Variable pitch coordinated optimization control method and system for wind turbine generator

Abstract

The invention discloses a variable-pitch coordinated optimization control method and system for a wind turbine generator, and relates to the technical field of wind turbine generator control, and the variable-pitch coordinated optimization control method comprises the following steps: obtaining operation characteristic data of the wind turbine generator; analyzing the obtained operation characteristic data of the wind turbine generator by using a factor analysis algorithm, and identifying key factors influencing the variable pitch performance of the wind turbine generator; predicting the variable pitch of the wind turbine generator at the future moment by using the variable pitch coordination control model to obtain a blade angle change trend; and dynamically adjusting the variable pitch angle of each blade based on the obtained blade angle change trend in combination with the real-time operation state of the wind turbine generator, and optimizing a variable pitch coordination control strategy. The variable-pitch coordinated control model is utilized to predict the variation trend of the blade angle, a scientific basis can be provided for variable-pitch angle adjustment at the future moment, the hysteresis of variable-pitch control is effectively reduced, and the control is more prospective and real-time adaptive.

Description

Technical Field

[0001] The present invention relates to the technical field of wind turbine generator set control, and in particular to a method and system for coordinated optimization control of variable pitch of a wind turbine generator set. Background Art

[0002] As two key control systems in large wind turbines, yaw control and pitch control play a vital role in improving wind power generation efficiency and ensuring the safety of unit operation. The yaw control system is mainly responsible for driving the nacelle to rotate so that the wind rotor always faces the wind direction to keep the plane where the wind rotor is located perpendicular to the wind direction, thereby maximizing the utilization rate of wind energy. In the actual working process, the wind direction will continue to change. The yaw control system needs to monitor the wind direction in real time through sensors and adjust the nacelle angle through the drive device to achieve dynamic tracking of wind energy resources. In addition, yaw control also needs to consider reducing the yaw frequency to extend the life of the equipment. At the same time, the pitch control system directly affects the wind energy utilization coefficient of the wind turbine by adjusting the pitch angle of the blades. At low wind speeds, pitch control optimizes the blade angle to capture wind energy to the maximum extent; at high wind speeds or exceeding the rated wind speed, the output power is limited by adjusting the pitch angle to ensure stable operation of the unit and prevent overload.

[0003] However, the existing technology is not convenient for extracting and identifying key factors affecting pitch performance from a large amount of operating characteristic data, cannot avoid redundant processing of non-critical variables, reduces the understanding of complex operating environments, is not convenient for predicting the changing trend of blade angles, cannot provide a scientific basis for future pitch angle adjustments, and is not convenient for precise pitch adjustment and coordinated control, thereby increasing the load fluctuations of wind turbine blades and transmission components, increasing mechanical stress and fatigue damage, and thereby shortening the service life of the equipment.

[0004] Currently, no effective solution has been proposed for the problems in the related technologies. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention proposes a method and system for coordinated optimization control of pitch changes of a wind turbine, which solves the problems raised in the above background technology, such as it is not convenient to extract and identify the key factors affecting the pitch change performance from a large amount of operating characteristic data, it cannot avoid redundant processing of non-critical variables, it reduces the understanding of complex operating environments, it is not convenient to predict the changing trend of the blade angle, it cannot provide a scientific basis for the adjustment of the pitch angle in the future, and it is not convenient for precise pitch adjustment and coordinated control, thereby increasing the load fluctuation of the wind turbine blades and transmission components, increasing mechanical stress and fatigue damage, and thus shortening the service life of the equipment.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0007] According to one aspect of the present invention, a pitch-changing coordinated optimization control method for a wind turbine is provided, the pitch-changing coordinated optimization control method comprising the following steps:

[0008] S1. Acquire real-time operation data of the wind turbine generator set, and pre-process the real-time operation data to obtain operation characteristic data of the wind turbine generator set;

[0009] S2. Based on the factor analysis algorithm, the obtained operation characteristic data of the wind turbine is analyzed by using the factor analysis algorithm to identify the key factors affecting the variable pitch performance of the wind turbine;

[0010] S3. Based on the identified key factors affecting the pitch performance of wind turbines, a pitch coordination control model is established, and the pitch coordination control model is used to predict the wind turbine pitch at future moments to obtain the blade angle change trend;

[0011] S4. Based on the obtained blade angle change trend and combined with the real-time operating status of the wind turbine, the pitch angle of each blade is dynamically adjusted, and the pitch coordination control strategy is optimized.

[0012] Further, based on the factor analysis algorithm, the obtained operation characteristic data of the wind turbine is analyzed by using the factor analysis algorithm to identify the key factors affecting the variable pitch performance of the wind turbine, including the following steps:

[0013] S21, obtaining a sample set of operation characteristic data of the wind turbine generator set, and classifying and arranging the sample set of operation characteristic data of the wind turbine generator set by using a factor analysis algorithm, and screening out a subset of the sample set of operation characteristic data that is consistent with the variable pitch performance characteristics of the wind turbine generator set;

[0014] S22. Based on the high-quality screening mechanism, further screening the screened subset of operation characteristic data samples, and selecting the operation characteristic data samples with the best pitch performance as high-quality individuals;

[0015] S23, using feature optimization algorithm to optimize and analyze high-quality individuals, and identify the operating characteristics that have the greatest impact on the variable pitch performance of the wind turbine;

[0016] S24. Based on the classification results and the optimization results of high-quality individuals, identify the key factors affecting the variable pitch performance of wind turbines.

[0017] Furthermore, based on the high-quality screening mechanism, the filtered subset of operation characteristic data samples is further screened, and the operation characteristic data samples with the best pitch performance are selected as high-quality individuals, including the following steps:

[0018] S221, collecting wind turbine operation characteristic data, sequentially collecting characteristic data at multiple operation moments, and using an association rule mining algorithm to analyze the association relationship between the wind turbine operation characteristic data, to obtain a subset of wind turbine operation characteristic data classified by characteristics;

[0019] S222, randomly selecting a number of operation feature data samples from the wind turbine operation feature data subset as a training set for a decision tree classifier, and calculating the entropy gain of the pitch performance-related features to obtain a preliminary decision tree classifier, and screening out key operation features related to the pitch performance;

[0020] S223, using an association rule mining algorithm to remove incomplete data points in the operation feature data set, performing a consistency check on the operation feature data with the highest degree of association, and screening out a subset of the operation feature data with the highest integrity;

[0021] S224, partitioning the operation feature data subset with the highest integrity into a number of operation feature data subsets of equal size, and using the several operation feature data subsets as initial training sets;

[0022] S225, training a number of operation feature data subsets, calculating the performance transition probability of key operation features under different operation states, and obtaining feature fingerprints related to the states;

[0023] S226, using several subsets of operation feature data as test sets for each other, and using a decision tree classifier to check the consistency of feature fingerprints. If the classification accuracy is greater than a preset threshold, retain the feature subset with the highest classification accuracy; otherwise, merge several subsets of operation feature data and retain the merged operation feature data;

[0024] S227, combining the decision tree classifier with the retained merged operation characteristic data, calculating the pitch performance classification probability, and if there is a disagreement between the classification results, incorporating the operation characteristic data into the performance state transfer matrix;

[0025] S228. When the classification results of the decision tree classifier are consistent with those of the state transfer matrix, new operating characteristic data are introduced and step S227 is repeated, the weight of the decision tree classifier is gradually reduced, and the operating characteristic data samples with the best variable pitch performance are screened out as high-quality individuals based on the results of the performance transfer probability.

[0026] Furthermore, the incomplete data points in the operation feature data set are removed by using the association rule mining algorithm, the consistency check is performed on the operation feature data with the highest degree of association, and the operation feature data subset with the highest integrity is screened out, which includes the following steps:

[0027] S2231, selecting a section of wind turbine operation time series data, and presetting a sliding window at the starting point to traverse the characteristic data points in the time series in sequence;

[0028] S2232: If the integrity index of the feature data point in the sliding window is greater than a preset threshold, the feature data point is determined to be incomplete data and added to the abnormal data set;

[0029] S2233, calculating the average distance between each feature data point in the sliding window and the center of the feature data space, and setting an association threshold; if the average distance is less than the association threshold, it is determined that the window contains abnormal feature data, and it is added to the abnormal data set;

[0030] S2234, the sliding window continues to move backward by one time step, replacing the initial feature data point in the sliding window, and repeating steps S2231 to S2233 until all feature data points in the time series are traversed, and finally an abnormal feature data set is output, and the time mark of the feature data point is recorded;

[0031] S2235. According to the classification rules of wind turbine operating characteristics, the operating characteristic data set is divided to generate multiple groups of operating characteristic subsets, the operating characteristic subset with the smallest data volume is deleted, and the operating characteristic subset with the largest data volume and the highest correlation is retained, and finally the operating characteristic data subset with the highest integrity is screened out.

[0032] Furthermore, the feature optimization algorithm is used to optimize and analyze the high-quality individuals, and the operation characteristics that have the greatest impact on the variable pitch performance of the wind turbine are identified, including the following steps:

[0033] S231, randomly generating an initial feature population from the operating feature data set of the wind turbine generator set, and initializing the parameters of the feature optimization algorithm;

[0034] S232, calculating the fitness of each characteristic individual in the initial characteristic population, and dividing each characteristic individual into a microhabitat according to a microhabitat classification method;

[0035] S233, for each characteristic individual in a microhabitat, generate a new characteristic population according to an optimization formula, and simulate the evolution of the characteristic population in the microhabitat;

[0036] S234, adjusting the distribution of feature populations according to the adaptive diffusion rule, so that high fitness features concentrate on influencing the target area, and low fitness features are gradually eliminated;

[0037] S235, checking whether the key operating feature that has the greatest impact on the pitch performance has been identified, if it has been found, terminating the feature optimization algorithm and outputting the optimal feature; otherwise, continuing the optimization;

[0038] S236, judging whether the maximum number of iterations has been reached, if so, terminating the feature optimization algorithm and outputting the optimal feature; otherwise, continuing the optimization;

[0039] S237, judging whether the current number of characteristic individuals exceeds the maximum number of characteristic populations, if so, continuing the optimization; otherwise, going to step S232;

[0040] S238, according to the competition rules within the niche, select a number of characteristic individuals from each niche as parents to enter the next iteration;

[0041] S239, repeating steps S232 to S238 until the maximum number of iterations is reached, and finally outputting the operating characteristics that have the greatest impact on the variable pitch performance of the wind turbine.

[0042] Furthermore, the optimization formula is:

[0043]

[0044] Where M X Represents the new feature population generated by the Xth feature individual;

[0045] H X Represents the fitness value of the Xth characteristic individual;

[0046] H max Indicates the maximum fitness in the current feature population;

[0047] H min Indicates the minimum fitness in the current feature population;

[0048] M max Indicates the maximum characteristic population size set;

[0049] M min Indicates the set minimum feature population size.

[0050] Furthermore, based on the identified key factors affecting the pitch performance of wind turbines, a pitch coordination control model is established, and the pitch coordination control model is used to predict the wind turbine pitch at future moments, and the blade angle change trend is obtained, which includes the following steps:

[0051] S31. Based on the grey correlation analysis method, the correlation between environmental factors and pitch performance is analyzed, the key factors with the greatest impact on pitch performance are screened out, the factors with the lowest correlation are eliminated, the key factor data set is reconstructed, and divided into a training set and a test set;

[0052] S32. Using variational mode decomposition technology, the time series signal corresponding to the key factor is decomposed into several intrinsic mode functions, each representing a subsequence with different frequency characteristics;

[0053] S33, for each decomposed subsequence, respectively construct a pitch coordination control submodel based on an extreme learning machine;

[0054] S34, using a parameter optimization algorithm to optimize the parameters of the variable pitch coordinated control sub-model of each sub-sequence, and constructing an optimized variable pitch coordinated control model;

[0055] S35, superimposing and reconstructing the prediction results of the variable pitch coordinated control sub-models of all sub-sequences, obtaining the prediction results of the blade angles of the wind turbines at future moments, and forming a blade angle change trend.

