The disturbance window marking submodule locates a fixed time window backward based on the start time point corresponding to each disturbance level in the wind speed disturbance level range, and assigns a number to each window, associates it with the corresponding disturbance level, establishes a window number sequence, and generates a disturbance window sequence number.

[0066] Based on the wind speed disturbance level range, the start time point of each marked disturbance level is statistically extracted. A fixed window length positioning operation is performed, defined here as 10 seconds. If node A01 detects a disturbance level of 3 at T+2 seconds, the window needs to be extended to T+11 seconds. The wind speed variability sequence from T+2 to T+11 seconds is extracted and assigned a unified disturbance response number, for example, D003. Then, the search continues for the next disturbance level change point. If a level 4 disturbance reappears at T+15 seconds, a new window is established from T+15 to T+24 seconds, numbered D004. The numbering method uses a decimal numbering pattern in chronological order. All disturbance response windows are stored sequentially in an array structure, each item containing the start time... The system includes information on time, end time, disturbance level, and window number. The window number must not be duplicated with the disturbance level; that is, the disturbance level is used for classification, and the window number is used for unique identification. The window number sequence is represented as follows: [D003, D004, D005…], corresponding to their respective disturbance levels and time ranges. The system also needs to check whether adjacent windows overlap. If there is a starting time interval less than the window length, the windows need to be merged and assigned a new number. For example, if two windows at T+10 seconds and T+15 seconds overlap, they are merged into a new window from T+10 to T+24 seconds, and the number is updated to D003, while D004 is discarded. All number changes need to be adjusted synchronously in the number sequence, and a number change log is output, which is the disturbance window sequence number.

[0067] Please see Figure 3 The load response measurement module includes:

[0068] The torque data extraction submodule locates the start and end time points corresponding to each disturbance response window based on the disturbance window sequence number, extracts the yaw side electronic control system torque sensor data sequence within the time period from the synchronization timestamp, performs time domain integrity filtering on the data, removes sampling anomalies and missing points, and obtains the disturbance torque data array;

[0069] Based on the disturbance response window sequence number, the first step is to obtain the start and end timestamp interval corresponding to the disturbance number. Let's assume the time period corresponding to disturbance response window D005 is from T=100 seconds to T=140 seconds. The system calls the synchronously recorded main control system timeline to extract the torque sensor data uploaded by the yaw-side electronic control system within this time period. The torque data is sampled at 0.01-second intervals, forming an original data sequence containing 4001 data points. Subsequently, the extracted sequence undergoes a data integrity check to determine if there are any missing samples, jumps, or outliers. If the difference between point 1001 and points before and after it exceeds a set threshold of 5 N·m, it is identified as a jump point and replaced with the average of adjacent points. Then, the entire sequence is continuously filtered, retaining only complete sequence segments without breaks. Let the continuous stable segment be from T=105 seconds to T=135 seconds, corresponding to 3001 valid data points. A continuous torque dataset within the disturbance response window is constructed. Simultaneously, the data within different disturbance response windows are formatted uniformly and stored using a two-dimensional array structure. The horizontal axis represents time points, and the vertical axis represents the disturbance response window number. The sample organization is shown in the table below:

[0070] Table 2. Disturbance response window D005 torque data segment table:

[0071] ;

[0072] As shown in Table 2, the data structure within the constructed disturbance response window has continuity and consistency, which facilitates subsequent periodic identification and time difference calculation operations, and finally obtains the disturbance torque data array.

[0073] The periodic trough identification submodule extracts each set of data window by window based on the disturbance torque data array, marks local minimum points using a differential comparison method, judges the trough position within a continuous period, extracts the timestamp of each periodic trough, compares the periodic time position with the theoretical period template, calculates the difference between each trough time and the theoretical periodic trough, and generates a periodic response time difference sequence.

