Wind turbine generator rotating speed control method based on LSTM closed-loop prediction and related device
By introducing an LSTM closed-loop prediction model into wind turbine units and combining it with a PID controller, feedforward control of wind speed and rotational speed is achieved. This solves the problem of slow dynamic response of traditional PID controllers in large wind turbine units, improves the accuracy and stability of rotational speed control, and reduces fatigue damage to mechanical components.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-01
- Publication Date
- 2026-04-07
AI Technical Summary
Traditional PID controllers cannot predict the changing trends of wind speed and rotational speed in large wind turbine units, resulting in problems such as slow dynamic response, increased overshoot, and aggravated fatigue load.
A wind turbine speed control method based on LSTM closed-loop prediction is adopted. By collecting real-time operating data of the wind turbine, the LSTM speed prediction model is used to output the future speed prediction sequence, generate a feedforward control compensation signal, and fuse it with the PID controller to generate the final control command, thereby realizing the active adjustment of the wind turbine.
It improves the operational stability and control precision of wind turbines under complex wind conditions, reduces fatigue damage to mechanical components, and enhances the accuracy and timeliness of speed prediction.
Smart Images
Figure CN121803399A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of wind power generation technology, and in particular to a wind turbine speed control method and related device based on LSTM closed-loop prediction. Background Technology
[0002] As the wind power industry develops towards larger and higher-power models, the rotor diameter and overall inertia of wind turbines have increased significantly. Against this backdrop, wind turbine speed control is no longer merely a matter of optimizing power generation efficiency, but a key technology directly related to the structural safety and service life of the entire unit. Due to the increased swept area of the rotor, the absorbed wind energy increases quadratically, while the turbine mass increases approximately cubically, resulting in enormous system inertia and slow dynamic response. Simultaneously, natural wind conditions exhibit strong randomness and turbulence, leading to frequent aerodynamic load abrupt changes. This contradiction between "rapid load changes" and "slow system response" is particularly pronounced under conditions of drastic wind speed fluctuations, making it extremely difficult to track the set speed. Increased speed overshoot or response delay can significantly increase the fatigue load on the turbine, even triggering system instability risks. Currently, the main control system of wind turbines generally adopts the traditional PID (Proportional-Integral-Derivative) controller as the core strategy for speed control. As a classic linear time-invariant (LTI) system design, the PID controller's control performance is highly dependent on an accurate system model and fixed control parameters. However, wind turbines are actually complex nonlinear and time-varying systems. Their dynamic speed characteristics are influenced by a combination of factors, including aerodynamic torque (nonlinear), transmission chain flexibility (time-varying), and turbulent wind speed (stochastic). Therefore, when traditional PID controllers are applied to large wind turbines, their pure feedback mechanism means they can only take corrective action after a speed deviation has occurred, failing to anticipate trends in wind speed and speed changes. Under load conditions with fluctuating wind speeds, this "reactive adjustment" mode causes the control system to constantly passively chase the target, easily leading to repeated speed oscillations. This not only degrades dynamic control performance but also significantly exacerbates fatigue damage to mechanical components, shortening the turbine's lifespan. Summary of the Invention
[0003] This invention provides a wind turbine speed control method and related device based on LSTM closed-loop prediction, which solves the defects of traditional PID speed control, which is unable to predict the trend of wind speed and speed changes due to inherent lag, resulting in slow dynamic response, increased overshoot and aggravated fatigue load in large wind turbines.
[0004] This invention provides a wind turbine speed control method based on LSTM closed-loop prediction, comprising: Collect real-time operating sequence data of the wind turbine, including wind speed, rotational speed, pitch angle, and generator torque; The wind speed, turbine speed, and blade pitch angle are input into a pre-trained LSTM speed prediction model, which outputs a speed prediction sequence for a future time domain. Based on the speed prediction sequence and the preset speed, a feedforward control compensation signal is generated; The real-time fan speed is compared with the preset speed to obtain the speed error, and the speed error is input into the PID controller to obtain the PID control output command; The confidence level is determined by comparing the predicted value of the speed prediction sequence with the real-time wind turbine speed. When the confidence level is passed, the feedforward control compensation signal and the PID control output command are fused to generate the final control command. The final control command is sent to the actuator of the wind turbine to adjust the operating status of the wind turbine.