[0056] Furthermore, the parameter optimization algorithm is used to optimize the parameters of the variable pitch coordinated control sub-model of each sub-sequence, and the optimized variable pitch coordinated control model is constructed, which includes the following steps:

[0057] S341, randomly generating a number of candidate parameter sets in the search space, each candidate parameter set corresponding to a solution of the propeller coordinated control sub-model;

[0058] S342, inputting each candidate parameter set into the corresponding pitch coordination control sub-model, calculating its fitness value, and evaluating the pros and cons of each candidate parameter;

[0059] S343, selecting a parameter solution with the best fitness value from the current candidate parameter set as the parameter solution with the best performance in the current iteration;

[0060] S344, adjusting the positions of other candidate parameter sets according to the update formula to make them close to the optimal parameter solution;

[0061] S345, calculating the fitness value of each candidate parameter set after the updated position, if the fitness value of any parameter solution is greater than the current optimal solution, then updating it to the new optimal parameter solution;

[0062] S346, for the candidate parameter solution with the lowest fitness value, remove it from the current candidate parameter set, and randomly generate a new parameter solution in the search space;

[0063] S347. Check whether the maximum number of iterations has been reached. If so, stop the optimization process, take the current optimal parameter solution as the final optimization result, and build an optimized variable pitch coordinated control model based on the final optimization result; otherwise, return to step S344 and continue the optimization.

[0064] Furthermore, the update formula is:

[0065] W a (m+1)=W a (m)+rand×(W ZY-W a (m)), a=1, 2, ... N;

[0066] Where W a (m+1) represents the position of the ath candidate parameter set at the m+1 iteration;

[0067] W a (m) represents the position of the ath candidate parameter set at the mth iteration;

[0068] rand represents a random number;

[0069] W ZY Represents the location of the optimal parameter solution in the search space;

[0070] ZY represents the optimal parameter solution;

[0071] N represents the total number of candidate parameter sets.

[0072] Furthermore, the factor analysis algorithm is used to analyze the obtained operation characteristic data of the wind turbine, and then the performance compliance index is obtained to determine whether to perform key factor identification; the performance compliance index is used to evaluate the degree of compliance between the operation characteristic data and the actual operation state of the wind turbine; the performance compliance index is obtained by the following method: E1, obtaining the classification accuracy of the decision tree classifier within a preset time period, and judging whether the classification accuracy is greater than the preset classification accuracy. If so, the feature data subset with the highest classification accuracy in the operation characteristic data is stored and E2 is executed, otherwise the feature data subsets in the operation characteristic data are merged and then stored; E2, obtaining the classification accuracy score, and at the same time, the performance transfer probability and pitch performance classification probability of the wind turbine within the preset time period are counted, and the performance transfer probability score and pitch performance classification probability score are obtained by combining the corresponding reference performance transfer probability and reference pitch performance classification probability; E3, from the preset Suppose that a preset performance compliance weight factor is obtained in a database, and a performance compliance index is obtained by combining the obtained classification accuracy score, performance transfer probability score and pitch performance classification probability score; the classification accuracy represents the ratio of the amount of operation feature data classified in a decision tree classifier to the total amount of operation feature data; the classification accuracy score represents the ratio of the classification accuracy to the preset classification accuracy; the performance transfer probability represents the probability that a wind turbine is transferred from one operating state to another; the performance transfer probability score represents the ratio of the absolute value of the difference between the performance transfer probability and the reference performance transfer probability to the reference performance transfer probability; the pitch performance classification probability score represents the ratio of the absolute value of the difference between the pitch performance classification probability and the reference pitch performance classification probability to the reference pitch performance classification probability; the performance compliance weight factor includes a classification accuracy score weight factor, a performance transfer probability score weight factor and a pitch performance classification probability score weight factor.

[0073] Furthermore, the use of the pitch coordination control model to predict the pitch of the wind turbine at a future moment also includes judging whether the wind turbine pitch prediction is completed based on the obtained pitch prediction accuracy index; the pitch prediction accuracy index is used to evaluate the accuracy and effectiveness of the pitch coordination control model in predicting the pitch of the wind turbine; the specific process of judging whether the wind turbine pitch prediction is completed based on the obtained pitch prediction accuracy index is: judging whether the obtained pitch prediction accuracy index is within the pitch prediction accuracy threshold range, if so, completing the wind turbine pitch prediction, otherwise adjusting the control parameters of the pitch coordination control model and reusing the pitch coordination control model to predict the wind turbine pitch at a future moment, until the obtained pitch prediction accuracy index is within the pitch prediction accuracy threshold range; the control parameters include prediction time step and pitch rate limit.

[0074] Furthermore, the pitch prediction accuracy index is obtained by the following method: when the obtained performance compliance index is within the performance compliance threshold range, the predicted pitch angle and predicted motor torque at the end of the prediction of the pitch coordination control model within a preset time period are obtained, and the actual pitch angle and actual motor torque of the wind turbine at the current moment are monitored; the pitch angle prediction score and the motor torque prediction score are obtained, and the pitch prediction accuracy index is obtained by combining the obtained performance compliance index and the preset environmental wind speed influencing factor in the database; the pitch angle prediction score represents the ratio of the absolute value of the difference between the predicted pitch angle and the actual pitch angle to the maximum allowable deviation of the pitch angle; the motor torque prediction score represents the ratio of the absolute value of the difference between the predicted motor torque and the actual motor torque to the maximum allowable deviation of the motor torque.

[0075] Furthermore, the dynamic adjustment of the pitch angle of each blade also includes obtaining a pitch coordination interference score to determine whether to optimize the pitch coordination control strategy; the pitch coordination interference score is used to evaluate the degree of interference with the pitch angle during dynamic adjustment under the current pitch coordination control strategy; the pitch coordination interference score is obtained by the following method: when the obtained pitch prediction accuracy index is within the pitch prediction accuracy threshold range, the response of the pitch command of each blade of the wind turbine is monitored in real time, and the average electromagnetic strength during the transmission process is determined to determine whether the average electromagnetic strength is less than the maximum allowable electromagnetic strength. If so, the electromagnetic interference coefficient is obtained, otherwise a preset prompt is given. Personnel adjust the electromagnetic shielding parameters of the current communication line, and the electromagnetic interference coefficient represents the ratio of the average electromagnetic intensity to the maximum allowable electromagnetic intensity; obtain the command response delay score and the command transmission delay score, and at the same time, combine the obtained electromagnetic interference coefficient and the pitch prediction accuracy index to obtain the pitch coordination interference score, the command response delay score represents the ratio of the difference between the actual response delay duration of the pitch command in the response process and the reference response delay duration to the reference response delay duration, and the command transmission delay score represents the ratio of the difference between the actual transmission delay duration of the pitch command in the transmission process and the reference transmission delay duration to the reference transmission delay duration.

[0076] According to another aspect of the present invention, a wind turbine generator system is provided with a coordinated optimization control system for changing the pitch of the wind turbine generator system. The coordinated optimization control system for changing the pitch of the wind turbine generator system comprises:

[0077] A data acquisition module is used to acquire the real-time operation data of the wind turbine generator set and pre-process the real-time operation data to obtain the operation characteristic data of the wind turbine generator set;

[0078] A factor analysis module is used to analyze the obtained operation characteristic data of the wind turbine using the factor analysis algorithm to identify key factors affecting the variable pitch performance of the wind turbine;

[0079] The model building module is used to establish a pitch coordination control model based on the key factors that affect the pitch performance of the wind turbine generator set identified, and use the pitch coordination control model to predict the pitch of the wind turbine generator set at future moments to obtain the blade angle change trend;

[0080] The coordination control module is used to dynamically adjust the pitch angle of each blade based on the obtained blade angle change trend and the real-time operating status of the wind turbine, and optimize the pitch coordination control strategy.

[0081] The beneficial effects of the present invention are:

[0082] 1. The present invention extracts and identifies key factors affecting pitch performance from a large amount of operating characteristic data, avoids redundant processing of non-critical variables, improves the understanding of complex operating environments, and uses a pitch coordination control model to predict the changing trend of blade angles, which can provide a scientific basis for pitch angle adjustment at future times, effectively reduce the lag of pitch control, make control more forward-looking and real-time adaptable, and dynamically adjust the blade pitch angle in combination with real-time operating status to ensure that the wind turbine can achieve optimal control under different wind speeds, wind directions and operating conditions. By optimizing the pitch coordination control strategy, the coordination and response speed of the system are improved, thereby achieving more efficient wind energy utilization and enhancing the pitch control capability of the wind turbine.

[0083] 2. The present invention classifies and organizes the operating characteristic data samples of the wind turbine set through a factor analysis algorithm, and can accurately screen out a subset of data samples that are consistent with the pitch performance characteristics. It further combines the high-quality screening mechanism and the feature optimization algorithm to identify the key factors affecting the pitch performance and reduce the interference of irrelevant features. Through the identification and optimization based on key factors, it can accurately capture the main influencing factors of the wind turbine pitch performance, improve the response speed and adjustment accuracy of the pitch system, and help optimize the pitch control strategy so that the unit can maintain efficient operation under different wind conditions.

[0084] 3. The present invention uses the grey correlation analysis method to screen out the key factors that affect the variable pitch performance and eliminate irrelevant factors to ensure the relevance and simplicity of the data. Subsequently, the variational mode decomposition technology is used to decompose the time series into subsequences of multi-frequency features, refine the data processing granularity, effectively separate the interference between complex features, and improve the prediction accuracy of future blade angle change trends. The parameters of each sub-model are iteratively adjusted through the parameter optimization algorithm, the optimal parameter set is selected and the low fitness solution is eliminated. The optimized variable pitch coordination control model is more adaptable and accurate, and the dynamic update and global search methods are used to avoid falling into the local optimum, thereby improving the optimization efficiency and the overall prediction ability of the variable pitch coordination control model.

[0085] 4. The present invention obtains the classification accuracy of the decision tree classifier within a preset time period. When the classification accuracy is greater than the preset classification accuracy, the classification accuracy score is obtained and the performance transfer probability and the pitch performance classification probability of the wind turbine within the preset time period are counted. Based on this, the performance transfer probability score and the pitch performance classification probability score are obtained. At the same time, the preset performance compliance weight factor is obtained in combination with the preset database to obtain the performance compliance index, thereby achieving more accurate acquisition of the performance compliance index, and further achieving a more accurate evaluation of the degree of compliance between the operating characteristic data and the actual operating status of the wind turbine.

[0086] 5. The present invention obtains the predicted pitch angle and predicted motor torque at the end of the prediction of the pitch coordination control model within a preset time period, and monitors the actual pitch angle and actual motor torque of the wind turbine at the current moment, thereby obtaining the pitch angle prediction score and the motor torque prediction score, and combining the preset environmental wind speed influencing factor in the database to obtain the pitch prediction accuracy index, thereby achieving more accurate acquisition of the pitch prediction accuracy index, and further achieving improved prediction accuracy and effectiveness of the pitch coordination control model.

[0087] 6. The present invention monitors the response of the pitch command of each blade of the wind turbine in real time and the average electromagnetic intensity during the transmission process. When the average electromagnetic intensity is less than the maximum allowable electromagnetic intensity, the command response delay score and the command transmission delay score are obtained. At the same time, the obtained electromagnetic interference coefficient and the pitch prediction accuracy index are combined to obtain the pitch coordination interference score, thereby achieving more accurate acquisition of the pitch coordination interference score, and further achieving a more accurate assessment of the degree of interference of the pitch angle during the dynamic adjustment process. BRIEF DESCRIPTION OF THE DRAWINGS

[0088] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0089] Figure 1 is a flow chart of a pitch coordination optimization control method for a wind turbine according to an embodiment of the present invention;

[0090] Figure 2 The present invention is a block diagram of a wind turbine pitch coordination optimization control system according to an embodiment of the present invention.