[0074] Based on the disturbance torque data array, the 3001 torque data points corresponding to window D005 are called. A point-by-point sliding comparison method is used to obtain the local minimum points in the torque waveform. Assuming the sliding window width is 5 points, the relationship between the center point and its adjacent two points is determined. If the center point is smaller than its four adjacent points, it is marked as a trough point. In the example, the torque value at T=106.20 seconds is 312.1 N·m, and the surrounding points are 312.5, 312.3, 312.4, and 312.6 N·m respectively. Therefore, 106.20 seconds is a trough point. Its timestamp is then recorded, and all trough time series are extracted, such as: 106.20, 108.05, 109.85, 1... At 11.65 seconds, the system calls the theoretical torque cycle set to 1.8 seconds, calculates the theoretical time points between adjacent troughs and compares them with the actual trough points to obtain the response time difference for each cycle. If the theoretical troughs are 106.00, 107.80, 109.60, and 111.40 seconds, the time differences are 0.20, 0.25, 0.25, and 0.25 seconds respectively, forming a cycle time difference sequence {0.20, 0.25, 0.25, 0.25}. This sequence is written into the record unit corresponding to window number D005 to form a structured response data table, where the time difference is the core metric used to determine the degree of cycle response deviation and generate the cycle response time difference sequence.

[0075] The response level mapping submodule extracts the maximum value of the time difference in each segment of the periodic response time difference sequence according to the disturbance response window. Combined with the preset torque response standard level range, the maximum time difference value is compared with each level interval. The level label is assigned according to the corresponding interval, and a one-to-one mapping relationship between the disturbance response window and the response level is established to generate torque level mapping labels.

[0076] For the periodic response time difference sequence, the response time difference groups of each disturbance response window are read sequentially. Let the sequence corresponding to window D005 be {0.20, 0.25, 0.25, 0.25}. The system extracts the maximum time difference value from it, taking 0.25 seconds as the representative index. Then, the response level standard interval is called, and the extracted value is matched with the level mapping table for judgment. The level mapping table is set as follows: 0–0.1 seconds is level 1, 0.1–0.2 seconds is level 2, 0.2–0.3 seconds is level 3, 0.3 seconds and above is level 4, and 0.25 seconds is the maximum time difference. Within the range of 0.2–0.3, the corresponding level label is L3, and this label is written into the D005 record field of the disturbance response window to establish a mapping relationship between the number D005 and L3. The above operation is repeated for all disturbance response windows to generate a complete response level matrix with the structure as follows: {D005: L3, D006: L2, D007: L4}. The result is stored in a one-dimensional dictionary format or a structured form, which can be used later to quickly locate the disturbance response intensity level and finally generate the torque level mapping label.

[0077] Please see Figure 4 The vibration trend summarization module includes:

[0078] The vibration data extraction submodule is based on torque level mapping labels. It sequentially locates the start and end time periods of each disturbance response window, selects the vibration acceleration sensor node corresponding to the wind turbine tower in the vertical direction, collects the acceleration sampling data sequence of the first 0.5 seconds and the last 0.5 seconds in each window, performs unified calibration on the sampling frequency, and generates acceleration data pairs before and after the disturbance.

[0079] Based on the torque level mapping labels, the start and end time intervals of the disturbance response window are analyzed one by one. Assuming label L3 corresponds to disturbance response window D007, with a time range of T=120.00 seconds to T=130.00 seconds, the system calls the sampling records of the tower vibration acceleration sensor node and extracts data from the vertical channel within the range of 0.5 seconds before and after the window, i.e., T=119.50 seconds to T=130.50 seconds. The sensor sampling frequency is set to 1000Hz, corresponding to 1000 data points per second. A total of 11000 data points will be generated during this time period. Subsequently, a frequency verification operation is performed on the extracted data to check each... If the time interval between data points is constant at 1ms, and there is an interval exceeding 2ms, it is judged as a timing anomaly and marked for removal. The system then executes the jump point removal process, replacing any point in the data sequence where the difference between any two points exceeds the set jump threshold of 2m / s². For example, if a single point suddenly increases by 15.3m / s² at T=122.173 seconds, with the values ​​of the points before and after being 9.6 and 9.5m / s², then the anomaly point is replaced with the average value before and after, i.e., (9.6+9.5) / 2=9.55m / s². The cleaned data segments are stored in the form of a two-dimensional array, with each group containing fields such as perturbation number, sampling time, and acceleration value, as shown in Table 3.

