System and method for optimizing control parameters of wind turbine generator
By integrating improved particle swarm optimization and gradient descent algorithms through a modular intelligent optimization system, wind turbine parameters are automatically adjusted, solving the problems of time-consuming and experience-dependent traditional manual tuning. This achieves efficient and reliable parameter optimization and improves wind energy utilization efficiency.
Patent Information
- Application Number
- CN202511015227.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-23
- Publication Date
- 2025-11-21
AI Technical Summary
Modern wind turbine control parameter tuning is time-consuming and relies on engineers' experience, making it difficult to meet the refined requirements of new standards such as IEC61400-27-2. Traditional manual tuning is inefficient and cannot fully unleash the potential of wind energy.
The intelligent optimization system adopts a modular design, including a core algorithm module, a parameter constraint module, a simulation processing module, and a UI interface module. It integrates an improved particle swarm optimization algorithm, gradient descent, and Adam algorithm, and evaluates parameter optimization through simulation models to achieve automated parameter adjustment and optimization.
Significantly reduce labor costs, improve parameter optimization efficiency, lower the professional experience threshold, ensure the global optimal solution, quickly find the optimal control parameters of wind turbine units, and improve wind energy utilization efficiency.
Smart Images
Figure CN120993726A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the technical field of wind turbine control parameter optimization, and in particular to a wind turbine control parameter optimization system and method based on intelligent optimization algorithms. Background Technology
[0002] Modern wind turbine control systems have evolved into complex, multi-level, multi-objective systems. At the electrical control level, vector control technology, with its decoupled torque and excitation current control characteristics, can achieve a maximum power point tracking accuracy of over 98%, making it the mainstream solution for megawatt-class wind turbines. As the core of the control system, the PID controller's parameter tuning quality directly affects the turbine's dynamic response characteristics. Wind turbines using optimized PID parameters can reduce drivetrain mechanical loads by 15%-20% and extend the lifespan of key components by 3-5 years. In terms of mechanical control, pitch control systems have evolved from simple proportional control to intelligent systems incorporating load observers and frequency domain vibration suppression.
[0003] However, wind turbine parameter tuning still faces significant challenges. Industry data shows that a single wind turbine has over 200 control parameters, and traditional manual tuning is time-consuming and relies heavily on engineers' experience. With the implementation of new standards such as IEC61400-27-2, the requirements for the precision of control systems will further increase. Only by breaking through the limitations of traditional manual tuning and building a closed-loop intelligent control system of "perception-decision-optimization" can the potential of wind energy be fully unleashed. Summary of the Invention
[0004] The purpose of this invention is to overcome the shortcomings of the prior art and propose a wind turbine control parameter optimization system and method based on intelligent optimization algorithms. It adopts a modular design concept and constructs an intelligent optimization system composed of five core components. The system architecture consists of an algorithm core module, a parameter constraint module, an optimal determination module, a simulation processing module, and a UI interface module. The modules work together to form a complete parameter optimization closed loop, reducing the time for wind turbine parameter debugging and improving work efficiency.
[0005] The objective of this invention is achieved through the following technical solution: a wind turbine control parameter optimization system based on an intelligent optimization algorithm, comprising:
[0006] The core algorithm module integrates an improved particle swarm optimization algorithm and a gradient descent combined with Adam algorithm, which is used to iteratively adjust parameters and obtain performance feedback by calling the simulation processing module to find the optimal parameters;
[0007] The parameter limiting module is used to ensure that the parameter update speed is within a preset reasonable range, and at the same time to determine whether the updated parameter meets the defined reasonable range;
[0008] The simulation processing module calls the preset wind turbine simulation model based on multiple sets of parallel input configurations and parameters, generates simulation load results, evaluates the parameters output by the core algorithm module by comparing the user-defined load target with the actual simulation load results, and obtains performance feedback.
[0009] The optimal determination module compares the simulation load results output by the simulation processing module to determine the local optimal parameters and the global optimal parameters, and obtains the optimization results of the current wind turbine control parameters.
[0010] The UI module is used to input configurations and simultaneously displays the iteration results of the core algorithm and parameter statistics in real time.