[0091] In the figure:

[0092] 1. Data acquisition module; 2. Factor analysis module; 3. Model building module; 4. Coordination control module. DETAILED DESCRIPTION

[0093] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0094] In the description of the present invention, unless otherwise specified, the meaning of "plurality" is two or more. In addition, the terms "first", "second", "third", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0095] According to an embodiment of the present invention, a method and system for coordinated optimization control of pitch changes of a wind turbine are provided.

[0096] The present invention is further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, according to the pitch coordination optimization control method of a wind turbine set according to an embodiment of the present invention, the pitch coordination optimization control method comprises the following steps:

[0097] S1. Acquire real-time operation data of the wind turbine generator set, and pre-process the real-time operation data to obtain operation characteristic data of the wind turbine generator set;

[0098] Specifically, real-time operation data includes:

[0099] 1) Environmental related data: wind speed, wind direction, ambient temperature, air density, humidity, etc.

[0100] 2) Unit operating status data: speed, power output, pitch angle, yaw angle, vibration data, temperature monitoring, etc.

[0101] 3) Operation data: control signals, fault signals, status indications, etc.

[0102] Specifically, the operating characteristic data includes:

[0103] 1) Performance-related characteristics: wind energy utilization coefficient, power coefficient, rated power proximity, energy output efficiency, etc.

[0104] 2) Control characteristics: dynamic change characteristics of pitch angle, yaw control frequency, speed change rate, etc.

[0105] 3) Operating status characteristics: load characteristics (load characteristics of the main shaft, blades, and tower), vibration characteristics (characteristic values ​​of high-frequency and low-frequency vibrations), temperature characteristics (temperature rise trend and stability of key components), etc.

[0106] It needs to be explained that the duplicate data, missing values ​​and outliers of the real-time operation data of the wind turbine are obtained, and the duplicate data, missing values ​​and outliers are denoised, filtered and smoothed; the unprocessed data in the real-time operation data are connected to generate a new data table, and different data tables are associated through external key values ​​to generate a complete data table to obtain an accurate operation data set; the principal component analysis method is used to extract physical signs from the obtained accurate operation data set to obtain the operation characteristic data of the wind turbine.

[0107] S2. Based on the factor analysis algorithm, the obtained operation characteristic data of the wind turbine is analyzed by using the factor analysis algorithm to identify the key factors affecting the variable pitch performance of the wind turbine;

[0108] Specifically, the key factors affecting the pitch performance of wind turbines include:

[0109] 1) Environmental factors: wind speed, wind direction deviation, air density, ambient temperature, etc.

[0110] 2) Equipment status factors: initial position of pitch angle, blade load, main shaft speed, power output, temperature of key components, vibration level, etc.

[0111] 3) Control parameter factors: pitch speed, control lag time, pitch frequency, wind energy utilization coefficient, power coefficient, yaw angle deviation, etc.

[0112] 4) Dynamic characteristic factors: wind speed change rate (rapid changes in wind speed (such as gusts) put forward higher dynamic adjustment requirements for the pitch system), blade fatigue characteristics (blade fatigue stress may limit the dynamic adjustment range of the pitch angle), system inertia (the mechanical and control inertia of the pitch system itself affects the response speed), etc.

[0113] S3. Based on the identified key factors affecting the pitch performance of wind turbines, a pitch coordination control model is established, and the pitch coordination control model is used to predict the pitch of wind turbines at future moments to obtain the blade angle change trend;

[0114] Specifically, the blade angle variation trend refers to the regularity of the wind turbine blade angle variation over time, and is used to describe the response of the pitch control system to the environment and operating conditions at different time points.

[0115] S4. Based on the obtained blade angle change trend and combined with the real-time operating status of the wind turbine, the pitch angle of each blade is dynamically adjusted, and the pitch coordination control strategy is optimized.

[0116] It should be explained that the predicted blade angle change trend is compared with the real-time operation data to identify the current operation deviation; based on the grey correlation analysis method, the matching degree between the wind condition change and the current blade angle is evaluated in real time to determine the adjustment priority; the adjustment factor based on the wind speed change rate, power deviation, and speed change is calculated to guide the optimization of the pitch angle; the angle of each blade is adjusted separately to adapt to the difference in wind conditions; by adjusting the angle, the wind energy utilization of each blade is balanced, and the tower vibration and load imbalance are reduced; at low wind speeds: increase the blade angle to capture more wind energy; near the rated wind speed: optimize the angle to maintain stable power output; at high wind speeds or above the rated wind speed: reduce the blade angle to limit power output and protect the equipment; the optimization of the pitch coordination control strategy includes:

[0117] 1) Establish optimization goals

[0118] Maximize wind energy utilization: ensure that the unit is always in the optimal efficiency range under different wind conditions.

[0119] Minimize mechanical loads: Reduce wind impact on blades, shafts and towers, and extend equipment life.

[0120] Power output stability: Reduce the impact of wind speed fluctuations on power generation.

[0121] 2) Dynamic Optimization of Parameters

[0122] Optimize control model parameters such as blade angular velocity, response time, adjustment range, etc. using real-time data input.

[0123] Apply genetic algorithm or particle swarm optimization algorithm to iteratively update control model parameters and improve response accuracy.

[0124] 3) Control logic adjustment

[0125] Layered control: Adjust the priority of control strategies according to different wind conditions (such as giving priority to protecting equipment when wind speed is too high).

[0126] Dynamic threshold adjustment: adjust the pitch angular velocity and minimum response time according to real-time wind conditions.

[0127] 4) Multi-objective coordination and implementation

[0128] Global coordination: Comprehensively consider the adjustment requirements of each blade and the overall power output target to avoid over-adjustment of a single blade.

[0129] Real-time feedback: Feedback the adjusted blade angle changes to the control system and recalculate the adjustment target for the next moment.

[0130] Fault-tolerant mechanism: When some sensors fail or data is abnormal, switch to redundant control logic to ensure stable operation of the system.

[0131] Preferably, based on a factor analysis algorithm, the obtained operation characteristic data of the wind turbine is analyzed using the factor analysis algorithm to identify key factors affecting the variable pitch performance of the wind turbine, including the following steps:

[0132] S21, obtaining a sample set of operation characteristic data of the wind turbine generator set, and classifying and arranging the sample set of operation characteristic data of the wind turbine generator set by using a factor analysis algorithm, and screening out a subset of the sample set of operation characteristic data that is consistent with the variable pitch performance characteristics of the wind turbine generator set;

[0133] S22. Based on the high-quality screening mechanism, further screening the screened subset of operation characteristic data samples, and selecting the operation characteristic data samples with the best pitch performance as high-quality individuals;

[0134] S23, using feature optimization algorithm to optimize and analyze high-quality individuals, and identify the operating characteristics that have the greatest impact on the variable pitch performance of the wind turbine;

[0135] S24. Based on the classification results and the optimization results of high-quality individuals, identify the key factors affecting the variable pitch performance of wind turbines.

[0136] It should be noted that the factor analysis algorithm is used to analyze the obtained operation characteristic data of the wind turbine, and then the performance compliance index is obtained to determine whether to perform key factor identification; the performance compliance index is used to evaluate the degree of compliance between the operation characteristic data and the actual operation state of the wind turbine; the performance compliance index is obtained by the following method: E1, obtain the classification accuracy of the decision tree classifier within a preset time period, and determine whether the classification accuracy is greater than the preset classification accuracy. If so, the feature data subset with the highest classification accuracy in the operation characteristic data is stored and E2 is executed, otherwise, the feature data subsets in the operation characteristic data are merged and then stored; E2, obtain the classification accuracy score (that is, Z in the restriction expression of the performance compliance index) t ), and at the same time, the performance transfer probability and pitch performance classification probability of the wind turbine in the preset time period are counted, and the performance transfer probability score (i.e., the Y in the restriction expression of the performance compliance index) is obtained by combining the corresponding reference performance transfer probability and reference pitch performance classification probability. t ) and the pitch performance classification probability score (i.e., the performance meets the index limit expression F t ); E3, obtain a preset performance compliance weight factor from a preset database, and combine the obtained classification accuracy score, performance transfer probability score and pitch performance classification probability score to obtain a performance compliance index.

[0137] The specific process for determining whether to perform key factor identification is as follows: determine whether the obtained performance compliance index is within the performance compliance threshold range (including the case where it is equal to the maximum and minimum values ​​of the historical performance compliance index); if so, identify the key factors affecting the wind turbine pitch performance; otherwise, prompt the preset personnel to perform shutdown operations and adjust the wind turbine pitch performance parameters (for example, by increasing the transmission ratio of the wind turbine pitch gearbox to optimize torque transmission, where the transmission ratio is the ratio of the number of wind turbine pitch driven gears to the number of active gears, which is used to measure the smoothness performance and efficiency of the motor torque).

[0138] Among them, the classification accuracy represents the ratio of the amount of operation feature data classified in the decision tree classifier to the total amount of operation feature data; the classification accuracy score represents the ratio of the classification accuracy to the preset classification accuracy (represented by the result of summing and averaging the historical classification accuracy of the historical operation feature data in the database in each historical time period); the performance transition probability represents the probability of the wind turbine set transferring from one operating state to another operating state; the performance transition probability score represents the ratio of the absolute value of the difference between the performance transition probability and the reference performance transition probability (represented by the result of summing and averaging the historical performance transition probabilities of the historical wind turbine sets in the database in each historical time period) to the reference performance transition probability; the pitch performance classification probability score represents the ratio of the absolute value of the difference between the pitch performance classification probability and the reference pitch performance classification probability (represented by the result of summing and averaging the historical pitch performance classification probabilities of the historical wind turbine sets in the database in each historical time period) to the reference pitch performance classification probability; the performance compliance weight factor includes the classification accuracy score weight factor (i.e., the β in the restriction expression of the performance compliance index); 1 ), performance transition probability score weight factor (i.e., the β in the restriction expression of the performance compliance index 2 ) and the pitch performance classification probability score weight factor (i.e., the β in the limiting expression of the performance compliance index) 3 ); pitch performance parameters include pitch angle (i.e. the rotation angle of the wind turbine blades relative to the ambient wind direction) and pitch motor torque.

[0139] The aforementioned database is a database for storing various types of setting data established before the design of the pitch coordination optimization control method of a wind turbine. The database includes but is not limited to preset classification accuracy, performance compliance weight factor, performance compliance threshold range, pitch prediction accuracy threshold range, pitch coordination interference allowable range and preset time period. The various numerical values ​​therein are directly set by technical personnel. The setting basis of the performance compliance threshold range can be determined according to the actual operating scenario of the wind turbine. For example, the performance compliance threshold range represents the range corresponding to the maximum and minimum values ​​of the historical performance compliance index of the wind turbine in the database in each historical time period. In addition, the various numerical values ​​in the database can be set and fine-tuned by technical personnel according to actual debugging.

[0140] The classification accuracy score weight factor, the performance transfer probability score weight factor and the pitch performance classification probability score weight factor are pre-set in the database. These weight factors reflect the degree of influence of the classification accuracy score, the performance transfer probability score and the pitch performance classification probability in the process of obtaining the performance compliance index. In practical applications, the weight factors corresponding to the classification accuracy score, the performance transfer probability score and the pitch performance classification probability can be directly retrieved from the preset database. This correspondence is achieved through a predefined mapping relationship. For example, specifically, this mapping relationship can be: in the process of obtaining the performance compliance index, each classification accuracy score, performance transfer probability score and pitch performance classification probability forms a one-to-one or many-to-one mapping set with a specific weight factor in the preset database. When it is needed, just input the real-time classification accuracy score, performance transfer probability score and pitch performance classification probability into this mapping set to obtain the corresponding weight factor. The sum of the classification accuracy score weight factor, the performance transfer probability score weight factor and the pitch performance classification probability score weight factor in this example is 1.

