Load reduction control method for wind turbine generator based on laser radar complex working condition recognition

Abstract

According to the laser radar complex working condition recognition-based load reduction control method for the wind turbine generator, a high-precision laser radar system is installed on the wind turbine generator, three-dimensional wind speed distribution in a wind field is monitored in real time, data are transmitted to a control system, and the real-time performance and accuracy of the data are ensured. Collected original data are preprocessed, a time sequence analysis method is adopted, the rules of wind speed and wind direction changing along with time are mined, and key features of sudden wind speed change and rapid wind direction change are captured. According to historical data, a fuzzy logic rule is defined, and various complex working conditions are accurately described. And designing an adaptive fuzzy logic controller (AFLC), and when the AFLC recognizes a complex working condition, immediately starting load reduction control logic to adjust operation parameters of the wind turbine generator. Meanwhile, the AFLC dynamically adjusts the load reduction threshold value according to the wind field condition, and it is ensured that complex working condition changes are responded in time. In the load reduction control process, the control effect is evaluated, and a fuzzy rule and a membership function in the AFLC are optimized according to the control effect.

Description

Technical Field

[0001] The present invention relates to the field of load shedding control of wind turbines, and in particular to a load shedding control method for wind turbines based on complex working condition identification by laser radar. Background Art

[0002] When a wind turbine encounters some extreme wind conditions, such as rapid changes in wind speed and wind direction, and extreme negative shear, the control stability will deteriorate. Rapid changes in wind speed and direction, especially near the rated wind speed, can easily cause overspeed and vibration failures in the wind turbine, while extreme negative shear can cause uneven loads on the wind rotor and make it easy for the blades to sweep the tower. When a wind turbine experiences these complex working conditions, it is very easy for extreme loads to occur, which can seriously lead to tower sweeps or even tower collapses. The feedforward control algorithm based on laser wind radar in the prior art is overly dependent on the accuracy of laser radar wind measurement and the accuracy of the theoretical model of the wind turbine, and has poor environmental adaptability; and the clearance control algorithm based on laser radar cannot respond to the tower sweep risk caused by complex working conditions in a timely manner due to control lag, so the safety protection capability of the prior art to deal with extremely complex working conditions is limited. Summary of the invention

[0003] The main purpose of the present invention is to provide a load reduction control method for a wind turbine based on complex working condition identification by laser radar, so as to solve the problem that the laser radar clearance control algorithm cannot promptly respond to the tower sweep risk caused by complex working conditions due to control lag.

[0004] In order to solve the above technical problems, the technical solution adopted by the present invention is: a load reduction control method for a wind turbine generator set based on complex working condition identification by laser radar, the method comprising: S1. Install a high-precision laser radar system on the wind turbine to monitor the three-dimensional wind speed distribution in the wind farm in real time with high frequency, covering instantaneous wind speed, wind direction and turbulence intensity parameters. The laser radar will transmit the collected data to the control system of the wind turbine in a timely manner to ensure the real-time and accuracy of the data; S2. Perform comprehensive preprocessing operations on the collected raw data. First, perform data cleaning to remove outliers and invalid data. Then, use advanced wavelet transform technology or other advanced filtering methods to remove noise interference while retaining useful wind farm information, providing a high-quality data foundation for subsequent complex working condition identification and load reduction control. S3. Use time series analysis methods to deeply explore the inherent laws and patterns of wind speed and wind direction changes over time. Through the analysis of time series data, it can keenly capture the key features of sudden changes in wind speed and rapid changes in wind direction, providing a strong basis for the accurate identification of complex working conditions; S4. Based on the historical data of wind speed and wind direction changes, define targeted fuzzy logic rules to accurately describe various types of complex working conditions; S5. Design a powerful adaptive fuzzy logic controller AFLC. The AFLC controller has the ability to automatically adjust its internal fuzzy rules and membership functions according to the actual situation of the current wind farm. AFLC can continuously improve its ability to identify complex working conditions by continuously learning and adapting to changes in the wind farm environment, providing a solid guarantee for the safe and stable operation of wind turbines. S6, AFLC successfully identified the complex operating conditions and immediately started the load reduction control logic, accurately adjusting the wind turbine operating parameters, including blade pitch angle and generator speed, according to the specific operating condition type identified; S7 and AFLC are not only responsible for identifying complex working conditions, but also can dynamically adjust the load reduction threshold according to the current wind farm conditions. When the detected wind speed change or wind direction change exceeds the dynamic threshold, the load reduction program is immediately started to ensure that the wind turbine can respond to changes in complex working conditions in a timely manner and ensure its safe and stable operation; When the system detects complex working conditions and reaches the dynamic threshold, it automatically implements load reduction measures; S8. During the load shedding control process, the operating data of the wind turbines, including vibration levels and power output parameters, are continuously collected. Through in-depth analysis of these data, the actual effect of the load shedding control is accurately evaluated, and each evaluation result is recorded in detail to provide valuable data support for subsequent analysis and optimization; S9. According to the actual results of each load reduction control, the fuzzy rules and membership functions in AFLC are adjusted and optimized in a targeted manner.

[0005] In the preferred solution, the specific steps of identifying complex working conditions by the time series analysis method in step S3 are: S31. Collect and pre-process wind speed and wind direction time series data, specifically: use the laser radar system installed on the wind turbine to continuously monitor the wind speed and wind direction in the wind farm, collect time series data at a frequency of 10-20 times per second, and record the wind speed value. and wind direction angle , the time span is ; The wavelet threshold denoising method is used to denoise the collected wind speed and wind direction time series data; Wind speed time series Perform wavelet transform to obtain wavelet coefficients ; The original signal is , after wavelet transform, ,in is the wavelet basis function, Indicates scale, Indicates time position; Choosing an appropriate threshold , process the wavelet coefficients, and the processed wavelet coefficients for: ; The denoised wind speed signal is obtained by inverse wavelet transform ; For wind direction time series The same method is used for denoising to obtain the denoised wind direction signal. ; S32, construct a fractional order chaotic time series model, specifically: for the time series , the fractional difference is defined as: ; in, , is the gamma function, and the denoised wind speed time series is calculated and wind direction time series Fractional difference of and ; Fractional differencing is used to capture the long memory and nonlinear characteristics of time series; Among them, the wind speed time series , construct the following fractional-order chaotic time series model: ; in, is the wind speed time delay, is the embedding dimension of wind speed, is the wind speed fractional parameter, is a nonlinear function, which is approximated by a neural network method; For wind direction time series , construct a similar fractional-order chaotic time series model: ; in, is the wind direction time delay, is the embedding dimension of wind direction, is the wind direction fractional parameter, is a nonlinear function; The fractional order chaotic time series model accurately describes the inherent laws and patterns of wind speed and wind direction changing with time; S33, calculate the fractional correlation dimension and fractional Lyapunov exponent key features, specifically: use the Grassberger-Procaccia algorithm to calculate the fractional correlation dimension , for the wind speed time series , define the correlation function for: ; in, is the number of wind speed data points, is the Heaviside function, when hour, ,otherwise ; Correlation dimension Calculated by the following formula: ; Wind direction time series The same method is used to calculate the fractional correlation dimension ; The fractional-order Lyapunov exponent calculation method includes: For the fractional-order chaotic system , calculate the fractional Lyapunov exponent , using the fractional-order extended version of the Wolf algorithm, it is calculated by tracking the trajectory changes of the system in the phase space; Includes: Selecting an initial point In phase space, calculate its The later evolution point ; Find with The nearest point , calculate the distance between them ; As time goes by, calculation ,in is from The point from which evolution comes; Calculate the average exponential growth rate ,For fractional-order systems, it is necessary to consider the influence of fractional-order derivatives and make appropriate adjustments to the above steps; The same calculation is performed on the fractional-order chaotic system of wind direction time series, and the fractional-order Lyapunov exponent is obtained: ; The fractional Lyapunov exponent is used to determine the chaotic characteristics and stability of the system; S34: Determine whether a complex working condition occurs according to the set threshold, specifically: set the wind speed fractional correlation dimension threshold according to historical data and experience and fractional Lyapunov exponent threshold ; Set the wind direction fractional correlation dimension threshold and fractional Lyapunov exponent threshold ; Complex working condition judgment: When the calculated wind speed fractional correlation dimension Exceeding the threshold or fractional Lyapunov exponent Exceeding the threshold When , it is considered that the wind speed presents complex working conditions; When the wind direction fractional correlation dimension Exceeding the threshold or fractional Lyapunov exponent Exceeding the threshold When , it is considered that the wind direction presents a complex working condition; If complex operating conditions occur in both wind speed and wind direction at the same time, it is judged that the entire wind turbine is in a complex operating condition.