[0011] Furthermore, the core algorithm module includes:
[0012] The improved particle swarm optimization algorithm incorporates all parameters into the algorithm optimization by increasing the particle dimension, and multiple particles can explore the solution space simultaneously. Its velocity update formula is as follows:
[0013]
[0014] After determining the particle update velocity, the particle position is updated using the following formula:
[0015]
[0016] In the above formula, w represents the inertia weight, used to balance global exploration and local development; c1 represents the strength of controlling the particle's movement towards the individual optimal position; c2 represents the strength of controlling the particle's movement towards the global optimal position; t represents the iteration round; and i represents the parameter index. This represents the i-th parameter in the t-th iteration; represents the update rate of the i-th parameter in the (t+1)-th iteration; pbest represents the locally optimal parameter.
[0017] Furthermore, the core algorithm module includes:
[0018] The gradient descent algorithm uses the gradient of the objective function with respect to the parameters, updates the parameters in the opposite direction of the gradient, and thus gradually approaches the minimum point of the objective function. The calculation formula is as follows:
[0019]
[0020] θ t+1 =θ t -ηg t
[0021] In the formula θ t This represents the current parameter value; η represents the learning rate, which controls the step size for each update. J(θ) represents the gradient operator with respect to the parameter θ. t ) represents the objective function with respect to parameter θ t The value at; g t This represents the gradient in iteration t.
[0022] Based on the gradient descent algorithm, the Adam algorithm is combined to automatically adjust the gradient learning rate, which can accelerate gradient descent, reduce oscillations, and speed up the finding of optimal parameters. At the same time, the learning rate of multiple parameters is automatically adjusted. The Adam parameter update method is as follows:
[0023] m t =β1m t-1 +(1-β1)g t
[0024] y t =β2y t-1 +(1-β2)g t 2
[0025]
[0026] In the formula, m t This represents the first-order estimate of the gradient, β1 represents the decay rate, and y t The second moment of the gradient is represented by β², the decay rate is represented by η, the global learning rate is represented by ∈, and g represents a constant to prevent the denominator from being zero. t This represents the gradient after t iterations.
[0027] Furthermore, the parameter limiting module includes:
[0028] A constraint function is constructed based on a reasonable range of user-defined parameters; during the iteration process, the following method is used to limit the parameter update rate:
[0029]
[0030] Where v represents the parameter update rate, and clip is the clipping function. When v is less than -v max Output -v max When v is greater than v max Then the output is v max ;v max Take 10%-12% of the upper and lower bounds of the parameter.
[0031] Furthermore, the optimal determination module includes:
[0032] When determining the optimal solution, it is necessary to make judgments at two levels: local optimal parameters and global optimal parameters;
[0033] In each iteration, all parameter configurations running in parallel generate corresponding simulation results. By comparing the outputs of these parallel conditions, the best-performing set of parameters in the current batch is selected as the local optimal parameters for this iteration.
[0034] As the optimization process progresses, each iteration produces a local optimum parameter. By comparing these local optimum parameters across iterations, the best combination of parameters in history is gradually selected, thereby determining the global optimum parameter in the entire optimization process.
[0035] A method for optimizing wind turbine control parameters based on intelligent optimization algorithms, implemented using the aforementioned wind turbine control parameter optimization system, includes the following steps:
[0036] (1) Configure basic files, which include fan operating condition files, control program files and related parameter files;
[0037] (2) Configure the optimization algorithm and its running parameters, call the optimization algorithm of the core algorithm module and configure its running parameters according to the optimization algorithm;
[0038] (3) Configure parameter attributes, parameter types, value ranges, and constraints;
[0039] (4) Set the target load for optimization;
[0040] (5) Call the simulation processing module to start the simulation and monitor the simulation information in real time;
[0041] (6) When the preset stop iteration optimization condition is reached, the optimal judgment module is called to output the optimization result, and the optimization result and its analysis report are displayed through the UI interface module.
[0042] Furthermore, step (2) includes:
[0043] Select an optimization algorithm and configure its operating parameters accordingly. For the improved particle swarm optimization algorithm, the operating parameters include the number of particles, the maximum number of iterations, the cognitive coefficient, the social coefficient, and the inertia weight. The number of particles is the number of particles participating in the optimization search. The maximum number of iterations is the maximum number of rounds the algorithm can run. The cognitive coefficient is the weight that adjusts the particle's movement towards its own historical best position. The social coefficient is the weight that adjusts the particle's movement towards the group's best position. The inertia weight controls the degree to which the particle maintains its previous speed. For the gradient descent algorithm, the operating parameters include the learning rate and step size of the gradient descent method.