[0005] According to the wind turbine speed control method based on LSTM closed-loop prediction provided by the present invention, when the confidence level judgment fails, the PID control output command is used as the control command.
[0006] According to the wind turbine speed control method based on LSTM closed-loop prediction provided by the present invention, the step of determining the confidence level based on the predicted value of the speed prediction sequence and the real-time wind turbine speed includes: Calculate the average absolute error between the predicted value of the speed prediction sequence in the most recent time window and the actual measured value of the real-time wind turbine speed at the corresponding moment; The mean absolute error is compared with a preset confidence threshold; If the mean absolute error is less than the confidence threshold, then the confidence level determination is passed; If the mean absolute error is greater than or equal to the confidence threshold, then the confidence judgment fails.
[0007] According to the wind turbine speed control method based on LSTM closed-loop prediction provided by the present invention, the LSTM speed prediction model is trained by the following method: The collected historical time-series data were normalized. A fixed-length window slides along the normalized time series data, and a segment of historical time series data captured within the window is used as an input sample. The actual rotation speed values at one or more times immediately following the window are used as the target output labels corresponding to the input samples. A training sample set is constructed based on the pairing relationship between multiple input samples and target output labels; The LSTM network is trained offline using the training sample set to obtain a pre-trained LSTM rotation speed prediction model.
[0008] According to the wind turbine speed control method based on LSTM closed-loop prediction provided by the present invention, the step of inputting the speed error into a PID controller to obtain a PID control output command includes: The product of the speed error and the proportional coefficient is calculated to obtain the proportional term instruction of the speed error; The integral term instruction of the speed error is obtained by multiplying the historical cumulative value of the speed error by the integral coefficient. The differential term of the speed error is obtained by multiplying the instantaneous rate of change of the speed error by the differential coefficient. The sum of the proportional term, the integral term, and the derivative term is used as the PID control output command.
[0009] According to the wind turbine speed control method based on LSTM closed-loop prediction provided by the present invention, the step of generating a feedforward control compensation signal based on the speed prediction sequence and the preset speed includes: Calculate the trend of the predicted rotational speed sequence relative to time to obtain the predicted rotational speed change rate; Based on the predicted rate of change of rotational speed and the preset rotational speed, the pitch rate compensation amount is calculated, and the pitch rate compensation amount is used as the feedforward control compensation signal.
[0010] According to the wind turbine speed control method based on LSTM closed-loop prediction provided by the present invention, the step of fusing the feedforward control compensation signal with the PID control output command to generate the final control command includes: The PID control output command is converted into the corresponding pitch rate reference command. The pitch rate compensation amount and the pitch rate reference command are algebraically superimposed to generate a fused pitch rate command. The fused pitch rate command is subjected to rate and position limiting processing to generate the final control command.
[0011] The present invention also provides a wind turbine speed control device based on LSTM closed-loop prediction, comprising: The data acquisition module is used to collect real-time operating sequence data of the wind turbine, including wind speed, rotational speed, pitch angle and generator torque. The first output module is used to input the wind speed, wind turbine speed and blade pitch angle into the pre-trained LSTM speed prediction model and output the speed prediction sequence for a future set time domain. The generation module is used to generate a feedforward control compensation signal based on the speed prediction sequence and the preset speed; The second output module is used to compare the collected real-time fan speed with the preset speed to obtain the speed error, and input the speed error into the PID controller to obtain the PID control output command; The fusion module is used to determine the confidence level based on the predicted value of the speed prediction sequence and the real-time wind turbine speed. When the confidence level is passed, the feedforward control compensation signal and the PID control output command are fused to generate the final control command. The adjustment module is used to send the control commands to the actuator of the wind turbine to adjust the operating state of the wind turbine.