[0006] In the preferred solution, the specific steps of defining the complex working condition fuzzy logic rules in step S4 are: S41, collect and pre-process historical data of wind speed and wind direction, and perform partition processing, specifically: collect historical data of wind speed and wind direction of wind turbines in different operating time periods, including wind speed values and wind direction angle ,At the same time, the corresponding wind turbine operating status information, such as vibration amplitude, power output, is recorded; The collected wind speed and direction data are partitioned and processed using equal-spaced partitions or adaptive partitioning methods based on data distribution; S42, constructing fuzzy sets of wind speed and wind direction, and defining a membership function, specifically: constructing a fuzzy set of wind speed for wind speed partitions; Define three fuzzy sets: low wind speed, medium wind speed and high wind speed; The speed value is , then its membership function belonging to the low wind speed fuzzy set is defined as: ; The membership function of the wind speed fuzzy set is: ; The membership function of the high wind speed fuzzy set is defined similarly; Through the construction of fuzzy sets, the precise wind speed value is converted into a fuzzy language description, and the recognition and judgment of wind conditions provide a basis for the subsequent definition of fuzzy logic rules. For wind direction partitioning, a fuzzy set of wind directions is constructed, including east wind (East_Wind), south wind (South_Wind), west wind (West_Wind) and north wind (North_Wind); Assume the wind direction angle is , then the membership function of the Dongfeng fuzzy set can be defined as: ; The membership functions of other wind direction fuzzy sets are defined similarly; S43, construct a fuzzy rough set information system, calculate the lower approximation and the upper approximation, specifically: Building an information system ,in is the domain, i.e. the set of all data samples; is a set of attributes, including wind speed attributes and wind direction attributes; is a collection of attribute values; is an information function that assigns a value to each attribute for each object; For a subset of attributes , defining the indiscernibility relation for: ; That is, if two objects have a subset of attributes have the same value on , then they are It is indistinguishable under the relation; For a fuzzy concept complex working condition and attribute subsets , the lower approximation and the upper approximation They are defined as: ; ; in, Is an object Fuzzy concept The membership degree of Is an object In the indiscernible relationship The equivalence class under ; S43, defining quantum logic gates, and generating quantum logic fuzzy rules according to the fuzzy rough set results and the fuzzy set, specifically: defining a set of quantum logic gates, used to represent the operations in the fuzzy logic rules, including a quantum AND gate (Quantum_AND) and a quantum OR gate (Quantum_OR); Assume two quantum states and , the quantum AND gate can be defined as: ; For the quantum OR gate, it can be defined as: ; According to the lower approximation and upper approximation calculated by fuzzy rough sets, as well as the fuzzy sets of wind speed and wind direction, quantum logic fuzzy rules are generated; The wind speed belongs to the high wind speed fuzzy set and the wind direction belongs to the east wind fuzzy set. It is considered that complex working conditions may occur, which can be expressed by quantum logic rules as follows: ; in, represents a quantum NOT gate, Indicates a quantum state where complex operating conditions may occur.

[0007] In the preferred embodiment, the specific steps of the AFLC controller in step S51 are: S51. Detailed classification of complex working conditions, including conditions of drastic wind speed changes, sudden wind direction changes, and extreme negative shear conditions. For conditions of drastic wind speed changes, define the wind speed change rate as exceeding a certain threshold. As the judgment condition of this working condition, When is the wind speed change, is the time interval; S52: For sudden wind direction changes, define the wind direction change angle exceeding the threshold in a short period of time. ,when When is the angle of wind direction change; For extreme negative shear conditions, the ratio of wind speed differences at different heights to the average wind speed exceeds the threshold. To judge, that is ,in and are the wind speeds at the top and bottom of the wind turbine, is the average wind speed; S53, mapping different working conditions with the operating parameters of the wind turbine, for the working conditions with drastic changes in wind speed, mainly adjusting the blade pitch angle and generator speed To reduce the load on wind turbines; A mapping relationship is established. When the wind speed change rate is large, the blade pitch angle is appropriately reduced and the generator speed is reduced, that is, , ,in and are the initial pitch angle and generator speed, and is the adjustment factor; For sudden changes in wind direction, the yaw angle of the wind turbine can be adjusted To reduce the lateral force of the wind rotor, ,in is the initial yaw angle, is the adjustment factor; For extreme negative shear conditions, the power output curve of the wind turbine can be optimized, and the load can be reduced by adjusting the excitation current and other parameters of the generator. The specific mapping relationship can be determined according to the actual situation. S54, AFLC includes input and output: input includes wind speed, wind direction, wind speed change rate, wind direction change angle wind field parameters, and current operating parameters of the wind turbine, including blade pitch angle, generator speed, and yaw angle; The output is the adjusted wind turbine operating parameters, including the adjusted blade pitch angle, generator speed, and yaw angle; S55. According to different complex working conditions and operating parameter ranges, a series of fuzzy rules are defined, specifically: “If the wind speed is high and the rate of change is large, then the blade pitch angle should be reduced much”, expressed in fuzzy language as “IF wind_speed is High AND wind_speed_change_rate is Large THEN blade_pitch_angleis Decrease_Much”; Determine the membership function of fuzzy rules through expert experience and historical data; The membership function for “high wind speed” can be defined as ,in is the wind speed, is the central value of high wind speed, is the standard deviation; S56, executing fuzzy reasoning and defuzzification of AFLC, performing fuzzy reasoning through fuzzy rules according to the input wind field parameters and operation parameters, and obtaining a fuzzy output result; Then, the centroid method is used to defuzzify the fuzzy output into an accurate operating parameter adjustment value. The centroid method formula is: ,in is the number of output fuzzy sets, It is The membership degree of the output fuzzy set is It is The central value of the output fuzzy set.

[0008] In the preferred solution, in steps S6-S7, the AFLC successfully identifies the complex working condition, and the specific method of the control logic of the AFLC is: S61, select appropriate wind field parameters as the basis for dynamic threshold adjustment, and the wind speed change rate, wind direction change angle, and wind turbine vibration amplitude parameters can be selected. and wind direction change angle As the main parameter for dynamic threshold adjustment; S62, constructing a dynamic threshold model, using a nonlinear function to construct the dynamic threshold model, ,in , , , are model parameters, which can be obtained by fitting historical data; By building a dynamic threshold model, the load reduction threshold can be dynamically adjusted according to the current wind farm conditions; S63, real-time monitoring of wind farm parameters and wind turbine operating status, continuous collection of wind speed change rate, wind direction change angle, vibration amplitude wind farm parameters, and wind turbine output power, speed operating status parameters; S64, update the model parameters according to the monitoring data, adopt the adaptive recursive least squares algorithm, and continuously update the parameters of the dynamic threshold model according to the real-time monitoring data. The recursive least squares formula is: , , ; in is the model parameter vector, is the time step, is the error between the actual value and the predicted value, is the input vector, It's the forgetting factor. is the identity matrix; S65. Calculate a dynamic threshold value according to the updated model parameters, and immediately start a load reduction procedure when the detected wind farm parameter change exceeds the dynamic threshold value.

[0009] In the preferred solution, in step S7, when the system detects a complex working condition and reaches a dynamic threshold, the specific method of automatically executing load reduction measures is as follows: S71: When the system detects a complex operating condition and reaches a dynamic threshold, it automatically executes load reduction measures and adjusts the operating parameters of the wind turbine according to the control logic designed in steps S51-S54. At the same time, other load reduction measures can be taken, including lowering the output power of wind turbines and adjusting the speed of wind rotors; S72. When adjusting operating parameters, use a gradual approach to avoid sudden and large adjustments that may cause impacts on wind turbines; For the adjustment of the blade pitch angle, a linear gradient method can be used, that is, ,in It's in time The blade pitch angle at is the initial pitch angle, is the adjustment rate; S73. Continuously and closely monitor the operating status of the wind turbine generator set, using various sensors, including vibration sensors, temperature sensors, and speed sensors, to monitor the operating status parameters of the wind turbine generator set in real time; The monitoring data will be transmitted to the control center so that the operation status of the wind turbine can be understood in a timely manner.

[0010] In the preferred solution, the specific method for evaluating the load reduction control effect in step S8 is: S81. Define evaluation indicators. Based on information geometry metrics, define evaluation indicators for load reduction control effects, including the vibration level reduction rate. and power output stability ; Vibration level reduction rate It can be defined as: ; in, is the probability density function of vibration acceleration before load reduction control, is the probability density function of vibration acceleration after load reduction control; Power output stability It can be defined as: ; S82. Calculate the actual effect of load reduction control based on the defined evaluation indicators, including the vibration level reduction rate Larger and more stable power output A higher value indicates better load shedding control effect; S83. Record the evaluation results of each load reduction control in detail, including the vibration level reduction rate, power output stability and other relevant parameters. The results may be recorded in a database or file system for subsequent analysis and query. S84. Use quantum annealing algorithm to analyze the recorded evaluation data to find the optimal load shedding control strategy. Quantum annealing algorithm is an optimization algorithm based on the principle of quantum mechanics, which can find the global optimal solution in a complex search space. First, the load reduction control problem is transformed into an optimization problem. The objective function can be the weighted sum of maximizing the vibration level reduction rate and the power output stability, that is, ,in and is the weight coefficient; Then, the quantum annealing algorithm is used for optimization. The core of the quantum annealing algorithm is to gradually reduce the energy of the system by simulating the annealing process of the quantum system and finally reach the global optimal state. In the load shedding control problem, the energy of the system can be regarded as the value of the objective function. By adjusting the operating parameters of the wind turbine, the energy of the system can be reduced, thereby finding the optimal load shedding control strategy.

