Fan control method, storage medium and electronic equipment
By constructing digital twin models of virtual wind farms and virtual wind turbines, and combining deep reinforcement learning algorithms to optimize wind turbine control strategies, the problems of non-physical output and insufficient personalized adaptation of wind turbine control algorithms are solved, and stable and efficient control of wind turbines under extreme operating conditions is achieved.
Patent Information
- Application Number
- CN202511303612.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-12
- Publication Date
- 2025-11-18
AI Technical Summary
Existing wind turbine control algorithms suffer from abnormal non-physical outputs and insufficient personalized adaptation, causing the control strategy to exceed the operating parameter limits under extreme conditions, and resulting in significant differences in control performance between different wind turbines.
A digital twin model, including a virtual wind farm and a virtual wind turbine, is constructed to simulate different wind conditions. The wind turbine control strategy model is optimized by combining deep reinforcement learning algorithms. The control parameters are optimized by the operating status of the virtual wind turbine under different wind conditions to ensure that the wind turbine meets the trade-off objectives of power generation efficiency, mechanical component load and grid dispatch under actual wind conditions.
It realizes the physical output and personalized adaptive control of wind turbines under various wind conditions, overcomes the defects of traditional algorithms, and ensures that wind turbines meet the comprehensive goals of power generation efficiency, mechanical component safety and grid response under actual wind conditions.
Smart Images
Figure CN120969044A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of wind turbine control, in particular to a wind turbine control method, a storage medium and an electronic device. BACKGROUND
[0002] The wind turbine is a system for converting the kinetic energy of wind into electric energy. Due to the high randomness and uncertainty of actual wind conditions, the wind turbine needs to meet multiple goals such as stable power generation efficiency, mechanical component load within the design range, and response speed of the wind turbine to grid dispatching during operation. However, these goals are in conflict if they must be met simultaneously, so the wind turbine needs to make real-time trade-offs.
[0003] Currently, the algorithm for realizing the above real-time trade-offs of the wind turbine completely relies on data operation results to realize the control of the wind turbine. The control strategy obtained under extreme working conditions may exceed the operating parameter limits of the wind turbine, and non-physical output may easily occur. In addition, the traditional algorithm does not consider individual differences between different wind turbines (such as blade aerodynamic asymmetry, gear box wear, etc.). If a unified control algorithm is used to control each wind turbine, the control performance of the same type of wind turbine will differ greatly. SUMMARY
[0004] The technical problem to be solved by the present application is that the control algorithm of the wind turbine in the prior art has abnormal non-physical output and insufficient individual adaptation, and thus a wind turbine control method, a storage medium and an electronic device are provided.
[0005] In a first aspect, the technical solution of the present application provides a wind turbine control method, comprising:
[0006] constructing a digital twin model and a wind turbine control strategy model, the digital twin model comprising a virtual wind farm and a virtual wind turbine, the virtual wind farm simulating different wind conditions, and the virtual wind turbine being constructed according to an actual wind turbine;
[0007] the wind turbine control strategy model is used to control the virtual wind turbine under different wind conditions to make the virtual wind turbine simulate the operating state under different wind conditions;
[0008] determining the power generation efficiency parameter, the mechanical component load parameter and the response speed parameter of the virtual wind turbine to grid dispatching of the virtual wind turbine according to the operating state of the virtual wind turbine under different wind conditions;
[0009] optimizing the regulation and control parameters in the wind turbine control strategy model according to the comprehensive index of the power generation efficiency parameter, the mechanical component load parameter and the response speed parameter;
[0010] loading the optimized wind turbine control strategy model into an actual wind turbine corresponding to the virtual wind turbine, for controlling the actual wind turbine according to actual wind conditions under an actual wind farm.
[0011] Preferably, the wind turbine control method, the digital twin model and the wind turbine control strategy model are constructed, the digital twin model comprises a virtual wind farm and a virtual wind turbine, the virtual wind farm simulates different wind conditions, and the virtual wind turbine is constructed according to an actual wind turbine: the different wind conditions include different combinations of wind direction and wind speed.
[0012] The wind turbine control strategy model is used to control the virtual wind turbine under different wind conditions, so that the virtual wind turbine simulates the operating state under different wind conditions: the operating state simulated by the virtual wind turbine includes multiple fault states.