[0044] Furthermore, step (3) includes:
[0045] Configure parameter attributes, set parameter types, and specify the data type of the parameters; set reasonable minimum and maximum values for the parameters, and the optimization process will strictly limit the parameters to search within these boundaries; at the same time, define the operating constraints that must be met during the optimization process. These constraints are implemented by calling the monitoring variables in the wind turbine simulation model or the preset evaluation function. Parameter combinations that violate the constraints are considered invalid solutions.
[0046] Furthermore, step (4) includes:
[0047] Set the target load to be optimized and define the objective function. The target load is a specific load index of the key components of the wind turbine that is minimized or maximized. The target load includes blade load, tower load and yaw system load. The objective function is the calculated value of the selected target load, which can be the maximum value, standard value and equivalent fatigue load value.
[0048] Furthermore, step (5) includes:
[0049] According to the configuration, the background wind turbine simulation model is called and iterative calculations are performed in combination with the preset optimization algorithm. In each iteration, one or more sets of parameter combinations are generated to drive the wind turbine simulation model to run and calculate the objective function value. The real-time information of the optimization process is dynamically displayed, including the current iteration number, the current optimal parameter combination and its corresponding objective function value, the distribution status of particles and warning or error messages.
[0050] Compared with the prior art, the present invention has the following advantages and beneficial effects:
[0051] 1. Significantly reduce labor costs: Eliminating the inefficient model of relying on repeated manual adjustments and tests significantly reduces high labor input costs.
[0052] 2. Improved optimization efficiency: The algorithm can run massive combinations of parameters in parallel, achieving efficient convergence of parameters within a set reasonable boundary. This ability to schedule parallel computing resources enables the optimization process to be completed quickly, saving time and resources.
[0053] 3. Lowering the professional experience threshold: Effectively reduces the reliance on in-depth professional experience in the field of parameter optimization. Even if users do not fully understand the specific physical or business meaning of the parameters, they can still efficiently and reliably locate the optimal or near-optimal parameter configuration with the help of the core algorithm module of this invention.
[0054] 4. Guaranteeing global optimal potential: The algorithm's systematic search strategy can traverse and cover the entire solution space. Theoretically, it ensures that, given sufficient computing resources and ample time, it is capable of discovering and approximating the theoretical optimal solution of the problem domain. This avoids the local optimum trap that may be caused by human experience and provides a theoretical basis for decision quality. Attached Figure Description
[0055] Figure 1 A schematic diagram of the interface for optimizing the control parameters of a wind turbine generator. Detailed Implementation
[0056] The present invention will be further described below with reference to specific embodiments.
[0057] Example 1
[0058] The wind turbine control parameter optimization system based on intelligent optimization algorithms provided in this embodiment includes:
[0059] 1) The core algorithm module integrates an improved particle swarm optimization algorithm and a gradient descent combined with Adam algorithm, which is used to iteratively adjust parameters and obtain performance feedback by calling the simulation processing module to find the optimal parameters.
[0060] The improved particle swarm optimization algorithm incorporates all parameters into the algorithm optimization by increasing the particle dimension, and multiple particles can explore the solution space simultaneously. Its velocity update formula is as follows:
[0061]
[0062] After determining the particle update velocity, the particle position is updated using the following formula:
[0063]
[0064] In the above formula, w represents the inertia weight, used to balance global exploration and local development; c1 represents the strength of controlling the particle's movement towards the individual optimal position; c2 represents the strength of controlling the particle's movement towards the global optimal position; t represents the iteration round; and i represents the parameter index. This represents the i-th parameter in the t-th iteration; represents the update rate of the i-th parameter in the (t+1)-th iteration; pbest represents the locally optimal parameter.
[0065] The gradient descent algorithm uses the gradient of the objective function with respect to the parameters, updates the parameters in the opposite direction of the gradient, and thus gradually approaches the minimum point of the objective function. The calculation formula is as follows:
[0066]
[0067] θ t+1 =θ t -ηg t
[0068] In the formula θ t This represents the current parameter value; η represents the learning rate, which controls the step size for each update. J(θ) represents the gradient operator with respect to the parameter θ.t ) represents the objective function with respect to parameter θ t The value at; g t This represents the gradient in iteration t.