[0080] Table 3. Vibration Acceleration Sampling Table for Disturbance D007:

[0081] ;

[0082] As shown in Table 3, the sampling segments before and after the disturbance constitute a continuous and clean vibration acceleration time series, and finally the acceleration data pairs before and after the disturbance are obtained.

[0083] The acceleration change calculation submodule performs mean calculation operation on each set of acceleration data before and after the disturbance, extracts the mean acceleration of the front window and the mean acceleration of the back window, calculates the direction of change by subtracting the front mean from the back mean, and marks the state as rising, falling, or no change according to the sign of the difference, and obtains the acceleration change direction label;

[0084] Based on the acceleration data before and after the disturbance, the mean of all sampled values ​​within the 0.5-second window is calculated sequentially. Let window D007 contain 500 acceleration samples with values ​​{9.45, 9.46, 9.48, ..., 9.51}, and the latter contain {9.54, 9.57, 9.56, ..., 9.59}. The mean of the first segment is calculated by dividing the sum by the number of samples. Assuming the sum of the accelerations in the first segment is 4718.5 m / s², its mean is 4718.5 ÷ 500 = 9.437 m / s². The sum of the accelerations in the latter segment is 4792.5 m / s², corresponding to the mean... The initial acceleration is 9.585 m / s². The two means are then subtracted, resulting in a difference Δa = 9.585 − 9.437 = 0.148 m / s². The system determines the direction of acceleration change based on the sign of this difference. If Δa > 0, it is classified as an upward trend; if Δa < 0, it is classified as a downward trend; and if Δa = 0, it is classified as no change. In this example, Δa = 0.148 > 0, so it is marked as an upward trend. This label is then bound to the disturbance number and written into the corresponding data table field structure. A structured output is generated according to a unified rule, ultimately outputting the acceleration change direction label.

[0085] The trend label construction submodule constructs a mapping relationship table structure based on the acceleration change direction label and the torque level mapping label corresponding to the disturbance response window. It assigns a unique identifier to each pair of disturbance level and change direction combinations, arranges the combination pairs in chronological order, and writes them into the label sequence to generate vibration trend label.

[0086] Based on the acceleration change direction label and the established torque level mapping label, the change direction and level value under each pair of disturbance response window numbers are extracted. The system's preset mapping rules are then called to map the combination pairs to standard trend labels. The combination method uses the disturbance level code and change direction to form a key value. For example, level L3 combined with the upward direction forms the key "L3_Upward". The corresponding identifier code for this key is looked up in the mapping table and set to TT12. After completing the label replacement, the corresponding disturbance response window D007 is marked as TT12. This combination and mapping operation is performed on all disturbance response windows. For example, D006 is combined as L2_Descending and mapped to TT05, and D008 is combined as L4_No Change and mapped to TT17. The system establishes a one-to-one correspondence table between disturbance response windows and trend labels, as shown in Table 4.

[0087] Table 4 Vibration Trend Label Mapping Table:

[0088] ;

[0089] As shown in Table 4, by constructing a combined label of disturbance level and acceleration change direction and performing standardized mapping, vibration trend identification labels are finally generated.

[0090] Please see Figure 5 The parameter correlation analysis module includes:

[0091] The directional trend extraction submodule analyzes the direction of change of three indicators corresponding to each disturbance response window: wind speed jump rate, torque period time difference, and vibration acceleration change, based on vibration trend identification tags. It determines the rising and falling states and converts them into three types of symbol labels: rising, falling, and no change, generating a multi-parameter change direction sequence.

[0092] Based on vibration trend labels, the system extracts three types of data associated with each disturbance response window: wind speed jump rate, torque period time difference, and vertical acceleration change. It then analyzes disturbance response window D201 line by line. Assuming the average wind speed jump rate in this window is 0.42 m / s² before the disturbance and 0.58 m / s² after, with a difference ΔV = +0.16 m / s², the wind speed jump direction is marked as upward. Similarly, the torque period time difference is 0.24 s before the disturbance and 0.29 s after the disturbance. If the difference ΔT = +0.05s, it is marked as an upward movement; the mean value of the acceleration in the front window is 9.52m / s², and in the back window it is 9.32m / s², and ΔA = −0.20m / s², it is marked as a downward movement. The three directions are upward, upward, and downward, respectively. After being converted into numerical symbols, they are assigned the values ​​+1, +1, and −1, respectively. The combination is represented as {+1, +1, −1}. This combination is encoded and written into the data structure. The symbol conversion of all disturbance response windows is completed, and finally, a multi-parameter change direction sequence is generated.