[0013] Preferably, the wind turbine control method, the virtual wind turbine is provided with physical mechanism and constraint information, including:
[0014] Aerodynamic constraints enable the virtual wind turbine to calculate the actual captured wind energy according to the wind conditions;
[0015] Transmission chain dynamics constraints ensure that the energy transmission process conforms to the law of mechanical motion in the process of converting wind energy into electrical energy by the virtual wind turbine;
[0016] Structural dynamics constraints enable the virtual wind turbine to simulate the stress of the actual wind turbine structure;
[0017] Actuator and safety constraints enable the virtual wind turbine to meet the limitation conditions of the mechanical performance of the real wind turbine in the process of executing control actions.
[0018] Preferably, the wind turbine control method, the wind turbine control strategy model is used to control the virtual wind turbine under different wind conditions, so that the virtual wind turbine simulates the operating state under different wind conditions, including:
[0019] The wind turbine control strategy model obtains the virtual wind speed and virtual wind direction in the virtual wind farm and the physical mechanism and constraint information of the virtual wind turbine;
[0020] The wind turbine control strategy model determines the operating parameters of the virtual wind turbine according to the virtual wind speed and the virtual wind direction, in combination with the physical mechanism and constraint information, and controls the virtual wind turbine according to the operating parameters, the operating parameters including a target pitch angle.
[0021] Preferably, the wind turbine control method, the comprehensive index according to the power generation efficiency parameter, the mechanical component load parameter and the response speed parameter is used to optimize the regulation and control parameters in the wind turbine control strategy model, including:
[0022] The comprehensive index is calculated as follows: Reward = w1 × ΔPower - w2 × (Load_Feedback) 2 -w3×(RPM-RPM_ref) 2 Where w1, w2, and w3 are weighting coefficients, ΔPower represents the power generation efficiency parameter, Load_Feedback represents the mechanical component load parameter, RPM represents the actual revolutions, RPM_ref represents the reference revolutions, and (RPM-RPM_ref) represents the response speed parameter; Reward represents the comprehensive index, and the larger the value, the better.
[0023] The control strategy model for the wind turbine includes weighting coefficients set in the following loss function: Total_Loss=Loss_power+λ1×Loss_load+λ2×Loss_grid; where λ1 and λ2 are weighting coefficients, Loss_power represents the power loss value, Loss_load represents the load loss value, Loss_grid represents the response speed value, and Total_Loss represents the total loss value, with smaller values being better.
[0024] During the process of adjusting the control parameters of the wind turbine control strategy model, the maximum value of the comprehensive index is located, and the control parameter corresponding to the maximum value is used as the optimization result of the control parameters in the wind turbine control strategy model.
[0025] Preferably, in the aforementioned wind turbine control method, the wind turbine control strategy model is trained in the following manner:
[0026] The chosen deep reinforcement learning algorithm model embedding physical information includes an input layer, hidden layers, an output layer, and a training and loss function layer, wherein:
[0027] The input layer receives input wind speed, rotor speed, generator speed, current blade pitch angle, and current power, calculates the tip speed ratio and available wind power, and outputs them.
[0028] The hidden layer receives the tip speed ratio and available wind power sent by the input layer, and calculates the target blade pitch angle;
[0029] The output layer outputs the target pitch angle;
[0030] The training and loss function layer calculates the total loss value when the wind turbine runs at the target pitch angle; and adjusts the network parameters in the input layer, the hidden layer, and the output layer in reverse according to the relationship between the total loss value and a set loss threshold.
[0031] The above steps are cycled until the total loss value is less than a set loss threshold or the cycle number reaches a set number of times.
[0032] The physical information embedded deep reinforcement learning algorithm model after the training process is completed is used as the fan control strategy model.
[0033] Preferably, in the fan control method, the input layer calculates the tip speed ratio λ and the available wind power P_available in the following manner:
[0034] λ = (ω_r x R) / V; wherein ω_r represents the rotor speed, V represents the wind speed, and R represents the blade radius;
[0035] P_available = 0.5 x p x π x R 2 x V 3 ; wherein p represents the air density;
[0036] The hidden layer determines the target pitch angle according to the tip speed ratio optimal control strategy in the low wind speed area and determines the target pitch angle according to the power non-exceeding limit control strategy in the high wind speed area.
[0037] In a second aspect, the technical scheme of the present application provides a computer readable storage medium, wherein the storage medium stores program information, and a computer reads the program information and executes the steps of the fan control method according to any one of the first aspect.