[0011] Mathematical formula: The Hamiltonian in the quantum annealing algorithm can be expressed as ,in is the initial Hamiltonian, is the Hamiltonian associated with the optimization problem. In the load shedding control problem, can be a simple energy function of the initial state, The objective function can be The negative number of .

[0012] In the preferred embodiment, the specific method for adjusting and optimizing the fuzzy rules and membership functions in AFLC in step S9 is as follows: S91. Collect and organize the wind turbine operation data of each load reduction control and establish performance evaluation indicators to measure the performance of AFLC; S92. Using the chaos immune algorithm to optimize the fuzzy rules and membership functions of AFLC, including encoding, initializing the population, calculating fitness, and chaos immune operation steps, specifically: Encode the fuzzy rules and membership functions of AFLC into a chromosome, and use binary coding or real number coding to encode the condition part and conclusion part of each fuzzy rule and the parameters of the membership function into the genes of the chromosome; AFLC includes at least three input variables: wind speed ,wind direction , vibration acceleration , the output variable is the blade pitch angle ; Fuzzy rules can be expressed as “IF is AND is AND is THEN is ",in , , , They are the fuzzy sets of wind speed, wind direction, vibration acceleration and blade pitch angle; A set of initial chromosomes is randomly generated, representing different combinations of fuzzy rules and membership functions; For each chromosome, calculate its fitness value, which can be determined based on the established performance evaluation index. The fitness function is defined as ,in , , are weight coefficients, which respectively indicate the importance of load reduction, stability, and response speed in the fitness function; The chromosomes are decoded into fuzzy rules and membership functions, and then the collected wind turbine operation data is input to simulate the operation of AFLC and calculate the load reduction degree. ,stability , Response speed , and then get the fitness value; The chaos immune algorithm is used for optimization, including chaos search and immune operation.

[0013] S92, applying the optimized fuzzy rules and membership functions to AFLC, the specific method is: according to the encoding method, converting the gene sequence in the optimal chromosome into specific fuzzy rules and membership function parameters; The AFLC can better adapt to different complex working conditions according to the optimized fuzzy rules and membership functions, and improve the load reduction control effect and safety of wind turbines.

[0014] The present invention provides a load reduction control method for a wind turbine based on complex working condition identification by laser radar, with enhanced real-time and accuracy: by installing a high-precision laser radar system and monitoring the three-dimensional wind speed distribution in the wind farm in real time at a high frequency, the real-time and accuracy of the data are ensured, and the ability to respond immediately to complex working conditions is improved.

[0015] Improved data quality: Comprehensive use of data cleaning, wavelet transform and other advanced filtering technologies effectively removes noise interference, retains valuable wind field information, and provides a high-quality data foundation for subsequent identification of complex working conditions.

[0016] Accurate identification of complex working conditions: Using time series analysis methods, we can deeply explore the inherent laws and patterns of wind speed and wind direction changes over time, and can keenly capture key features such as sudden changes in wind speed and rapid changes in wind direction, thereby improving the accuracy and reliability of complex working condition identification.

[0017] Adaptive control enhancement: An adaptive fuzzy logic controller (AFLC) is designed, which can automatically adjust its internal fuzzy rules and membership functions according to the actual situation of the current wind farm. By continuously learning and adapting to changes in the wind farm environment, it continuously improves the ability to identify complex working conditions.

[0018] Optimization of load shedding control strategy: AFLC is not only responsible for identifying complex working conditions, but also dynamically adjusts the load shedding threshold according to the current wind farm conditions, ensuring that the wind turbine can respond to changes in complex working conditions in a timely manner and ensure its safe and stable operation. In addition, effective load shedding control is achieved by accurately adjusting the operating parameters of the wind turbine, such as blade pitch angle and generator speed.

[0019] Continuous performance monitoring and optimization: During the load shedding control process, the wind turbine operating data is continuously collected and analyzed in depth to accurately evaluate the actual effect of the load shedding control. Each evaluation result is recorded in detail, providing valuable data support for subsequent analysis and optimization, thereby continuously improving the overall performance and adaptability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The present invention will be further described below in conjunction with the accompanying drawings and embodiments: Figure 1 is a flow chart of the load shedding control method of the present invention; DETAILED DESCRIPTION Example 1 like Figure 1 As shown, a load reduction control method for a wind turbine generator set based on complex working condition identification by laser radar, the method comprising: S1. Install a high-precision laser radar system on the wind turbine to monitor the three-dimensional wind speed distribution in the wind farm in real time with high frequency, covering instantaneous wind speed, wind direction and turbulence intensity parameters. The laser radar will transmit the collected data to the control system of the wind turbine in a timely manner to ensure the real-time and accuracy of the data; S2. Perform comprehensive preprocessing operations on the collected raw data. First, perform data cleaning to remove outliers and invalid data. Then, use advanced wavelet transform technology or other advanced filtering methods to remove noise interference while retaining useful wind farm information, providing a high-quality data foundation for subsequent complex working condition identification and load reduction control. S3. Use time series analysis methods to deeply explore the inherent laws and patterns of wind speed and wind direction changes over time. Through the analysis of time series data, it can keenly capture the key features of sudden changes in wind speed and rapid changes in wind direction, providing a strong basis for the accurate identification of complex working conditions; S4. Based on the historical data of wind speed and wind direction changes, define targeted fuzzy logic rules to accurately describe various types of complex working conditions; S5. Design a powerful adaptive fuzzy logic controller AFLC. The AFLC controller has the ability to automatically adjust its internal fuzzy rules and membership functions according to the actual situation of the current wind farm. AFLC can continuously improve its ability to identify complex working conditions by continuously learning and adapting to changes in the wind farm environment, providing a solid guarantee for the safe and stable operation of wind turbines. S6, AFLC successfully identified the complex operating conditions and immediately started the load reduction control logic, accurately adjusting the wind turbine operating parameters, including blade pitch angle and generator speed, according to the specific operating condition type identified; S7 and AFLC are not only responsible for identifying complex working conditions, but also can dynamically adjust the load reduction threshold according to the current wind farm conditions. When the detected wind speed change or wind direction change exceeds the dynamic threshold, the load reduction program is immediately started to ensure that the wind turbine can respond to changes in complex working conditions in a timely manner and ensure its safe and stable operation; When the system detects complex working conditions and reaches the dynamic threshold, it automatically implements load reduction measures; S8. During the load shedding control process, the operating data of the wind turbines, including vibration levels and power output parameters, are continuously collected. Through in-depth analysis of these data, the actual effect of the load shedding control is accurately evaluated, and each evaluation result is recorded in detail to provide valuable data support for subsequent analysis and optimization; S9. According to the actual results of each load reduction control, the fuzzy rules and membership functions in AFLC are adjusted and optimized in a targeted manner.

[0021] Example 2 Further illustrate with reference to Example 1, Figure 1 As shown, the specific steps are as follows: Collect and preprocess wind speed and direction time series data.

[0022] Construct a fractional-order chaotic time series model.

[0023] Compute key features such as fractional correlation dimension and fractional Lyapunov exponents.

[0024] Determine whether a complex working condition occurs based on the set threshold.

[0025] The specific steps of the time series analysis method in step S3 to identify complex working conditions are: S31. Collect and pre-process wind speed and wind direction time series data, specifically: use the laser radar system installed on the wind turbine to continuously monitor the wind speed and wind direction in the wind farm, collect time series data at a frequency of 10-20 times per second, and record the wind speed value. and wind direction angle , the time span is ; The wavelet threshold denoising method is used to denoise the collected wind speed and wind direction time series data; Wind speed time series Perform wavelet transform to obtain wavelet coefficients ; The original signal is , after wavelet transform, ,in is the wavelet basis function, Indicates scale, Indicates time position; Choosing an appropriate threshold , process the wavelet coefficients, and the processed wavelet coefficients for: ; The denoised wind speed signal is obtained by inverse wavelet transform ; For wind direction time series The same method is used for denoising to obtain the denoised wind direction signal. ; Remove noise interference, improve data quality, and provide an accurate data basis for subsequent analysis.

[0026] S32, construct a fractional order chaotic time series model, specifically: for the time series , the fractional difference is defined as: ; in, , is the gamma function, and the denoised wind speed time series is calculated and wind direction time series Fractional difference of and ; Uses: Fractional differencing can better capture the long memory and nonlinear characteristics of time series.