[0093] The direction combination statistics submodule combines the wind speed jump direction, torque response direction, and vibration trend direction within each disturbance response window based on the multi-parameter change direction sequence. It then filters and judges combinations based on directional consistency, calculating the frequency of combinations where all three directions are completely consistent, using the following formula:

[0094] ;

[0095] The cumulative contribution value of directional consistency is obtained through calculation, and the frequency of occurrence of various directional combinations is counted to obtain a frequency distribution table of directional combinations; among which, Indicates the first The direction of wind speed jump rate change within each disturbance response window. This indicates the direction of the torque period difference change corresponding to the disturbance response window. This indicates the direction of vibration trend change within the disturbance response window. The three parameter value ranges are set as follows: increase = +1, decrease = -1, and no change = 0. The cumulative contribution to synchronization consistency across all directions;

[0096] Based on the multi-parameter change direction sequence, for each group of three-element direction combinations {wind speed direction} Torque direction Vibration direction The system performs consistency calculations, assuming the disturbance response window D201 direction group is {+1, +1, -1}. The system sequentially calculates the absolute values ​​of the pairwise differences between the three directions and sums them. Taking the reciprocal of this value, the contribution value is 1 / 4 = 0.25. The sum of the cumulative consistency frequency values ​​within all disturbance response windows is obtained according to the following formula. Then, each type of direction combination is classified, such as {+1, +1, +1}, {−1, −1, −1}, {+1, +1, −1}, etc. The number of occurrences is counted and recorded as the frequency. A structured distribution table is constructed as shown in Table 5.

[0097] Table 5. Frequency Statistics of Multi-Parameter Directional Combinations:

[0098] ;

[0099] As shown in Table 5, each group of directional combination mappings is independently encoded, and its cumulative frequency is counted to obtain the directional combination frequency distribution table.

[0100] formula The operational logic is based on a measurement mechanism of directional consistency among three parameters, and the three difference terms in the formula... , , These represent the directional differences between the wind speed jump direction and the torque response direction, the torque response direction and the vibration trend direction, and the vibration trend direction and the wind speed jump direction, respectively. The absolute values ​​of these directional differences are standardized to unsigned numbers to reflect whether the two directions are consistent. If the two directions are consistent, the difference is 0; if they are inconsistent, the difference is 1 or 2. The sum of the three differences varies within the range [0, 6]. The smaller the sum, the more consistent the three directions are. The reciprocal of this sum is then used as the consistency contribution value, reflecting the degree of synchronous change in the disturbance response window. The total consistency score is obtained by summing the reciprocals of all disturbance response windows. Therefore, the summation term uses addition to aggregate all directional differences, uses absolute value to measure the sign consistency between individual pairs of directions, and uses the reciprocal of the overall sum to measure the degree of synchronization, thus fully expressing the cooperative directional relationship between the three types of parameters.

[0101] The cumulative frequency value of directional consistency represents the overall degree to which the changing directions of the three parameters—wind speed variability, torque period difference, and vibration trend—are synchronous and consistent across all disturbance response windows. The larger the value, the more disturbance response windows show that the changing directions of the three parameters tend to be consistent, i.e., all three simultaneously exhibit an upward, downward, or unchanged directional state. This value is achieved by accumulating the directional consistency scores of the three parameters in each window. If the three directions are completely consistent in a certain window, then that window contributes the most to the frequency value. If there are differences in the three directions, then the contribution value of that window decreases. Therefore, this cumulative frequency value reflects the multi-parameter linkage and coordination of the entire system in the disturbance response process and is a core quantitative indicator for measuring the consistency of the impact of wind speed disturbances on structural response.