[0012] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the wind turbine speed control method based on LSTM closed-loop prediction as described in any of the preceding claims.
[0013] The present invention also provides a non-transitory readable storage medium storing a computer program thereon, wherein the computer program, when executed by a processor, implements the wind turbine speed control method based on LSTM closed-loop prediction as described above.
[0014] This invention provides a wind turbine speed control method and related apparatus based on LSTM closed-loop prediction. The method involves collecting real-time operating sequence data of the wind turbine, including wind speed, turbine speed, pitch angle, and generator torque. This data is then input into a pre-trained LSTM speed prediction model, which outputs a predicted speed sequence for a future time domain. A feedforward control compensation signal is generated based on the predicted speed sequence and a preset speed. The collected real-time turbine speed is compared with the preset speed to obtain a speed error, which is then input into a PID controller to generate a PID control output command. A confidence level is determined based on the predicted value of the speed prediction sequence and the real-time turbine speed. If the confidence level is passed, the feedforward control compensation signal and the PID control output command are fused to generate a final control command. This control command is then sent to the wind turbine's actuator to adjust the wind turbine's operating state. Compared to traditional PID control, this LSTM closed-loop prediction speed control method has stronger nonlinear modeling capabilities and better time-series feature extraction, enabling it to more accurately capture the dynamic changes in the wind turbine's operating state. In complex and variable wind conditions, the LSTM model automatically extracts key features through deep learning from historical data, avoiding the control lag and deviation problems caused by the reliance on fixed parameters and linear assumptions in traditional PID control. Furthermore, the long-term memory mechanism of the LSTM network enables it to effectively handle long-period dependencies in time-series data, significantly improving the accuracy and timeliness of speed prediction. Combined with a closed-loop protection control strategy, it can quickly respond to and suppress speed deviations when dealing with load conditions characterized by repeated fluctuations in wind speed and engine speed, reducing system overshoot and settling time, further enhancing the operational stability and control accuracy of wind turbines under complex wind conditions and reducing fatigue loads. Attached Figure Description
[0015] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0016] Figure 1 This is one of the flowcharts of the wind turbine speed control method based on LSTM closed-loop prediction provided in the embodiments of the present invention; Figure 2 This is the second flowchart of the wind turbine speed control method based on LSTM closed-loop prediction provided in the embodiments of the present invention; Figure 3 This is a functional structure diagram of the wind turbine speed control device based on LSTM closed-loop prediction provided in an embodiment of the present invention. Figure 4 This is a functional structure diagram of the electronic device provided in the embodiments of the present invention. Detailed Implementation
[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.
[0018] Figure 1 A flowchart of the wind turbine speed control method based on LSTM closed-loop prediction provided in an embodiment of the present invention is shown below. Figure 1 As shown, the wind turbine speed control method based on LSTM closed-loop prediction provided in this embodiment of the invention includes: Step 101: Collect real-time operating sequence data of the wind turbine, including but not limited to wind speed, rotational speed, pitch angle and generator torque; Step 102: Input the wind speed, rotational speed, blade pitch angle and generator torque into the pre-trained LSTM rotational speed prediction model, and output the rotational speed prediction sequence for the future set time domain; Step 103: Generate a feedforward control compensation signal based on the predicted rotational speed sequence and the preset rotational speed; Step 104: Compare the collected real-time fan speed with the preset speed to obtain the speed error, and input the speed error into the PID controller to obtain the PID control output command; In this embodiment of the invention, the real-time wind turbine speed collected is used to update the input of the LSTM speed prediction model to achieve closed-loop prediction.
[0019] Step 105: Calculate the confidence level based on the predicted value of the speed prediction sequence and the real-time fan speed. When the confidence level is passed, fuse the feedforward control compensation signal with the PID control output command to generate the final control command. Step 106: Send the final control command to the actuator of the wind turbine to adjust the operating status of the wind turbine.
