A method and system for online monitoring of a pitch bearing state

Abstract

The application provides a kind of variable pitch bearing state online monitoring method and system, it is related to variable pitch bearing technical field, including the following steps: wind speed prediction device calculates the wind intensity coefficient of each variable pitch bearing in each predicted time period and marks corresponding predicted time period;State prediction device predicts each predicted time period according to historical operation data and wind intensity coefficient, obtains predicted state data set;Wind farm planning total health function is constructed, and global optimal maintenance path is generated by iterative algorithm.The above method can quantize the nonlinear influence of wind speed, so that the system can predict the risk in real time.Iterative algorithm generates global optimal maintenance path to make wind farm planning total health function obtain optimal solution, and realize the economic maximum extension of bearing life.

Description

Technical Field

[0001] This invention relates to pitch bearings, specifically to a method and system for online monitoring of the condition of pitch bearings. Background Technology

[0002] As a core transmission component of the pitch system in wind turbines, the lifespan of pitch bearings is directly affected by the amplitude and frequency characteristics caused by dynamic wind loads. Wind speed fluctuations cause nonlinear changes in the aerodynamic forces acting on the blades, generating alternating inertial loads, which in turn excite bearing vibration. When the amplitude increases or the frequency approaches the system's natural frequency, resonance effects will occur, accelerating raceway fatigue spalling and bolt loosening. Therefore, online monitoring of the pitch bearing's condition is necessary.

[0003] However, existing monitoring technologies have significant drawbacks: traditional methods, through vibration RMS analysis, can detect bearing damage, but they ignore the dynamic correlation between wind speed and vibration parameters, lack predictability, and cannot generate optimal maintenance paths through prediction. Summary of the Invention

[0004] To overcome the existing technical problems, the present invention provides a method and system for online monitoring of pitch bearing condition that can generate optimal maintenance paths.

[0005] The present invention adopts the following technical solution.

[0006] A method for online monitoring of pitch bearing condition includes multiple pitch bearings, each pitch bearing being equipped with a condition monitoring device, a condition prediction device, and a database; It also includes the following steps: The wind speed prediction device calculates the wind intensity coefficient of each pitch bearing in each prediction time period and marks the corresponding prediction time period. The wind intensity coefficient is then synchronized to the cloud and each state prediction device. The pitch bearing is monitored by a condition monitoring device, and the bearing condition dataset is stored in a database and uploaded to the cloud. The database stores historical operating data. The condition prediction device makes predictions for each prediction period based on the historical operating data and the wind intensity coefficient, and the predicted condition dataset is uploaded to the cloud. The cloud merges the bearing status dataset into the corresponding historical dataset of the pitch bearing. Based on the predicted status dataset, the historical dataset, and the wind intensity coefficient, a wind farm planning overall health function is constructed. A globally optimal maintenance path is generated through an iterative algorithm to ensure that the wind farm planning overall health function obtains the optimal solution.

[0007] A further improvement of the present invention is that the historical dataset includes maintenance time data, multiple historical time periods, and historical average wind intensity, historical average amplitude data, and historical average frequency data, each corresponding to a historical time period. Maintenance time data includes the last maintenance timestamp, the last scheduled maintenance timestamp, the last maintenance delay duration, and the average maintenance duration. The last maintenance delay duration is the time between the last scheduled maintenance timestamp and the last maintenance timestamp. The specific steps for constructing the overall health function for wind farm planning include: S11. Select the historical time period between the last maintenance time stamp and the last maintenance time stamp, and integrate them into a delayed time period. Calculate the actual risk level and actual dynamic penalty amplification coefficient of the pitch bearing in the delayed time period based on the historical dataset. Repeat this step to obtain the actual risk level and actual dynamic penalty amplification coefficient of each pitch bearing. S12. Calculate the equipment life coefficient for each pitch bearing based on the actual risk level, the actual dynamic penalty amplification factor, and the duration of the previous maintenance delay. If this is the first maintenance, the equipment life coefficient is equal to the preset initial health degradation value. ; S13. Calculate the predicted risk level and predicted dynamic penalty amplification coefficient of the pitch bearing in the current iteration based on the predicted state dataset, and obtain the overall health function of the wind farm planning in the current iteration by combining the previous maintenance timestamp, equipment life coefficient, and average maintenance duration.

[0008] A further improvement to this invention is that the expression for the overall health function of wind farm planning is as follows: , , in, It is the first The health degradation value of a pitch bearing It is a weighting factor for equipment aging. It is the current timestamp. This is the timestamp from the last maintenance. It is the equipment lifespan factor. It is a weighting factor for delayed maintenance. It is the duration between the current timestamp and the timestamp of this maintenance due date. It is the predictive dynamic penalty amplification factor. It is a predictive risk weighting factor. It predicts the degree of risk.

[0009] A further improvement to the present invention includes the following specific steps for calculating the wind intensity coefficient of each pitch bearing in each prediction time period and marking the corresponding prediction time period using a wind speed prediction device: The wind speed prediction device predicts the future average wind speed and future average wind direction for each prediction period. The wind speed prediction device is equipped with a layout diagram with multiple pitch bearing position data and layout direction data. The pitch bearing located at the tail end of the future average wind direction is selected as the wind speed intensity benchmark. The wind intensity coefficient of each pitch bearing in each forecast time period is calculated and the corresponding forecast time period is marked. The expression of the wind intensity coefficient is as follows: , in, It is the wind intensity coefficient of the i-th pitch bearing in the q-th prediction time period. It is the future average wind speed for the q-th forecast period. It is the dynamic wind force attenuation coefficient of the i-th pitch bearing relative to the wind speed intensity benchmark. It is the angle between the future average wind direction in the q-th prediction time period and the arrangement direction of the i-th pitch bearing.