[0141] Specifically, the specific restriction expression of the performance compliance index is:

[0142]

[0143]

[0144] Where t is the number of the preset time period, t=1,2,...,T, T is the total number of preset time periods, e is a natural constant, FU t It indicates the performance compliance index of the operating characteristic data in the tth preset time period, β 1 Represents the classification accuracy score weight factor, Z t Indicates the classification accuracy score of the running feature data in the tth preset time period, Z1 t Indicates the classification accuracy of the running feature data in the tth preset time period, Z1 0 Represents the preset classification accuracy, β 2 represents the performance transition probability score weight factor, Y t Indicates the performance transition probability score of the operating characteristic data in the tth preset time period, Y1 t represents the performance transition probability of the operating characteristic data in the tth preset time period, Y1 0 represents the reference performance transition probability, β 3 represents the probability score weight factor of pitch performance classification, F t F1 represents the probability score of pitch performance classification of the operating characteristic data in the tth preset time period, trepresents the probability of pitch performance classification of the operating characteristic data in the tth preset time period, F1 0 represents the reference pitch performance classification probability.

[0145] It should be understood that the performance compliance index increases with the increase of the classification accuracy score, and decreases with the increase of the performance transfer probability score and the pitch performance classification probability score. Among them, the classification accuracy score increases with the increase of the classification accuracy, and the performance transfer probability score increases with the performance transfer probability deviation (i.e. |Y1 t -Y1 0 |) increases, and the probability score of pitch performance classification increases with the pitch performance classification probability deviation (i.e. |F1 t -F1 0 |) increases with the increase of.

[0146] It should be noted that the classification accuracy score also indirectly affects the values ​​of the performance transfer probability score and the pitch performance classification probability score. When the classification accuracy score increases, the accuracy of the operation characteristic data analysis is improved, which helps the pitch coordination control model to more accurately predict the future state transfer of the blade angle, thereby reducing the deviation of the performance transfer probability, that is, reducing the performance transfer probability score.

[0147] Similarly, the increase in the classification accuracy score means that the pitch coordination control model can more accurately judge the changes in the wind turbine pitch performance. This accuracy improves the confidence of the pitch coordination control model in the pitch performance classification, thereby reducing the pitch performance classification probability score. By considering the above indirect influence mechanism, it is helpful to more clearly understand the indirect influence of the classification accuracy score on the performance transfer probability score and the pitch performance classification probability score, which in turn affects the value of the performance compliance index, and achieves a more accurate evaluation of the degree of compliance between the operating characteristic data and the actual operating status of the wind turbine, thereby achieving an improvement in the accuracy and reliability of the prediction of the change trend of the wind turbine blade angle, and effectively reducing the hysteresis of the wind turbine pitch control.

[0148] It should be explained that the wind turbine operating characteristic data such as wind speed, wind direction, blade angle, main shaft speed, power output, ambient temperature, vibration level, etc. are obtained to form an operating characteristic data sample set; the data samples are divided into two categories related to or unrelated to the pitch performance, and the correlation between each feature and the pitch performance index (such as pitch response time, power output stability) is calculated based on the factor analysis algorithm. Through the correlation threshold, the operating characteristic data subset that is highly correlated with the pitch performance is screened out; the classified data samples include:

[0149] Relevant subsets: wind speed, blade angle, spindle speed, power output, etc.

[0150] Non-relevant subsets: ambient humidity, noise level, etc.

[0151] Set screening criteria and measure the pitch performance:

[0152] Minimize power deviation: the deviation between actual power and rated power during operation.

[0153] Blade angle response speed: the average time for blade angle adjustment.

[0154] System stability: failure rate of the pitch system. Select sample data from relevant subsets. Score the pitch performance index of each sample. Sort by score and select the sample with the highest score as the high-quality individual. Identify the characteristic variables that have the greatest impact on the pitch performance, use the feature optimization algorithm to calculate the contribution weight of each operating feature to the pitch performance index, set the threshold according to the contribution weight, filter out the features that have a greater impact on the pitch performance, and combine the classification results of step S21 with the feature weight optimization results of step S23. Compare the characteristic performance of high-quality individuals, confirm the effectiveness of key factors, and determine the key factors that ultimately affect the pitch performance of wind turbines:

[0155] External environmental factors: wind speed, wind speed change rate.

[0156] Equipment status factors: blade angle, spindle speed, vibration level.

[0157] Specifically, the factor analysis algorithm is an improved butterfly optimization algorithm. On the basis of the traditional butterfly optimization algorithm, a high-quality screening mechanism (i.e., association rule algorithm) is introduced. The association relationship between the operating characteristic data of wind turbines is found through the association rule mining method. The C4.5 decision tree method (i.e., decision tree classifier) ​​is used to achieve the preliminary classification of the operating characteristic data. The improved butterfly optimization algorithm can better achieve the balance between local search and global search. Secondly, the feature optimization algorithm (i.e., the adaptive invasive weed optimization algorithm based on microhabitats) is used to optimize and analyze high-quality individuals to further improve the accuracy of the recognition results. The traditional butterfly optimization algorithm is a heuristic swarm intelligence optimization algorithm, and its core strategy is to imitate the foraging and mating behavior of butterflies. It is assumed that all butterflies can emit a certain fragrance to attract each other and approach each other, and the fragrance concentration is related to the fitness function.

[0158] The traditional invasive weed optimization algorithm simulates the powerful colonization and domination ability of weeds by imitating the growth, reproduction, diffusion and competition of weeds in the field. In the invasive weed optimization algorithm, as the number of iterations increases, the spatial distribution range of the next generation of seeds will gradually decrease, which can ensure that the algorithm has a strong global search ability in the early stage, and a strong local search ability in the later stage. However, this also leads to the insufficient local search ability of the invasive weed optimization algorithm in the early stage, and lack of population diversity in the later stage. In order to obtain a better optimization effect, the present invention adds the microhabitat idea and the adaptive mechanism to the traditional invasive weed optimization algorithm, and obtains an adaptive weed invasion algorithm based on the microhabitat idea. The microhabitat idea is used to increase the population diversity of the algorithm; the periodic operator and the adaptive algorithm are introduced into the adaptive mechanism, so that the standard deviation of the spatial diffusion of the weed individual not only changes with the number of iterations, but also can change dynamically according to the parameters of the periodic operator and the fitness value of the individual.

[0159] Specifically, the microhabitat concept comes from biology and refers to a living environment under a specific environment. During their evolution, organisms generally live together with the same species as themselves and reproduce together. Each generation of individuals is divided into several categories according to their fitness values, and each category can represent a microhabitat. The present invention uses the classification competition characteristics of the microhabitat to perform population reproduction, thereby increasing population diversity and improving the global optimization ability of the algorithm.

[0160] Preferably, based on the high-quality screening mechanism, further screening the screened subset of operating characteristic data samples, and selecting the operating characteristic data samples with the best pitch performance as high-quality individuals comprises the following steps:

[0161] S221, collecting wind turbine operation characteristic data, sequentially collecting characteristic data at multiple operation moments, and using an association rule mining algorithm (i.e., Apriori algorithm) to analyze the association relationship between the wind turbine operation characteristic data, to obtain a subset of wind turbine operation characteristic data classified by characteristics;

[0162] Specifically, the Apriori algorithm is a classic association rule mining algorithm used to discover frequent itemsets and association rules from a data set. The algorithm is based on the important property that all subsets of frequent itemsets are also frequent, which effectively reduces the search space.

[0163] S222, randomly selecting a number of operation feature data samples from the wind turbine operation feature data subset as a training set for a decision tree classifier, and calculating the entropy gain of the pitch performance-related features to obtain a preliminary decision tree classifier, and screening out key operation features related to the pitch performance;

[0164] S223, using an association rule mining algorithm to remove incomplete data points in the operation feature data set, performing a consistency check on the operation feature data with the highest degree of association, and screening out a subset of the operation feature data with the highest integrity;

[0165] S224, partitioning the operation feature data subset with the highest integrity into a number of operation feature data subsets of equal size, and using the several operation feature data subsets as initial training sets;

[0166] S225, training a number of operation feature data subsets, calculating the performance transition probability of key operation features under different operation states, and obtaining feature fingerprints related to the states;

[0167] S226, using several subsets of operation feature data as test sets for each other, and using a decision tree classifier to check the consistency of feature fingerprints. If the classification accuracy is greater than a preset threshold, retain the feature subset with the highest classification accuracy; otherwise, merge several subsets of operation feature data and retain the merged operation feature data;

[0168] S227, combining the decision tree classifier with the retained merged operation characteristic data, calculating the pitch performance classification probability, and if there is a disagreement between the classification results, incorporating the operation characteristic data into the performance state transfer matrix;

[0169] S228. When the classification results of the decision tree classifier are consistent with those of the state transfer matrix, new operating characteristic data are introduced and step S227 is repeated, the weight of the decision tree classifier is gradually reduced, and the operating characteristic data samples with the best variable pitch performance are screened out as high-quality individuals based on the results of the performance transfer probability.

[0170] It needs to be explained that, in order to facilitate understanding of the above technical solution of the present invention, the present invention will be described in detail below based on the high-quality screening mechanism in the actual process, to further screen the screened subset of operating characteristic data samples, and select the operating characteristic data samples with the best variable pitch performance as high-quality individuals.

[0171] Step 1: Data collection and correlation analysis

[0172] 1) Data collection:

[0173] The collected operating characteristic data include: wind speed, blade angle, main shaft speed, power output, vibration level, and ambient temperature.

[0174] The sampling data at each time point is in the form of: data point = [wind speed: 10.5m / s, blade angle: 12°, main shaft speed: 1400rpm, power output: 850kW, vibration: 0.02g].

[0175] 2) Attribute association mining:

[0176] Use the Apriori algorithm to calculate the support and confidence between features:

[0177] Association rule example: If wind speed > 12m / s and blade angle < 15°, then power output > 900kW, confidence = 85%. Correlation between wind speed and main shaft speed: 0.78.

[0178] Classification results: Subsets related to pitch performance, such as wind speed, blade angle, and power output, are filtered out.

[0179] Step 2: Preliminary decision tree classifier

[0180] 1) Randomly select training set: Randomly select 500 data samples from the subset as the training set, and the remaining data as the test set.

[0181] 2) Entropy gain calculation:

[0182] Assuming wind speed is a classification feature, the initial entropy F(D) is:

[0183]

[0184] In the formula, F(D) represents the information entropy of the data set D; k represents the total number of categories in the data set; p i Represents the probability of the i-th class data sample.

[0185] The conditional entropy after wind speed splitting is F(D|A): F(D|A)=0.6×0.81+0.4×0.72=0.774.

[0186] Information gain: Gain(A)=F(D)-F(D|A)=1.0-0.774=0.226.

[0187] Repeating the above calculations, we find that wind speed, blade angle, and power output are the important characteristics.

[0188] 3) Preliminary classifier results: A preliminary decision tree was constructed with wind speed, blade angle and power output as the main splitting nodes.

[0189] Step 3: Association rules and completeness screening

[0190] 1) Remove incomplete data points: Use association rules to filter and remove data points with missing values ​​or confidence levels below 70%, retaining 400 complete data points.

[0191] 2)Consistency check:

[0192] Check the consistency of the correlation between wind speed and power output: the standard deviation of the data distribution is σ = 0.15. Samples with consistency greater than 95% are retained.

[0193] Step 4: Partitioning and Training

[0194] 1) Data partitioning: Divide the 400 data into 4 subsets, each with 100 samples.

[0195] 1) Training initial subset:

[0196] Use each subset to train a decision tree and calculate the performance transition probability of the key features. For example:

[0197] The transition probability P of wind speed from 10.5 m / s to 11.2 m / s is 0.85.