[0027] Among them, the wind speed time series , construct the following fractional-order chaotic time series model: ; in, is the wind speed time delay, is the embedding dimension of wind speed, is the wind speed fractional parameter, is a nonlinear function, which is approximated by a neural network method; For wind direction time series , construct a similar fractional-order chaotic time series model: ; in, is the wind direction time delay, is the embedding dimension of wind direction, is the wind direction fractional parameter, is a nonlinear function; Usefulness: This model can more accurately describe the inherent laws and patterns of wind speed and direction changes over time.

[0028] S33, calculate the fractional correlation dimension and fractional Lyapunov exponent key features, specifically: use the Grassberger-Procaccia algorithm to calculate the fractional correlation dimension , for the wind speed time series , define the correlation function for: ; in, is the number of wind speed data points, is the Heaviside function, when hour, ,otherwise ; Correlation dimension Calculated by the following formula: ; Wind direction time series The same method is used to calculate the fractional correlation dimension ; The fractional correlation dimension can reflect the complexity and chaotic characteristics of the time series in the fractional sense, and is helpful to capture the key characteristics of wind speed and direction.

[0029] The fractional-order Lyapunov exponent calculation method includes: For the fractional-order chaotic system , calculate the fractional Lyapunov exponent , using the fractional-order extended version of the Wolf algorithm, it is calculated by tracking the trajectory changes of the system in the phase space; Includes: Selecting an initial point In phase space, calculate its The later evolution point ; Find with The nearest point , calculate the distance between them ; As time goes by, calculation ,in is from The point from which evolution comes; Calculate the average exponential growth rate ,For fractional-order systems, it is necessary to consider the influence of fractional-order derivatives and make appropriate adjustments to the above steps; The same calculation is performed on the fractional-order chaotic system of wind direction time series, and the fractional-order Lyapunov exponent is obtained: ; The fractional Lyapunov exponent can determine the chaotic characteristics and stability of the system, and is of great significance for identifying sudden changes and rapid transitions in wind speed and direction.

[0030] S34: Determine whether a complex working condition occurs according to the set threshold, specifically: set the wind speed fractional correlation dimension threshold according to historical data and experience and fractional Lyapunov exponent threshold ; Set the wind direction fractional correlation dimension threshold and fractional Lyapunov exponent threshold ; Complex working condition judgment: When the calculated wind speed fractional correlation dimension Exceeding the threshold or fractional Lyapunov exponent Exceeding the threshold When , it is considered that the wind speed presents complex working conditions; When the wind direction fractional correlation dimension Exceeding the threshold or fractional Lyapunov exponent Exceeding the threshold When , it is considered that the wind direction presents a complex working condition; If complex operating conditions occur in both wind speed and wind direction at the same time, it is judged that the entire wind turbine is in a complex operating condition.

[0031] Example 3 Further illustrate with reference to Example 1, Figure 1 As shown, the specific steps are as follows: Collect and preprocess historical data of wind speed and direction for partition processing.

[0032] Construct fuzzy sets of wind speed and wind direction and define membership functions.

[0033] Construct a fuzzy rough set information system and calculate the lower approximation and upper approximation.

[0034] Define quantum logic gates and generate quantum logic fuzzy rules based on fuzzy rough set results and fuzzy sets.

[0035] The specific steps of defining the fuzzy logic rules for complex working conditions in step S4 are: S41, collect and pre-process historical data of wind speed and wind direction, and perform partition processing, specifically: collect historical data of wind speed and wind direction of wind turbines in different operating time periods, including wind speed values and wind direction angle ,At the same time, the corresponding wind turbine operating status information, such as vibration amplitude, power output, is recorded; The time span should cover a variety of wind conditions to ensure data diversity and representativeness.

[0036] For example, in a wind farm, data on wind turbines in different seasons and weather conditions are collected over a year, including light winds and variable wind directions in spring, strong winds and stable wind directions in summer, strong winds and rotating wind directions in autumn, and cold and strong winds in winter.

[0037] The collected wind speed and direction data are partitioned and processed using equal-spaced partitions or adaptive partitioning methods based on data distribution; For example, wind speed data can be divided into several intervals, such as , , , ,in is the division threshold. Assume that according to the actual data, the wind speed is divided into three intervals: low wind speed interval , medium wind speed range , high wind speed range .

[0038] Similar partitioning is also performed on the wind direction angle data, such as , , , For example, the wind direction angle is divided into four intervals: east wind interval , South Wind Zone , westerly zone , North Wind Zone .

[0039] S42, constructing fuzzy sets of wind speed and wind direction, and defining a membership function, specifically: constructing a fuzzy set of wind speed for wind speed partitions; Define three fuzzy sets: low wind speed, medium wind speed and high wind speed; The speed value is , then its membership function belonging to the low wind speed fuzzy set is defined as: ; The membership function of the wind speed fuzzy set is: ; The membership function of the high wind speed fuzzy set is defined similarly; Practical application scenario: In a specific wind farm, when the wind speed is 4m / s, according to the defined membership function, it completely belongs to the low wind speed fuzzy set. When the wind speed is 7m / s, its membership in the low wind speed fuzzy set is 0.6, and its membership in the medium wind speed fuzzy set is 0.4.

[0040] Through the construction of fuzzy sets, the precise wind speed value is converted into a fuzzy language description, and the recognition and judgment of wind conditions provide a basis for the subsequent definition of fuzzy logic rules. For wind direction partitioning, a fuzzy set of wind directions is constructed, including east wind (East_Wind), south wind (South_Wind), west wind (West_Wind) and north wind (North_Wind); Assume the wind direction angle is , then the membership function of the Dongfeng fuzzy set can be defined as: ; The membership functions of other wind direction fuzzy sets are defined similarly; Practical application scenario: When the wind direction angle is 45 degrees, it completely belongs to the east wind fuzzy set. When the wind direction angle is 135 degrees, it completely belongs to the south wind fuzzy set.

[0041] S43, construct a fuzzy rough set information system, calculate the lower approximation and the upper approximation, specifically: Building an information system ,in is the domain, i.e. the set of all data samples; is a set of attributes, including wind speed attributes and wind direction attributes; is a collection of attribute values; is an information function that assigns a value to each attribute for each object; For example, suppose there are 1000 data samples, each of which contains two attribute values: wind speed and wind direction. Then the domain There are 1000 elements. , a collection of attribute values Determined according to the scope of actual data.

[0042] For a subset of attributes , defining the indiscernibility relation for: ; That is, if two objects have a subset of attributes have the same value on , then they are It is indistinguishable under the relation; If the wind speed and wind direction of two data samples are very close, they may be considered as similar cases under the indistinguishability relationship, which helps to determine the possible range of complex working conditions.

[0043] For a fuzzy concept complex working condition and attribute subsets , the lower approximation and the upper approximation They are defined as: ; ; in, Is an object A fuzzy concept The membership degree of Is an object In the indiscernible relationship The equivalence class under ; Practical application scenario: Assume that the fuzzy concept of complex working conditions is extremely high wind speed and drastic wind direction changes. By calculating the lower approximation and the upper approximation, it is possible to determine which data samples definitely belong to complex working conditions and which data samples may belong to complex working conditions. For example, if a data sample has a very high wind speed and a large wind direction change, and is in an indistinguishable relationship with other samples of similar conditions, then it may belong to the upper approximation set of complex working conditions.

[0044] S43, defining quantum logic gates, and generating quantum logic fuzzy rules according to the fuzzy rough set results and the fuzzy set, specifically: defining a set of quantum logic gates, used to represent the operations in the fuzzy logic rules, including a quantum AND gate (Quantum_AND) and a quantum OR gate (Quantum_OR); Assume two quantum states and , the quantum AND gate can be defined as: ; For the quantum OR gate, it can be defined as: ; In the identification of complex working conditions of wind turbines, the quantum AND gate can indicate that the wind speed and wind direction meet certain conditions at the same time and are considered complex working conditions. For example, when the wind speed belongs to the high wind speed fuzzy set and the wind direction belongs to the east wind fuzzy set, the operation of the quantum AND gate can determine whether this situation meets the conditions of the complex working condition.

[0045] According to the lower approximation and upper approximation calculated by fuzzy rough sets, as well as the fuzzy sets of wind speed and wind direction, quantum logic fuzzy rules are generated; The wind speed belongs to the high wind speed fuzzy set and the wind direction belongs to the east wind fuzzy set. It is considered that complex working conditions may occur, which can be expressed by quantum logic rules as follows: ; in, represents a quantum NOT gate, Indicates a quantum state where complex operating conditions may occur.

[0046] In a specific wind farm, if complex operating conditions are likely to occur based on quantum logic rules, the wind turbine control system can take load reduction measures in advance to ensure the safe operation of the wind turbine, for example, adjusting the blade pitch angle, reducing the generator speed, etc.