[0069] Based on the gradient descent algorithm, the Adam algorithm is combined to automatically adjust the gradient learning rate, which can accelerate gradient descent, reduce oscillations, and speed up the finding of optimal parameters. At the same time, the learning rate of multiple parameters is automatically adjusted. The Adam parameter update method is as follows:
[0070] m t =β1m t-1 +(1-β1)g t
[0071] y t =β2y t-1 +(1-β2)g t 2
[0072]
[0073] In the formula, m t This represents the first-order estimate of the gradient, β1 represents the decay rate, and y t The second moment of the gradient is represented by β², the decay rate is represented by η, the global learning rate is represented by ∈, and g represents a constant to prevent the denominator from being zero. t This represents the gradient after t iterations.
[0074] 2) Parameter limiting module, used to ensure that the parameter update speed is within a preset reasonable range, and at the same time to determine whether the updated parameter meets the defined reasonable range;
[0075] A constraint function is constructed based on a reasonable range of user-defined parameters; during the iteration process, the following method is used to limit the parameter update rate:
[0076]
[0077] Where v represents the parameter update rate, and clip is the clipping function. When v is less than -v max Output -v max When v is greater than v max Then the output is v max ;v max Take 10%-12% of the upper and lower bounds of the parameter.
[0078] 3) The simulation processing module calls the preset wind turbine simulation model based on multiple sets of parallel input configurations and parameters, generates simulation load results, evaluates the parameters output by the core algorithm module by comparing the user-defined load target with the actual simulation load results, and obtains performance feedback.
[0079] 4) The optimal determination module compares the simulation load results output by the simulation processing module to determine the local optimal parameters and the global optimal parameters, and obtains the optimization results of the current wind turbine control parameters.
[0080] When determining the optimal solution, it is necessary to make judgments at two levels: local optimal parameters and global optimal parameters.
[0081] In each iteration, all parameter configurations running in parallel generate corresponding simulation results. By comparing the outputs of these parallel conditions, the best-performing set of parameters in the current batch is selected as the local optimal parameters for this iteration.
[0082] As the optimization process progresses, each iteration produces a local optimum parameter. By comparing these local optimum parameters across iterations, the best combination of parameters in history is gradually selected, thereby determining the global optimum parameter in the entire optimization process.
[0083] UI interface module, see Figure 1 As shown, it is used to input configuration, and at the same time displays the iteration results of the core algorithm and the statistical information of the parameters in real time.
[0084] Example 2
[0085] This embodiment discloses a method for optimizing wind turbine control parameters based on an intelligent optimization algorithm, implemented using the wind turbine control parameter optimization system described in Embodiment 1, and includes the following steps:
[0086] (1) Configure basic files, which include fan operating condition files, control program files and related parameter files;
[0087] Users first need to access the UI interface, see [link / reference] Figure 1 As shown, load or create the necessary configuration files. These files are the basic inputs to the optimization process and specifically include:
[0088] Wind turbine operating condition file: describes the environmental conditions under which the wind turbine operates, such as wind speed sequence, turbulence intensity, wind direction, etc., as well as the wind turbine model, such as blade length, hub radius, tower height, etc.
[0089] Control program file: Contains the source code or executable model of the wind turbine control logic to be optimized, such as PID controllers, pitch control strategies, torque control strategies, etc. The optimization process will adjust the parameters specified in this program.
[0090] Parameter file: A file containing a list of parameters to be optimized, initial values, or historical optimization results can be preloaded for easy reuse of settings.
[0091] (2) Configure the optimization algorithm and its running parameters, call the optimization algorithm of the core algorithm module and configure its running parameters according to the optimization algorithm;
[0092] See Figure 1 As shown, in Figure 1 The area is labeled "Hyperparameter Configuration" for configuration.
[0093] Select an optimization algorithm: Choose an applicable optimization algorithm from the drop-down menu or options. In this example, the improved particle swarm optimization (PSO) algorithm is selected as an example.
[0094] Configure algorithm hyperparameters: Configure the running parameters according to the selected algorithm.