[0020] In this embodiment of the invention, when the confidence level judgment fails, the PID control output instruction is used as the control instruction.
[0021] This invention utilizes a deep learning model based on a Long Short-Term Memory Network (LSTM) for wind turbine speed prediction and protection. By analyzing historical and real-time multi-dimensional time-series data such as wind speed, wind speed, and blade pitch angle, a nonlinear mapping relationship of the wind turbine's dynamic behavior is established. The model can capture long-term dependent features, predict future wind speed trends, and trigger control actions in advance before actual overspeed occurs.
[0022] LSTM (Laser-Semiconductor Memory) networks are a type of recurrent neural network (RNN) that processes input data by iterating through time steps and updating the RNN state. The RNN state contains information remembered from all previous time steps. LSTM neural networks can be used to predict longer time series and subsequent values of the series, such as 10-30 seconds, using previous time steps as input. The model can be continuously trained and optimized to achieve high prediction accuracy and dynamic response capabilities. During wind turbine operation, this method can predict the trend of speed changes caused by wind speed fluctuations in advance and dynamically adjust key control variables such as pitch angle and generator torque based on the prediction results, achieving proactive control of wind turbine speed. Under load conditions with repeated fluctuations in wind speed and speed, proactive compensation adjustments can be made in advance based on deviations from set values, significantly improving the stability of speed operation, reducing fatigue damage to mechanical components, and improving the overall operating efficiency and confidence of the wind turbine.
[0023] When traditional PID controllers are applied to large wind turbines, their pure feedback mechanism means they can only take corrective action after a speed deviation has occurred, failing to anticipate changes in wind speed and engine speed. Under load conditions with fluctuating wind speeds, this "reactive adjustment" mode causes the control system to constantly passively chase the target, easily leading to repeated speed oscillations. This not only degrades dynamic control performance but also significantly exacerbates fatigue damage to mechanical components, shortening the turbine's lifespan.
[0024] The wind turbine speed control method based on LSTM closed-loop prediction provided in this invention collects real-time operating sequence data of the wind turbine, including wind speed, speed, pitch angle, and generator torque. The wind speed, speed, pitch angle, and generator torque are input into a pre-trained LSTM speed prediction model, which outputs a speed prediction sequence for a future set time domain. Based on the speed prediction sequence and a preset speed, a feedforward control compensation signal is generated. The collected real-time wind turbine speed is compared with the preset speed to obtain the speed error, which is then input into a PID controller to obtain a PID control output command. A confidence level is determined based on the predicted value of the speed prediction sequence and the real-time wind turbine speed. If the confidence level is passed, the feedforward control compensation signal and the PID control output command are fused to generate a final control command. The control command is sent to the actuator of the wind turbine to adjust the operating state of the wind turbine. Compared with traditional PID control, the LSTM closed-loop prediction speed control method of this invention has stronger nonlinear modeling capabilities and advantages in time-series feature extraction, and can more accurately capture the dynamic changes in the operating state of the wind turbine. In complex and variable wind conditions, the LSTM model automatically extracts key features through deep learning from historical data, avoiding the control lag and deviation problems caused by the reliance on fixed parameters and linear assumptions in traditional PID control. Furthermore, the long-term memory mechanism of the LSTM network enables it to effectively handle long-period dependencies in time-series data, significantly improving the accuracy and timeliness of speed prediction. Combined with a closed-loop protection control strategy, it can quickly respond to and suppress speed deviations when dealing with load conditions characterized by repeated fluctuations in wind speed and engine speed, reducing system overshoot and settling time, further enhancing the operational stability and control accuracy of wind turbines under complex wind conditions and reducing fatigue loads.