[0198] The transition probability P of the spindle speed from 1400 rpm to 1420 rpm is 0.9.

[0199] Step 5: Feature fingerprint calculation

[0200] 1) Feature fingerprint definition: The fingerprint of each operating state consists of the probability distribution of key features: fingerprint = [wind speed distribution, blade angle distribution, power distribution].

[0201] 2) Calculate fingerprint:

[0202] Example fingerprint:

[0203] Wind speed distribution: [10.0, 10.5, 11.0], corresponding probability [0.3, 0.5, 0.2].

[0204] Blade angle distribution: [10°, 12°, 14°], corresponding probability [0.4, 0.4, 0.2].

[0205] Step 6: Testing and merging

[0206] 1) Test set mutual inspection:

[0207] Each subset is used as the other's test set to calculate the classification accuracy. For example:

[0208] The classification accuracy of subset A is 90%, and the classification accuracy of subset B is 85%.

[0209] 2) Classifier optimization: If the classification accuracy is lower than 80%, merge the subsets and retrain. The number of samples after merging: 200, and the classification accuracy is improved to 92%.

[0210] Step 7: State transfer matrix and consistency check

[0211] 1) State transfer matrix:

[0212] Define the transition probability matrix of each operating state:

[0213]

[0214] Each row of the transition probability matrix corresponds to an initial state, each column corresponds to the target state after the transition, and the sum of the rows is 1, ensuring that the sum of the probabilities does not exceed 1. The transition probability of wind speed from low to high is 85%.

[0215] Example explanation:

[0216] First line: Transfer from state 1 to:

[0217] The probability of state 1 is 0.85.

[0218] The probability of state 2 is 0.10.

[0219] The probability of state 3 is 0.05.

[0220] Line 2: Transfer from state 2 to:

[0221] The probability of state 1 is 0.15.

[0222] The probability of state 2 is 0.80.

[0223] The probability of state 3 is 0.05.

[0224] Line 3: Transfer from state 3 to:

[0225] The probability of state 1 is 0.05.

[0226] The probability of state 2 is 0.15.

[0227] The probability of state 3 is 0.80.

[0228] Step 8: Screening high-quality individuals

[0229] 1) Introducing new data:

[0230] Add the newly added data into the training set and repeat step S227.

[0231] The weight of the decision tree is gradually reduced, and the performance transition probability dominates the final screening.

[0232] 2) High-quality samples:

[0233] Finally, the best performing high-quality individual samples were screened out: high-quality individual = [wind speed: 10.5m / s, blade angle: 12°, main shaft speed: 1400rpm, power output: 850kW].

[0234] Preferably, using an association rule mining algorithm to remove incomplete data points in the operation feature data set, performing a consistency check on the operation feature data with the highest degree of association, and screening out a subset of the operation feature data with the highest integrity includes the following steps:

[0235] S2231, selecting a section of wind turbine operation time series data, and presetting a sliding window at the starting point to traverse the characteristic data points in the time series in sequence;

[0236] S2232: If the integrity index of the feature data point in the sliding window is greater than a preset threshold, the feature data point is determined to be incomplete data and added to the abnormal data set;

[0237] S2233, calculating the average distance between each feature data point in the sliding window and the center of the feature data space, and setting an association threshold; if the average distance is less than the association threshold, it is determined that the window contains abnormal feature data, and it is added to the abnormal data set;

[0238] S2234, the sliding window continues to move backward by one time step, replacing the initial feature data point in the sliding window, and repeating steps S2231 to S2233 until all feature data points in the time series are traversed, and finally an abnormal feature data set is output, and the time mark of the feature data point is recorded;

[0239] S2235. According to the classification rules of wind turbine operating characteristics, the operating characteristic data set is divided to generate multiple groups of operating characteristic subsets, the operating characteristic subset with the smallest data volume is deleted, and the operating characteristic subset with the largest data volume and the highest correlation is retained, and finally the operating characteristic data subset with the highest integrity is screened out.

[0240] It should be explained that a period of 24 hours of continuous running time series data is selected; the sliding window size is set to 1 hour, that is, the time period for each analysis is 1 hour; starting from the starting point of the data, the entire time series is traversed in steps of 1 hour; completeness indicators are set, such as the missing rate of data points, the proportion of outliers, etc.; preset thresholds, for example, data points with a missing rate of more than 10% are considered incomplete; for each sliding window, its completeness indicator is calculated. If the indicator exceeds the preset threshold, the data points in the window are judged as incomplete and recorded; the spatial center of each feature data in the sliding window is calculated, such as the mean or median; the average distance from each data point in the window to the spatial center of the feature data is calculated; an association threshold is set, for example, an average distance less than 0.5 standard deviation is considered normal; if the average distance is less than the association threshold, it is determined that the window contains normal feature data; if the average distance is greater than the association threshold, it is determined that it contains abnormal feature data and recorded; after each analysis, the sliding window moves backward one time step; steps S2231 to S2233 are repeated until all time series are traversed; finally, an abnormal feature data set is output, and the time stamp of each data point is recorded; according to the wind turbine operating feature classification rules, the data set is divided into multiple subsets; the operating feature subset with the smallest data volume is deleted to reduce the impact of noise; the operating feature subset with the largest data volume and the highest degree of correlation is retained; and the operating feature data subset with the highest integrity and related to the pitch performance is screened out.

[0241] Preferably, optimizing and analyzing the high-quality individuals using a feature optimization algorithm to identify the operating features that have the greatest impact on the pitch performance of the wind turbine generator system includes the following steps:

[0242] S231, randomly generating an initial feature population from the operating feature data set of the wind turbine generator set, and initializing the parameters of the feature optimization algorithm;

[0243] S232, calculating the fitness of each characteristic individual in the initial characteristic population, and dividing each characteristic individual into a microhabitat according to a microhabitat classification method;

[0244] S233, for each characteristic individual in a microhabitat, generate a new characteristic population according to an optimization formula, and simulate the evolution of the characteristic population in the microhabitat;

[0245] S234, adjusting the distribution of feature populations according to the adaptive diffusion rule, so that high fitness features concentrate on influencing the target area, and low fitness features are gradually eliminated;

[0246] S235, checking whether the key operating feature that has the greatest impact on the pitch performance has been identified, if it has been found, terminating the feature optimization algorithm and outputting the optimal feature; otherwise, continuing the optimization;

[0247] S236, judging whether the maximum number of iterations has been reached, if so, terminating the feature optimization algorithm and outputting the optimal feature; otherwise, continuing the optimization;

[0248] S237, judging whether the current number of characteristic individuals exceeds the maximum number of characteristic populations, if so, continuing the optimization; otherwise, going to step S232;

[0249] S238, according to the competition rules within the niche, select a number of characteristic individuals from each niche as parents to enter the next iteration;

[0250] S239, repeating steps S232 to S238 until the maximum number of iterations is reached, and finally outputting the operating characteristics that have the greatest impact on the variable pitch performance of the wind turbine.

[0251] It should be explained that the initial feature population is randomly generated from the operating feature data set of the wind turbine, and each feature individual contains the values ​​of features such as wind speed, blade angle, and main shaft speed; the key parameters of the feature optimization algorithm are set, such as: population size (such as 100 feature individuals); the maximum number of iterations (such as 50 times); the threshold of the adaptive diffusion rule and the microhabitat classification method; the fitness is calculated for each feature individual, and the fitness measures the degree of influence of the feature on the pitch performance, such as the correlation with the pitch response time or power output; the population is divided into multiple microhabitats according to the similarity of the feature individuals, and each microhabitat represents a specific feature distribution area; in each microhabitat, a new feature population is generated through operations such as crossover and mutation, so that the distribution of the feature population is closer to the actual optimization target; the evolution process of the feature population in the microhabitat is simulated, so that the high fitness features are continuously enhanced and the low fitness features are gradually eliminated; the distribution of the feature population is dynamically adjusted. For example, let the high fitness features concentrate on affecting the target area, while the low fitness features gradually reduce their contribution to the population; check whether the key operating features that have the greatest impact on the pitch performance have been identified in the current feature population; if found, end the feature optimization algorithm and output the key features; if the maximum number of iterations has been reached (such as 50 times), end the feature optimization algorithm and output the key features with the highest current fitness; check whether the current number of feature individuals exceeds the preset maximum population number (such as 200); if exceeded, randomly eliminate low fitness features and keep the population size within a reasonable range; in each microhabitat, select several feature individuals with the highest fitness as parents through competition rules; the parents enter the next iteration to generate a new feature population; starting from step S232, recalculate the fitness, divide the microhabitat, generate a new population, adjust the distribution, and gradually optimize the feature population; if the maximum number of iterations has not been reached, continue to optimize; when the maximum number of iterations is reached or the optimal fitness feature is found, output the key operating features that have the greatest impact on the pitch performance.

[0252] Preferably, the optimization formula is:

[0253]

[0254] Where M X Represents the new feature population generated by the Xth feature individual;

[0255] H X Represents the fitness value of the Xth characteristic individual;

[0256] H max Indicates the maximum fitness in the current feature population;

[0257] H min Indicates the minimum fitness in the current feature population;

[0258] M max Indicates the maximum characteristic population size set;

[0259] M min Indicates the set minimum feature population size.

[0260] Preferably, based on the identified key factors affecting the pitch performance of the wind turbine, a pitch coordination control model is established, and the pitch coordination control model is used to predict the wind turbine pitch at a future moment, and obtaining the blade angle change trend includes the following steps:

[0261] S31. Based on the grey correlation analysis method, the correlation between environmental factors and pitch performance is analyzed, the key factors with the greatest impact on pitch performance are screened out, the factors with the lowest correlation are eliminated, the key factor data set is reconstructed, and divided into a training set and a test set;

[0262] S32. Using variational mode decomposition technology, the time series signal corresponding to the key factor is decomposed into several intrinsic mode functions, each representing a subsequence with different frequency characteristics;

[0263] S33, for each decomposed subsequence, respectively construct a pitch coordination control submodel based on an extreme learning machine;

[0264] S34, using a parameter optimization algorithm to optimize the parameters of the variable pitch coordinated control sub-model of each sub-sequence, and constructing an optimized variable pitch coordinated control model;

[0265] S35. The prediction results of the variable pitch coordinated control sub-models of all sub-sequences are superimposed and reconstructed to obtain the prediction results of the blade angles of the wind turbines at future moments, thereby forming a blade angle change trend.

[0266] It should be noted that the pitch of the wind turbine at a future moment is predicted using the pitch coordination control model, and then it also includes determining whether the pitch prediction of the wind turbine is completed based on the obtained pitch prediction accuracy index; the pitch prediction accuracy index is used to evaluate the accuracy and effectiveness of the pitch coordination control model in predicting the pitch of the wind turbine; the specific process of determining whether the pitch prediction of the wind turbine is completed based on the obtained pitch prediction accuracy index is as follows: determine whether the obtained pitch prediction accuracy index is within the pitch prediction accuracy threshold range (including the cases of being equal to the maximum and minimum values of the historical pitch prediction accuracy index), if so, the pitch prediction of the wind turbine is completed, otherwise, adjust the control parameters of the pitch coordination control model (including the prediction time step and the pitch rate limit) (for example, by reducing the prediction time step parameter of the control interface in the pitch coordination control model to reduce the prediction time step of the pitch coordination control model to ensure the accuracy of the prediction of the pitch coordination control model; under strong wind conditions, increase the pitch rate limit parameter of the control interface in the pitch coordination control model to ensure the safe operation of the wind turbine) and reuse the pitch coordination control model to predict the pitch of the wind turbine at a future moment until the obtained pitch prediction accuracy index is within the pitch prediction accuracy threshold range.