[0038] In a third aspect, the technical scheme of the present application provides a computer program product, which includes computer programs / instructions, and the computer programs / instructions are executed by a processor to realize the steps of the fan control method according to any one of the first aspect.
[0039] In a fourth aspect, the technical scheme of the present application provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory, and the processor executes the computer program to realize the steps of the fan control method according to any one of the first aspect.
[0040] Compared with the prior art, the above technical scheme provided by the present application has the following technical effects:
[0041] The wind turbine control method, storage medium, and electronic equipment provided in this application construct a virtual wind turbine corresponding to an actual wind turbine, and control the virtual wind turbine's operating state under simulated different wind conditions in a virtual wind farm. The virtual wind turbine is controlled using a wind turbine control strategy model. Based on the control results, it is determined whether the virtual wind turbine's operating state meets the trade-off objectives of its power generation efficiency parameters, mechanical component load parameters, and response speed parameters to grid dispatch. During the control process, the control parameters in the wind turbine control strategy model are optimized based on these parameters. The optimized wind turbine control strategy model can control the virtual wind turbine's operating state under various wind conditions to meet the trade-off objectives. Only then is the wind turbine control strategy model loaded into the corresponding actual wind turbine. Because the virtual wind turbine is constructed based on the actual wind turbine, if the virtual wind turbine can meet the trade-off objectives, the actual wind turbine, under the control of this model, can also meet the trade-off objectives under various actual wind conditions. Furthermore, the obtained outputs are all physical outputs verified by the virtual wind turbine, and are completely personalized adaptive control based on each actual wind turbine, overcoming the shortcomings of traditional algorithms. Attached Figure Description
[0042] Figure 1 This is a flowchart of a fan control method according to an embodiment of the present invention;
[0043] Figure 2 This is a schematic diagram of the structure of a fan control system according to an embodiment of the present invention;
[0044] Figure 3 This is a schematic diagram of the hardware connection relationship of an electronic device that performs the fan control method according to an embodiment of the present invention. Detailed Implementation
[0045] The specific embodiments of this application will be further described below with reference to the accompanying drawings.
[0046] It is readily understood that, based on the technical solution of this application, various structural and implementation methods can be interchanged by those skilled in the art without altering the essential spirit of this application. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this application and should not be considered as the entirety of this application or as limitations or restrictions on the technical solution of the application.
[0047] This embodiment provides a fan control method, such as Figure 1 As shown, it includes the following steps:
[0048] S100: Construct a digital twin model and a wind turbine control strategy model. The digital twin model includes a virtual wind farm and a virtual wind turbine. The virtual wind farm simulates different wind conditions, and the virtual wind turbine is constructed based on the actual wind turbine.
[0049] like Figure 2As shown, the digital twin model is arranged on an industrial computer, and the virtual wind farm is simulated by a professional wind farm simulation tool (such as GH Bladed, FAST) or a self-developed simulation module, with "multi-physical field coupling simulation" as the core, to reproduce the wind condition characteristics of the actual wind farm. The virtual wind turbine is driven by a physical mechanism, and is based on the design drawings, component parameters (such as blade size, transmission chain inertia, actuator performance), and measured data (such as historical running speed, torque, load) of the actual wind turbine, to construct a high-fidelity model coupled with multiple subsystems. The main purpose of this step is to simulate the operation of the actual wind turbine by the virtual wind turbine, and therefore, the accuracy of the constructed results can be verified by tests, for example, collecting the running data (such as speed, power, and variable pitch angle) of the actual wind turbine under typical wind conditions (such as wind speed of 10 m / s), simulating the virtual wind turbine under the same wind conditions, comparing the deviation (such as power deviation ≤ 5%, speed deviation ≤ 2%) between the "virtual output data" and the "real measured data", adjusting the model parameters (such as Cp curve correction coefficient, transmission chain damping coefficient) to reduce the deviation, and ensuring the consistency of the dynamic response of the virtual wind turbine with that of the actual wind turbine.
[0050] The wind turbine control strategy model is used to perform virtual control, which can use a deep learning algorithm, etc., and is obtained after training and optimization of the internal parameters of the virtual wind turbine.
[0051] S200: The wind turbine control strategy model is used to control the virtual wind turbine under different wind conditions, so as to simulate the running state of the virtual wind turbine under different wind conditions.
[0052] This step simulates the running state of the virtual wind turbine under different wind conditions through real-time interaction between the wind turbine control strategy model and the digital twin model.