[0010] A further improvement of the present invention is that the historical operating data includes multiple historical time periods, as well as historical average wind intensity, historical average amplitude data and historical average frequency data, each corresponding to a historical time period. The predicted state data includes the prediction period, wind intensity coefficient, predicted average amplitude data, and predicted average frequency data; The specific steps by which the state prediction device makes predictions for each prediction period based on historical operating data and wind intensity coefficient to obtain the predicted state dataset include: S21. If the wind intensity coefficient is greater than the cut-out wind speed threshold, then the predicted state data marked as shutdown is obtained. S22. If the wind intensity coefficient is less than the cut-in wind speed threshold, then the predicted state data marked as non-start is obtained. S23. If the wind intensity coefficient is between the cut-out wind speed threshold and the cut-in wind speed threshold, calculate the pre-selected reference range based on the wind intensity coefficient and the preset similar intensity range, select several historical average wind intensities that fall within the pre-selected reference range, extract the wind intensity coefficient, the historical average wind intensity, and the historical average amplitude data and historical average frequency data corresponding to the historical average wind intensity, and integrate them into the historical reference dataset. Calculate the predicted state data based on the entire historical reference dataset. S24. Repeat steps S21 to S23 for each prediction time period to obtain multiple prediction state data that correspond one-to-one with the prediction time period, and integrate them into the prediction state dataset.

[0011] A further improvement to the present invention, after selecting several historical average wind intensities falling within a pre-selected reference range, includes the following step: If the number of historical average wind intensities falling within the pre-selected reference range exceeds the preset historical reference quantity value, then the difference between the current timestamp and the timestamp of each historical average wind intensity is calculated. If it is greater than the preset spatiotemporal reference threshold, then the historical average wind intensity is not selected. After integration into the historical reference dataset, the following steps are also included: extracting the timestamp of the historical average wind intensity and merging it into the corresponding historical reference dataset; The expression for calculating the predicted state data based on the entire historical reference dataset is as follows: , in, It is the first The predicted value for each prediction dataset, where N is the number of historical reference datasets. It is the time decay weight coefficient of the nth historical reference dataset. It is a time decay weighting factor. It is the timestamp of the nth historical reference dataset. It is the current timestamp. It is a dynamic reference time threshold. It is the dynamic decay weight of the nth historical reference dataset. It is the first Distance decay factor for each predicted data point It is the difference between the historical average wind intensity and the wind intensity coefficient in the nth historical reference dataset. It is the nth historical reference dataset and the nth The predicted data is the same as the historical values ​​of the same type.

[0012] A further improvement to the present invention is that the expression for the actual risk level in step S11 is as follows: ， ， in, It refers to the actual level of risk. It is the number of historical time periods within the delayed time period. It is the first The historical average wind intensity for a given historical period It is the first Historical amplitude and frequency influence coefficients for a given historical time period It is the first Historical average vibration frequency data for a historical period, It is a safe vibration frequency value. It is the first Historical average amplitude data for a given historical period This is the safe amplitude value. It is the amplitude and frequency weighting factor; If the actual risk level is less than the preset healthy operating value, the actual dynamic penalty amplification factor is 0. If the actual risk level is greater than the preset healthy operating value, the expression for the actual dynamic penalty amplification factor is as follows: , in, It is a penalty amplification weighting factor. These are healthy operating values; The expression for calculating the equipment life factor for each pitch bearing is as follows: , in, It is the equipment life coefficient of the i-th pitch bearing. This is the equipment life coefficient of the i-th pitch bearing at the time of the last maintenance. If this is the first maintenance, then this item is equal to the preset initial health degradation value. , It is the delay time since the last maintenance of the i-th pitch bearing. It represents the actual risk level of the i-th pitch bearing. is the actual dynamic penalty amplification factor for the i-th pitch bearing, and M is the number of pitch bearings.

[0013] A further improvement to the present invention is that the expression for predicting the degree of risk in step 13 is as follows: ， ， in, It is about predicting the degree of risk. It is the number of time periods in the current iteration prediction. It is the first Wind intensity coefficient for a given forecast period It is the first The influence coefficient of predicted amplitude and frequency for each prediction time period. It is the first The predicted average vibration frequency data for each prediction time period, It is a safe vibration frequency value. It is the first The predicted average amplitude data for each prediction period. This is the safe amplitude value. It is the amplitude and frequency weighting factor; If the predicted risk level is less than the preset healthy operating value, the predicted dynamic penalty amplification factor is 0. If the predicted risk level is greater than the preset healthy operating value, the expression for the predicted dynamic penalty amplification factor is as follows: , in, It is a penalty amplification weighting factor. This is a healthy operating value.

[0014] A further improvement to this invention, the specific steps for generating the globally optimal maintenance path through an iterative algorithm, include: S31. Establish a virtual aging center based on the wind intensity coefficient and location data of each pitch bearing. The expression for the virtual aging center is as follows: , Where M is the number of pitch bearings. It is the health degradation value of the i-th pitch bearing in the current iteration. It is the position of the i-th pitch bearing; S32. Perform a global search iteratively based on the virtual aging center up to any pitch bearing. The expression for the global search is as follows: , , in, It is the path position in iteration t+1. It is the current position in the t-th iteration. To the c-th pitch bearing position The cost value is calculated using the cost function. It represents the path position of the pitch bearing with the lowest generation value in the t-th iteration. This is the current location. To the c-th pitch bearing position Path length, This is the average maintenance time for the c-th pitch bearing. It is the health degradation value of the c-th pitch bearing in the current iteration. , , These are cost weighting factors; The travel time to the pitch bearing is calculated based on its path length, and the daily working time is calculated by combining this with the average maintenance time of the pitch bearing. If the daily working time is less than a preset working time threshold, a local search is iterated to any pitch bearing, and the sum of the travel time, average maintenance time, and daily working time for that pitch bearing is calculated, covering the daily working time. The expression for the local search is as follows: , If the daily working hours exceed the preset working hours threshold, record whether the work for the day is completed at the end of the day. If completed, perform a global search iteration to any pitch bearing the next day. If not completed, convert the remaining working hours into the working hours for the next day and perform a local search iteration to any pitch bearing.

[0015] This invention also proposes an online monitoring system for the condition of a pitch bearing, which utilizes the method described above, including: The wind speed prediction device calculates the wind intensity coefficient of each pitch bearing in each prediction time period and marks the corresponding prediction time period, and synchronizes the wind intensity coefficient to the cloud and each state prediction device. The status monitoring device is used to receive the wind intensity coefficient marked with the corresponding prediction time period from the wind speed prediction device, and is used to monitor the pitch bearing, obtain the bearing status dataset, store it in the database, and upload it to the cloud. The database stores historical operating data and transmits the historical operating data to the status prediction device. The state prediction device makes predictions for each prediction period based on historical operating data and wind intensity coefficient, obtains the predicted state dataset, and uploads it to the cloud. In the cloud, the bearing status dataset is merged into the historical dataset corresponding to the pitch bearing. The predicted status dataset, historical dataset, pitch bearing location data and wind intensity coefficient are used to construct the overall health function of the wind farm planning. The globally optimal maintenance path is generated through an iterative algorithm so that the overall health function of the wind farm planning obtains the optimal solution.

