An Electromechanical-Electromagnetic Adaptive Co-modeling Method for Wind Turbines

Abstract

This invention discloses an electromechanical-electromagnetic adaptive collaborative modeling method for wind turbine generators, belonging to the field of wind power engineering technology. The method includes: obtaining an initial model / model combination from a pre-constructed multi-precision model library based on the simulation objective; wherein the multi-precision model library includes electromechanical models, electromagnetic models, and hybrid models of electromechanical and electromagnetic models; and, based on the initial model / model combination, using the Lüpertzfox optimization algorithm, obtaining the model / model combination most suitable for the simulation objective from the pre-constructed multi-precision model library. This invention overcomes the shortcomings of traditional wind turbine generator modeling methods, such as fixed modeling accuracy and simplistic models.

Description

Technical Field

[0001] This invention relates to the field of wind power engineering technology, and in particular to a method for adaptive collaborative modeling of electromechanical and electromagnetic components of wind turbine generators. Background Technology

[0002] Wind turbines are typical multi-physics systems with strong coupling, involving complex interactions of aerodynamic loads, mechanical transmission, electromagnetic conversion, power electronics and control. Traditional modeling methods face a dilemma: (1) Although electromechanical models (lumped parameters, equivalent circuits) are computationally efficient, they ignore spatial electromagnetic field distribution, local saturation and high-frequency losses, making it difficult to accurately capture key phenomena such as generator demagnetization and converter switching transients; (2) Although electromagnetic models have high physical fidelity, they consume huge computational resources and cannot support dynamic optimization and real-time control of the entire machine.

[0003] Existing research often employs a single model with fixed precision, leading to an imbalance between efficiency and accuracy in full-condition simulations: simple, low-precision models exhibit distortion in fault ride-through and vibration analysis, while simple, high-precision models are unsuitable for wind farm-level scheduling or control parameter tuning. An adaptive modeling architecture is urgently needed to dynamically balance the computational burden while ensuring the accuracy of key features. Summary of the Invention

[0004] The purpose of this invention is to provide an electromechanical-electromagnetic adaptive collaborative modeling method for wind turbine generators. This method establishes a multi-precision model library based on model complexity and relevant parameters of different components. Then, it selects models / model combinations from the multi-precision model library according to the simulation objective, and uses the Lüpertz fox optimization algorithm for iterative calculations to select the model / model combination most suitable for the simulation objective. This invention is achieved through the following technical solutions.

[0005] This invention introduces an electromechanical-electromagnetic adaptive collaborative modeling method for wind turbine generators, including the following:

[0006] Based on the simulation objectives, an initial model / model combination is obtained from a pre-built multi-precision model library; the multi-precision model library includes electromechanical models, electromagnetic models, and hybrid models of electromechanical and electromagnetic models;

[0007] Based on the initial model / model combination, the Lüper Fox optimization algorithm is used to obtain the model / model combination that best matches the simulation target from a pre-built multi-precision model library.

[0008] In practical applications, traditional modeling methods for wind turbines suffer from computational burdens or insufficient accuracy. This invention overcomes these shortcomings by fusing electromechanical and electromagnetic models. Specifically, it first establishes a multi-precision model library covering system and component levels based on model complexity and relevant parameters. Then, it selects operating conditions based on the simulation objectives, derives specific dominant physical factors from these conditions, and selects appropriate models / model combinations based on these dominant physical factors. Finally, it optimizes this selection using the Lüpertz algorithm. If accuracy or computational burden issues remain, optimization continues until the computational cost is minimized while meeting accuracy requirements.

[0009] Optionally, a multi-precision model library can be built, including:

[0010] Based on model complexity, a system-level multi-precision model library covering system levels is established. Among them, the system-level models are used to analyze wind power systems. The system-level models include electromechanical models and hybrid electromechanical and electromagnetic models. The electromechanical models include the ultra-simplified electromechanical model L0, and the hybrid electromechanical and electromagnetic models include the simplified electromechanical-electromagnetic model L1 and the high-fidelity electromechanical-electromagnetic model L2.

[0011] Based on generator and converter parameters, a component-level multi-precision model library covering the component level is established. The component-level models are used to analyze the components in the wind power system. The component-level models include electromagnetic models, which include generator and converter models.