[0267] The pitch prediction accuracy index is obtained through the following method: when the obtained performance compliance index is within the performance compliance threshold range, obtain the predicted pitch angle and predicted motor torque at the end of the prediction within a preset time period by the pitch coordination control model, and simultaneously monitor the actual pitch angle (measured by the potentiometer in the angle measurement device) and the actual motor torque (measured by the torque tester) of the wind turbine at the current moment; obtain the pitch angle prediction score (i.e., J in the limiting expression of the pitch prediction accuracy index) t ) and the motor torque prediction score (i.e., N in the limiting expression of the pitch prediction accuracy index) t ), and at the same time, combine the obtained performance compliance index (i.e., FU in the limiting expression of the pitch prediction accuracy index) t ) and the pre-set environmental wind speed influence factor in the database (i.e., μ in the limiting expression of the pitch prediction accuracy index) to obtain the pitch prediction accuracy index; the pitch angle prediction score represents the ratio of the absolute value of the difference between the predicted pitch angle and the actual pitch angle to the maximum allowable deviation of the pitch angle; the motor torque prediction score represents the ratio of the absolute value of the difference between the predicted motor torque and the actual motor torque to the maximum allowable deviation of the motor torque.

[0268] Among them, the pitch prediction accuracy threshold range represents the range corresponding to the maximum and minimum values ​​of the historical pitch prediction accuracy index of the pitch coordination control model in the database in each historical time period; the environmental wind speed impact factor is stored in the database, which represents the numerical value corresponding to the degree of influence of the environmental wind speed on the pitch angle. When the environmental wind speed impact factor needs to be obtained, the current environmental wind speed measured by the anemometer is input into a pre-set mapping set, and the environmental wind speed impact factor corresponding to the current environmental score is retrieved according to the mapping relationship in the mapping set. This mapping relationship is a one-to-one correspondence between the historical environmental score and the environmental impact factor (that is, each environmental wind speed corresponds to an impact factor) or many-to-one (that is, multiple environmental wind speeds share an impact factor). The value range of the environmental wind speed impact factor in this example is between 0 and 1 (including 0 and 1).

[0269] Specifically, the specific limiting expression of the pitch prediction accuracy index is:

[0270]

[0271] Where t is the number of the preset time period, t = 1, 2, ..., T, T is the total number of preset time periods, e is a natural constant, CE t It represents the pitch prediction accuracy index of the pitch coordination control model at the end of the prediction in the tth preset time period, FU t represents the performance compliance index of the operating characteristic data in the tth preset time period, ΔFU represents the performance compliance threshold range, μ represents the environmental wind speed impact factor, J t represents the pitch angle prediction score of the pitch coordination control model at the end of the prediction in the tth preset time period, J2 t represents the predicted pitch angle at the end of the prediction of the pitch coordination control model in the tth preset time period, J1 t It represents the actual pitch angle of the wind turbine corresponding to the end of the prediction of the pitch coordination control model in the tth preset time period, ΔJ max Indicates the maximum allowable deviation of the pitch angle (indicates the maximum value of the historical pitch angle deviation of each blade of the historical wind turbine in the database within the historical time period), N t N2 represents the motor torque prediction score at the end of the prediction of the pitch coordinated control model in the tth preset time period, t N1 represents the predicted motor torque at the end of the prediction of the pitch coordinated control model in the tth preset time period, t It represents the actual motor torque of the wind turbine corresponding to the end of the prediction of the pitch coordination control model in the tth preset time period, ΔN max Indicates the maximum allowable deviation of the motor torque (indicates the maximum value of the historical motor torque deviation of the historical wind turbines in the database in each historical time period).

[0272] It should be understood that the pitch prediction accuracy index increases with the increase of the performance compliance index, and decreases with the increase of the pitch angle prediction score and the motor torque prediction score. The pitch angle prediction score increases with the pitch angle deviation (i.e. |J2 t -J1 t |) increases, and the motor torque prediction score increases with the motor torque deviation (i.e. |N2 t -N1 t |) increases with the increase of.

[0273] It should be noted that the performance compliance index also indirectly affects the values ​​of the pitch angle prediction score and the motor torque prediction score. When the performance compliance index increases, it means that the operation of the wind turbine is closer to its optimal state, that is, the wind turbine can more effectively capture wind energy and convert it into electrical energy. This efficient operation is often accompanied by more precise pitch control and more stable motor torque output, thereby reducing the pitch angle deviation and motor torque deviation (that is, reducing the pitch angle prediction score and motor torque prediction score).

[0274] The pitch angle prediction score also indirectly affects the value of the motor torque prediction score. There is a mechanical coupling effect between the wind turbine's pitch system and the motor torque. The change in pitch angle will directly affect the force on the blades, and then affect the output of the motor torque. Therefore, when the pitch angle prediction is more accurate, the motor torque prediction will also be more accurate, that is, when the pitch angle prediction score decreases, the corresponding motor torque prediction score will also decrease accordingly. By considering the above indirect influence mechanism, it is helpful to more comprehensively and accurately understand the mechanical coupling effect and interdependence between the pitch angle and the motor torque, and then realize the more accurate prediction of the change trend of each blade of the wind turbine by the pitch coordination control model, effectively solving the hysteresis of the wind turbine pitch control.

[0275] Specifically, the variable pitch coordinated control model is a kernel extreme learning machine (KELM), which is a new type of artificial intelligence prediction model. It introduces the kernel function idea on the basis of the extreme learning machine (ELM), thereby effectively overcoming the problem of low output stability caused by the random generation of initial weights and thresholds in ELM, and improving the output stability. However, its prediction performance is usually affected by parameter selection. Therefore, the present invention uses variational mode decomposition to decompose the actual time series signal, and then adds gray correlation analysis to perform correlation analysis on the input and output variables of the variable pitch coordinated control model to increase the correlation of the data. Based on the KLEM model, a parameter optimization algorithm (i.e., a black hole optimization algorithm) is used to optimize its parameters. Variational mode decomposition (VMD) is a signal processing technology that is widely used to analyze non-stationary and nonlinear signals. It decomposes the signal into a set of intrinsic mode functions (IMFs) with different center frequencies through a variational method. These IMFs represent different frequency components in the signal and can be used for feature extraction, predictive modeling, and pattern recognition.

[0276] It should be explained that the environmental factors (such as wind speed, wind direction, air density, temperature) and operating characteristics (such as blade angle, power output, speed) data of wind turbines are collected during operation; the correlation between each factor and pitch performance is calculated by grey correlation analysis method; the factors with the highest correlation (such as wind speed and wind direction) are screened out and the factors with lower correlation (such as temperature) are eliminated; the data set is reconstructed using the screened key factors and divided into training set and test set; training set: used to train the pitch coordination control model; test set: used to evaluate model performance; the time series data of key factors (such as wind speed, wind direction, blade angle) are input into the variational mode decomposition algorithm; decomposed into several intrinsic mode functions (IMFs), each subsequence represents different frequency characteristics:

[0277] High-frequency signals: such as instantaneous wind speed changes.

[0278] Low-frequency signals: such as long-term trend changes.

[0279] Extract the time characteristics (such as fluctuation amplitude and trend) of each subsequence to provide input for subsequent model construction; for each decomposed subsequence, select the extreme learning machine (ELM) as the submodel. The extreme learning machine has fast training and strong generalization capabilities, and is suitable for processing nonlinear time series; independently train a variable pitch coordinated control submodel for each subsequence to predict future subsequence changes; use the parameter optimization algorithm to adjust the hyperparameters of each submodel; output the optimized submodel, and each submodel predicts the subsequence data of a specific frequency; superimpose the prediction results of all submodels according to the frequency components of the original signal to reconstruct the complete time series; the reconstructed time series is the predicted value of the blade angle change at the future moment; draw a blade angle change trend chart based on the reconstruction results to show the changes in the future period of time. Trend analysis: If the blade angle rises rapidly in a short period of time, the controller response speed may need to be increased; if the angle changes relatively smoothly, the system operation is stable.

[0280] Preferably, using a parameter optimization algorithm to optimize the parameters of the pitch coordination control submodel of each subsequence, and constructing the optimized pitch coordination control model includes the following steps:

[0281] S341, randomly generating a number of candidate parameter sets in the search space, each candidate parameter set corresponding to a solution of the propeller coordinated control sub-model;

[0282] S342, input each candidate parameter set (i.e., each star in the black hole optimization algorithm) into the corresponding pitch coordination control sub-model, calculate its fitness value, and evaluate the pros and cons of each candidate parameter;

[0283] S343, selecting a parameter solution with the best fitness value from the current candidate parameter set as the parameter solution with the best performance in the current iteration (i.e., the black hole in the black hole optimization algorithm);

[0284] S344, adjusting the positions of other candidate parameter sets according to the update formula to make them close to the optimal parameter solution;

[0285] S345, calculating the fitness value of each candidate parameter set after the updated position, if the fitness value of any parameter solution is greater than the current optimal solution, then updating it to the new optimal parameter solution;

[0286] S346, for the candidate parameter solution with the lowest fitness value, remove it from the current candidate parameter set, and randomly generate a new parameter solution in the search space;

[0287] S347. Check whether the maximum number of iterations is reached. If the maximum number of iterations is reached, stop the optimization process, take the current optimal parameter solution as the final optimization result, and construct an optimized pitch coordination control model based on the final optimization result; otherwise, return to step S344 and continue the optimization.

[0288] Specifically, the parameter optimization algorithm is the black hole optimization algorithm. The idea of the algorithm more completely describes the general characteristics of the black hole phenomenon in nature. Based on these characteristics of the black hole optimization algorithm, the optimization search principle of the black hole optimization algorithm is relatively simple and easy to implement. The black hole optimization algorithm mainly simulates the actual black hole phenomenon, randomly arranges a certain number of stars in a certain search space, and determines and evaluates the fitness function of each star in the search space through statistical means, and selects a star with the best fitness value as the black hole. The boundary of this black hole is regarded as the area where the current global optimal solution is located, and the black hole itself is regarded as the current global optimal solution.

[0289] It should be explained that the parameter range of the pitch coordination control sub-model is determined, such as the upper and lower limits of the learning rate, regularization coefficient, etc.; several groups of parameters (such as 50 groups) are randomly generated within the search space, and each group of parameters defines a solution of the sub-model; each group of candidate parameters is input into the corresponding pitch coordination control sub-model; using the training data, calculate the prediction error or objective function value of the model as the fitness value; the fitness value measures the quality of the parameter set for the model prediction performance; the higher the fitness value, the better the parameter set and the smaller the prediction error of the model; select the solution with the highest fitness value from the current candidate parameter set as the best parameter set in the current iteration; according to the update formula (such as the optimization rule based on distance weight), move other candidate parameters closer to the optimal solution; through appropriate random perturbations, ensure that the search covers the global range and avoid falling into local optima; input each updated candidate parameter set into the model one by one and calculate its new fitness value; if the fitness value of any candidate parameter set exceeds the current optimal solution, update it as the new optimal solution; remove the candidate parameter solution with the lowest fitness value from the current candidate parameter set; randomly generate a new parameter set within the search space and supplement it to the candidate parameter set to maintain the population size; if the maximum number of iterations (such as 100 times) is reached, stop the optimization process; take the current optimal parameter set as the final optimization result and construct an optimized pitch coordination control model; return to step S344 and continue the optimization.

[0290] Preferably, the update formula is:

[0291] W a (m + 1)=W a (m)+rand×(W ZY -W a (m)), a = 1, 2,... N;

[0292] Where W a (m+1) represents the position of the ath candidate parameter set at the m+1 iteration;

[0293] W a (m) represents the position of the ath candidate parameter set (i.e., the star in the black hole optimization algorithm) at the mth iteration;

[0294] rand represents a random number between [0, 1];

[0295] W ZY Indicates the position of the optimal parameter solution in the search space (i.e., the black hole position in the black hole optimization algorithm);

[0296] ZY represents the optimal parameter solution;

[0297] N represents the total number of candidate parameter sets.