[0016] The beneficial effects of this invention are as follows: Traditional vibration RMS analysis can only detect current bearing damage, while this solution predicts the wind intensity coefficient through a wind speed prediction device and quantifies the nonlinear influence of wind speed, enabling the system to predict the vibration life risk of the pitch bearing under specific wind speeds in real time. An iterative algorithm generates a globally optimal maintenance path to ensure the optimal solution for the overall health function of the wind farm planning, thereby maximizing the economic extension of bearing life. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the main process of the present invention; Figure 2 This is a schematic diagram illustrating the calculation of the wind intensity coefficient of this invention. Detailed Implementation

[0019] The accompanying drawings are for illustrative purposes only and should not be construed as limiting the scope of this patent. To better illustrate this embodiment, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the average size of the product.

[0020] It will be understood by those skilled in the art that certain well-known structures and their descriptions may be omitted in the accompanying drawings. The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0021] Reference Figures 1 to 2 As can be seen, an online monitoring method for the condition of pitch bearings includes multiple pitch bearings, each of which is equipped with a condition monitoring device, a condition prediction device, and a database. It also includes the following steps: The wind speed prediction device calculates the wind intensity coefficient of each pitch bearing in each prediction time period and marks the corresponding prediction time period. The wind intensity coefficient is then synchronized to the cloud and each state prediction device. A further improvement to the present invention includes the following specific steps for calculating the wind intensity coefficient of each pitch bearing in each prediction time period and marking the corresponding prediction time period using a wind speed prediction device: The wind speed prediction device predicts the future average wind speed and future average wind direction for each prediction period. The wind speed prediction device is equipped with a layout diagram with multiple pitch bearing position data and layout direction data. The pitch bearing located at the tail end of the future average wind direction is selected as the wind speed intensity benchmark. The wind intensity coefficient of each pitch bearing in each forecast time period is calculated and the corresponding forecast time period is marked. The expression of the wind intensity coefficient is as follows: , in, It is the wind intensity coefficient of the i-th pitch bearing in the q-th prediction time period. It is the future average wind speed for the q-th forecast period. It is the dynamic wind force attenuation coefficient of the i-th pitch bearing relative to the wind speed intensity benchmark. It is the angle between the future average wind direction in the q-th prediction time period and the arrangement direction of the i-th pitch bearing.

[0022] Each prediction time period is determined based on the predicted time range. For example, if the maintenance of all pitch bearings takes a week from start to finish, then in conventional prediction methods, the prediction time period needs to be at least one week (the overall maintenance path needs to be planned in advance). The longer the prediction time is, the lower the prediction accuracy. For predicting the average wind speed and average wind direction for the next week, it is recommended that each prediction time period be 3 to 6 hours. However, since the globally optimal maintenance path in this application is calculated through an iterative algorithm, the next pitch bearing to be maintained can be calculated during the maintenance of the i-th pitch bearing. In this case, the prediction time period basically only needs to be within one day. Under this premise, it is recommended that each prediction time period be one hour, which greatly improves the prediction accuracy.

[0023] The orientation data refers to the direction in which the pitch bearing corresponds to the wind turbine blades. Specifically, for example... Figure 2 The wind turbine located in the upper right corner is at the tail end of the expected wind direction; therefore, the wind turbine in the upper right corner serves as the wind speed intensity benchmark, i.e., the pitch bearing in the upper right corner serves as the wind speed intensity benchmark. By calculating the wind intensity coefficient, the effective wind speed blowing across the wind turbine blades can be effectively determined.

[0024] For dynamic wind attenuation coefficient The specific value will change depending on the wind speed intensity benchmark, because this value is calculated using historical operating data. If no historical data is available for reference, then the current pitch bearing dynamic wind attenuation coefficient will be used. The default value is 1. The dynamic wind attenuation coefficient is calculated by selecting historical average wind intensity with similar historical average wind direction from the historical operating data of this pitch bearing. It is important to note that the wind speed prediction device used to predict future average wind speed and direction is not installed on the wind turbine, while the anemometer used to detect real-time wind speed is installed on the wind turbine. The historical average wind intensity is calculated based on the wind speed measurements from the anemometer.

[0025] As a specific implementation method, if the i-th pitch bearing has the same historical average wind intensity as the pitch bearing used as the wind speed intensity benchmark, then the dynamic wind force attenuation coefficient is... The value is equal to 1. Additionally, it's important to note that for wind turbines in the same area, there is usually a prevailing wind direction (this is why wind turbines are installed there). Therefore, their arrangement directions are not only generally consistent, but their future average wind direction is also generally consistent. Consequently, the pitch bearings used as the benchmark for wind speed intensity are also generally fixed, thus affecting the dynamic wind attenuation coefficient. The value remains basically unchanged, so there is no need to use a large amount of historical data for calculation.

[0026] The pitch bearing is monitored by a condition monitoring device, and the bearing condition dataset is stored in a database and uploaded to the cloud. The database stores historical operating data. The condition prediction device makes predictions for each prediction period based on the historical operating data and the wind intensity coefficient, and the predicted condition dataset is uploaded to the cloud. A further improvement of the present invention is that the historical operating data includes multiple historical time periods, as well as historical average wind intensity, historical average amplitude data and historical average frequency data, each corresponding to a historical time period. The predicted state data includes the prediction period, wind intensity coefficient, predicted average amplitude data, and predicted average frequency data; The specific steps by which the state prediction device makes predictions for each prediction period based on historical operating data and wind intensity coefficient to obtain the predicted state dataset include: S21. If the wind intensity coefficient is greater than the cut-out wind speed threshold, then the predicted state data marked as shutdown is obtained. S22. If the wind intensity coefficient is less than the cut-in wind speed threshold, then the predicted state data marked as non-start is obtained. S23. If the wind intensity coefficient is between the cut-out wind speed threshold and the cut-in wind speed threshold, calculate the pre-selected reference range based on the wind intensity coefficient and the preset similar intensity range, select several historical average wind intensities that fall within the pre-selected reference range, extract the wind intensity coefficient, the historical average wind intensity, and the historical average amplitude data and historical average frequency data corresponding to the historical average wind intensity, and integrate them into the historical reference dataset. Calculate the predicted state data based on the entire historical reference dataset. S24. Repeat steps S21 to S23 for each prediction time period to obtain multiple prediction state data that correspond one-to-one with the prediction time period, and integrate them into the prediction state dataset.