[0025] Based on any of the above embodiments, the step of determining the confidence level based on the predicted value of the speed prediction sequence and the real-time wind turbine speed includes: Step 201: Calculate the average absolute error between the predicted value of the speed prediction sequence in the most recent time window and the actual measured value of the real-time wind turbine speed at the corresponding moment; Step 202: Compare the mean absolute error with a preset confidence threshold; Step 203: If the mean absolute error is less than the confidence threshold, then the confidence level determination is passed; Step 204: If the mean absolute error is greater than or equal to the confidence threshold, then the confidence judgment fails.
[0026] This invention utilizes an LSTM model to predict the timing pattern of rotational speed, generating feedforward control commands before disturbances affect the system, thus overcoming the lag of large inertia systems.
[0027] Based on any of the above embodiments, the LSTM speed prediction model training method includes: Step 301: Normalize the collected historical time-series data; Step 302: Slide a window of fixed length along the normalized time series data, take a segment of historical time series data captured within the window as an input sample, and take the actual rotation speed values of one or more times immediately following the window as the target output label corresponding to the input sample. Step 303: Construct a training sample set based on the pairing relationship between multiple input samples and target output labels; Step 304: Use the training sample set to train the LSTM network offline to obtain a pre-trained LSTM rotation speed prediction model.
[0028] The embodiments of the present invention directly construct dynamic models of complex nonlinear systems by learning from multivariate data rather than physical formulas, thus avoiding the simplification and uncertainty of mechanism modeling.
[0029] Based on any of the above embodiments, the step of inputting the speed error into the PID controller to obtain a PID control output command includes: Step 401: Calculate the product of the speed error and the proportional coefficient to obtain the proportional term of the speed error; Step 402: Calculate the product of the historical cumulative value of the speed error and the integral coefficient to obtain the integral term of the speed error; Step 403: Calculate the product of the instantaneous rate of change of the speed error and the differential coefficient to obtain the differential term of the speed error; Step 404: The sum of the proportional term instruction, the integral term instruction, and the derivative term instruction is taken as the PID control output instruction.
[0030] Based on any of the above embodiments, the step of generating a feedforward control compensation signal based on the speed prediction sequence and the preset speed includes: Step 501: Calculate the trend of the predicted rotational speed sequence relative to time to obtain the predicted rotational speed change rate; Step 502: Based on the predicted speed change rate and the preset speed, calculate the pitch rate compensation amount, and use the pitch rate compensation amount as the feedforward control compensation signal.
[0031] Based on any of the above embodiments, the step of fusing the feedforward control compensation signal with the PID control output command to generate the final control command includes: Step 601: Convert the PID control output command into the corresponding pitch rate reference command. Step 602: Algebraically superimpose the pitch rate compensation amount and the pitch rate reference command to generate a fused pitch rate command. Step 603: Perform rate and position limiting processing on the fused pitch rate command to generate the final control command.
[0032] The embodiments of the present invention utilize the LSTM open-loop model to predict rotational speed. When dealing with load conditions where wind speed and rotational speed fluctuate repeatedly, the technology can quickly respond and suppress rotational speed deviation, thereby reducing system overshoot and settling time.
[0033] The flow chart of the wind turbine speed control method based on LSTM closed-loop prediction provided by this invention is as follows: Figure 2 As shown, a training sample set is constructed by collecting real-time operating data such as wind speed, engine speed, blade pitch angle, and generator torque, and normalizing the data. Next, an LSTM network structure is designed, including the number of nodes in the input, hidden, and output layers. A sliding window method is used to reconstruct the time-series data, enhancing the model's ability to capture dynamic characteristics. The LSTM model is trained offline using the sample set data, and hyperparameters are adjusted using a validation set to ensure good generalization ability and prediction accuracy. Then, the trained LSTM closed-loop prediction model is embedded in the actual control system of the wind turbine. Real-time data generated by the wind turbine is input to the LSTM prediction model, which can output a real-time prediction of the engine speed change trend in the next few seconds. The prediction result is used as a feedforward signal input to the controller. Finally, the controller calculates the confidence interval based on the input feedforward engine speed (future 10-30s) and the measured data. If the calculated MAE is <3%, the model prediction result is considered reliable, and the feedforward PID fusion control strategy can continue to be executed; otherwise, the traditional PID control strategy is executed to ensure system stability and safety. The feedforward PID fusion control strategy calculates the rate of change between the input feedforward speed sequence and the set speed, then calculates the pitch rate compensation rate of change, and incorporates and corrects the traditional PID calculation results to form the feedforward PID fusion control strategy. This strategy can quickly respond to and suppress speed deviations, reducing system overshoot and settling time when dealing with load conditions characterized by repeated fluctuations in wind speed and engine speed. This further improves the operational stability and control accuracy of wind turbines under complex wind conditions and reduces fatigue loads.