[0102] The combined feature recognition submodule sorts the direction combination frequency distribution table from high to low, extracts the first and second items, parses the direction state combination pattern of the three indicators corresponding to the combination, assigns a unique code to establish a mapping relationship, and generates a multi-parameter synchronous combination type.

[0103] Based on the frequency distribution table of directional combinations, the system sorts the combinations from highest to lowest frequency and extracts the combination with the highest frequency as C3, which corresponds to three directions: wind speed increase, torque increase, and vibration decrease. This combination constitutes the most frequently occurring cooperative change pattern during the disturbance response process. The system assigns a standardized label to it, coded as MC-Type03, and maps and binds it to its corresponding three directions, writing it into the multi-parameter identification output field. Finally, the mapping sequence structure is established as {C3: MC-Type03, C1: MC-Type01, C2: MC-Type02}. The system uses the code MC-Type03 as the identifier of the feature combination with the strongest correlation in the current stage, and finally generates the multi-parameter synchronous combination type.

[0104] Please see Figure 6 The status level determination module includes:

[0105] The template matching submodule is based on the multi-parameter synchronous combination type. It performs item-by-item matching between the combination code of each disturbance response window and the built-in offshore wind power operation and maintenance experience template. It judges according to structural consistency, extracts the successfully matched combination records and the risk level segment to which they belong in the template, and generates a combination matching level index table.

[0106] Based on the multi-parameter synchronous combination type, the current disturbance response window number and its corresponding combination code are first parsed. Let the combination type corresponding to window D012 be MC-Type03. The system reads the offshore wind power operation and maintenance experience template table. The template has stored the correspondence between historically identified combination types and their operation and maintenance classification levels. Let MC-Type03 correspond to level II in the template, indicating a medium risk level. The matching method adopts a string primary key comparison mechanism to ensure unique matching. Each combination type has a matching fault tolerance verification logic. If there is no completely consistent record of the combination identifier in the template, the system marks the combination as "unconverged" and does not participate in the subsequent level mapping process. If the matching is successful, the level segment corresponding to the combination is extracted and an index relationship between the combination and the level is established. In multiple disturbance response windows, such as D013 mapped to MC-Type02 with a classification level of I, and D014 mapped to MC-Type05 with a classification level of IV, the successfully matched combinations and their levels are registered in the structured table with three items: combination code, window number, and level number, as shown in Table 6.

[0107] Table 6: Combination Matching Level Index Table

[0108] ;

[0109] As shown in Table 6, the mapping relationship between the disturbance response window and the level segment is obtained through the matching operation, and finally the combined matching level index table is generated.

[0110] The level mapping submodule classifies the disturbance response window number according to the level index table and the definition rules of the risk segment, sets the level identifier code such as L1 to L4, and maps the disturbance response window to the corresponding level label to generate the disturbance window status level label.

[0111] Based on the combined matching level index table, the system extracts the level segment to which each disturbance response window belongs and calls the preset level mapping rules to convert the level segment into a standard level label. The rules are set as follows: segment I corresponds to label L1, segment II corresponds to label L2, segment III corresponds to label L3, and segment IV corresponds to label L4. For example, window D012 belongs to segment II and corresponds to label L2. The system sequentially performs the level label assignment operation on all successfully matched windows and binds the disturbance response window number with the level label to the output table structure, which is {window number: D012, level label: L2}. If a window fails to match due to combination failure, no value is assigned. This process constructs a level output sequence on a per-unit basis for the disturbance response window, forming a bidirectional mapping table between the window number and the standard level. At the same time, the system performs an integrity check on the assigned data, removes duplicate or erroneous records, and outputs a sorted structure after verification, finally generating the disturbance window status level label.

[0112] The output module combines the disturbance window status level label with the data receiving channel of the operation and maintenance scheduling platform to package and encapsulate each disturbance response window number and its level label in sequence, and pushes them to generate the offshore wind turbine status assessment results.

