Permanent magnet synchronous wind turbine generator efficiency optimal control method and system based on data driving
By optimizing the control strategy of permanent magnet synchronous wind turbines using a data-driven neural network model, the problem of ineffective converter loss reduction was solved, resulting in improved system efficiency, extended equipment lifespan, and efficient control adaptable to complex environments.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- GUODIAN XIANGSHAN OFFSHORE WIND POWER CO LTD
- Filing Date
- 2026-03-20
- Publication Date
- 2026-05-15
AI Technical Summary
Existing control strategies for permanent magnet synchronous wind turbines have failed to effectively reduce converter losses, resulting in limited improvement in overall system efficiency. Furthermore, traditional methods may increase converter losses when pursuing the minimization of motor losses, and there is a lack of effective loss optimization solutions.
A data-driven multilayer feedforward neural network is used to fit the relationship between converter losses and the d-axis component of torque current. By establishing an accurate model, the control strategy is optimized to minimize converter losses. The neural network model is combined with traditional control algorithms to form a complete efficiency-optimal control link.
It significantly improves the overall power generation efficiency of wind turbine units, reduces the operating heat generation of converters, extends equipment life, increases economic benefits and energy utilization efficiency, and adapts to complex wind farm environments.
Smart Images
Figure CN122052631A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of wind power generation technology and relates to a data-driven method and system for optimal efficiency control of permanent magnet synchronous wind turbine generators. Background Technology
[0002] Wind power generation, as a mature new energy technology, plays a crucial role in energy transition. Among them, permanent magnet synchronous wind turbines are widely used in the wind power field due to their advantages of high efficiency, high reliability, and simple structure. Currently, most permanent magnet synchronous wind turbines use dual-pulse-width modulation back-to-back grid-connected converters for grid connection, and the turbine-side converters commonly employ zero-d-axis current control strategies or field weakening control strategies to achieve maximum power point tracking. However, existing control strategies lack consideration for changes in wind turbine converter losses, directly affecting the power generation efficiency and output power of the wind power grid-connected system. Therefore, inventing a control strategy that can reduce converter losses in permanent magnet wind power generation systems is of great significance for improving the overall efficiency of wind power generation systems and increasing economic benefits.
[0003] Existing control strategies have a relatively singular control objective, focusing solely on maximizing power output without fully considering the impact of the control strategy on converter losses. This results in ineffective control of overall system losses. Some control methods aimed at reducing motor losses, such as minimum motor loss control, actually increase converter losses, thereby raising the total system losses and contradicting the initial goal of improving wind power system efficiency. Existing research on reducing converter losses largely focuses on lowering the switching frequency, lacking effective solutions for minimizing converter losses through optimizing the control strategy itself. This makes it difficult to meet the urgent needs of wind turbines to increase power and efficiency. Furthermore, the physical relationship between converter losses and the d-axis component of torque current is still unclear, and traditional model-based derivations cannot provide reliable evidence for optimizing control strategies.
[0004] The aforementioned problems limit the efficiency improvement of permanent magnet wind power systems, thus restricting the economic benefits and energy utilization efficiency of wind power projects. Summary of the Invention
[0005] In view of this, the purpose of this invention is to provide a data-driven optimal efficiency control method and system for permanent magnet synchronous wind turbine generators. Addressing the problems of existing maximum power point tracking control strategies for permanent magnet wind power systems not taking into account converter losses, the difficulty of accurately correlating converter losses and current parameters using traditional modeling methods, and the limited improvement in overall system efficiency, this invention provides an optimal efficiency control method for permanent magnet wind power systems based on neural network fitting to minimize converter losses. By establishing a precise correlation model between converter losses and the d-axis components of torque and current, the optimal value of the d-axis component of torque and current is determined when converter losses are minimized, thereby improving the overall efficiency of the wind power system while ensuring the reliability and adaptability of the control strategy.