[0034] This invention provides a wind turbine speed control method based on LSTM closed-loop prediction. By introducing LSTM closed-loop predictive control, it can compensate for speed change trends in advance, solving the control lag problem of traditional PID control due to wind turbine overspeed caused by large inertia and gusts. A closed-loop predictive model is constructed based on LSTM. By learning the temporal characteristics of historical wind speed and speed data, it achieves accurate prediction of future speed trends (e.g., 10-30 seconds). The prediction result is then used as a feedforward signal into the control loop, forming a predictive-driven control strategy. This compensates for the lag in PID response, achieving fast, stable, and accurate speed tracking control. For operating conditions with drastic wind speed fluctuations, it can effectively reduce overshoot and response delay, improve speed stability, and reduce fatigue load.
[0035] The following describes the wind turbine speed control device based on LSTM closed-loop prediction provided by the present invention. The wind turbine speed control device based on LSTM closed-loop prediction described below can be referred to in correspondence with the wind turbine speed control method based on LSTM closed-loop prediction described above.
[0036] Figure 3 The functional structure diagram of the wind turbine speed control device based on LSTM closed-loop prediction provided in the embodiments of the present invention is as follows: Figure 3 As shown, the wind turbine speed control device based on LSTM closed-loop prediction provided in this embodiment of the invention includes: The data acquisition module 301 is used to acquire real-time operating sequence data of the wind turbine, including wind speed, turbine speed and blade pitch angle. The first output module 302 is used to input the wind speed, wind turbine speed and blade pitch angle into the pre-trained LSTM speed prediction model and output the speed prediction sequence for a future set time domain. The generation module 303 is used to generate a feedforward control compensation signal based on the speed prediction sequence and the preset speed; The second output module 304 is used to compare the collected real-time fan speed with the preset speed to obtain the speed error, and input the speed error into the PID controller to obtain the PID control output command; The fusion module 305 is used to determine the confidence level based on the predicted value of the speed prediction sequence and the real-time wind turbine speed. When the confidence level is passed, the feedforward control compensation signal and the PID control output command are fused to generate the final control command. The adjustment module 306 is used to send the control command to the actuator of the wind turbine to adjust the operating state of the wind turbine.
[0037] The wind turbine speed control device based on LSTM closed-loop prediction provided in this invention collects real-time operating sequence data of the wind turbine, including wind speed, speed, pitch angle, and generator torque. The wind speed, speed, pitch angle, and generator torque are input into a pre-trained LSTM speed prediction model, which outputs a speed prediction sequence for a future set time domain. Based on the speed prediction sequence and a preset speed, a feedforward control compensation signal is generated. The collected real-time wind turbine speed is compared with the preset speed to obtain the speed error, which is then input into a PID controller to obtain a PID control output command. A confidence level is determined based on the predicted value of the speed prediction sequence and the real-time wind turbine speed. If the confidence level is passed, the feedforward control compensation signal and the PID control output command are fused to generate a final control command. The control command is sent to the actuator of the wind turbine to adjust the operating state of the wind turbine. Compared with traditional PID control, the LSTM closed-loop prediction speed control in this invention has stronger nonlinear modeling capabilities and advantages in time-series feature extraction, enabling it to more accurately capture the dynamic changes in the operating state of the wind turbine. In complex and variable wind conditions, the LSTM model automatically extracts key features through deep learning from historical data, avoiding the control lag and deviation problems caused by the reliance on fixed parameters and linear assumptions in traditional PID control. Furthermore, the long-term memory mechanism of the LSTM network enables it to effectively handle long-period dependencies in time-series data, significantly improving the accuracy and timeliness of speed prediction. Combined with a closed-loop protection control strategy, it can quickly respond to and suppress speed deviations when dealing with load conditions characterized by repeated fluctuations in wind speed and engine speed, reducing system overshoot and settling time, further enhancing the operational stability and control accuracy of wind turbines under complex wind conditions and reducing fatigue loads.