[0113] Based on the disturbance window status level label, each disturbance response window number and its corresponding level label are packaged into a structured data frame according to the interface protocol. Each frame contains three data fields: window number, level label, and corresponding combination code. The system encodes and compresses the packaged data and writes it into a buffer queue, then pushes it to the data receiving interface of the operation and maintenance scheduling platform to ensure that the window identification information and assessment level are transmitted synchronously. Example data frames are: {D012, L2, MC-Type03}, {D013, L1, MC-Type02}, etc. All data are arranged in ascending order by window number and include a timestamp field to form a complete output structure. At the same time, output logs are generated for platform backtracking and fault tracking records. The final output result is the offshore wind turbine status assessment result.

[0114] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A condition assessment system for offshore wind turbines based on multi-parameter correlation fusion, characterized in that, include: The disturbance feature recognition module collects second-level wind speed data from the wind speed sensor at the front end of the offshore wind turbine blade, calculates the wind speed jump rate, classifies the wind speed disturbance interval, marks the disturbance response window and assigns a corresponding number to the window, and obtains the disturbance window sequence number. The load response measurement module acquires synchronous torque sensor data based on the disturbance window sequence number, filters continuous periodic torque waveforms within the disturbance response window, identifies the occurrence time of periodic troughs, compares with the theoretical period to obtain the time difference sequence, selects the maximum time difference value within each disturbance window and maps it to the torque response standard level range to obtain the torque level mapping label. The vibration trend summarization module extracts the vertical vibration data of the tower vibration acceleration sensor within the corresponding disturbance response window based on the torque level mapping label, calculates the average acceleration change, constructs a vibration variation identifier using the change direction and disturbance level, and generates a vibration trend identifier label. The parameter correlation analysis module statistically analyzes three items in the disturbance response window corresponding to the vibration trend label: wind speed jump rate, torque cycle time difference, and vibration change label. It identifies whether the rising and falling directions of these three items are in the same synchronous direction in the current disturbance response window, records the cumulative number of each direction combination category, extracts the highest combination feature by frequency, and outputs the multi-parameter synchronous combination type. The status level determination module maps the current disturbance response window with a status identifier based on the multi-parameter synchronization combination type, outputs the status assessment level and assessment label of the current disturbance response window, and obtains the status assessment result of the offshore wind turbine. Specifically, the marked disturbance response window is configured to match a fixed window backward from the start time point corresponding to each disturbance level; The parameter correlation analysis module includes: The directional trend extraction submodule, based on the vibration trend identification label, analyzes the changing direction of three indicators corresponding to each disturbance response window: wind speed jump rate, torque period time difference, and vibration acceleration change. It determines the rising and falling states and converts them into three types of symbol labels: rising, falling, and no change, generating a multi-parameter change direction sequence. The direction combination statistics submodule combines the wind speed jump direction, torque response direction, and vibration trend direction within each disturbance response window based on the multi-parameter change direction sequence. It also filters and judges based on direction consistency, calculates the frequency of combinations where the three directions are completely consistent, calculates the cumulative contribution value of direction consistency, and counts the frequency of occurrence of various direction combinations to obtain a direction combination frequency distribution table. The combined feature recognition submodule sorts the first and second items according to the frequency distribution table of the direction combination from high to low, analyzes the direction state combination pattern of the three indicators corresponding to the combination, assigns a unique code to establish a mapping relationship, and generates a multi-parameter synchronous combination type.

2. The offshore wind turbine condition assessment system based on multi-parameter correlation fusion according to claim 1, characterized in that, The disturbance window sequence number includes the disturbance level number, disturbance start time, and disturbance response window number. The torque level mapping label includes the maximum period time difference value, torque response level, and period matching label. The vibration trend identification label includes the acceleration change direction, disturbance level corresponding category, and vibration change combination identifier. The multi-parameter synchronization combination type includes the wind speed jump rate change direction, torque period time difference difference change direction, and vibration change direction combination category. The offshore wind turbine status assessment result includes the status level label, assessment label, and mapping type matching status.