[0047] Example 4 Further illustrate with reference to Example 1, Figure 1 As shown, the specific steps of the AFLC controller in step S51 are: S51. Detailed classification of complex working conditions, including conditions of drastic wind speed changes, sudden wind direction changes, and extreme negative shear conditions. For conditions of drastic wind speed changes, define the wind speed change rate as exceeding a certain threshold. As the judgment condition of this working condition, When is the wind speed change, is the time interval; In an offshore wind farm, when strong winds suddenly hit, the wind speed increases from 10m / s to 25m / s in a short period of time, with a time interval of 5 seconds. Assume that the wind speed change rate threshold is set to is 3m / s², and the wind speed change rate calculated at this time is , just meeting the judgment conditions for the condition of drastic wind speed changes.

[0048] S52: For sudden wind direction changes, define the wind direction change angle exceeding the threshold in a short period of time. ,when When is the angle of wind direction change; In a mountainous wind farm, the wind direction was originally easterly. Due to the influence of the terrain, the wind direction suddenly changed to southerly within 10 minutes, and the wind direction change angle was 90 degrees. Assume that the wind direction change angle threshold is set to It is 60 degrees, which meets the judgment conditions for sudden change of wind direction.

[0049] For extreme negative shear conditions, the ratio of wind speed differences at different heights to the average wind speed exceeds the threshold. To judge, that is ,in and are the wind speeds at the top and bottom of the wind turbine, is the average wind speed; In a high-tower wind turbine, the wind speed at the top is measured to be 15 m / s, the wind speed at the bottom is 10 m / s, and the average wind speed is 12.5 m / s. Assume that the extreme negative shear condition threshold is set is 0.2, and the calculated ratio is , exceeds the threshold and is judged as an extreme negative shear condition.

[0050] S53, mapping different working conditions with the operating parameters of the wind turbine, for the working conditions with drastic changes in wind speed, mainly adjusting the blade pitch angle and generator speed To reduce the load on wind turbines; A mapping relationship is established. When the wind speed change rate is large, the blade pitch angle is appropriately reduced and the generator speed is reduced, that is, , ,in and are the initial pitch angle and generator speed, and is the adjustment factor; In the above offshore wind farm, under the condition of drastic wind speed changes, it is assumed that the initial blade pitch angle is 15 degrees, the generator speed is 1500rpm. After measurement and calculation, the wind speed change 15m / s, adjustment coefficient is 0.5, is 10. Then the adjusted blade pitch angle is calculated according to the formula: The adjusted generator speed is rpm.

[0051] For sudden changes in wind direction, the yaw angle of the wind turbine can be adjusted To reduce the lateral force of the wind rotor, ,in is the initial yaw angle, is the adjustment factor; In the case of sudden wind direction changes in mountainous wind farms, assuming that the initial yaw angle is 0 degrees, adjust the coefficient is 0.5. Since the wind direction changes by 90 degrees, the adjusted yaw angle calculated by the formula is Spend For extreme negative shear conditions, the power output curve of the wind turbine can be optimized, and the load can be reduced by adjusting the excitation current and other parameters of the generator. The specific mapping relationship can be determined according to the actual situation. In the extreme negative shear conditions of high-tower wind turbines, the excitation current of the generator can be adjusted according to the wind speed differences at different heights to make the power output more stable. For example, when the wind speed difference between the top and the bottom is large, the excitation current can be appropriately reduced to reduce the power output to reduce the load on the wind turbine.

[0052] S54, AFLC includes input and output: input includes wind speed, wind direction, wind speed change rate, wind direction change angle wind field parameters, and current operating parameters of the wind turbine, including blade pitch angle, generator speed, and yaw angle; At the control center of a large wind farm, AFLC receives data from wind speed sensors, wind direction sensors, wind turbine sensors and other equipment, including real-time wind speed of 12m / s, wind direction of northwest, wind speed change rate of 2m / s², wind direction change angle of 15 degrees, and current wind turbine blade pitch angle of 10 degrees, generator speed of 1200rpm, yaw angle of 30 degrees and other parameters.

[0053] The output is the adjusted wind turbine operating parameters, including the adjusted blade pitch angle, generator speed, and yaw angle; After fuzzy reasoning and defuzzification based on the input parameters, AFLC outputs the adjusted parameters such as blade pitch angle of 8 degrees, generator speed of 1150rpm, yaw angle of 35 degrees, etc., to achieve load reduction control of the wind turbine.

[0054] S55. According to different complex working conditions and operating parameter ranges, a series of fuzzy rules are defined, specifically: “If the wind speed is high and the rate of change is large, then the blade pitch angle should be reduced much”, expressed in fuzzy language as “IF wind_speed is High AND wind_speed_change_rate is Large THEN blade_pitch_angleis Decrease_Much”; In a strong wind weather, the wind speed reaches 20m / s and the wind speed change rate is 5m / s². According to the fuzzy rule, the wind speed is judged to be high and the wind speed change rate is judged to be large. Therefore, the AFLC outputs a decision that the blade pitch angle should be reduced a lot.

[0055] Determine the membership function of fuzzy rules through expert experience and historical data; The membership function for “high wind speed” can be defined as ,in is the wind speed, is the central value of high wind speed, is the standard deviation; By analyzing a large amount of historical data, the central value of high wind speed is determined is 18m / s, standard deviation When the real-time wind speed is 19m / s, the membership function is substituted to calculate the high wind speed, which further supports the judgment of the fuzzy rule.

[0056] S56, executing fuzzy reasoning and defuzzification of AFLC, performing fuzzy reasoning through fuzzy rules according to the input wind field parameters and operation parameters, and obtaining a fuzzy output result; The parameters such as wind speed, wind direction, wind speed change rate, wind direction change angle are used as inputs, and reasoning is performed according to fuzzy rules. For example, if the input wind speed is high and the wind direction change angle is large, the fuzzy output results such as the blade pitch angle should be reduced and the yaw angle should be adjusted can be obtained through fuzzy rule reasoning.

[0057] Then the centroid method is used to defuzzify the fuzzy output into an accurate operating parameter adjustment value. The centroid method formula is: ,in is the number of output fuzzy sets, It is The membership degree of the output fuzzy set is It is The central value of the output fuzzy set.

[0058] AFLC can automatically adjust its internal fuzzy rules and membership functions according to the actual situation of the current wind farm, realize precise control of the operating parameters of wind turbines, and improve the effect and adaptability of load reduction control.

[0059] Example 5 Further illustrate with reference to Example 1, Figure 1 As shown, in steps S6-S7, the AFLC successfully identifies the complex working condition, and the specific method of the control logic of the AFLC is: S61, select appropriate wind field parameters as the basis for dynamic threshold adjustment, and the wind speed change rate, wind direction change angle, and wind turbine vibration amplitude parameters can be selected. and wind direction change angle As the main parameter for dynamic threshold adjustment; S62, constructing a dynamic threshold model, using a nonlinear function to construct the dynamic threshold model, ,in , , , are model parameters, which can be obtained by fitting historical data; Uses: By constructing a dynamic threshold model, the load reduction threshold can be dynamically adjusted according to the current wind farm conditions, thereby improving the wind turbine's ability to respond to complex working conditions. S63, real-time monitoring of wind farm parameters and wind turbine operating status, continuous collection of wind speed change rate, wind direction change angle, vibration amplitude wind farm parameters, and wind turbine output power, speed operating status parameters; S64, update the model parameters according to the monitoring data, adopt the adaptive recursive least squares algorithm, and continuously update the parameters of the dynamic threshold model according to the real-time monitoring data. The recursive least squares formula is: , , ; in is the model parameter vector, is the time step, is the error between the actual value and the predicted value, is the input vector, It's the forgetting factor. is the identity matrix; S65. Calculate a dynamic threshold value according to the updated model parameters, and immediately start a load reduction procedure when the detected wind farm parameter change exceeds the dynamic threshold value.

[0060] Uses: By updating model parameters in real time and dynamically adjusting the load reduction threshold, it can ensure that wind turbines can respond to changes in complex working conditions in a timely manner and ensure their safe and stable operation.

[0061] In the preferred solution, in step S7, when the system detects a complex working condition and reaches a dynamic threshold, the specific method of automatically executing load reduction measures is as follows: S71: When the system detects a complex operating condition and reaches a dynamic threshold, it automatically executes load reduction measures and adjusts the operating parameters of the wind turbine according to the control logic designed in steps S51-S54. At the same time, other load reduction measures can be taken, including lowering the output power of wind turbines and adjusting the speed of wind rotors; S72. When adjusting operating parameters, use a gradual approach to avoid sudden and large adjustments that may cause impacts on wind turbines; For the adjustment of the blade pitch angle, a linear gradient method can be used, that is, ,in It's in time The blade pitch angle at is the initial pitch angle, is the adjustment rate; Uses: Implementing load reduction measures can effectively reduce the load on wind turbines and ensure their safe operation under complex working conditions. The gradual approach can ensure a smooth transition during the load reduction process and avoid unnecessary impact and damage to wind turbines.