[0006] To achieve the above objectives, the present invention provides the following technical solution: A data-driven method for optimal efficiency control of permanent magnet synchronous wind turbine generators includes the following steps: S1: Based on the simulation platform, build a basic simulation model of the permanent magnet wind power system, simulate the system operation conditions under different wind speeds, different loads, and different control parameters, collect the converter loss data PCloss and the torque current d-axis component data iTd corresponding to each operating condition, and obtain no less than 2000 sets of valid data samples. S2: Converter loss data PCloss and torque current d-axis component data collected from S1 The data is cleaned to remove outliers and noisy data, and then normalized to map it to the (0,1) interval. The normalization formula is as follows:
[0007] in, For the normalized data, x For current data, The minimum value in a set of data. The maximum value in a set of data; S3: A multi-layer feedforward neural network is used as the fitting model. The neural network includes an input layer, at least one hidden layer, and an output layer. The number of nodes in the input layer is determined according to the d-axis component of the torque current. The characteristic dimensions are determined, and the predicted values of strain gauge losses for the output layer nodes are obtained. S4: Divide the preprocessed data from S2 into training, validation, and test sets in a ratio of 7:1.5:1.5. Use the mean squared error (MSE) as the loss function and train the neural network using the backpropagation algorithm and the Adam optimizer. S5: L2 regularization and early stopping are introduced during the training process of S4 to prevent overfitting; S6: Input the test set data into the neural network trained in S5, calculate the error between the predicted converter loss and the actual converter loss, and form a fitting model of converter loss-torque current d-axis component when the error is less than 1.5%. S7: Employs the optimal tip speed ratio method to achieve maximum power point tracking (MPPT) for permanent magnet wind turbines, and collects wind speed data in real time. v Wind turbine speed data ω Calculate the actual tip speed ratio λ With the optimal tip speed ratio λ opt Compare and calculate the reference speed based on the deviation. ω The wind turbine speed is adjusted by the speed controller, and the reference speed is set. ωThe deviation from the actual rotational speed is input to a proportional-integral (PI) controller, which outputs a q-axis command current. i q ; S8: Input the initial value of the d-axis component of the torque current under real-time operating conditions into the converter loss-torque current d-axis component fitting model formed in S6 to predict the converter loss. Use an ergonomic search algorithm to traverse candidate torque current d-axis component values under the current constraint of the permanent magnet synchronous motor, and select the candidate value with the minimum converter loss as the optimal value of the torque current d-axis component with a step size of 0.01 times the rated current. ; S9: Convert the q-axis command current obtained from S7 i q The optimal value of the d-axis component of the torque current obtained from S8 As the input to the current controller, a proportional-integral (PI) control algorithm is used, based on the actual d-axis current. i d q-axis current i q Calculate the d-axis voltage reference value based on the deviation from the commanded current. V d q-axis voltage reference value V q ; S10: Obtain the d-axis voltage reference value from S9 V d q-axis voltage reference value V q Input the space vector pulse width modulation (SVPWM) modulation module to generate the pulse width modulation (PWM) drive signal for the machine-side converter; S11: Integrates the data acquisition and preprocessing, neural network fitting, maximum power point tracking, loss minimization current optimization, current control, and space vector pulse width modulation (SVPWM) modules of S1~S10 to form a complete efficiency-optimized control link for permanent magnet wind power systems, enabling real-time data interaction among the modules.
[0008] Furthermore, in S1, the simulation platform is the MATLAB / Simulink platform, and the data samples cover operating scenarios with wind speeds of 3~15m / s and load rates of 10%~100%.
[0009] Furthermore, in S3, the number of hidden layer neurons is set by trial and error or empirical formulas and adjusted according to model performance during training.
[0010] Furthermore, in S4, the number of training iterations is 100 to 200 rounds, and training stops if the loss on the validation set does not decrease for 10 consecutive rounds.
[0011] Furthermore, in step S6, if the error is not less than 1.5%, the number of hidden layer neurons or the learning rate of the Adam optimizer are adjusted, and steps S4 to S5 are re-executed until the error is less than 1.5%.
[0012] Furthermore, in S8, the traversal search algorithm uses the initial value of the d-axis component of the torque current as a reference, and traverses the candidate torque current d-axis component values with a step size of 0.01 times the rated current under the current constraint condition of the permanent magnet synchronous motor.
[0013] Furthermore, in S10, the space vector pulse width modulation (SVPWM) module calculates the conduction time of each switching transistor based on the voltage space vector distribution law.