[0298] According to another embodiment of the present invention, Figure 2 As shown, a wind turbine generator set pitch coordination optimization control system is also provided, and the pitch coordination optimization control system includes:

[0299] The data acquisition module 1 is used to acquire the real-time operation data of the wind turbine generator set and pre-process the real-time operation data to obtain the operation characteristic data of the wind turbine generator set;

[0300] Factor analysis module 2 is used to analyze the obtained operation characteristic data of the wind turbine using the factor analysis algorithm to identify the key factors affecting the variable pitch performance of the wind turbine;

[0301] Model building module 3 is used to establish a pitch coordination control model based on the identified key factors affecting the pitch performance of the wind turbine, and use the pitch coordination control model to predict the pitch of the wind turbine at a future moment to obtain the blade angle change trend;

[0302] The coordination control module 4 is used to dynamically adjust the pitch angle of each blade based on the obtained blade angle variation trend and in combination with the real-time operation status of the wind turbine generator set, and optimize the pitch coordination control strategy.

[0303] It should be noted that the pitch angle of each blade is adjusted dynamically, and then the pitch coordination interference score is obtained to determine whether to optimize the pitch coordination control strategy (determine whether the obtained pitch coordination interference score is within the allowable range of pitch coordination interference. If so, it indicates that the current pitch coordination control strategy is effective. Otherwise, it indicates that the interference degree of the pitch angle at the current moment is not within the expected allowable range. At this time, it is necessary to prompt the preset personnel to optimize the pitch coordination control strategy); the pitch coordination interference score is used to evaluate the interference degree of the pitch angle during the dynamic adjustment process under the current pitch coordination control strategy; the pitch coordination interference score is obtained by the following method: when the obtained pitch prediction accuracy index is in the pitch prediction accuracy index, the pitch prediction accuracy index is in the pitch prediction accuracy index. When the accuracy threshold is within the range, the average electromagnetic intensity during the response and transmission (response refers to the duration corresponding to the process of each blade of the wind turbine set starting to dynamically adjust the blade angle after receiving the pitch command, and transmission refers to the duration corresponding to the process of the pitch command being transmitted from the pitch coordination control model to each blade of the wind turbine set) of the pitch command (referring to the command for adjusting the blade angle issued by the pitch coordination optimization control system of the wind turbine set to each blade of the wind turbine set) of each blade of the wind turbine set (responding and transmitting together within a preset time period) is monitored in real time, and it is determined whether the average electromagnetic intensity is less than the maximum allowable electromagnetic intensity. If so, the electromagnetic interference coefficient (i.e., C in the limiting expression of the pitch coordination interference score) is obtained. t ), otherwise the preset personnel are prompted to inspect the current communication line, and then re-adjust dynamically and re-obtain the average electromagnetic strength (measured by an electromagnetic field strength measuring instrument) until the re-obtained average electromagnetic strength is less than the maximum allowable electromagnetic strength. The electromagnetic interference coefficient represents the ratio of the average electromagnetic strength to the maximum allowable electromagnetic strength; obtain the command response delay score (i.e., the X in the limiting expression of the pitch coordination interference score) t ) and the command transmission delay fraction (i.e., S in the limiting expression of the pitch coordination interference fraction) t ), and at the same time combined with the obtained electromagnetic interference coefficient and pitch prediction accuracy index (i.e., CE in the limiting expression of pitch coordination interference score) t ) to obtain the pitch coordination interference score, the command response delay score represents the ratio of the difference between the actual response delay duration of the pitch command in the response process (recorded in real time by the timer) and the reference response delay duration to the reference response delay duration, and the command transmission delay score represents the ratio of the difference between the actual transmission delay duration of the pitch command in the transmission process (recorded in real time by the timer) and the reference transmission delay duration to the reference transmission delay duration.

[0304] Specifically, the specific limiting expression of the pitch coordination interference fraction is:

[0305]

[0306] Where t is the number of the preset time period (only one pitch change command is responded and transmitted in a preset time period), t = 1, 2, ..., T, T is the total number of preset time periods, e is a natural constant, XIE t It represents the pitch coordination interference fraction of each blade of the wind turbine during the dynamic adjustment of the pitch angle in the tth preset time period, CE t represents the pitch prediction accuracy index of the pitch coordination control model at the end of the prediction in the tth preset time period, ΔCE represents the pitch prediction accuracy threshold range, and X t Indicates the command response delay fraction of the pitch command of each blade of the wind turbine in the tth preset time period, X1 t Indicates the actual response delay time of the pitch change command of each blade of the wind turbine in the tth preset time period, X1 0 represents the reference response delay time (represented by the sum and average of the historical response delay times of the pitch change instructions of the wind turbines in the database during the historical response process), S t S1 represents the transmission delay fraction of the pitch command of each blade of the wind turbine during the transmission process of the pitch command in the tth preset time period, t Indicates the actual transmission delay time of the pitch change command of each blade of the wind turbine when the transmission ends within the tth preset time period, S1 0 represents the reference transmission delay time (represented by the sum and average of the historical transmission delay time of the pitch change command of the wind turbine in the database during the historical transmission process), C t It represents the electromagnetic interference coefficient during the response and transmission of the pitch change command of each blade of the wind turbine in the tth preset time period, It represents the average electromagnetic intensity of the pitch command of each blade of the wind turbine during the response and transmission process in the tth preset time period, C1 max Indicates the maximum allowable electromagnetic intensity (indicates the maximum value of the historical electromagnetic intensity of the pitch change command of the wind turbine in the database during the historical response and transmission process).

[0307] The pitch prediction accuracy threshold range is 1.5 to 2.5, the maximum allowable electromagnetic intensity is 0.15, and the change statistics of the pitch coordination interference score are shown in Table 1:

[0308] Table 1 Statistics of changes in pitch coordination interference scores

[0309]

[0310] It should be understood that, from the first and second rows of data in Table 1, it can be seen that the pitch coordination interference score decreases with the increase of the pitch prediction accuracy index, and from the third, fourth and fifth rows of data in Table 1, it can be seen that the pitch coordination interference score increases with the increase of the command response delay score, the command transmission delay score and the electromagnetic interference coefficient, among which the command response delay score increases with the increase of the actual response delay duration, and the command transmission delay score increases with the increase of the actual transmission delay duration.

[0311] It should be noted that the pitch prediction accuracy index also indirectly affects the value of the electromagnetic interference coefficient. When the pitch prediction accuracy index increases, it means that the prediction performance of the pitch coordination control model is enhanced, which increases the response rate and transmission rate of the pitch command, reduces the electromagnetic interference caused by the prediction error, and helps to reduce the electromagnetic interference score. This reflects that the wind turbine is operating in a more stable electromagnetic environment.

[0312] The command response delay score also indirectly affects the value of the command transmission delay score. When the command response delay score increases, it means that the response speed of the pitch coordination control model to the pitch command slows down. This may be due to the response delay caused by the congestion of the communication network. At this time, the prediction performance of the pitch coordination control model will be reduced, thereby affecting the stability and efficiency of the prediction, and then leading to an increase in the waiting time of the pitch command during the transmission process. This is because the pitch command needs to be processed and responded to in a timely manner during the transmission process. Therefore, when the command response delay score increases, the corresponding command transmission delay score also increases.

[0313] By considering the above-mentioned indirect impact mechanism, it is helpful to prompt the preset personnel to discover problems in time and take corresponding solutions, improve the predictive performance of the variable pitch coordinated control system, reduce the reduction in wind turbine operating efficiency caused by response delay and transmission delay, and help promote the sustainable and healthy development of the wind power industry.

[0314] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A wind turbine pitch coordination optimization control method, characterized in that: The pitch-changing coordinated optimization control method comprises the following steps: S1. Acquire real-time operation data of the wind turbine generator set, and pre-process the real-time operation data to obtain operation characteristic data of the wind turbine generator set; S2. Based on the factor analysis algorithm, the obtained operation characteristic data of the wind turbine is analyzed by using the factor analysis algorithm to identify the key factors affecting the variable pitch performance of the wind turbine; S3. Based on the identified key factors affecting the pitch performance of wind turbines, a pitch coordination control model is established, and the pitch coordination control model is used to predict the pitch of wind turbines at future moments to obtain the blade angle change trend; S4. Based on the obtained blade angle change trend and combined with the real-time operating status of the wind turbine, the pitch angle of each blade is dynamically adjusted, and the pitch coordination control strategy is optimized.

2. A wind turbine pitch coordination optimization control method according to claim 1, characterized in that: The method of analyzing the obtained operation characteristic data of the wind turbine using the factor analysis algorithm to identify the key factors affecting the variable pitch performance of the wind turbine includes the following steps: S21, obtaining a sample set of operation characteristic data of the wind turbine generator set, and classifying and arranging the sample set of operation characteristic data of the wind turbine generator set by using a factor analysis algorithm, and screening out a subset of the sample set of operation characteristic data that is consistent with the variable pitch performance characteristics of the wind turbine generator set; S22. Based on the high-quality screening mechanism, further screening the screened subset of operation characteristic data samples, and selecting the operation characteristic data samples with the best pitch performance as high-quality individuals; S23, using feature optimization algorithm to optimize and analyze high-quality individuals, and identify the operating characteristics that have the greatest impact on the variable pitch performance of the wind turbine; S24. Based on the classification results and the optimization results of high-quality individuals, identify the key factors affecting the variable pitch performance of wind turbines.

3. A wind turbine pitch coordination optimization control method according to claim 2, characterized in that: The method of further screening the screened subset of operation characteristic data samples based on the high-quality screening mechanism and selecting the operation characteristic data samples with the best pitch performance as high-quality individuals comprises the following steps: S221, collecting wind turbine operation characteristic data, sequentially collecting characteristic data at multiple operation moments, and using an association rule mining algorithm to analyze the association relationship between the wind turbine operation characteristic data, to obtain a subset of wind turbine operation characteristic data classified by characteristics; S222, randomly selecting a number of operation feature data samples from the wind turbine operation feature data subset as a training set for a decision tree classifier, and calculating the entropy gain of the pitch performance-related features to obtain a preliminary decision tree classifier, and screening out key operation features related to the pitch performance; S223, using an association rule mining algorithm to remove incomplete data points in the operation feature data set, performing a consistency check on the operation feature data with the highest degree of association, and screening out a subset of the operation feature data with the highest integrity; S224, partitioning the operation feature data subset with the highest integrity into a number of operation feature data subsets of equal size, and using the several operation feature data subsets as initial training sets; S225, training a number of operation feature data subsets, calculating the performance transition probability of key operation features under different operation states, and obtaining feature fingerprints related to the states; S226, using several subsets of operation feature data as test sets for each other, and using a decision tree classifier to check the consistency of feature fingerprints. If the classification accuracy is greater than a preset threshold, retain the feature subset with the highest classification accuracy; otherwise, merge several subsets of operation feature data and retain the merged operation feature data; S227, combining the decision tree classifier with the retained merged operation characteristic data, calculating the pitch performance classification probability, and if there is a disagreement between the classification results, incorporating the operation characteristic data into the performance state transfer matrix; S228. When the classification results of the decision tree classifier are consistent with those of the state transfer matrix, new operating characteristic data are introduced and step S227 is repeated, the weight of the decision tree classifier is gradually reduced, and the operating characteristic data samples with the best variable pitch performance are screened out as high-quality individuals based on the results of the performance transfer probability.