[0027] Specifically, the cut-in wind speed refers to the minimum wind speed at which a wind turbine can start and begin generating electricity. The cut-out wind speed refers to the maximum wind speed at which a wind turbine will stop operating for safety reasons. Therefore, once the cut-out wind speed threshold is exceeded, the pitch bearing will be in a shutdown state. Similarly, once it is below the cut-in wind speed threshold, the pitch bearing will be in a non-starting state. When the pitch bearing is in a shutdown or non-starting state, the dynamic penalty amplification factor is predicted in the subsequent calculation of the maintenance level. The value is zero, predicting the level of risk. The value is also zero, a significant decrease. The numerical value of health decline.

[0028] A further improvement to the present invention, after selecting several historical average wind intensities falling within a pre-selected reference range, includes the following step: If the number of historical average wind intensities falling within the pre-selected reference range exceeds the preset historical reference quantity value, then the difference between the current timestamp and the timestamp of each historical average wind intensity is calculated. If it is greater than the preset spatiotemporal reference threshold, then the historical average wind intensity is not selected. Since the condition of pitch bearings deteriorates with age, older historical average wind intensities should be discarded when a sufficient number of historical average wind intensities fall within the pre-selected reference range. Specifically, the number of historical references is five, and the spatiotemporal reference threshold is one month. It's important to note that if a high wind intensity coefficient results in a low number of historical average wind intensities falling within the pre-selected reference range, that historical average wind intensity will still be selected even if the difference between the timestamp of that historical average wind intensity and the current timestamp exceeds the preset spatiotemporal reference threshold. This is because, with a high wind intensity coefficient, the impact of the wind intensity coefficient outweighs the impact of the pitch bearing's service life.

[0029] After integration into the historical reference dataset, the following steps are also included: extracting the timestamp of the historical average wind intensity and merging it into the corresponding historical reference dataset; The expression for calculating the predicted state data based on the entire historical reference dataset is as follows: , in, It is the first The predicted value for each prediction dataset, where N is the number of historical reference datasets. It is the time decay weight coefficient of the nth historical reference dataset. It is a time decay weighting factor. It is the timestamp of the nth historical reference dataset. It is the current timestamp. It is a dynamic reference time threshold. It is the dynamic decay weight of the nth historical reference dataset. It is the first Distance decay factor for each predicted data point It is the difference between the historical average wind intensity and the wind intensity coefficient in the nth historical reference dataset. It is the nth historical reference dataset and the nth The predicted data is the same as the historical values ​​of the same type.

[0030] When the number of historical reference datasets is less than or equal to the preset number of historical references, and the wind intensity coefficient is greater than the full-load wind speed, the dynamic reference time threshold is... This is equal to the oldest timestamp in the historical reference dataset, thus ensuring that the time decay weighting coefficient is the same for each historical reference dataset. The reason for requiring the wind intensity coefficient to be greater than the full-load wind speed is that when the wind intensity coefficient is between the cut-in wind speed and the full-load wind speed, the power increases with wind speed in an approximately cubic relationship, resulting in drastic fluctuations in the corresponding amplitude and frequency data. In this case, the influence of the pitch bearing's service life is similar to the influence caused by the wind intensity coefficient. For other cases, the dynamic reference time threshold... It equals the latest timestamp in the historical reference dataset.

[0031] Time decay weighting factor The setting is used to optimize the numerical value, facilitating data processing by the system. Therefore, the time decay weight factor... The numerical value can be determined based on the difference between the oldest timestamp and the latest timestamp in the historical reference dataset. Specifically, the unit of timestamp is days, assuming the oldest timestamp differs from the latest timestamp by one month, and a decay weighting factor is used. The value can be 1 / 15, making The minimum value is 0.135, which facilitates data processing by the system. It can be set to 1.

[0032] The cloud merges the bearing status dataset into the corresponding historical dataset of the pitch bearing. Based on the predicted status dataset, the historical dataset, and the wind intensity coefficient, a wind farm planning overall health function is constructed. A globally optimal maintenance path is generated through an iterative algorithm to ensure that the wind farm planning overall health function obtains the optimal solution.

[0033] Traditional vibration RMS analysis can only detect current bearing damage, while this solution predicts the wind intensity coefficient using a wind speed prediction device and quantifies the nonlinear effects of wind speed, enabling the system to predict the vibration life risk of the pitch bearing under specific wind speeds in real time. An iterative algorithm generates a globally optimal maintenance path to ensure the optimal solution for the overall health function of the wind farm planning, thereby maximizing the economic extension of bearing life.

[0034] A further improvement of the present invention is that the historical dataset includes maintenance time data, multiple historical time periods, and historical average wind intensity, historical average amplitude data, and historical average frequency data, each corresponding to a historical time period. Maintenance time data includes the last maintenance timestamp, the last scheduled maintenance timestamp, the last maintenance delay duration, and the average maintenance duration. The last maintenance delay duration is the time between the last scheduled maintenance timestamp and the last maintenance timestamp. The specific steps for constructing the overall health function for wind farm planning include: S11. Select the historical time period between the last maintenance time stamp and the last maintenance time stamp, and integrate them into a delayed time period. Calculate the actual risk level and actual dynamic penalty amplification coefficient of the pitch bearing in the delayed time period based on the historical dataset. Repeat this step to obtain the actual risk level and actual dynamic penalty amplification coefficient of each pitch bearing. Due to factors such as manpower shortages, economic efficiency, and the characteristics of pitch bearings, there is often a long period of time after all pitch bearings have been fully maintained without the need for further maintenance. If some of these bearings are maintained again during this period, it can lead to over-maintenance. For example, new grease can react with old grease, forming acidic substances that corrode the bearing raceways. Therefore, wind turbines in the same area are typically maintained using a periodic, uniform maintenance approach. However, as the lifespan of the pitch bearings decreases, the frequency of maintenance increases. For instance, routine maintenance is usually every six months, but after several years of use, it may only be necessary every two to three months. Therefore, if maintenance is performed directly according to the installation sequence of the wind turbines, the condition of each pitch bearing cannot be fully considered. For example, the earliest installed wind turbines are not started for a much longer period of time because the wind speed is not high enough to reach the cut-off wind speed. This results in the pitch bearings having a shorter lifespan than other wind turbines. Meanwhile, the last installed wind turbines are almost always operating at full wind speed. The high temperature generated by the rotation of their pitch bearings causes the lubricating oil viscosity to decrease, reducing the protection of the bearings and accelerating the oxidation and evaporation of the lubricating oil. If maintenance is performed according to the installation sequence of the wind turbines, the last installed wind turbine, which should be prioritized for maintenance, will be the last to be maintained.