[0014] A data-driven optimal control system for permanent magnet synchronous wind turbines includes a data acquisition and preprocessing module, a neural network fitting module, a maximum power point tracking module, a loss minimization current optimization module, a current control module, and a space vector pulse width modulation (SVPWM) module, which are sequentially connected by data interaction. The data acquisition and preprocessing module is used to acquire converter loss data (PCloss) and torque current d-axis component data. After cleaning and normalization, the output is sent to the neural network fitting module; The neural network fitting module is used to form a converter loss-torque current d-axis component fitting model with an error of less than 1.5% based on a multi-layer forward feedback neural network, and output it to the loss minimization current optimization module. The maximum power point tracking module is used to output the q-axis command current using the optimal tip speed ratio method. i q To the current control module; The loss minimization current optimization module is used to input the initial value of the real-time torque current d-axis component into the converter loss-torque current d-axis component fitting model, and use a traversal search algorithm to select the optimal value of the torque current d-axis component under current constraints. And output to the current control module; The current control module is used to control the current according to the q-axis command. i q Optimal value of d-axis component of torque current and actual d-axis current i d q-axis current i qThe d-axis voltage reference value is calculated using a proportional-integral (PI) control algorithm. V d q-axis voltage reference value V q And output to the Space Vector Pulse Width Modulation (SVPWM) module; The space vector pulse width modulation (SVPWM) module is used to adjust the d-axis voltage reference value. V d q-axis voltage reference value V q The pulse width modulation (PWM) drive signal for the generator-side converter is generated.
[0015] Furthermore, the neural network fitting module employs mean squared error (MSE) as the loss function, backpropagation algorithm, Adam optimizer, L2 regularization, and early stopping method during the training process.
[0016] Furthermore, the loss minimization current optimization module adopts a traversal search algorithm with a step size of 0.01 times the rated current.
[0017] Furthermore, the normalization formula for the data acquisition and preprocessing module is as follows:
[0018] in, For the normalized data, x For current data, The minimum value in a set of data. The maximum value in a set of data; Furthermore, the maximum power point tracking module collects wind speed data in real time. v Wind turbine speed data ω Calculate the actual tip speed ratio λ With the optimal tip speed ratio λ opt In comparison, the deviation is used to output the q-axis command current through a proportional-integral (PI) controller. i q .
[0019] The beneficial effects of this invention are as follows: (1) This invention establishes a high-precision neural network fitting model through a data-driven approach, accurately describing the nonlinear relationship between converter losses and the d-axis component of torque current, and minimizing converter losses under the premise of ensuring maximum power point tracking, thereby significantly improving the overall power generation efficiency of the permanent magnet synchronous wind turbine grid-connected system.
[0020] (2) The converter loss model constructed in this invention maintains stable accuracy over a wide range of wind speeds and loads. It does not require repeated training or parameter adjustment for different operating scenarios, and has strong adaptability and robustness. It can adapt to the complex environment of frequent wind speed fluctuations and variable loads in actual wind farms, reducing the workload of on-site commissioning and maintenance.
[0021] (3) This invention achieves effective control of converter losses by optimizing the d-axis component of torque current, reducing the heat generated during converter operation, protecting key power devices such as insulated gate bipolar transistors from premature performance degradation, extending the overall service life of the equipment, and reducing maintenance frequency and component replacement costs.
[0022] (4) This invention organically integrates maximum power point tracking control and converter loss minimization control, avoiding the contradiction in traditional methods where simply pursuing the minimization of motor losses leads to an increase in converter losses, thereby achieving a comprehensive reduction in total system losses and further improving the economic benefits and energy utilization efficiency of wind turbine units.
[0023] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description
[0024] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein: Figure 1 The graph shows the test results for neural network fitting. Figure 2 Flowchart of the converter loss model fitted by a neural network; Figure 3 Power characteristics of wind turbine generators; Figure 4 The diagram shows the optimal control strategy for permanent magnet wind turbine generators with the lowest converter losses. Figure 5 A comparison chart of converter losses under different control methods; Figure 6 A comparison chart showing the efficiency results of different control methods. Detailed Implementation
[0025] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.
[0026] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual pictures. They should not be construed as limiting the invention. To better illustrate the embodiments of the invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.
[0027] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," "front," and "rear" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present invention. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.
[0028] Example 1 This embodiment provides a complete implementation process of a data-driven method for optimal efficiency control of permanent magnet synchronous wind turbine generators, including the following steps: First, a basic simulation model of a 6.2MW permanent magnet wind power system was built based on the MATLAB / Simulink platform. This model includes a wind turbine aerodynamics module, a permanent magnet synchronous generator module, a dual-pulse-width modulation back-to-back grid-connected converter module, a turbine-side vector control module, and a grid-side control module. The simulated wind speed range is 3m / s to 15m / s, and the load rate range is 10% to 100%. The main control parameters that vary are the d-axis components of torque and current. i Td The step size is set to 0.01 times the rated current. Through batch simulation runs, at least 2000 sets of valid data samples are collected. Each set of samples includes the converter loss data (PCloss) and the corresponding torque and current d-axis component data under the current operating condition. i Td This ensures that the data covers typical operating scenarios of the system.