4. A wind turbine pitch coordination optimization control method according to claim 3, characterized in that: The method of using the association rule mining algorithm to remove incomplete data points in the operation feature data set, performing consistency check on the operation feature data with the highest degree of association, and screening out the operation feature data subset with the highest integrity includes the following steps: S2231, selecting a section of wind turbine operation time series data, and presetting a sliding window at the starting point to traverse the characteristic data points in the time series in sequence; S2232: If the integrity index of the feature data point in the sliding window is greater than a preset threshold, the feature data point is determined to be incomplete data and added to the abnormal data set; S2233, calculating the average distance between each feature data point in the sliding window and the center of the feature data space, and setting an association threshold; if the average distance is less than the association threshold, it is determined that the window contains abnormal feature data, and it is added to the abnormal data set; S2234, the sliding window continues to move backward by one time step, replacing the initial feature data point in the sliding window, and repeating steps S2231 to S2233 until all feature data points in the time series are traversed, and finally an abnormal feature data set is output, and the time mark of the feature data point is recorded; S2235. According to the classification rules of wind turbine operating characteristics, the operating characteristic data set is divided to generate multiple groups of operating characteristic subsets, the operating characteristic subset with the smallest data volume is deleted, and the operating characteristic subset with the largest data volume and the highest correlation is retained, and finally the operating characteristic data subset with the highest integrity is screened out.

5. A wind turbine pitch coordination optimization control method according to claim 4, characterized in that: The method of optimizing and analyzing high-quality individuals using a feature optimization algorithm to identify the operating features that have the greatest impact on the variable pitch performance of a wind turbine generator system includes the following steps: S231, randomly generating an initial feature population from the operating feature data set of the wind turbine generator set, and initializing the parameters of the feature optimization algorithm; S232, calculating the fitness of each characteristic individual in the initial characteristic population, and dividing each characteristic individual into a microhabitat according to a microhabitat classification method; S233, for each characteristic individual in a microhabitat, generate a new characteristic population according to an optimization formula, and simulate the evolution of the characteristic population in the microhabitat; S234, adjusting the distribution of feature populations according to the adaptive diffusion rule, so that high fitness features concentrate on influencing the target area, and low fitness features are gradually eliminated; S235, checking whether the key operating feature that has the greatest impact on the pitch performance has been identified, if it has been found, terminating the feature optimization algorithm and outputting the optimal feature; otherwise, continuing the optimization; S236, judging whether the maximum number of iterations has been reached, if so, terminating the feature optimization algorithm and outputting the optimal feature; otherwise, continuing the optimization; S237, judging whether the current number of characteristic individuals exceeds the maximum number of characteristic populations, if so, continuing the optimization; otherwise, going to step S232; S238, according to the competition rules within the niche, select a number of characteristic individuals from each niche as parents to enter the next iteration; S239, repeating steps S232 to S238 until the maximum number of iterations is reached, and finally outputting the operating characteristics that have the greatest impact on the variable pitch performance of the wind turbine.

6. A wind turbine pitch coordination optimization control method according to claim 5, characterized in that: The optimization formula is: Where M X Represents the new feature population generated by the Xth feature individual; H X Represents the fitness value of the Xth characteristic individual; H max Indicates the maximum fitness in the current feature population; H min Indicates the minimum fitness in the current feature population; M max Indicates the maximum characteristic population size set; M min Indicates the set minimum feature population size.

7. The method for coordinated optimization control of pitch change of a wind turbine according to claim 1, characterized in that: The method of establishing a pitch coordination control model based on the identified key factors affecting the pitch performance of the wind turbine generator set, and using the pitch coordination control model to predict the pitch of the wind turbine generator set at a future moment to obtain the blade angle change trend includes the following steps: S31. Based on the grey correlation analysis method, the correlation between environmental factors and pitch performance is analyzed, the key factors with the greatest impact on pitch performance are screened out, the factors with the lowest correlation are eliminated, the key factor data set is reconstructed, and divided into a training set and a test set; S32. Using variational mode decomposition technology, the time series signal corresponding to the key factor is decomposed into several intrinsic mode functions, each representing a subsequence with different frequency characteristics; S33, for each decomposed subsequence, respectively construct a pitch coordination control submodel based on an extreme learning machine; S34, using a parameter optimization algorithm to optimize the parameters of the variable pitch coordinated control sub-model of each sub-sequence, and constructing an optimized variable pitch coordinated control model; S35, superimposing and reconstructing the prediction results of the variable pitch coordinated control sub-models of all sub-sequences, obtaining the prediction results of the blade angles of the wind turbines at future moments, and forming a blade angle change trend.

8. A wind turbine pitch coordination optimization control method according to claim 7, characterized in that: The method of optimizing the parameters of the pitch coordination control submodel of each subsequence by using a parameter optimization algorithm and constructing an optimized pitch coordination control model comprises the following steps: S341, randomly generating a number of candidate parameter sets in the search space, each candidate parameter set corresponding to a solution of the propeller coordinated control sub-model; S342, inputting each candidate parameter set into the corresponding pitch coordination control sub-model, calculating its fitness value, and evaluating the pros and cons of each candidate parameter; S343, selecting a parameter solution with the best fitness value from the current candidate parameter set as the parameter solution with the best performance in the current iteration; S344, adjusting the positions of other candidate parameter sets according to the update formula to make them close to the optimal parameter solution; S345, calculating the fitness value of each candidate parameter set after the updated position, if the fitness value of any parameter solution is greater than the current optimal solution, then updating it to the new optimal parameter solution; S346, for the candidate parameter solution with the lowest fitness value, remove it from the current candidate parameter set, and randomly generate a new parameter solution in the search space; S347. Check whether the maximum number of iterations has been reached. If so, stop the optimization process, take the current optimal parameter solution as the final optimization result, and build an optimized variable pitch coordinated control model based on the final optimization result; otherwise, return to step S344 and continue the optimization.

9. A wind turbine pitch coordination optimization control method according to claim 8, characterized in that: The update formula is: W a (m+1)=W a (m)+rand×(W ZY -W a (m)),a=1,2,…N; Where W a (m+1) represents the position of the ath candidate parameter set at the m+1 iteration; W a (m) represents the position of the ath candidate parameter set at the mth iteration; rand represents a random number; W ZY Represents the location of the optimal parameter solution in the search space; ZY represents the optimal parameter solution; N represents the total number of candidate parameter sets.

10. The method for coordinated optimization control of pitch change of a wind turbine according to claim 1, characterized in that: The step of analyzing the obtained operating characteristic data of the wind turbine using a factor analysis algorithm further includes obtaining a performance compliance index to determine whether to perform key factor identification; The performance compliance index is used to evaluate the degree of compliance between the operating characteristic data and the actual operating state of the wind turbine generator set; The performance compliance index is obtained by the following method: E1, obtain the classification accuracy of the decision tree classifier within a preset time period, and determine whether the classification accuracy is greater than the preset classification accuracy. If so, store the feature data subset with the highest classification accuracy in the running feature data and execute E2, otherwise merge the feature data subsets in the running feature data and then store them; E2, obtain the classification accuracy score, and at the same time count the performance transfer probability and pitch performance classification probability of the wind turbine in the preset time period, and respectively combine the corresponding reference performance transfer probability and reference pitch performance classification probability to obtain the performance transfer probability score and pitch performance classification probability score; E3, obtaining a preset performance compliance weight factor from a preset database, and combining the obtained classification accuracy score, performance transfer probability score and pitch performance classification probability score to obtain a performance compliance index; The classification accuracy represents the ratio of the amount of operation feature data classified in the decision tree classifier to the total amount of operation feature data; The classification accuracy score represents the ratio of the classification accuracy to the preset classification accuracy; The performance transition probability represents the probability of the wind turbine generator system transitioning from one operating state to another operating state; The performance transition probability score represents the ratio of the absolute value of the difference between the performance transition probability and the reference performance transition probability to the reference performance transition probability; The pitch performance classification probability score represents the ratio of the absolute value of the difference between the pitch performance classification probability and the reference pitch performance classification probability to the reference pitch performance classification probability; The performance compliance weight factors include a classification accuracy score weight factor, a performance transfer probability score weight factor, and a pitch performance classification probability score weight factor.

11. The method for coordinated optimization control of pitch change of a wind turbine according to claim 1, characterized in that: The pitch change coordination control model is used to predict the wind turbine pitch change at a future moment, and then the pitch change prediction of the wind turbine is determined based on the obtained pitch change prediction accuracy index. The pitch prediction accuracy index is used to evaluate the accuracy and effectiveness of the pitch coordination control model in predicting the pitch conditions of wind turbines; The specific process of judging whether the wind turbine pitch prediction is completed based on the acquired pitch prediction accuracy index is as follows: Determine whether the obtained pitch prediction accuracy index is within the pitch prediction accuracy threshold range. If so, complete the wind turbine pitch prediction. Otherwise, adjust the control parameters of the pitch coordination control model and reuse the pitch coordination control model to predict the wind turbine pitch at future times until the obtained pitch prediction accuracy index is within the pitch prediction accuracy threshold range. The control parameters include prediction time step and pitch rate limit.

12. A wind turbine pitch coordination optimization control method according to claim 11, characterized in that: The pitch prediction accuracy index is obtained by the following method: When the obtained performance compliance index is within the performance compliance threshold range, the predicted pitch angle and predicted motor torque at the end of the prediction of the pitch coordination control model within the preset time period are obtained, and the actual pitch angle and actual motor torque of the wind turbine at the current moment are monitored at the same time; Obtain the pitch angle prediction score and the motor torque prediction score, and combine the obtained performance compliance index and the ambient wind speed influence factor preset in the database to obtain the pitch prediction accuracy index; The pitch angle prediction score represents the ratio of the absolute value of the difference between the predicted pitch angle and the actual pitch angle to the maximum allowable deviation of the pitch angle; The motor torque prediction score represents a ratio of an absolute value of a difference between the predicted motor torque and the actual motor torque to a maximum allowable deviation of the motor torque.

13. The method for coordinated optimization control of pitch change of a wind turbine according to claim 1, characterized in that: The dynamically adjusting the pitch angle of each blade also includes obtaining a pitch coordination interference score to determine whether to optimize the pitch coordination control strategy; The pitch coordination interference score is used to evaluate the degree of interference of the pitch angle during the dynamic adjustment process under the current pitch coordination control strategy; The pitch coordination interference score is obtained by the following method: When the acquired pitch prediction accuracy index is within the pitch prediction accuracy threshold range, the response of the pitch command of each blade of the wind turbine and the average electromagnetic intensity during the transmission process are monitored in real time to determine whether the average electromagnetic intensity is less than the maximum allowable electromagnetic intensity. If so, the electromagnetic interference coefficient is obtained, otherwise the preset personnel is prompted to adjust the electromagnetic shielding parameters of the current communication line, and the electromagnetic interference coefficient represents the ratio of the average electromagnetic intensity to the maximum allowable electromagnetic intensity; The command response delay score and the command transmission delay score are obtained, and the pitch coordination interference score is obtained by combining the obtained electromagnetic interference coefficient and the pitch prediction accuracy index. The command response delay score represents the ratio of the difference between the actual response delay duration of the pitch command in the response process and the reference response delay duration to the reference response delay duration. The command transmission delay score represents the ratio of the difference between the actual transmission delay duration of the pitch command in the transmission process and the reference transmission delay duration to the reference transmission delay duration.

14. A wind turbine pitch coordination optimization control system, used to implement the wind turbine pitch coordination optimization control method according to any one of claims 1 to 13, characterized in that: The pitch coordination optimization control system includes: A data acquisition module is used to acquire the real-time operation data of the wind turbine generator set and pre-process the real-time operation data to obtain the operation characteristic data of the wind turbine generator set; A factor analysis module is used to analyze the obtained operation characteristic data of the wind turbine using the factor analysis algorithm to identify key factors affecting the variable pitch performance of the wind turbine; The model building module is used to establish a pitch coordination control model based on the identified key factors affecting the pitch performance of the wind turbine, and use the pitch coordination control model to predict the wind turbine pitch at future moments to obtain the blade angle change trend; the coordination control module is used to dynamically adjust the pitch angle of each blade based on the obtained blade angle change trend and combined with the real-time operating status of the wind turbine, and optimize the pitch coordination control strategy.

Images