[0035] Therefore, during the maintenance process, some pitch bearings may be delayed in their maintenance due to the maintenance of other pitch bearings, resulting in additional losses for the pitch bearings. When constructing the overall health function for wind farm planning, this additional loss needs to be fully considered.

[0036] A further improvement to the present invention is that the expression for the actual risk level in step S11 is as follows: ， ， in, It refers to the actual level of risk. It is the number of historical time periods within the delayed time period. It is the first The historical average wind intensity for a given historical period It is the first Historical amplitude and frequency influence coefficients for a given historical time period It is the first Historical average vibration frequency data for a historical period, It is a safe vibration frequency value. It is the first Historical average amplitude data for a given historical period This is the safe amplitude value. It is the amplitude and frequency weighting factor; If the actual risk level is less than the preset healthy operating value, the actual dynamic penalty amplification factor is 0. If the actual risk level is greater than the preset healthy operating value, the expression for the actual dynamic penalty amplification factor is as follows: , in, It is a penalty amplification weighting factor. These are healthy operating values; Specifically, the preset healthy operating value is equal to the full-load wind speed. This is because, although the pitch system fine-tunes the blade angle according to control system commands to maintain stable output, when the actual risk level is high, the mechanical and aerodynamic loads on the unit increase significantly. Furthermore, frequent blade angle adjustments are required, leading to continuous contact and sliding between the rollers and raceways inside the bearing, resulting in increased wear and fretting corrosion, and consequently, a more severe loss of the overall lifespan of the pitch bearing. Specifically, the penalty amplification weighting factor... The value is 1.5.

[0037] S12. Calculate the equipment life coefficient for each pitch bearing based on the actual risk level, the actual dynamic penalty amplification factor, and the duration of the previous maintenance delay. If this is the first maintenance, the equipment life coefficient is equal to the preset initial health degradation value. ; Since amplitude and frequency vibrations can cause significant damage to pitch bearings, calculating the historical amplitude and frequency influence coefficient based on historical average frequency and amplitude data can effectively provide a safety margin for the actual risk level. Specifically, The value is 1. If the pitch bearing does not have a previous maintenance delay period, the actual risk level is considered to be 0.

[0038] The expression for calculating the equipment life factor for each pitch bearing is as follows: , in, It is the equipment life coefficient of the i-th pitch bearing. This is the equipment life coefficient of the i-th pitch bearing at the time of the last maintenance. If this is the first maintenance, then this item is equal to the preset initial health degradation value. , It is the delay time since the last maintenance of the i-th pitch bearing. It represents the actual risk level of the i-th pitch bearing. is the actual dynamic penalty amplification factor for the i-th pitch bearing, and M is the number of pitch bearings.

[0039] Last maintenance delay The unit is days. The greater the actual risk level, the larger the actual dynamic penalty amplification factor, which can increase the degree of change in the life coefficient of the pitch bearing equipment, thereby indirectly increasing the health degradation value and making it a priority for maintenance (the optimal solution of the wind farm planning total health function means maximizing the health degradation value of the entire wind farm's wind turbines). (Minimum total). Using a normalized formula, the changes in the equipment life coefficient of all pitch bearings can be comprehensively balanced, preventing a sharp decrease in the equipment life coefficient due to increased actual risk. Specifically, the initial health degradation value... The initial health degradation value is determined based on the number of maintenance cycles throughout the wind turbine's lifespan. For example, in the "Wind Turbine Blade - Full-Scale Structure Test" and actual verification, the number of maintenance cycles throughout the entire lifespan is nearly 150 times. It can be set to 200.

[0040] S13. Calculate the predicted risk level and predicted dynamic penalty amplification coefficient of the pitch bearing in the current iteration based on the predicted state dataset, and obtain the overall health function of the wind farm planning in the current iteration by combining the previous maintenance timestamp, equipment life coefficient, and average maintenance duration.

[0041] A further improvement to the present invention is that the expression for predicting the degree of risk in step 13 is as follows: ， ， in, It is about predicting the degree of risk. It is the number of time periods in the current iteration prediction. It is the first Wind intensity coefficient for a given forecast period It is the first The influence coefficient of predicted amplitude and frequency for each prediction time period. It is the first The predicted average vibration frequency data for each prediction time period, It is a safe vibration frequency value. It is the first The predicted average amplitude data for each prediction period. This is the safe amplitude value. It is the amplitude and frequency weighting factor; Specifically, if the plan is to maintain the i-th pitch bearing today, an iterative cost algorithm is used simultaneously to select the next pitch bearing to be maintained. Before calculating the iterative cost algorithm, the wind speed prediction device updates the wind intensity coefficient for the day. Among the unmaintained pitch bearings, the longest dynamic average maintenance time is selected. For example, if the longest pitch bearing requires maintenance every day, then the number of iterative prediction time periods is the prediction time period within one day. In one embodiment of the present invention, the prediction time period is one hour, therefore the number of iterative prediction time periods is... indivual.

[0042] If the predicted risk level is less than the preset healthy operating value, the predicted dynamic penalty amplification factor is 0. If the predicted risk level is greater than the preset healthy operating value, the expression for the predicted dynamic penalty amplification factor is as follows: , in, It is a penalty amplification weighting factor. This is a healthy operating value.

[0043] A further improvement to this invention is that the expression for the overall health function of wind farm planning is as follows: , , in, It is the first The health degradation value of a pitch bearing It is a weighting factor for equipment aging. It is the current timestamp. This is the timestamp from the last maintenance. It is the equipment lifespan factor. It is a weighting factor for delayed maintenance. It is the duration between the current timestamp and the timestamp of this maintenance due date. It is the predictive dynamic penalty amplification factor. It is a predictive risk weighting factor. It predicts the degree of risk.