[0029] Secondly, the collected raw data is cleaned to remove outliers and noise caused by simulation fluctuations. Then, normalization is applied to map all data to the 0-1 range; the normalization formula is as follows: ,in For the normalized data, x For current data, The minimum value in a set of data. The maximum value in a set of data.
[0030] Next, a multi-layer feedforward neural network is constructed as the fitting model. The input layer has one node, corresponding to the d-axis component of the torque current. i Td The system uses a single feature dimension; it has two hidden layers, with 10 neurons in the first hidden layer and 5 neurons in the second hidden layer; the output layer has 1 node, corresponding to the predicted loss value of the strain gauge. The hidden layer uses the ReLU function, and the output layer uses a linear activation function.
[0031] The preprocessed data was randomly divided into training, validation, and test sets in a 7:1.5:1.5 ratio. Mean Squared Error (MSE) was used as the loss function, and training was performed using backpropagation combined with the Adam optimizer, with an initial learning rate of 0.001. L2 regularization was introduced during training, with a weight decay coefficient of 0.0001. Early stopping was also employed; training was stopped if the validation set loss did not decrease for 10 consecutive iterations. The maximum number of training iterations was set to 200.
[0032] After training, the test set is input into the model to calculate the relative error between the predicted converter loss and the actual converter loss. If the error is greater than 1.5%, the number of hidden layer neurons or the learning rate is adjusted, and retraining is performed until the error is less than 1.5%. The final converter loss-torque current d-axis component fitting model is as follows: Figure 1 and Figure 2 As shown.
[0033] In the real-time control phase, the optimal tip speed ratio method is first used to achieve maximum power point tracking. Wind speed data is then collected in real time. v Wind turbine speed data ω Calculate the actual tip speed ratio Where R is the rotor radius. The actual tip speed ratio... λ Compared with the optimal tip speed ratio λ opt If a comparison is made and a deviation exists, a reference speed is calculated based on the relationship between the tip speed ratio and the rotational speed. ω The rotor speed is adjusted by a proportional-integral (PI) controller on the outer loop of the speed control loop. The speed deviation is input into the PI controller, which outputs the q-axis command current. i q The power characteristics of wind turbines are as follows: Figure 3 As shown in the figure, the power curves at different wind speeds and the trajectory of the maximum power point corresponding to the optimal tip speed ratio are displayed.
[0034] Subsequently, the initial values of the d-axis component of the torque current under the current operating conditions are input into the trained neural network model to predict the corresponding inductor losses. A traversal search algorithm is employed, using the initial values as a baseline, to traverse candidate torque current d-axis component values within the permanent magnet synchronous motor current constraint circle, with a step size of 0.01 times the rated current. For each candidate value traversed, the corresponding loss is calculated using the neural network model. Finally, the candidate value with the minimum loss is selected as the optimal value for the torque current d-axis component. i Td .
[0035] q-axis command current i q and the optimal value of the d-axis component of torque current i Td The input current inner loop controller uses a proportional-integral (PI) control algorithm, based on the actual d-axis current. i d q-axis current i q Calculate the d-axis voltage reference value based on the deviation from the command value. V d and q-axis voltage reference value V q .
[0036] Finally, the d-axis voltage reference value V d and q-axis voltage reference value V q The input space vector pulse width modulation (SVPWM) module calculates the conduction time of each switch based on the voltage space vector distribution law, and generates the pulse width modulation (PWM) drive signal for the machine-side converter.
[0037] All the above modules are integrated to form a complete and optimally efficient control link, such as Figure 4 As shown. Figure 4 This diagram illustrates the optimal control strategy for permanent magnet wind turbines that minimizes converter losses. The wind speed is displayed from top to bottom in the diagram. v and rotational speed ωInput to the maximum power point tracking module, output q-axis command current. i q Simultaneously, the initial value of the d-axis component of the torque current is input into the neural network fitting model, and after traversal search and optimization, the optimal value of the d-axis component of the torque current is output. i Td Both input current controller and output voltage reference value V d and V q The signal is then fed into the Space Vector Pulse Width Modulation (SVPWM) module, which ultimately generates a pulse width modulation (PWM) signal to control the machine-side converter.