[0062] S73. Continuously and closely monitor the operating status of the wind turbine generator set, using various sensors, including vibration sensors, temperature sensors, and speed sensors, to monitor the operating status parameters of the wind turbine generator set in real time; The monitoring data will be transmitted to the control center so that the operation status of the wind turbine can be understood in a timely manner.

[0063] Continue to closely monitor the operating status of wind turbines.

[0064] Use various sensors, such as vibration sensors, temperature sensors, speed sensors, etc., to monitor the operating status parameters of wind turbines in real time.

[0065] A real-time monitoring system is established to transmit monitoring data to the control center so as to timely understand the operating status of the wind turbines.

[0066] Make feedback adjustments based on the monitoring results.

[0067] If the wind turbine operating status is monitored to be abnormal, such as excessive vibration amplitude, high temperature, unstable speed, etc., feedback adjustment will be made immediately.

[0068] The intensity and method of load reduction measures can be adjusted, or other emergency measures can be taken, such as shutdown for maintenance.

[0069] Example 6 Further illustrate with reference to Example 1, Figure 1 As shown, the specific method for evaluating the load reduction control effect in step S8 is: S81. Define evaluation indicators. Based on information geometry metrics, define evaluation indicators for load reduction control effects, including the vibration level reduction rate. and power output stability ; Vibration level reduction rate It can be defined as: ; in, is the probability density function of vibration acceleration before load reduction control, is the probability density function of vibration acceleration after load reduction control; Power output stability It can be defined as: ; S82. Calculate the actual effect of load reduction control based on the defined evaluation indicators, including the vibration level reduction rate Larger and more stable power output A higher value indicates better load shedding control effect; Usefulness: By defining clear evaluation indicators, the actual effect of load shedding control can be quantitatively evaluated, providing a basis for subsequent analysis and optimization.

[0070] S83. Record the evaluation results of each load reduction control in detail, including the vibration level reduction rate, power output stability and other relevant parameters. The results may be recorded in a database or file system for subsequent analysis and query. S84. Use quantum annealing algorithm to analyze the recorded evaluation data to find the optimal load shedding control strategy. Quantum annealing algorithm is an optimization algorithm based on the principle of quantum mechanics, which can find the global optimal solution in a complex search space. First, the load reduction control problem is transformed into an optimization problem. The objective function can be the weighted sum of maximizing the vibration level reduction rate and the power output stability, that is, ,in and is the weight coefficient; Then, the quantum annealing algorithm is used for optimization. The core of the quantum annealing algorithm is to gradually reduce the energy of the system by simulating the annealing process of the quantum system and finally reach the global optimal state. In the load shedding control problem, the energy of the system can be regarded as the value of the objective function. By adjusting the operating parameters of the wind turbine, the energy of the system can be reduced, thereby finding the optimal load shedding control strategy.

[0071] Mathematical formula: The Hamiltonian in the quantum annealing algorithm can be expressed as ,in is the initial Hamiltonian, is the Hamiltonian associated with the optimization problem. In the load shedding control problem, can be a simple energy function of the initial state, The objective function can be Negative number of .

[0072] Recording the evaluation results can provide data support for subsequent analysis and optimization. The quantum annealing algorithm can find the optimal load reduction control strategy from a large amount of evaluation data to improve the performance and safety of wind turbines.

[0073] Example 7 Further illustrate with reference to Example 1, Figure 1 As shown, the specific steps are as follows: Collect and organize the wind turbine operation data for each load reduction control.

[0074] Performance evaluation indicators are established to measure the performance of AFLC.

[0075] The chaotic immune algorithm is used to optimize the fuzzy rules and membership functions of AFLC, including the steps of encoding, initializing population, calculating fitness, and chaotic immune operation.

[0076] The optimized fuzzy rules and membership functions are applied to AFLC.

[0077] The specific method for adjusting and optimizing the fuzzy rules and membership functions in AFLC in step S9 is as follows: Collect wind turbine operation data during each load reduction control process, including wind speed, wind direction, vibration level, power output and other parameters, as well as the corresponding AFLC output control parameter adjustment values. For example, over a period of time, record the wind speed under different wind conditions. ,wind direction , vibration acceleration , Power Output , and the blade pitch angle after AFLC adjustment , generator speed wait; The collected data are sorted and classified, and grouped according to different complex working conditions. The data of the working condition with drastic wind speed changes are grouped into one group, the data of the working condition with sudden wind direction changes are grouped into one group, the data of the working condition with extreme negative shear conditions are grouped into one group, etc.

[0078] S91. Collect and organize the wind turbine operation data of each load reduction control and establish performance evaluation indicators to measure the performance of AFLC; Defining load reduction indicators is the ratio of the load difference of the wind turbine before and after load reduction control to the load before load reduction control, that is, ,in is the load before load shedding control, It is the load after load reduction control.

[0079] Stability index It can be measured by the standard deviation of vibration acceleration; Right now ,in is the number of data points, is the vibration acceleration in time The value of is the average value of the vibration acceleration Response speed index It can be measured by the time from the detection of complex operating conditions to the time the wind turbine reaches a stable state.

[0080] These indicators can quantitatively evaluate the performance of AFLC and provide a basis for subsequent optimization.

[0081] S92. Using the chaos immune algorithm to optimize the fuzzy rules and membership functions of AFLC, including encoding, initializing the population, calculating fitness, and chaos immune operation steps, specifically: Encode the fuzzy rules and membership functions of AFLC into a chromosome, and use binary coding or real number coding to encode the condition part and conclusion part of each fuzzy rule and the parameters of the membership function into the genes of the chromosome; AFLC includes at least three input variables: wind speed ,wind direction , vibration acceleration , the output variable is the blade pitch angle ; Fuzzy rules can be expressed as “IF is AND is AND is THEN is ",in , , , They are the fuzzy sets of wind speed, wind direction, vibration acceleration and blade pitch angle; A set of initial chromosomes is randomly generated, representing different combinations of fuzzy rules and membership functions; For each chromosome, calculate its fitness value, which can be determined based on the established performance evaluation index. The fitness function is defined as ,in , , are weight coefficients, which respectively indicate the importance of load reduction, stability, and response speed in the fitness function; The chromosomes are decoded into fuzzy rules and membership functions, and then the collected wind turbine operation data is input to simulate the operation of AFLC and calculate the load reduction degree. ,stability , Response speed , and then get the fitness value; The chaos immune algorithm is used for optimization, including chaos search and immune operation.

[0082] (1) Chaotic search: Use chaotic mapping to generate chaotic sequences, perform chaotic perturbations on the current optimal chromosome, and expand the search range. For example, you can use Logistic chaotic mapping ,in is the control parameter, usually , . Map the chaotic sequence to the gene space of the chromosome to obtain a new chromosome.

[0083] (2) Immune operation: including immune selection, cloning, mutation and immune memory. Immune selection is to select a part of excellent chromosomes according to the fitness value to enter the next round of iteration. Cloning is to copy the selected chromosomes to increase their number in the population. Mutation is to randomly mutate the cloned chromosomes to increase the diversity of the population. Immune memory is to preserve the best chromosomes in each generation to prevent the loss of excellent solutions.

[0084] Logistic chaos mapping formula: , which is used to generate chaotic sequences, perform chaotic perturbations on chromosomes, and expand the search range.

[0085] - Fitness function formula: , its purpose is to calculate the fitness value of the chromosome according to the performance evaluation index, and to select excellent chromosomes for evolution.

[0086] Repeat steps S91 and S92 until the stopping condition is met, such as reaching the maximum number of iterations or the fitness value no longer increases. S92, applying the optimized fuzzy rules and membership functions to AFLC, the specific method is: according to the encoding method, converting the gene sequence in the optimal chromosome into specific fuzzy rules and membership function parameters; The AFLC can better adapt to different complex working conditions according to the optimized fuzzy rules and membership functions, and improve the load reduction control effect and safety of wind turbines.

[0087] The above embodiments are only preferred technical solutions of the present invention and should not be regarded as limiting the present invention. The protection scope of the present invention shall be the technical solutions recorded in the claims, including the same replacement solutions of the technical features in the technical solutions recorded in the claims. That is, the same replacement improvements within this scope are also within the protection scope of the present invention.