[0038] Figure 4 An example is a schematic diagram of the physical structure of a communication device, such as... Figure 4As shown, the communication device may include: a processor 410, a communication interface 420, a memory 430, and a communication bus 440, wherein the processor 410, the communication interface 420, and the memory 430 communicate with each other through the communication bus 440. The memory 430 includes a computer program, an operating system, and acquired data. The processor 410 can call the logic instructions in the memory 430 to execute a wind turbine speed control method based on LSTM closed-loop prediction. The method includes: acquiring real-time operating sequence data of the wind turbine, including wind speed, speed, pitch angle, and generator torque; inputting the wind speed, speed, pitch angle, and generator torque into a pre-trained LSTM speed prediction model and outputting a speed prediction sequence for a future set time domain; generating a feedforward control compensation signal based on the speed prediction sequence and a preset speed; comparing the acquired real-time wind turbine speed with the preset speed to obtain a speed error, and inputting the speed error into a PID controller to obtain a PID control output command; performing a confidence judgment based on the predicted value of the speed prediction sequence and the real-time wind turbine speed; when the confidence judgment passes, fusing the feedforward control compensation signal and the PID control output command to generate a final control command; and sending the control command to the actuator of the wind turbine to adjust the operating state of the wind turbine.
[0039] Furthermore, the logical instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to related technologies, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0040] On the other hand, the present invention also provides a non-transient readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the LSTM closed-loop prediction-based wind turbine speed control method provided by the above methods. The method includes: acquiring real-time operating sequence data of the wind turbine, the time sequence data including wind speed, turbine speed, and pitch angle; inputting the wind speed, turbine speed, and pitch angle into a pre-trained LSTM speed prediction model, and outputting a speed prediction sequence for a future set time domain; generating a feedforward control compensation signal based on the speed prediction sequence and a preset speed; comparing the acquired real-time turbine speed with the preset speed to obtain a speed error, and inputting the speed error into a PID controller to obtain a PID control output command; performing a confidence judgment based on the predicted value of the speed prediction sequence and the real-time turbine speed; when the confidence judgment passes, fusing the feedforward control compensation signal and the PID control output command to generate a final control command; and sending the control command to the actuator of the wind turbine to adjust the operating state of the wind turbine.
[0041] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.
[0042] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of software products. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.
[0043] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A wind turbine speed control method based on LSTM closed-loop prediction, characterized in that, include: Collect real-time operating sequence data of the wind turbine, including wind speed, rotational speed, pitch angle and generator torque; The wind speed, rotational speed, blade pitch angle, and generator torque are input into a pre-trained LSTM rotational speed prediction model, which outputs a rotational speed prediction sequence for a future time domain. Based on the speed prediction sequence and the preset speed, a feedforward control compensation signal is generated; The real-time fan speed is compared with the preset speed to obtain the speed error, and the speed error is input into the PID controller to obtain the PID control output command; The confidence level is determined by comparing the predicted value of the speed prediction sequence with the real-time wind turbine speed. When the confidence level is passed, the feedforward control compensation signal and the PID control output command are fused to generate the final control command. The final control command is sent to the actuator of the wind turbine to adjust the operating status of the wind turbine.