[0038] Example 2 This embodiment, based on Embodiment 1, verifies and compares the proposed method, covering the application effect of the method described in claim 1 in a real system.
[0039] A complete 6.2MW permanent magnet wind power grid-connected system was built based on the MATLAB / Simulink platform. The grid-side converter adopted space vector pulse width modulation (SVPWM) modulation and was connected to a 690V grid. Three control methods were implemented: the method of this invention, the traditional zero d-axis current control method, and the minimum motor loss control method. Simulations were conducted and compared under the same random wind speed sequence.
[0040] Converter losses, for example Figure 5 As shown. Figure 5 The graph compares converter losses using different control methods. The horizontal axis represents simulation time, and the vertical axis represents instantaneous converter loss. The three curves in the graph correspond to the method of this invention, zero d-axis current control, and minimum motor loss control, respectively. The converter loss curve of the method of this invention is consistently below the other two curves, indicating that the converter loss is significantly lower than that of the other two methods.
[0041] System efficiency comparison Figure 6 As shown. Figure 6 The graph compares the efficiency results of different control methods. The horizontal axis represents simulation time, and the vertical axis represents the overall system efficiency. The three curves correspond to the method of this invention, zero d-axis current control, and minimum motor loss control, respectively. The efficiency curve of the method of this invention is always at the top, indicating that the overall system efficiency is higher than the other two traditional methods, thus improving the overall efficiency of the permanent magnet wind power grid-connected system.
[0042] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
Claims
1. A data-driven optimal control method for permanent magnet synchronous wind turbine generators, characterized in that: Includes the following steps: S1: Based on the simulation platform, build a basic simulation model of the permanent magnet wind power system, simulate the system operation conditions under different wind speeds, different loads, and different control parameters, collect the converter loss data PCloss and the torque current d-axis component data iTd corresponding to each operating condition, and obtain no less than 2000 sets of valid data samples. S2: Converter loss data PCloss and torque current d-axis component data collected from S1 The data is cleaned to remove outliers and noisy data, and then normalized to map it to the (0,1) interval. The normalization formula is as follows: in, For the normalized data, x For current data, The minimum value in a set of data. The maximum value in a set of data; S3: A multi-layer feedforward neural network is used as the fitting model. The neural network includes an input layer, at least one hidden layer, and an output layer. The number of nodes in the input layer is determined according to the d-axis component of the torque current. The characteristic dimensions are determined, and the predicted values of strain gauge losses for the output layer nodes are obtained. S4: Divide the preprocessed data from S2 into training, validation, and test sets in a ratio of 7:1.5:1.
5. Use the mean squared error (MSE) as the loss function and train the neural network using the backpropagation algorithm and the Adam optimizer. S5: L2 regularization and early stopping are introduced during the training process of S4 to prevent overfitting; S6: Input the test set data into the neural network trained in S5, calculate the error between the predicted converter loss and the actual converter loss, and form a fitting model of converter loss-torque current d-axis component when the error is less than 1.5%. S7: Employs the optimal tip speed ratio method to achieve maximum power point tracking (MPPT) for permanent magnet wind turbines, and collects wind speed data in real time. v Wind turbine speed data ω Calculate the actual tip speed ratio λ With the optimal tip speed ratio λ opt Compare and calculate the reference speed based on the deviation. ω The wind turbine speed is adjusted by the speed controller, and the reference speed is set. ω The deviation from the actual rotational speed is input to a proportional-integral (PI) controller, which outputs a q-axis command current. i q ; S8: Input the initial value of the d-axis component of the torque current under real-time operating conditions into the converter loss-torque current d-axis component fitting model formed in S6 to predict the converter loss. Use an ergonomic search algorithm to traverse candidate torque current d-axis component values under the current constraint of the permanent magnet synchronous motor, and select the candidate value with the minimum converter loss as the optimal value of the torque current d-axis component with a step size of 0.01 times the rated current. ; S9: Convert the q-axis command current obtained from S7 i q The optimal value of the d-axis component of the torque current obtained from S8 As the input to the current controller, a proportional-integral (PI) control algorithm is used, based on the actual d-axis current. i d q-axis current i q Calculate the d-axis voltage reference value based on the deviation from the commanded current. V d q-axis voltage reference value V q ; S10: Obtain the d-axis voltage reference value from S9 V d q-axis voltage reference value V q Input the space vector pulse width modulation (SVPWM) modulation module to generate the pulse width modulation (PWM) drive signal for the machine-side converter; S11: Integrates the data acquisition and preprocessing, neural network fitting, maximum power point tracking, loss minimization current optimization, current control, and space vector pulse width modulation (SVPWM) modules of S1~S10 to form a complete efficiency-optimized control link for permanent magnet wind power systems, enabling real-time data interaction among the modules.