Claims

1. A load reduction control method for a wind turbine generator set based on complex working condition identification by laser radar, characterized in that: The method includes: S1. Install a high-precision laser radar system on the wind turbine to monitor the three-dimensional wind speed distribution in the wind farm in real time with high frequency, covering instantaneous wind speed, wind direction and turbulence intensity parameters. The laser radar will transmit the collected data to the control system of the wind turbine in a timely manner to ensure the real-time and accuracy of the data; S2. Perform comprehensive preprocessing operations on the collected raw data. First, perform data cleaning to remove outliers and invalid data. Then, use advanced wavelet transform technology or other advanced filtering methods to remove noise interference while retaining useful wind farm information, providing a high-quality data foundation for subsequent complex working condition identification and load reduction control. S3. Use time series analysis methods to deeply explore the inherent laws and patterns of wind speed and wind direction changes over time. Through the analysis of time series data, it can keenly capture the key features of sudden changes in wind speed and rapid changes in wind direction, providing a strong basis for the accurate identification of complex working conditions; S4. Based on the historical data of wind speed and wind direction changes, define targeted fuzzy logic rules to accurately describe various types of complex working conditions; S5. Design a powerful adaptive fuzzy logic controller AFLC. The AFLC controller has the ability to automatically adjust its internal fuzzy rules and membership functions according to the actual situation of the current wind farm. AFLC can continuously improve its ability to identify complex working conditions by continuously learning and adapting to changes in the wind farm environment, providing a solid guarantee for the safe and stable operation of wind turbines. S6, AFLC successfully identified the complex operating conditions and immediately started the load reduction control logic, accurately adjusting the wind turbine operating parameters, including blade pitch angle and generator speed, according to the specific operating condition type identified; S7 and AFLC are not only responsible for identifying complex working conditions, but also can dynamically adjust the load reduction threshold according to the current wind farm conditions. When the detected wind speed change or wind direction change exceeds the dynamic threshold, the load reduction program is immediately started to ensure that the wind turbine can respond to changes in complex working conditions in a timely manner and ensure its safe and stable operation; When the system detects complex working conditions and reaches the dynamic threshold, it automatically implements load reduction measures; S8. During the load shedding control process, the operating data of the wind turbines, including vibration levels and power output parameters, are continuously collected. Through in-depth analysis of these data, the actual effect of the load shedding control is accurately evaluated, and each evaluation result is recorded in detail to provide valuable data support for subsequent analysis and optimization; S9. According to the actual results of each load reduction control, the fuzzy rules and membership functions in AFLC are adjusted and optimized in a targeted manner.

2. According to claim 1, a load reduction control method for a wind turbine generator set based on complex working condition identification by laser radar is characterized by: The specific steps of the time series analysis method in step S3 to identify complex working conditions are: S31. Collect and pre-process wind speed and wind direction time series data, specifically: use the laser radar system installed on the wind turbine to continuously monitor the wind speed and wind direction in the wind farm, collect time series data at a frequency of 10-20 times per second, and record the wind speed value. and wind direction angle , the time span is ; The wavelet threshold denoising method is used to denoise the collected wind speed and wind direction time series data; Wind speed time series Perform wavelet transform to obtain wavelet coefficients ; The original signal is , after wavelet transform, ,in is the wavelet basis function, Indicates scale, Indicates time position; Choosing an appropriate threshold , process the wavelet coefficients, and the processed wavelet coefficients for: ； The denoised wind speed signal is obtained by inverse wavelet transform ; For wind direction time series The same method is used for denoising to obtain the denoised wind direction signal. ; S32, construct a fractional order chaotic time series model, specifically: for the time series , the fractional difference is defined as: ； in, , is the gamma function, and the denoised wind speed time series is calculated and wind direction time series Fractional difference of and ; Fractional differencing is used to capture the long memory and nonlinear characteristics of time series; Among them, the wind speed time series , construct the following fractional-order chaotic time series model: ； in, is the wind speed time delay, is the embedding dimension of wind speed, is the wind speed fractional parameter, is a nonlinear function, which is approximated by a neural network method; For wind direction time series , construct a similar fractional-order chaotic time series model: ； in, is the wind direction time delay, is the embedding dimension of wind direction, is the wind direction fractional parameter, is a nonlinear function; The fractional order chaotic time series model accurately describes the inherent laws and patterns of wind speed and wind direction changing with time; S33, calculate the fractional correlation dimension and fractional Lyapunov exponent key features, specifically: use the Grassberger-Procaccia algorithm to calculate the fractional correlation dimension , for the wind speed time series , define the correlation function for: ； in, is the number of wind speed data points, is the Heaviside function, when hour, ,otherwise ; Correlation dimension Calculated by the following formula: ； Wind direction time series The same method is used to calculate the fractional correlation dimension ; The fractional-order Lyapunov exponent calculation method includes: For the fractional-order chaotic system , calculate the fractional Lyapunov exponent , using the fractional-order extended version of the Wolf algorithm, it is calculated by tracking the trajectory changes of the system in the phase space; Includes: Selecting an initial point In phase space, calculate its The later evolution point ; Find with The nearest point , calculate the distance between them ; As time goes by, calculation ,in is from The point from which evolution comes; Calculate the average exponential growth rate ,For fractional-order systems, it is necessary to consider the influence of fractional-order derivatives and make appropriate adjustments to the above steps; The same calculation is performed on the fractional-order chaotic system of wind direction time series, and the fractional-order Lyapunov exponent is obtained: ; The fractional Lyapunov exponent is used to determine the chaotic characteristics and stability of the system; S34: Determine whether a complex working condition occurs according to the set threshold, specifically: set the wind speed fractional correlation dimension threshold according to historical data and experience and fractional Lyapunov exponent threshold ; Set the wind direction fractional correlation dimension threshold and fractional Lyapunov exponent threshold ; Complex working condition judgment: When the calculated wind speed fractional correlation dimension Exceeding the threshold or fractional Lyapunov exponent Exceeding the threshold When , it is considered that the wind speed presents complex working conditions; When the wind direction fractional correlation dimension Exceeding the threshold or fractional Lyapunov exponent Exceeding the threshold When , it is considered that the wind direction presents a complex working condition; If complex operating conditions occur in both wind speed and wind direction at the same time, it is judged that the entire wind turbine is in a complex operating condition.

3. According to claim 1, a load reduction control method for a wind turbine generator set based on complex working condition identification by laser radar is characterized by: The specific steps of defining the fuzzy logic rules for complex working conditions in step S4 are: S41, collect and pre-process historical data of wind speed and wind direction, and perform partition processing, specifically: collect historical data of wind speed and wind direction of wind turbines in different operating time periods, including wind speed values and wind direction angle ,At the same time, the corresponding wind turbine operating status information, such as vibration amplitude, power output, is recorded; The collected wind speed and direction data are partitioned and processed using equal-spaced partitions or adaptive partitioning methods based on data distribution; S42, constructing fuzzy sets of wind speed and wind direction, and defining a membership function, specifically: constructing a fuzzy set of wind speed for wind speed partitions; Define three fuzzy sets: low wind speed, medium wind speed and high wind speed; The speed value is , then its membership function belonging to the low wind speed fuzzy set is defined as: ； The membership function of the wind speed fuzzy set is: ； The membership function of the high wind speed fuzzy set is defined similarly; Through the construction of fuzzy sets, the precise wind speed value is converted into a fuzzy language description, and the recognition and judgment of wind conditions provide a basis for the subsequent definition of fuzzy logic rules. For wind direction partitioning, a fuzzy set of wind directions is constructed, including east wind, south wind, west wind and north wind; Assume the wind direction angle is , then the membership function of the Dongfeng fuzzy set can be defined as: ； The membership functions of other wind direction fuzzy sets are defined similarly; S43, construct a fuzzy rough set information system, calculate the lower approximation and the upper approximation, specifically: Building an information system ,in is the domain, i.e. the set of all data samples; is a set of attributes, including wind speed attributes and wind direction attributes; is a collection of attribute values; is an information function that assigns a value to each attribute for each object; For a subset of attributes , defining the indiscernibility relation for: ； That is, if two objects have a subset of attributes have the same value on , then they are It is indistinguishable under the relation; For a fuzzy concept complex working condition and attribute subsets , the lower approximation and the upper approximation They are defined as: ； ； in, Is an object A fuzzy concept The membership degree of Is an object In the indiscernible relationship The equivalence class under ; S43, defining quantum logic gates, and generating quantum logic fuzzy rules according to the fuzzy rough set results and the fuzzy set, specifically: defining a set of quantum logic gates for representing operations in fuzzy logic rules, including quantum AND gates and quantum OR gates; Assume two quantum states and , the quantum AND gate can be defined as: ； For the quantum OR gate, it can be defined as: ； According to the lower approximation and upper approximation calculated by fuzzy rough sets, as well as the fuzzy sets of wind speed and wind direction, quantum logic fuzzy rules are generated; The wind speed belongs to the high wind speed fuzzy set and the wind direction belongs to the east wind fuzzy set. It is considered that complex working conditions may occur, which can be expressed by quantum logic rules as follows: ； in, represents a quantum NOT gate, Indicates a quantum state where complex operating conditions may occur.