[0044] Using this health degradation value calculation formula can fully take into account the time interval since the last maintenance of the pitch bearing, the equipment life coefficient, and the length of time since the current maintenance was delayed. Based on the predicted dynamic penalty amplification factor and the predicted risk level, pitch bearings with longer maintenance intervals, lower equipment life coefficients, and higher predicted risk levels due to maintenance delays are prioritized for maintenance. Specifically, The values ​​can be referenced from the cut-out wind speed and the cut-in wind speed, such as , It is 0.5, due to the initial health decay value. Set to 200, therefore The default value is 66. This section uses months as the unit. When the equipment lifespan factor decreases, The value can also be increased accordingly. For example, when the pitch bearing is maintained less frequently (e.g., once every six months on average) and less frequently (e.g., once every three months on average), the value can be increased accordingly. The value can be adjusted to 132, and can be set manually or automatically according to the equipment lifespan coefficient. For example, when the equipment lifespan coefficient is 199... The value is 99. Of course, the most accurate way to determine the maintenance frequency of the pitch bearing is by manual judgment during the last maintenance.

[0045] At this point, it is important to note that obtaining the optimal solution for the overall health function of the wind farm planning means achieving the optimal health degradation value for all wind turbines in the entire wind farm. The sum is the lowest.

[0046] A further improvement to this invention, the specific steps for generating the globally optimal maintenance path through an iterative algorithm, include: S31. Establish a virtual aging center based on the wind intensity coefficient and location data of each pitch bearing. The expression for the virtual aging center is as follows: , Where M is the number of pitch bearings. It is the health degradation value of the i-th pitch bearing in the current iteration. It is the position of the i-th pitch bearing; By establishing this virtual aging center, health decline values... At higher levels, its The greater the weight of a position, the more likely it is to decay towards a healthy value. At the highest point, the virtual aging center is updated once after each maintenance of a pitch bearing, thus obtaining a new virtual aging center.

[0047] S32. Perform a global search iteratively based on the virtual aging center up to any pitch bearing. The expression for the global search is as follows: , , in, It is the path position in iteration t+1. It is the current position in the t-th iteration. To the c-th pitch bearing position The cost value is calculated using the cost function. It represents the path position of the pitch bearing with the lowest generation value in the t-th iteration. This is the current location. To the c-th pitch bearing position Path length, This is the average maintenance time for the c-th pitch bearing. It is the health degradation value of the c-th pitch bearing in the current iteration. , , These are cost weighting factors; Global search enables maintenance personnel to approach the health degradation value. Maintenance and repair are carried out in the area with the highest average. Of course, based on the cost function settings, if the health degradation value of the pitch bearing in the virtual aging center... The value is low, while the health degradation value of the surrounding pitch bearings is low. A high value likely indicates a priority to maintain health degradation levels. The pitch bearing has a relatively high rate of change. The random disturbance number r is a random number between 0 and 1, and its disturbance intensity can be reduced after squaring. The logic of this formula is that the cost value of each pitch bearing is calculated with the virtual aging center as the origin, and the pitch bearing with the lowest cost value is selected as the next maintenance target. The lower the cost value, the greater the influence of the position of the c-th pitch bearing, that is, the iteration proceeds towards the position of the pitch bearing with the lowest cost value.

[0048] The formula for calculating the cost function is as follows: the closer the c-th pitch bearing is to the current position, the shorter the average maintenance time of that pitch bearing, and the lower the health degradation value. The higher the value, the lower its cost, thus requiring priority maintenance. Specifically, Using kilometers as the unit, Use 2, Using days as the unit, Using 0.5, Using 6, thereby increasing the health decline value. The influence weight is determined by the fact that if there is no pitch bearing with a cost less than 1, and at least two pitch bearings have a cost equal to 1, then any pitch bearing is selected.

[0049] The travel time to the pitch bearing is calculated based on its path length, and the daily working time is calculated by combining this with the average maintenance time of the pitch bearing. If the daily working time is less than a preset working time threshold, a local search is iterated to any pitch bearing, and the sum of the travel time, average maintenance time, and daily working time for that pitch bearing is calculated, covering the daily working time. The expression for the local search is as follows: , If the daily working hours exceed the preset working hours threshold, record whether the work for the day is completed at the end of the day. If completed, perform a global search iteration to any pitch bearing the next day. If not completed, convert the remaining working hours into the working hours for the next day and perform a local search iteration to any pitch bearing.

[0050] Unlike global search, the core of local search is to calculate the cost of each pitch bearing based on the current position, select the pitch bearing with the lowest cost to calculate the next maintenance position, and the lower the cost, the greater the influence of the position of the c-th pitch bearing.

[0051] The advantage of generating a globally optimal maintenance path through an iterative algorithm lies in its interruptibility. It allows for the recalculation of the next pitch bearing requiring maintenance after a sudden pause in maintenance. Furthermore, it comprehensively considers the distance to each pitch bearing, the time required for maintenance of each bearing, and the health degradation value of each bearing. This allows the overall health function of the wind farm planning to obtain the optimal solution, meaning the maintenance planning maximizes the economic benefits of maintenance. Furthermore, in the technical solution of this invention, maintenance path planning can be comprehensively performed simultaneously within the algorithm for both the six-month maintenance interval for newly installed pitch bearings and the three-month maintenance interval for pitch bearings with a certain age. As mentioned above, each pitch bearing... The values ​​can be manually set based on the manual judgment during the last maintenance to adjust the health degradation value of each pitch bearing. Weighting of maintenance intervals. Health degradation value after a pitch bearing undergoes maintenance. The first term is approximately equal to zero, and the second term is also equal to zero because there is no lag. Its health decay value... The value is much lower than that of other pitch bearings, so other pitch bearings will be prioritized for maintenance. This allows the iterative algorithm to run continuously for 365 days a year, resulting in the optimal maintenance strategy for the entire wind farm. As a further improvement of this invention, to avoid damaging the lifespan of the same pitch bearing by performing multiple maintenance procedures in a short period, maintenance is not performed when the second term equals 0.

[0052] Furthermore, the iterative algorithm can consider the maintenance work status of each shift. If the maintenance work is not completed on a given day, it will continue to maintain the pitch bearing the next day, starting a local search based on the current position. If the work is completed on a given day, the algorithm can start from the location of the virtual aging center the next day. It's important to note that when performing a global search, the current position of the cost function... This refers to their starting point.