2. The wind turbine speed control method based on LSTM closed-loop prediction according to claim 1, characterized in that, If the confidence level judgment fails, the PID control output command is used as the final control command.
3. The wind turbine speed control method based on LSTM closed-loop prediction according to claim 1, characterized in that, The step of determining the confidence level based on the predicted values of the speed prediction sequence and the real-time wind turbine speed includes: Calculate the average absolute error between the predicted value of the speed prediction sequence in the most recent time window and the actual measured value of the real-time wind turbine speed at the corresponding moment; The mean absolute error is compared with a preset confidence threshold; If the mean absolute error is less than the confidence threshold, then the confidence level determination is passed; If the mean absolute error is greater than or equal to the confidence threshold, then the confidence judgment fails.
4. The wind turbine speed control method based on LSTM closed-loop prediction according to claim 1, characterized in that, The LSTM speed prediction model was trained using the following method: The collected historical time-series data were normalized. A fixed-length window slides along the normalized time series data, and a segment of historical time series data captured within the window is used as an input sample. The actual rotation speed values at one or more times immediately following the window are used as the target output labels corresponding to the input samples. A training sample set is constructed based on the pairing relationship between multiple input samples and target output labels; The LSTM network is trained offline using the training sample set to obtain a pre-trained LSTM rotation speed prediction model.
5. The wind turbine speed control method based on LSTM closed-loop prediction according to claim 1, characterized in that, The step of inputting the speed error into the PID controller to obtain the PID control output command includes: The product of the speed error and the proportional coefficient is calculated to obtain the proportional term instruction of the speed error; The integral term instruction of the speed error is obtained by multiplying the historical cumulative value of the speed error by the integral coefficient. The differential term of the speed error is obtained by multiplying the instantaneous rate of change of the speed error by the differential coefficient. The sum of the proportional term, the integral term, and the derivative term is used as the PID control output command.
6. The wind turbine speed control method based on LSTM closed-loop prediction according to claim 1, characterized in that, The step of generating a feedforward control compensation signal based on the speed prediction sequence and the preset speed includes: Calculate the trend of the predicted rotational speed sequence relative to time to obtain the predicted rotational speed change rate; Based on the predicted rate of change of rotational speed and the preset rotational speed, the pitch rate compensation amount is calculated, and the pitch rate compensation amount is used as the feedforward control compensation signal.
7. The wind turbine speed control method based on LSTM closed-loop prediction according to claim 6, characterized in that, The step of fusing the feedforward control compensation signal with the PID control output command to generate the final control command includes: The PID control output command is converted into the corresponding pitch rate reference command. The pitch rate compensation amount and the pitch rate reference command are algebraically superimposed to generate a fused pitch rate command. The fused pitch rate command is subjected to rate and position limiting processing to generate the final control command.
8. A wind turbine speed control device based on LSTM closed-loop prediction, characterized in that, include: The data acquisition module is used to collect real-time operating sequence data of the wind turbine, including wind speed, rotational speed, pitch angle and generator torque. The first output module is used to input the wind speed, rotational speed, blade pitch angle and generator torque into the pre-trained LSTM rotational speed prediction model and output the rotational speed prediction sequence for a future set time domain. The generation module is used to generate a feedforward control compensation signal based on the speed prediction sequence and the preset speed; The second output module is used to compare the collected real-time fan speed with the preset speed to obtain the speed error, and input the speed error into the PID controller to obtain the PID control output command; The fusion module is used to determine the confidence level based on the predicted value of the speed prediction sequence and the real-time wind turbine speed. When the confidence level is passed, the feedforward control compensation signal and the PID control output command are fused to generate the final control command. The adjustment module is used to send the final control command to the actuator of the wind turbine to adjust the operating state of the wind turbine.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the wind turbine speed control method based on LSTM closed-loop prediction as described in any one of claims 1 to 7.
10. A non-transitory readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the wind turbine speed control method based on LSTM closed-loop prediction as described in any one of claims 1 to 7.