4. According to claim 1, a load reduction control method for a wind turbine generator set based on complex working condition identification by laser radar, characterized in that: The specific steps of the AFLC controller in step S51 are: S51. Detailed classification of complex working conditions, including conditions of drastic wind speed changes, sudden wind direction changes, and extreme negative shear conditions. For conditions of drastic wind speed changes, define the wind speed change rate as exceeding a certain threshold. As the judgment condition of this working condition, When is the wind speed change, is the time interval; S52: For sudden wind direction changes, define the wind direction change angle exceeding the threshold in a short period of time. ,when When is the angle of wind direction change; For extreme negative shear conditions, the ratio of wind speed differences at different heights to the average wind speed exceeds the threshold. To judge, that is ,in and are the wind speeds at the top and bottom of the wind turbine, is the average wind speed; S53, mapping different working conditions with the operating parameters of the wind turbine, for the working conditions with drastic changes in wind speed, mainly adjusting the blade pitch angle and generator speed To reduce the load on wind turbines; A mapping relationship is established. When the wind speed change rate is large, the blade pitch angle is appropriately reduced and the generator speed is reduced, that is, , ,in and are the initial pitch angle and generator speed, and is the adjustment factor; For sudden changes in wind direction, the yaw angle of the wind turbine can be adjusted To reduce the lateral force of the wind rotor, ,in is the initial yaw angle, is the adjustment factor; For extreme negative shear conditions, the power output curve of the wind turbine can be optimized and the load can be reduced by adjusting the excitation current parameters of the generator. The specific mapping relationship can be determined according to the actual situation. S54, AFLC includes input and output: input includes wind speed, wind direction, wind speed change rate, wind direction change angle wind field parameters, and current operating parameters of the wind turbine, including blade pitch angle, generator speed, and yaw angle; The output is the adjusted wind turbine operating parameters, including the adjusted blade pitch angle, generator speed, and yaw angle; S55. According to different complex working conditions and operating parameter ranges, a series of fuzzy rules are defined, specifically: "If the wind speed is high and the rate of change is large, then the blade pitch angle should be reduced much", expressed in fuzzy language as "IF wind_speed is High AND wind_speed_change_rate is Large THEN blade_pitch_angle isDecrease_Much"; Determine the membership function of fuzzy rules through expert experience and historical data; The membership function for "high wind speed" can be defined as ,in is the wind speed, is the central value of high wind speed, is the standard deviation; S56, executing fuzzy reasoning and defuzzification of AFLC, performing fuzzy reasoning through fuzzy rules according to the input wind field parameters and operation parameters, and obtaining a fuzzy output result; Then, the centroid method is used to defuzzify the fuzzy output into an accurate operating parameter adjustment value. The centroid method formula is: ,in is the number of output fuzzy sets, It is The membership degree of the output fuzzy set is It is The central value of the output fuzzy set.

5. The load reduction control method for a wind turbine generator set based on complex working condition identification by laser radar according to claim 4 is characterized by: In steps S6-S7, the AFLC successfully identifies the complex working condition, and the specific method of the control logic of the AFLC is: S61, select appropriate wind field parameters as the basis for dynamic threshold adjustment, and the wind speed change rate, wind direction change angle, and wind turbine vibration amplitude parameters can be selected. and wind direction change angle As the main parameter for dynamic threshold adjustment; S62, constructing a dynamic threshold model, using a nonlinear function to construct the dynamic threshold model, ,in , , , are model parameters, which can be obtained by fitting historical data; By building a dynamic threshold model, the load reduction threshold can be dynamically adjusted according to the current wind farm conditions; S63, real-time monitoring of wind farm parameters and wind turbine operating status, continuous collection of wind speed change rate, wind direction change angle, vibration amplitude wind farm parameters, and wind turbine output power, speed operating status parameters; S64, update the model parameters according to the monitoring data, adopt the adaptive recursive least squares algorithm, and continuously update the parameters of the dynamic threshold model according to the real-time monitoring data. The recursive least squares formula is: ， ， ； in is the model parameter vector, is the time step, is the error between the actual value and the predicted value, is the input vector, It's the forgetting factor. is the identity matrix; S65. Calculate a dynamic threshold value according to the updated model parameters, and immediately start a load reduction procedure when the detected wind farm parameter change exceeds the dynamic threshold value.

6. According to claim 5, a load reduction control method for a wind turbine generator set based on complex working condition identification by laser radar is characterized by: In step S7, when the system detects a complex working condition and reaches a dynamic threshold, the specific method of automatically executing load reduction measures is as follows: S71: When the system detects a complex operating condition and reaches a dynamic threshold, it automatically executes load reduction measures and adjusts the operating parameters of the wind turbine according to the control logic designed in steps S51-S54. At the same time, other load reduction measures can be taken, including lowering the output power of wind turbines and adjusting the speed of wind rotors; S72. When adjusting operating parameters, use a gradual approach to avoid sudden and large adjustments that may cause impacts on wind turbines; For the adjustment of the blade pitch angle, a linear gradient method can be used, that is, ,in It's in time The blade pitch angle at is the initial pitch angle, is the adjustment rate; S73. Continuously and closely monitor the operating status of the wind turbine generator set, using various sensors, including vibration sensors, temperature sensors, and speed sensors, to monitor the operating status parameters of the wind turbine generator set in real time; The monitoring data will be transmitted to the control center so that the operation status of the wind turbine can be understood in a timely manner.

7. The load reduction control method for a wind turbine generator set based on complex working condition identification by laser radar according to claim 1 is characterized by: The specific method for evaluating the load reduction control effect in step S8 is: S81. Define evaluation indicators. Based on information geometry metrics, define evaluation indicators for load reduction control effects, including the vibration level reduction rate. and power output stability ; Vibration level reduction rate It can be defined as: ； in, is the probability density function of vibration acceleration before load reduction control, is the probability density function of vibration acceleration after load reduction control; Power output stability It can be defined as: ； S82. Calculate the actual effect of load reduction control based on the defined evaluation indicators, including the vibration level reduction rate Larger and more stable power output A higher value indicates better load shedding control effect; S83. Record the evaluation results of each load reduction control in detail, including the vibration level reduction rate, power output stability and other relevant parameters. The results may be recorded in a database or file system for subsequent analysis and query. S84. Use quantum annealing algorithm to analyze the recorded evaluation data to find the optimal load shedding control strategy. Quantum annealing algorithm is an optimization algorithm based on the principle of quantum mechanics, which can find the global optimal solution in a complex search space. First, the load reduction control problem is transformed into an optimization problem. The objective function can be the weighted sum of maximizing the vibration level reduction rate and the power output stability, that is, ,in and is the weight coefficient; Then, the quantum annealing algorithm is used for optimization. The core of the quantum annealing algorithm is to gradually reduce the energy of the system by simulating the annealing process of the quantum system and finally reach the global optimal state. In the load shedding control problem, the energy of the system can be regarded as the value of the objective function. By adjusting the operating parameters of the wind turbine, the energy of the system can be reduced, thereby finding the optimal load shedding control strategy.

8. The load reduction control method for a wind turbine generator set based on complex working condition identification by laser radar according to claim 1 is characterized by: The specific method for adjusting and optimizing the fuzzy rules and membership functions in AFLC in step S9 is as follows: S91. Collect and organize the wind turbine operation data of each load reduction control and establish performance evaluation indicators to measure the performance of AFLC; S92. Using the chaos immune algorithm to optimize the fuzzy rules and membership functions of AFLC, including encoding, initializing the population, calculating fitness, and chaos immune operation steps, specifically: Encode the fuzzy rules and membership functions of AFLC into a chromosome, and use binary coding or real number coding to encode the condition part and conclusion part of each fuzzy rule and the parameters of the membership function into the genes of the chromosome; AFLC includes at least three input variables: wind speed ,wind direction , vibration acceleration , the output variable is the blade pitch angle ; Fuzzy rules can be expressed as "IF is AND is AND is THEN is ",in , , , They are the fuzzy sets of wind speed, wind direction, vibration acceleration and blade pitch angle; A set of initial chromosomes is randomly generated, representing different combinations of fuzzy rules and membership functions; For each chromosome, calculate its fitness value, which can be determined based on the established performance evaluation index. The fitness function is defined as ,in , , are weight coefficients, which respectively indicate the importance of load reduction, stability, and response speed in the fitness function; The chromosomes are decoded into fuzzy rules and membership functions, and then the collected wind turbine operation data is input to simulate the operation of AFLC and calculate the load reduction degree. ,stability , Response speed , and then get the fitness value; Adopt chaos immune algorithm for optimization; including chaos search and immune operation; S92, applying the optimized fuzzy rules and membership functions to AFLC, the specific method is: according to the encoding method, converting the gene sequence in the optimal chromosome into specific fuzzy rules and membership function parameters; The AFLC can better adapt to different complex working conditions according to the optimized fuzzy rules and membership functions, and improve the load reduction control effect and safety of wind turbines.

Images