[0095] For the PSO algorithm, the key hyperparameters include:
[0096] Number of particles: Sets the number of particles participating in the optimization search, which can be set to 20, 50, or 100. The number of particles directly affects the breadth of the search and the consumption of computational resources.
[0097] Maximum number of iterations: Set the maximum number of iterations the algorithm can run, 100 times.
[0098] Cognitive coefficient (c1) and social coefficient (c2): These are the weights that adjust the movement of a particle toward its own historical best position and the group's best position, respectively. c1 = 2.0, c2 = 2.0.
[0099] Inertia weight (w): controls the degree to which the particle maintains its previous velocity, w = 0.8, or adopts a linear decreasing strategy.
[0100] (3) Configure parameter attributes, parameter types, value ranges, and constraints;
[0101] See Figure 1 As shown, in Figure 1 The area marked "Parameter Configuration to be Optimized" is configured in detail.
[0102] Add / Select Parameters: Explicitly specify which parameters in the control program file need optimization. Parameters can be added by manually entering their names or by importing parameters from a parameter file for automatic identification.
[0103] Configuration parameter properties: For each parameter to be optimized, the following properties need to be set:
[0104] Parameter type: Specify the data type of the parameter, such as floating-point number or integer.
[0105] Boundaries: Define the allowed minimum and maximum values for the parameter. The range of the parameter PitchAngle is set to [0.1, 5.0]. The optimization process will strictly limit the parameter search within these boundaries.
[0106] Constraints: Define operational constraints that must be met during the optimization process, such as the generator speed not exceeding 115% of the rated speed. These constraints are implemented by calling monitoring variables in the simulation model or preset evaluation functions. Parameter combinations that violate the constraints will be considered invalid solutions and will be automatically avoided or penalized.
[0107] Initial values: Initial guesses for the parameters can be provided, either a single value or a set. For the PSO algorithm, if not provided, particle positions will be randomly initialized within the boundaries.
[0108] (4) Set the target load for optimization;
[0109] See Figure 1 As shown, in Figure 1 Configure the area as "Target Load Settings".
[0110] Select Target Components and Load Types: Users specify the target loads for optimization, aiming to minimize or maximize specific load parameters of critical wind turbine components. These parameters are derived from the output of the simulation model. Common target loads include blade loads, tower loads, and yaw system loads. Other targets include power generation, minimum headroom, and maximum load under specific operating conditions. Users can select one or more targets for multi-objective optimization.
[0111] Define the objective function: The objective function is a calculated value of a selected load index, such as the maximum value, standard value, or equivalent fatigue load value. Under the DLC1.2 condition, minimize the maximum value of the load at the blade root.
[0112] (5) Call the simulation processing module to start the simulation and monitor the simulation information in real time;
[0113] According to the configuration, the background wind turbine simulation model is called and iterative calculations are performed in combination with the preset optimization algorithm. In each iteration, one or more sets of parameter combinations are generated to drive the wind turbine simulation model to run and calculate the objective function value. The real-time information of the optimization process is dynamically displayed, including the current iteration number, the current optimal parameter combination and its corresponding objective function value, the distribution status of particles and warning or error messages.
[0114] (6) When the preset stop iteration optimization condition is reached, the optimal judgment module is called to output the optimization result, and the optimization result and its analysis report are displayed through the UI interface module.
[0115] The optimization process ends when the preset stopping conditions are met, such as reaching the maximum number of iterations, the convergence threshold, or the target load.
[0116] Statistical Information Display: The final statistical information and optimization results are displayed centrally. Figure 1The area is labeled "Statistics". These statistics include:
[0117] Optimal parameter combination: The parameter values that provide the best performance, i.e., the minimum / maximum objective function value.
[0118] Optimal target value: The objective function value corresponding to the optimal combination of parameters, such as the minimized load value.
[0119] Load statistics: Under the optimal parameter combination, detailed statistical results of all load indicators of interest to the user, including the target load and other relevant loads, such as maximum, minimum, average, standard deviation, and equivalent fatigue load. This helps to assess whether optimization has any other negative impacts.
[0120] Convergence curve: Displays the trend of the objective function value as the number of iterations, intuitively showing the optimization process and convergence status.
[0121] Users can export the above statistical information as an analysis report for subsequent applications or analysis.