[0053] S300: According to the running state of the virtual wind turbine under different wind conditions, the power generation efficiency parameter, mechanical component load parameter, and response speed parameter of the virtual wind turbine to grid scheduling are determined.
[0054] Based on the running data of the virtual wind turbine in S200, three types of key parameters, i.e., power generation efficiency, mechanical load, and grid response, are extracted and quantified to form a quantitative index system of control strategy performance.
[0055] S400: optimizing the regulating parameters in the wind turbine control strategy model according to the comprehensive indexes of the power generation efficiency parameter, the mechanical component load parameter and the response speed parameter. The power generation efficiency parameter is used to measure the ability of the control strategy to capture wind energy under different wind conditions, which can be calculated based on the power output data of the virtual wind turbine and the available wind power data of the virtual wind farm. The mechanical component load parameter is used to evaluate the protection ability of the control strategy to the wind turbine structure, which can be calculated based on the load data of the virtual wind turbine structure dynamics subsystem, focusing on the "fatigue load" and "limit load" of the key components. The response speed parameter is used to evaluate the flexibility of the control strategy under the grid dispatching scenario, which can be calculated based on the response data of the virtual wind turbine to the "grid power instruction".
[0056] S500: loading the optimized wind turbine control strategy model into the actual wind turbine corresponding to the virtual wind turbine for controlling the actual wind turbine under actual wind conditions according to actual wind conditions.
[0057] In specific implementation, the optimized wind turbine control strategy model of S400 can be processed in a lightweight manner, compatible with the actual wind turbine PLC control host, and loaded on site to ensure stable and safe control of the actual wind turbine in the real wind farm.
[0058] Specifically, since the PLC control host of the actual wind turbine has lower computing power than the industrial computer, the optimized wind turbine control strategy model needs to be processed in a lightweight manner, retaining the core control logic and removing redundant calculation modules. For example, model distillation algorithm is used to realize lightweight, knowledge distillation technology is used, the optimized complex model is used as the "teacher model", a "student model" with smaller parameter size is trained, the core decision-making ability is retained by minimizing the deviation between the "student model output" and the "teacher model output", and redundant parameters with absolute value less than a threshold are deleted to reduce the model size, which can be usually compressed to 1 / 3-1 / 5 of the original model.
[0059] Further, the above scheme can further include a subsequent step of continuous optimization to ensure long-term stable operation of the optimized model. For example, the running data (wind speed, rotating speed, power, load, control instruction) of the actual wind turbine are collected regularly and uploaded to the industrial computer; based on the above data, the optimized model is updated in the digital twin environment, and the updated model is repeated in the lightweight-loading process of S500 to be placed in the actual wind turbine, realizing continuous improvement of the model iteration.
[0060] The above embodiments of the present application control the running state of the virtual wind turbine under different simulated wind conditions in the virtual wind field by constructing a virtual wind turbine corresponding to an actual wind turbine. The virtual wind turbine is controlled by a wind turbine control strategy model. According to the control result, it is determined whether the running state of the virtual wind turbine can meet the trade-off target of the power generation efficiency parameter, the mechanical component load parameter and the response speed parameter of the virtual wind turbine to the power grid scheduling. In the control process, the regulation and control parameters in the wind turbine control strategy model are optimized according to the above parameters. The optimized wind turbine control strategy model can control the running state of the virtual wind turbine under each different wind condition to meet the trade-off target. Then, the wind turbine control strategy model is loaded into the corresponding actual wind turbine. Since the virtual wind turbine is constructed by comparing the actual wind turbine, the actual wind turbine can meet the trade-off target under various actual wind conditions under the control of the model, and the physical output verified by the virtual wind turbine is obtained, and the individualized adaptive control is completely performed according to each actual wind turbine, which overcomes the defects of the traditional algorithm.