[0053] This invention also proposes an online monitoring system for the condition of a pitch bearing, which utilizes the method described above, including: The wind speed prediction device calculates the wind intensity coefficient of each pitch bearing in each prediction time period and marks the corresponding prediction time period, and synchronizes the wind intensity coefficient to the cloud and each state prediction device. The status monitoring device is used to receive the wind intensity coefficient marked with the corresponding prediction time period from the wind speed prediction device, and is used to monitor the pitch bearing, obtain the bearing status dataset, store it in the database, and upload it to the cloud. The database stores historical operating data and transmits the historical operating data to the status prediction device. The state prediction device makes predictions for each prediction period based on historical operating data and wind intensity coefficient, obtains the predicted state dataset, and uploads it to the cloud. In the cloud, the bearing status dataset is merged into the historical dataset corresponding to the pitch bearing. The predicted status dataset, historical dataset, pitch bearing location data and wind intensity coefficient are used to construct the overall health function of the wind farm planning. The globally optimal maintenance path is generated through an iterative algorithm so that the overall health function of the wind farm planning obtains the optimal solution.

[0054] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art can make other variations or modifications based on the above description. It is neither necessary nor possible to exhaustively describe all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A method for online monitoring of the condition of a pitch bearing, characterized in that, It includes multiple pitch bearings, each of which is equipped with a condition monitoring device, a condition prediction device, and a database; It also includes the following steps: The wind speed prediction device calculates the wind intensity coefficient of each pitch bearing in each prediction time period and marks the corresponding prediction time period. The wind intensity coefficient is then synchronized to the cloud and each state prediction device. The pitch bearing is monitored by a condition monitoring device, and the bearing condition dataset is stored in a database and uploaded to the cloud. The database stores historical operating data. The condition prediction device makes predictions for each prediction period based on the historical operating data and the wind intensity coefficient, and the predicted condition dataset is uploaded to the cloud. The cloud merges the bearing status dataset into the corresponding historical dataset of the pitch bearing. Based on the predicted status dataset, the historical dataset, and the wind intensity coefficient, a wind farm planning overall health function is constructed. A globally optimal maintenance path is generated through an iterative algorithm to ensure that the wind farm planning overall health function obtains the optimal solution.

2. The method for online monitoring of pitch bearing condition according to claim 1, characterized in that, The historical dataset includes maintenance time data, multiple historical time periods, and historical average wind intensity, historical average amplitude data, and historical average frequency data, each corresponding to a specific historical time period. Maintenance time data includes the last maintenance timestamp, the last scheduled maintenance timestamp, the last maintenance delay duration, and the average maintenance duration. The last maintenance delay duration is the time between the last scheduled maintenance timestamp and the last maintenance timestamp. The specific steps for constructing the overall health function for wind farm planning include: S11. Select the historical time period between the last maintenance time stamp and the last maintenance time stamp, and integrate them into a delayed time period. Calculate the actual risk level and actual dynamic penalty amplification coefficient of the pitch bearing in the delayed time period based on the historical dataset. Repeat this step to obtain the actual risk level and actual dynamic penalty amplification coefficient of each pitch bearing. S12. Calculate the equipment life coefficient for each pitch bearing based on the actual risk level, the actual dynamic penalty amplification factor, and the duration of the previous maintenance delay. If this is the first maintenance, the equipment life coefficient is equal to the preset initial health degradation value. ; S13. Calculate the predicted risk level and predicted dynamic penalty amplification coefficient of the pitch bearing in the current iteration based on the predicted state dataset, and obtain the overall health function of the wind farm planning in the current iteration by combining the previous maintenance timestamp, equipment life coefficient, and average maintenance duration.

3. The method for online monitoring of pitch bearing condition according to claim 2, characterized in that, The expression for the overall health function of wind farm planning is as follows: ， ， in, It is the first The health degradation value of a pitch bearing It is a weighting factor for equipment aging. It is the current timestamp. This is the timestamp from the last maintenance. It is the equipment lifespan factor. It is a weighting factor for delayed maintenance. It is the duration between the current timestamp and the timestamp of this maintenance due date. It is the predictive dynamic penalty amplification factor. It is a predictive risk weighting factor. It is a prediction of the degree of risk.

4. The method for online monitoring of pitch bearing condition according to claim 1, characterized in that, The specific steps for calculating the wind intensity coefficient of each pitch bearing in each prediction time period and marking the corresponding prediction time period using a wind speed prediction device include: The wind speed prediction device predicts the future average wind speed and future average wind direction for each prediction period. The wind speed prediction device is equipped with a layout diagram with multiple pitch bearing position data and layout direction data. The pitch bearing located at the tail end of the future average wind direction is selected as the wind speed intensity benchmark. The wind intensity coefficient of each pitch bearing in each forecast time period is calculated and the corresponding forecast time period is marked. The expression of the wind intensity coefficient is as follows: ， in, It is the wind intensity coefficient of the i-th pitch bearing in the q-th prediction time period. It is the future average wind speed for the q-th forecast period. It is the dynamic wind force attenuation coefficient of the i-th pitch bearing relative to the wind speed intensity benchmark. It is the angle between the future average wind direction in the q-th prediction time period and the arrangement direction of the i-th pitch bearing.

5. The method for online monitoring of pitch bearing condition according to claim 1, characterized in that, Historical operational data includes multiple historical time periods, as well as historical average wind intensity, historical average amplitude data, and historical average frequency data, each corresponding to a specific historical time period. The predicted state data includes the prediction period, wind intensity coefficient, predicted average amplitude data, and predicted average frequency data; The specific steps by which the state prediction device makes predictions for each prediction period based on historical operating data and wind intensity coefficient to obtain the predicted state dataset include: S21. If the wind intensity coefficient is greater than the cut-out wind speed threshold, then the predicted state data marked as shutdown is obtained. S22. If the wind intensity coefficient is less than the cut-in wind speed threshold, then the predicted state data marked as non-start is obtained. S23. If the wind intensity coefficient is between the cut-out wind speed threshold and the cut-in wind speed threshold, then calculate the pre-selected reference range based on the wind intensity coefficient and the preset similar intensity range, select several historical average wind intensities that fall within the pre-selected reference range, extract the wind intensity coefficient, the historical average wind intensity, and the historical average amplitude data and historical average frequency data corresponding to the historical average wind intensity, and integrate them into the historical reference dataset. Calculate the predicted state data based on the entire historical reference dataset. S24. Repeat steps S21 to S23 for each prediction time period to obtain multiple prediction state data that correspond one-to-one with the prediction time period, and integrate them into the prediction state dataset.