[0122] In summary, this invention guides users through complex configurations via a user interface, significantly lowering the barrier to entry. Its automated optimization process replaces traditional manual trial and error, significantly saving labor costs and time. The improved particle swarm optimization algorithm effectively explores the entire parameter solution space, approximating the theoretical optimal solution within the limits of computing power and time. Boundary and constraint settings ensure the safety and feasibility of the optimization results. Real-time monitoring and detailed statistical information provide process transparency and result reliability.
[0123] Example 3
[0124] This embodiment discloses a non-transitory computer-readable medium storing instructions that, when executed by a processor, perform the steps of the wind turbine control parameter optimization method based on intelligent optimization algorithm as described in Embodiment 2.
[0125] In this embodiment, the non-transitory computer-readable medium can be a disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), USB flash drive, portable hard drive, etc.
[0126] Example 4
[0127] This embodiment discloses a computing device, including a processor and a memory for storing processor-executable programs. When the processor executes the program stored in the memory, it implements the wind turbine control parameter optimization method based on intelligent optimization algorithm described in Embodiment 2.
[0128] The computing device described in this embodiment may be a desktop computer, laptop computer, smartphone, PDA handheld terminal, tablet computer, programmable logic controller (PLC), or other terminal device with processor function.
[0129] The above-described embodiments are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Therefore, any changes made in accordance with the shape and principle of the present invention should be covered within the protection scope of the present invention.
Claims
1. A wind turbine control parameter optimization system based on intelligent optimization algorithms, characterized in that, include: The core algorithm module integrates an improved particle swarm optimization algorithm and a gradient descent combined with Adam algorithm, which is used to iteratively adjust parameters and obtain performance feedback by calling the simulation processing module to find the optimal parameters; The parameter limiting module is used to ensure that the parameter update speed is within a preset reasonable range, and at the same time to determine whether the updated parameter meets the defined reasonable range; The simulation processing module calls the preset wind turbine simulation model based on multiple sets of parallel input configurations and parameters, generates simulation load results, evaluates the parameters output by the core algorithm module by comparing the user-defined load target with the actual simulation load results, and obtains performance feedback. The optimal determination module compares the simulation load results output by the simulation processing module to determine the local optimal parameters and the global optimal parameters, and obtains the optimization results of the current wind turbine control parameters. The UI module is used to input configurations and simultaneously displays the iteration results of the core algorithm and parameter statistics in real time.
2. The wind turbine control parameter optimization system based on intelligent optimization algorithm according to claim 1, characterized in that, The core algorithm module includes: The improved particle swarm optimization algorithm incorporates all parameters into the algorithm optimization by increasing the particle dimension, and multiple particles can explore the solution space simultaneously. Its velocity update formula is as follows: After determining the particle update velocity, the particle position is updated using the following formula: In the above formula, w represents the inertia weight, used to balance global exploration and local development; c1 represents the strength of controlling the particle's movement towards the individual optimal position; c2 represents the strength of controlling the particle's movement towards the global optimal position; t represents the iteration round; and i represents the parameter index. This represents the i-th parameter in the t-th iteration; represents the update rate of the i-th parameter in the (t+1)-th iteration; pbest represents the locally optimal parameter.
3. The wind turbine control parameter optimization system based on intelligent optimization algorithm according to claim 1, characterized in that, The core algorithm module includes: The gradient descent algorithm uses the gradient of the objective function with respect to the parameters, updates the parameters in the opposite direction of the gradient, and thus gradually approaches the minimum point of the objective function. The calculation formula is as follows: i t+1 =θ t -ηg t In the formula θ t This represents the current parameter value; η represents the learning rate, which controls the step size for each update. J(θ) represents the gradient operator with respect to the parameter θ. t ) represents the objective function with respect to parameter θ t The value at; g t This represents the gradient in iteration t. Based on the gradient descent algorithm, the Adam algorithm is combined to automatically adjust the gradient learning rate, which can accelerate gradient descent, reduce oscillations, and speed up the finding of optimal parameters. At the same time, the learning rate of multiple parameters is automatically adjusted. The Adam parameter update method is as follows: m t =β1m t-1 +(1-β1)g t y t =β2y t-1 +(1-β2)g t 2 In the formula, m t This represents the first-order estimate of the gradient, β1 represents the decay rate, and y t The second moment of the gradient is represented by β², the decay rate is represented by η, the global learning rate is represented by ∈, and g represents a constant to prevent the denominator from being zero. t This represents the gradient after t iterations.