3. The offshore wind turbine condition assessment system based on multi-parameter correlation fusion according to claim 1, characterized in that, The disturbance feature recognition module includes: The data acquisition submodule collects second-level wind speed data from the wind speed sensor nodes of the front blades of the offshore wind turbine, obtains the raw wind speed value sequence at each moment, performs time series alignment processing on the data, unifies the timestamp format and data frequency, and generates wind speed time series data. The jump rate calculation submodule calculates the wind speed difference between adjacent moments based on the wind speed time series data, and performs division operation at unit time intervals to obtain the wind speed jump rate value sequence for the corresponding time period. Based on the absolute value of the wind speed jump rate and the preset disturbance classification threshold, it makes hourly judgments and classifies the levels to generate wind speed disturbance level intervals. The disturbance window marking submodule locates a fixed time window backward based on the start time point corresponding to each disturbance level in the wind speed disturbance level range, and assigns a number to each window, associates it with the corresponding disturbance level, establishes a window number sequence, and generates a disturbance window sequence number.

4. The offshore wind turbine condition assessment system based on multi-parameter correlation fusion according to claim 1, characterized in that, The load response measurement module includes: The torque data extraction submodule locates the start and end time points corresponding to each disturbance response window based on the disturbance window sequence number, extracts the yaw side electronic control system torque sensor data sequence within the time period from the synchronization timestamp, performs time domain integrity filtering on the data, removes sampling anomalies and missing points, and obtains the disturbance torque data array. The periodic trough identification submodule extracts each group of data window by window based on the disturbance torque data array, marks local minimum points using a differential comparison method, judges the trough position within a continuous period, extracts the timestamp of each periodic trough, compares the periodic time position with the theoretical period template, calculates the difference between each trough time and the theoretical periodic trough, and generates a periodic response time difference sequence. The response level mapping submodule extracts the maximum value of the time difference in each segment of the periodic response time difference sequence according to the disturbance response window. Combined with the preset torque response standard level range, the maximum time difference value is compared with each level interval. The level label is assigned according to the corresponding interval, and a one-to-one mapping relationship between the disturbance response window and the response level is established to generate torque level mapping labels.

5. The offshore wind turbine condition assessment system based on multi-parameter correlation fusion according to claim 1, characterized in that, The vibration trend summarization module includes: The vibration data extraction submodule, based on the torque level mapping label, sequentially locates the start and end time periods of each disturbance response window, selects the vibration acceleration sensor node corresponding to the wind turbine tower in the vertical direction, collects the acceleration sampling data sequence before and after each window, performs unified calibration on the sampling frequency, and generates acceleration data pairs before and after the disturbance. The acceleration change calculation submodule performs mean calculation operation on each set of data based on the acceleration data before and after the disturbance, extracts the mean acceleration of the front window and the mean acceleration of the back window, calculates the direction of change by subtracting the front mean from the back mean, and marks the difference as rising, falling, or no change state, respectively, and obtains the acceleration change direction label. The trend label construction submodule constructs a mapping relationship table structure based on the acceleration change direction label and the torque level mapping label in the corresponding disturbance response window. It assigns a unique identifier to each pair of disturbance level and change direction combinations, arranges the combination pairs in chronological order, and writes them into the label sequence to generate vibration trend label.

6. The offshore wind turbine condition assessment system based on multi-parameter correlation fusion according to claim 1, characterized in that, The formula for calculating the cumulative contribution value of directional consistency is as follows: ； in, This indicates the direction of the wind speed jump rate change in the i-th disturbance response window. This indicates the direction of the torque period difference change corresponding to the disturbance response window. This indicates the direction of vibration trend change within the disturbance response window. The cumulative contribution to synchronization consistency across all directions.

7. The offshore wind turbine condition assessment system based on multi-parameter correlation fusion according to claim 1, characterized in that, The status level determination module includes: Based on the multi-parameter synchronization combination type, the template matching submodule performs item-by-item matching between the combination code of each disturbance response window and the built-in offshore wind power operation and maintenance experience template, judges according to structural consistency, extracts the successfully matched combination records and the risk level segment to which they belong in the template, and generates a combination matching level index table. The level mapping submodule classifies the disturbance response window number according to the level index table of the combination matching level, according to the definition rules of the risk segment, sets the level identifier code, and maps the disturbance response window to the corresponding level label to generate the disturbance window status level label. The output module combines the disturbance window status level labels, packages and encapsulates each disturbance response window number and its level label in sequence, and pushes it to the data receiving channel of the operation and maintenance scheduling platform to generate the offshore wind turbine status assessment results.

Images