[0061] Preferably, in step S100 of the above scheme, when the virtual wind field simulates different wind conditions, the different wind conditions include different combinations of wind direction and wind speed. The extreme wind condition refers to a special wind environment that poses a significant threat to the structural safety, operation stability and power generation performance of the wind turbine beyond the normal design working condition of the wind turbine. The core feature of this type of wind condition is that the wind speed, wind direction or turbulence intensity far exceeds the normal range, which may directly cause the wind turbine to shut down, component damage, etc. The running state simulated by the virtual wind turbine includes a plurality of fault states. The fault states can include 12 types of faults such as blade jamming fault, generator overload fault, controller communication fault, etc. Correspondingly, the fault response control logic is pre-embedded in the wind turbine control strategy model to ensure that the virtual wind turbine does not directly crash during subsequent fault simulation, thereby providing a basic fault tolerance capability for subsequent control of the actual wind turbine. For example, the deviation between the 1st blade pitch angle feedback and the command is 3°, which is determined as jamming; the command "increase the pitch angle of the 2nd and 3rd blades by 2°" is output to balance the wind wheel stress; at the same time, the generator torque is reduced by 5% to avoid overspeed of the wind wheel; the structural dynamics subsystem of the virtual wind turbine calculates the change of the blade load (the bending moment of the jammed blade increases by 10%, and the bending moment of the other blades decreases by 5%). Through the above simulation, the control strategy optimization for all working conditions can be realized for various wind conditions and various faults.
[0062] Preferably, in the above scheme, the virtual wind turbine is provided with physical mechanism and constraint information, including: aerodynamic constraints, enabling the virtual wind turbine to calculate the actual captured wind energy according to the wind conditions; enabling the virtual wind turbine to calculate the actual captured wind energy according to the wind speed, rotational speed, and variable pitch angle through the Cp curve, avoiding the virtual power generation efficiency from deviating from the physical law. Transmission chain dynamics constraints, the virtual wind turbine ensures that the energy transmission process (wind rotor → main shaft → gearbox → generator) conforms to the mechanical motion law in the process of converting wind energy into electrical energy; for example, simulating torque impact when the rotational speed fluctuates, avoiding the virtual transmission chain from being "never damaged". Structural dynamics constraints enable the virtual wind turbine to simulate the actual wind turbine structure stress (such as tower vibration, blade fatigue load), and ensure that the virtual wind turbine can be used to test the structural safety; actuator and safety constraints enable the virtual wind turbine to meet the limitation conditions of the mechanical performance of the real wind turbine in the process of executing control actions (such as pitch speed, torque adjustment), for example, the virtual pitch system cannot instantaneously turn from 0° to 90°, ensuring that the control strategy can be directly applied to the actual wind turbine after being verified in the virtual environment. The above four physical mechanisms and constraints are the core of the virtual wind turbine, ensuring that it can accurately reproduce the behavior of the actual wind turbine. This scheme can achieve active smooth output torque, reduce shaft torsional vibration, suppress tower front and rear vibration and blade swing, thereby significantly reducing fatigue load and extending the service life of main components (such as main shaft, gearbox, blade) of the wind turbine by more than 20%.
[0063] Further preferably, the wind turbine control strategy model in step S200 is used to control the virtual wind turbine under different wind conditions, so that the virtual wind turbine simulates the operating state under different wind conditions, including:
[0064] S201: The wind turbine control strategy model obtains the virtual wind speed and virtual wind direction in the virtual wind farm, and the physical mechanism and constraint information of the virtual wind turbine.
[0065] In this step, in addition to the basic virtual wind speed (including turbulence intensity, wind speed shear) and virtual wind direction (including wind direction angle change rate), real-time state parameters of the virtual wind turbine can also be obtained according to actual needs, such as nacelle azimuth angle, impeller rotational speed, real-time pitch angle of blade, transmission shaft torque; generator output power, converter DC side voltage, grid frequency; air density, temperature, air pressure, etc.
[0066] S202: The wind turbine control strategy model determines the operating parameters of the virtual wind turbine according to the virtual wind speed and the virtual wind direction, in combination with the physical mechanism and constraint information, and controls the virtual wind turbine according to the operating parameters, the operating parameters including target pitch angle.
[0067] In this step, according to the aerodynamic model: based on the blade element momentum theory (BEM) to calculate the aerodynamic load (such as lift, drag, thrust coefficient) under different wind speed, pitch angle, rotating speed; According to the transmission chain dynamics model: simulate the elastic deformation and damping characteristics of the gearbox, transmission shaft, establish the torque transmission relationship; According to the generator efficiency model: combined with the efficiency curve of the generator, determine the optimal power generation efficiency interval. In addition, according to the constraint information, such as the pitch angle adjustment range (usually-5°~30°), the maximum rotating speed of the impeller (such as 18rpm), the cabin yaw angle speed limit (such as 0.5° / s); The upper limit of the generator output power (to avoid overload), the maximum current threshold of the converter; The shutdown threshold under the limit wind speed (such as cut-out wind speed 25m / s), the blade load peak limit (to prevent structural fatigue) and the like can be obtained.