4. The wind turbine control parameter optimization system based on intelligent optimization algorithm according to claim 1, characterized in that, The parameter limiting module includes: A constraint function is constructed based on a reasonable range of user-defined parameters; during the iteration process, the following method is used to limit the parameter update rate: Where v represents the parameter update rate, and clip is the clipping function. When v is less than -v max Output -v max When v is greater than v max Then the output is v max ;v max Take 10%-12% of the upper and lower bounds of the parameter.
5. The wind turbine control parameter optimization system based on intelligent optimization algorithm according to claim 1, characterized in that, The optimal determination module includes: When determining the optimal solution, it is necessary to make judgments at two levels: local optimal parameters and global optimal parameters; In each iteration, all parameter configurations running in parallel will generate corresponding simulation results; by comparing the outputs of these parallel conditions, the best-performing set of parameters in the current batch is selected as the local optimal parameters for this iteration. As the optimization process progresses, each iteration produces a local optimum parameter. By comparing these local optimum parameters across iterations, the best combination of parameters in history is gradually selected, thereby determining the global optimum parameter in the entire optimization process.
6. A method for optimizing control parameters of a wind turbine based on an intelligent optimization algorithm, characterized in that, The wind turbine control parameter optimization system based on any one of claims 1-5 includes the following steps: (1) Configure basic files, which include fan operating condition files, control program files and related parameter files; (2) Configure the optimization algorithm and its running parameters, call the optimization algorithm of the core algorithm module and configure its running parameters according to the optimization algorithm; (3) Configure parameter attributes, parameter types, value ranges, and constraints; (4) Set the target load for optimization; (5) Call the simulation processing module to start the simulation and monitor the simulation information in real time; (6) When the preset stop iteration optimization condition is reached, the optimal judgment module is called to output the optimization result, and the optimization result and its analysis report are displayed through the UI interface module.
7. The method for optimizing wind turbine control parameters based on intelligent optimization algorithms according to claim 6, characterized in that, Step (2) includes: Select an optimization algorithm and configure its operating parameters accordingly. For the improved particle swarm optimization algorithm, the operating parameters include the number of particles, the maximum number of iterations, the cognitive coefficient, the social coefficient, and the inertia weight. The number of particles is the number of particles participating in the optimization search. The maximum number of iterations is the maximum number of rounds the algorithm can run. The cognitive coefficient is the weight that adjusts the particle's movement towards its own historical best position. The social coefficient is the weight that adjusts the particle's movement towards the group's best position. The inertia weight controls the degree to which the particle maintains its previous speed. For the gradient descent algorithm, the operating parameters include the learning rate and step size of the gradient descent method.
8. The method for optimizing wind turbine control parameters based on intelligent optimization algorithms according to claim 6, characterized in that, Step (3) includes: Configure parameter attributes, set parameter types, and specify the data type of the parameters; set reasonable minimum and maximum values for the parameters, and the optimization process will strictly limit the parameters to search within these boundaries; at the same time, define the operating constraints that must be met during the optimization process. These constraints are implemented by calling the monitoring variables in the wind turbine simulation model or the preset evaluation function. Parameter combinations that violate the constraints are considered invalid solutions.
9. The method for optimizing wind turbine control parameters based on intelligent optimization algorithms according to claim 6, characterized in that, Step (4) includes: Set the target load to be optimized and define the objective function. The target load is a specific load index of the key components of the wind turbine that is minimized or maximized. The target load includes blade load, tower load and yaw system load. The objective function is the calculated value of the selected target load, which can be the maximum value, standard value and equivalent fatigue load value.
10. The method for optimizing wind turbine control parameters based on intelligent optimization algorithms according to claim 6, characterized in that, Step (5) includes: According to the configuration, the background wind turbine simulation model is called and iterative calculations are performed in combination with the preset optimization algorithm. In each iteration, one or more sets of parameter combinations are generated to drive the wind turbine simulation model to run and calculate the objective function value. The real-time information of the optimization process is dynamically displayed, including the current iteration number, the current optimal parameter combination and its corresponding objective function value, the distribution status of particles and warning or error messages.