2. The data-driven optimal control method for permanent magnet synchronous wind turbine generators according to claim 1, characterized in that: In S1, the simulation platform is the MATLAB / Simulink platform, and the data samples cover operating scenarios with wind speeds of 3~15m / s and load rates of 10%~100%.
3. The data-driven optimal control method for permanent magnet synchronous wind turbine generators according to claim 1, characterized in that: In S3, the number of hidden layer neurons is set by trial and error or empirical formulas and adjusted according to model performance during training.
4. The data-driven optimal control method for permanent magnet synchronous wind turbine generators according to claim 1, characterized in that: In S4, the number of training iterations is 100 to 200 rounds. If the loss on the validation set does not decrease for 10 consecutive rounds, the training is stopped.
5. The data-driven optimal control method for permanent magnet synchronous wind turbine generators according to claim 1, characterized in that: In step S6, if the error is not less than 1.5%, the number of hidden layer neurons or the learning rate of the Adam optimizer are adjusted, and steps S4 to S5 are re-executed until the error is less than 1.5%.
6. The data-driven optimal control method for permanent magnet synchronous wind turbine generators according to claim 1, characterized in that: In S8, the traversal search algorithm takes the initial value of the d-axis component of the torque current as a reference, and traverses the candidate torque current d-axis component values with a step size of 0.01 times the rated current under the current constraint condition of the permanent magnet synchronous motor.
7. The data-driven optimal control method for permanent magnet synchronous wind turbine generators according to claim 1, characterized in that: In step S10, the space vector pulse width modulation (SVPWM) module calculates the conduction time of each switch based on the voltage space vector distribution law.
8. A data-driven optimal control system for permanent magnet synchronous wind turbine generators, characterized in that: It includes a data acquisition and preprocessing module, a neural network fitting module, a maximum power point tracking module, a loss minimization current optimization module, a current control module, and a space vector pulse width modulation (SVPWM) module, which are connected in sequence for data interaction. The data acquisition and preprocessing module is used to acquire converter loss data (PCloss) and torque current d-axis component data. After cleaning and normalization, the output is sent to the neural network fitting module; The neural network fitting module is used to form a converter loss-torque current d-axis component fitting model with an error of less than 1.5% based on a multi-layer forward feedback neural network, and output it to the loss minimization current optimization module. The maximum power tracking module is used to output the q-axis command current using the optimal tip speed ratio method. i q To the current control module; The loss minimization current optimization module is used to input the initial value of the real-time torque current d-axis component into the converter loss-torque current d-axis component fitting model, and use a traversal search algorithm to select the optimal value of the torque current d-axis component under current constraints. And output to the current control module; The current control module is used to control the current according to the q-axis command. i q Optimal value of d-axis component of torque current and actual d-axis current i d q-axis current i q The d-axis voltage reference value is calculated using a proportional-integral (PI) control algorithm. V d q-axis voltage reference value V q And output to the Space Vector Pulse Width Modulation (SVPWM) module; The space vector pulse width modulation (SVPWM) module is used to adjust the d-axis voltage reference value. V d q-axis voltage reference value V q The pulse width modulation (PWM) drive signal for the generator-side converter is generated.
9. The data-driven optimal control system for permanent magnet synchronous wind turbine generators according to claim 8, characterized in that: The neural network fitting module uses mean squared error (MSE) as the loss function, backpropagation algorithm, Adam optimizer, L2 regularization, and early stopping method during the training process. The current optimization module for minimizing losses uses a traversal search algorithm with a step size of 0.01 times the rated current.
10. The data-driven optimal control system for permanent magnet synchronous wind turbine generators according to claim 8, characterized in that: The normalization formula for the data acquisition and preprocessing module is as follows: in, For the normalized data, x For current data, The minimum value in a set of data. The maximum value in a set of data; The maximum power tracking module collects wind speed data in real time. v Wind turbine speed data ω Calculate the actual tip speed ratio λ With the optimal tip speed ratio λ opt In comparison, the deviation is used to output the q-axis command current through a proportional-integral (PI) controller. i q .