[0068] The fan control strategy model can determine the target pitch angle according to the hierarchical control strategy, for example, in the low wind speed area (such as 3~12m / s): taking the maximum wind energy capture as the target, the power maximization is realized through the optimal tip speed ratio (TSR) control; In the high wind speed area (such as 12~25m / s): taking the constant power output as the target, the input energy is limited through the pitch angle adjustment to protect the safety of the unit.
[0069] Further preferably, in step S300, the optimization of the regulation and control parameters in the fan control strategy model according to the comprehensive index of the power generation efficiency parameter, the mechanical component load parameter and the response speed parameter comprises:
[0070] The comprehensive index is calculated in the following manner: Reward=w1×ΔPower-w2×(Load_Feedback)2-w3×(RPM-RPM_ref)2; wherein w1, w2 and w3 are weight coefficients, ΔPower represents the power generation efficiency parameter, Load_Feedback represents the mechanical component load parameter, which can be calculated by (Load-Load_limit), Load represents the load, Load_limit represents the load limit value; RPM represents the actual rotating number, RPM_ref represents the reference rotating number, (RPM-RPM_ref) represents the response speed parameter, which can also be replaced by (Speed-Speed_limit), Speed represents the impeller rotating speed, Speed_limit represents the design limit value of the impeller rotating speed; Reward represents the comprehensive index, the larger the value is, the better.
[0071] The control parameter in the fan control strategy model includes a weight coefficient set in a loss function: Total_Loss = Loss_power + λ1 x Loss_load + λ2 x Loss_grid; wherein λ1 and λ2 are weight coefficients, Loss_power represents a power loss value, Loss_load represents a load loss value, Loss_grid represents a response speed value, and Total_Loss represents a total loss value, and the smaller the value is, the better. In the above loss function, the power loss can be estimated by using the target pitch angle and the current pitch angle output by the network, estimating the power P_est through a physical formula, calculating MSE(P_est, P_measured), and P_measured is the measured power. The load loss can be calculated by establishing a simple load model (such as the relationship between thrust and pitch angle and wind speed), calculating the estimated value of the thrust, and punishing the value max(0, Thrust_est-Thrust_limit) that exceeds the safety threshold. Thrust_est is the estimated value of the thrust, and Thrust_limit is the safety threshold. The response speed can be calculated according to the difference between the target pitch angle and the actual pitch angle. The maximum value of the comprehensive index is located in the process of adjusting the control parameter of the fan control strategy model, and the control parameter corresponding to the maximum value is used as the optimization result of the control parameter in the fan control strategy model.
[0072] Preferably, the fan control strategy model is trained by the following method:
[0073] A deep reinforcement learning algorithm model embedded with physical information, i.e., a neural network, is selected, and in the training loss function of the neural network, in addition to the conventional "data loss" (such as mean square error), a "physical loss" term is added, and the physical loss term corresponds to the comprehensive index.
[0074] It includes an input layer, a hidden layer, an output layer, and a training and loss function layer, wherein:
[0075] The input layer receives input wind speed, rotor speed, generator speed, current pitch angle, current power, calculates tip speed ratio and available wind power and outputs; further, the above inputs can also include wind direction, impeller speed, generator speed, hub height, air density, tip speed ratio λ, turbulence intensity, wind shear coefficient, wind turbine component load, etc. Specifically, the input layer calculates the tip speed ratio λ and the available wind power P available by: λ = (ω r x R) / V; wherein: ω r represents rotor speed, V represents wind speed, and R represents blade radius; P available = 0.5 x p x p x R2 x V3; wherein, p represents air density; the hidden layer receives the tip speed ratio and the available wind power sent by the input layer, and calculates a target pitch angle; specifically, the hidden layer uses an activation function to realize through multiple fully connected layers (such as 256, 128 neurons), which learn complex patterns from data and physical characteristics, and the hidden layer determines the target pitch angle according to the control strategy optimal for the tip speed ratio in the low wind speed area, and determines the target pitch angle according to the power not over-limit control strategy in the high wind speed area. In addition, the output can also include target generator torque, control action, value function, etc., which are set according to actual needs. The output layer can select a linear activation function to output the target pitch angle. The training and loss function layer calculates the total loss value when the wind turbine runs at the target pitch angle; the network parameters in the input layer, the hidden layer and the output layer are adjusted reversely according to the relationship between the total loss value and the set loss threshold. The above steps are repeated until the total loss value is less than the set loss threshold or the number of cycles reaches the set number of times. The deep reinforcement learning algorithm model with physical information embedded after completing the above training process is used as the wind turbine control strategy model. Through the above scheme of the present application, the grid dispatching instruction (such as power limit, frequency response) can be used as one of the optimization targets, so that the wind turbine changes from traditional "passive response" to "active support". It can smooth power fluctuations, reduce grid connection impact and improve the grid connection performance of wind farms as friendly power sources.