6. The method for online monitoring of pitch bearing condition according to claim 5, characterized in that, After selecting several historical average wind intensities falling within the pre-selected reference range, the following steps are also included: If the number of historical average wind intensities falling within the pre-selected reference range exceeds the preset historical reference quantity value, then the difference between the current timestamp and the timestamp of each historical average wind intensity is calculated. If it is greater than the preset spatiotemporal reference threshold, then the historical average wind intensity is not selected. After integration into the historical reference dataset, the following steps are also included: extracting the timestamp of the historical average wind intensity and merging it into the corresponding historical reference dataset; The expression for calculating the predicted state data based on the entire historical reference dataset is as follows: ， in, It is the first The predicted value for each prediction dataset, where N is the number of historical reference datasets. It is the time decay weight coefficient of the nth historical reference dataset. It is a time decay weighting factor. It is the timestamp of the nth historical reference dataset. It is the current timestamp. It is a dynamic reference time threshold. It is the dynamic decay weight of the nth historical reference dataset. It is the first Distance decay factor for each predicted data point It is the difference between the historical average wind intensity and the wind intensity coefficient in the nth historical reference dataset. It is the nth historical reference dataset and the nth The predicted data is the same as the historical values ​​of the same type.

7. The method for online monitoring of pitch bearing condition according to claim 2, characterized in that, The expression for the actual risk level in step S11 is as follows: ， ， in, It refers to the actual level of risk. It is the number of historical time periods within the delayed time period. It is the first The historical average wind intensity for a given historical period It is the first Historical amplitude and frequency influence coefficients for a given historical time period It is the first Historical average vibration frequency data for a historical period, It is a safe vibration frequency value. It is the first Historical average amplitude data for a given historical period This is the safe amplitude value. It is the amplitude and frequency weighting factor; If the actual risk level is less than the preset healthy operating value, the actual dynamic penalty amplification factor is 0. If the actual risk level is greater than the preset healthy operating value, the expression for the actual dynamic penalty amplification factor is as follows: ， in, It is a penalty amplification weighting factor. These are healthy operating values; The expression for calculating the equipment life factor for each pitch bearing is as follows: ， in, It is the equipment life coefficient of the i-th pitch bearing. This is the equipment life coefficient of the i-th pitch bearing at the time of the last maintenance. If this is the first maintenance, then this item is equal to the preset initial health degradation value. , It is the delay time since the last maintenance of the i-th pitch bearing. It represents the actual risk level of the i-th pitch bearing. is the actual dynamic penalty amplification factor for the i-th pitch bearing, and M is the number of pitch bearings.

8. The method for online monitoring of pitch bearing condition according to claim 2, characterized in that, The expression for predicting the level of risk in step 13 is as follows: ， ， in, It is about predicting the degree of risk. It is the number of time periods in the current iteration prediction. It is the first Wind intensity coefficient for a given forecast period It is the first The influence coefficient of predicted amplitude and frequency for each prediction time period. It is the first The predicted average vibration frequency data for each prediction time period, It is a safe vibration frequency value. It is the first The predicted average amplitude data for each prediction period. This is the safe amplitude value. It is the amplitude and frequency weighting factor; If the predicted risk level is less than the preset healthy operating value, the predicted dynamic penalty amplification factor is 0. If the predicted risk level is greater than the preset healthy operating value, the expression for the predicted dynamic penalty amplification factor is as follows: ， in, It is a penalty amplification weighting factor. This is a healthy operating value.

9. The method for online monitoring of the condition of a pitch bearing according to claim 3, characterized in that, The specific steps for generating the globally optimal maintenance path using an iterative algorithm include: S31. Establish a virtual aging center based on the wind intensity coefficient and location data of each pitch bearing. The expression for the virtual aging center is as follows: ， Where M is the number of pitch bearings. It is the health degradation value of the i-th pitch bearing in the current iteration. It is the position of the i-th pitch bearing; S32. Perform a global search iteratively based on the virtual aging center up to any pitch bearing. The expression for the global search is as follows: ， ， in, It is the path position in iteration t+1. It is the current position in the t-th iteration. To the c-th pitch bearing position The cost value is calculated using the cost function. It represents the path position of the pitch bearing with the lowest generation value in the t-th iteration. This is the current location. To the c-th pitch bearing position Path length, This is the average maintenance time for the c-th pitch bearing. It is the health degradation value of the c-th pitch bearing in the current iteration. , , These are cost weighting factors; The travel time to the pitch bearing is calculated based on its path length, and the daily working time is calculated by combining this with the average maintenance time of the pitch bearing. If the daily working time is less than a preset working time threshold, a local search is iterated to any pitch bearing, and the sum of the travel time, average maintenance time, and daily working time for that pitch bearing is calculated, covering the daily working time. The expression for the local search is as follows: ， If the daily working hours exceed the preset working hours threshold, record whether the work for the day is completed at the end of the day. If completed, perform a global search iteration to any pitch bearing the next day. If not completed, convert the remaining working hours into the working hours for the next day and perform a local search iteration to any pitch bearing.

10. An online monitoring system for the condition of a pitch bearing, characterized in that, The method described using any one of claims 1-9 includes: The wind speed prediction device calculates the wind intensity coefficient of each pitch bearing in each prediction time period and marks the corresponding prediction time period, and synchronizes the wind intensity coefficient to the cloud and each state prediction device. The status monitoring device is used to receive the wind intensity coefficient marked with the corresponding prediction time period from the wind speed prediction device, and is used to monitor the pitch bearing, obtain the bearing status dataset, store it in the database, and upload it to the cloud. The database stores historical operating data and transmits the historical operating data to the status prediction device. The state prediction device makes predictions for each prediction period based on historical operating data and wind intensity coefficient, obtains the predicted state dataset, and uploads it to the cloud. In the cloud, the bearing status dataset is merged into the historical dataset corresponding to the pitch bearing. The predicted status dataset, historical dataset, pitch bearing location data and wind intensity coefficient are used to construct the overall health function of the wind farm planning. The globally optimal maintenance path is generated through an iterative algorithm so that the overall health function of the wind farm planning obtains the optimal solution.

Images