[0076] The embodiment of the present application provides a computer readable storage medium, wherein the storage medium stores program information, and the computer reads the program information to execute the steps of the wind turbine control method in any one of the above schemes.
[0077] The embodiment of the present application provides a computer program product, comprising a computer program / instruction, which, when executed by a processor, realizes the steps of the wind turbine control method in any one of the above schemes.
[0078] The embodiment of the present application also provides an electronic device, such as Figure 3As shown, the electronic device includes at least one processor 31 and at least one memory 32, at least one of the memories 32 stores program information, and at least one of the processors 31 reads the program information and executes the fan control method according to any of the above method embodiments. The device can also include an input device 33 and an output device 34. The processor 31, the memory 32, the input device 33 and the output device 34 can be communicatively connected. The memory 32, as a non-volatile computer readable storage medium, can be used to store non-volatile software programs, non-volatile computer executable programs and modules. The processor 31 executes various functional applications and data processing by running the non-volatile software programs, instructions and modules stored in the memory 32, that is, implements the fan control method provided in any of the above solutions. The memory 32 can include a program storage area and a data storage area, wherein the program storage area can store an operating system and at least one application required by a function; the data storage area can store data created according to the use of the fan control method, etc. In addition, the memory 32 can include a high-speed random access memory, and can also include a non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 32 can optionally include a memory remotely arranged with respect to the processor 31, and these remote memories can be connected to the device executing the fan control method through a network. Examples of the above network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof. The input device 33 can receive input user clicks and generate signal inputs related to user settings and function control of the fan control method. The output device 34 can include a display device such as a display screen. When the one or more modules are stored in the memory 32 and are run by the one or more processors 31, the fan control method in any of the above method embodiments is executed.
[0079] According to the needs, the above technical solutions can be combined to achieve the best technical effect.
[0080] The above is only the principles and preferred embodiments of the present application. It should be noted that for those skilled in the art, on the basis of the principles of the present application, a number of other variants can also be made, which should also be considered as the protection scope of the present application.
Claims
1. A method of controlling a fan, characterized by, The method comprises the following steps: constructing a digital twin model and a wind turbine control strategy model, the digital twin model comprising a virtual wind farm and a virtual wind turbine, the virtual wind farm simulating different wind conditions, and the virtual wind turbine being constructed according to an actual wind turbine; the wind turbine control strategy model being used to control the virtual wind turbine under different wind conditions so that the virtual wind turbine simulates the operating state under different wind conditions; determining the power generation efficiency parameter, the mechanical component load parameter and the response speed parameter of the virtual wind turbine according to the operating state of the virtual wind turbine under different wind conditions; optimizing the regulation and control parameters in the wind turbine control strategy model according to the comprehensive index of the power generation efficiency parameter, the mechanical component load parameter and the response speed parameter; loading the optimized wind turbine control strategy model into the actual wind turbine corresponding to the virtual wind turbine, and using the wind turbine control strategy model to control the actual wind turbine according to the actual wind conditions in the actual wind farm.
2. The wind turbine control method according to claim 1, wherein: in the step of constructing a digital twin model and a wind turbine control strategy model, the digital twin model comprising a virtual wind farm and a virtual wind turbine, the virtual wind farm simulating different wind conditions, and the virtual wind turbine being constructed according to an actual wind turbine, the different wind conditions comprise different combinations of wind direction and wind speed; in the step of using the wind turbine control strategy model to control the virtual wind turbine under different wind conditions so that the virtual wind turbine simulates the operating state under different wind conditions, the virtual wind turbine simulates the operating state under different wind conditions, and the operating state simulated by the virtual wind turbine comprises multiple fault states.
3. The fan control method of claim 2, wherein The virtual wind turbine is provided with physical mechanism and constraint information, comprising: aerodynamic constraint, which enables the virtual wind turbine to calculate the actual captured wind energy according to the wind conditions; transmission chain dynamics constraint, which ensures that the energy transmission process conforms to the law of mechanical movement in the process of converting wind energy into electrical energy by the virtual wind turbine; structural dynamics constraint, which enables the virtual wind turbine to simulate the stress of the actual wind turbine structure; actuator and safety constraint, which enables the virtual wind turbine to meet the limitation conditions of the mechanical properties of the real wind turbine in the process of executing control actions.
4. The fan control method according to claim 3, wherein The wind turbine control strategy model is used to control the virtual wind turbine under different wind conditions so that the virtual wind turbine simulates the operating state under different wind conditions, comprising: the wind turbine control strategy model acquires the virtual wind speed and virtual wind direction in the virtual wind farm and the physical mechanism and constraint information of the virtual wind turbine; the wind turbine control strategy model determines the operating parameters of the virtual wind turbine according to the virtual wind speed and the virtual wind direction in combination with the physical mechanism and constraint information and controls the virtual wind turbine according to the operating parameters, the operating parameters comprising a target pitch angle.
5. The fan control method of claim 2, wherein The optimization of the regulation and control parameters in the wind turbine control strategy model according to the comprehensive index of the power generation efficiency parameter, the mechanical component load parameter and the response speed parameter comprises: The comprehensive index is calculated as follows: Reward = w1 × ΔPower - w2 × (Load_Feedback) 2 -w3×(RPM-RPM_ref) 2 Where w1, w2, and w3 are weighting coefficients, ΔPower represents the power generation efficiency parameter, Load_Feedback represents the mechanical component load parameter, RPM represents the actual revolutions, RPM_ref represents the reference revolutions, and (RPM-RPM_ref) represents the response speed parameter; Reward represents the comprehensive index, and the larger the value, the better. The regulation parameter in the fan control strategy model includes a weight coefficient set in a loss function: Total_Loss = Loss_power + λ1 x Loss_load + λ2 x Loss_grid; wherein λ1 and λ2 are weight coefficients, Loss_power represents a power loss value, Loss_load represents a load loss value, Loss_grid represents a response speed value, and Total_Loss represents a total loss value, and the smaller the value is, the better it is; The maximum value of the comprehensive index is located in the process of adjusting the regulation parameter of the fan control strategy model, and the regulation parameter corresponding to the maximum value is taken as the optimization result of the regulation parameter in the fan control strategy model.
6. The fan control method of claim 5, wherein The fan control strategy model is trained in the following manner: A deep reinforcement learning algorithm model with physical information embedding is selected, which includes an input layer, a hidden layer, an output layer, and a training and loss function layer, wherein: The input layer receives input wind speed, rotor speed, generator speed, current pitch angle, and current power, calculates tip speed ratio and available wind power, and outputs; The hidden layer receives the tip speed ratio and the available wind power sent by the input layer, and calculates a target pitch angle; The output layer outputs the target pitch angle; The training and loss function layer calculates the total loss value of the fan running at the target pitch angle; and reversely adjusts the network parameters in the input layer, the hidden layer, and the output layer according to the relationship between the total loss value and the set loss threshold value; The above steps are repeated until the total loss value is less than the set loss threshold value or the number of cycles reaches the set number of times; The deep reinforcement learning algorithm model with physical information embedding that completes the above training process is taken as the fan control strategy model.
7. The fan control method according to claim 6, wherein: The input layer calculates the tip speed ratio λ and the available wind power P_available in the following manner: λ = (ω_r x R) / V; wherein ω_r represents rotor speed, V represents wind speed, and R represents blade radius; P available = 0.5 x p x p x R 2 x V 3 ; wherein p represents air density; The hidden layer determines the target pitch angle according to the control strategy of optimal tip speed ratio in the low wind speed area, and determines the target pitch angle according to the power non-exceeding limit control strategy in the high wind speed area.
8. A computer-readable storage medium, characterized in that, The storage medium stores program information, and the computer reads the program information and executes the steps of the fan control method according to any one of claims 1-7.
9. A computer program product, characterised in that, The computer program / instruction is characterized in that the computer program / instruction is executed by the processor to realize the steps of the fan control method according to any one of claims 1-7.
10. An electronic device comprising a memory, a processor, and a computer program stored on the memory, wherein the computer program, when executed by the processor, is arranged to perform the method of any one of claims 1 to 9. The processor executes the computer program to realize the steps of the fan control method according to any one of claims 1-7.