[0012] In this invention, the models in the component-level multi-precision model library listed above are only a portion. The models in the component-level multi-precision model library are mainly electromagnetic models, selected as needed. The system-level multi-precision model library consists of electromechanical-electromagnetic hybrid models (excluding L0) categorized by precision.

[0013] Optionally, the generator-related parameters include frequency, active power, reactive power, stator voltage, stator current, rotor voltage, rotor current, power factor, and generator temperature; the converter-related parameters include rated power, voltage range, conversion efficiency, protection level, and cooling method.

[0014] Optionally, based on the simulation objective, an initial model / model combination can be obtained from a pre-built multi-precision model library, including:

[0015] Based on the simulation objectives, the operating conditions of the wind power system are determined. The operating conditions include wind speed type, grid status and control mode. Wind speed type includes steady wind, turbulent wind and extreme gusts. Grid status includes normal, voltage sag, frequency fluctuation and fault. Control mode includes maximum power point tracking, power limiting and fault ride-through.

[0016] Based on the operating conditions, determine the physical factors that dominate the operating conditions;

[0017] The initial model / model combination is determined based on the physical factors of the dominant operating conditions.

[0018] The simulation objectives include annual power generation estimation, grid-connected harmonic analysis, drivetrain fatigue life, low-voltage ride-through capability, and permanent magnet generator demagnetization.

[0019] Optionally, based on the initial model / model combination, the Lüperfox optimization algorithm is used to obtain the model / model combination that best matches the simulation target from a pre-built multi-precision model library, including:

[0020] If the initial model is obtained, the initial hunting position in the initialization phase of the Lüper Fox optimization algorithm is used to simulate the initial model. Depending on the type of hunting position, during the daytime or nighttime hunting phase, the hunting position is iteratively updated by checking whether the Lüper Fox's eyes and ears rotate, until the updated hunting position meets the preset requirements. Among these steps, optimizing the hunting position involves reselecting a model from the multi-precision model library.

[0021] The system-level multi-precision model library corresponds to the daytime hunting stage, and the component-level multi-precision model library corresponds to the nighttime hunting stage.

[0022] If the initial model combination is obtained, for each model in the model combination, the initial hunting position of the model combination is simulated by the Lüper Fox optimization algorithm in the initialization stage. Depending on the type of hunting position, during the daytime hunting stage or the nighttime hunting stage, the hunting position is iteratively updated by checking whether the Lüper Fox's eyes and ears rotate, until the updated hunting position meets the preset requirements.

[0023] The formula for generating the initial hunting position in the initialization phase of the Lüper Fox optimization algorithm is as follows:

[0024] (1)

[0025] In the formula, i represents the index of the Lüper fox, j represents the search dimension, and rand represents a random number in the range [0,1]. Let represent the initial hunting position of the i-th Lüper fox in dimension j. This indicates the lower bound of the search domain in search dimension j. This indicates the upper bound of the search domain in search dimension j.

[0026] Optionally, during the daytime hunting phase, when the Lüper fox's eyes do not rotate, the formula for updating the hunting position is:

[0027] (2),

[0028] In the formula, k is the number of iterations. During the daytime hunting phase, the hunting position of the i-th fox in the (k+1)th iteration is determined by its eyes not rotating. During the daytime hunting phase, the Lüper fox, in the kth iteration, randomly selects the optimal hunting position by hunting without rotating its eyes. During the daytime hunting phase, the hunting position of the i-th fox in the k-th iteration is determined by its eyes not rotating. For random numbers different from rand, is the global optimal position vector, p is a random value in the range [0,1], s is the visual power, and h is the auditory power;

[0029] During the daytime hunting phase, when the Lüper fox's eyes rotate, the formula for updating the hunting position is:

[0030] (3)

[0031] In the formula, During the daytime hunting phase, the i-th fox determines its hunting position by rotating its eyes in the (k+1)th iteration. During the daytime hunting phase, the hunting position of the i-th fox after its eyes have rotated. To limit the step size of the random walk, The angle of eye rotation;

[0032] During the daytime hunting phase, assuming the Lüper fox's ears do not rotate, the formula for updating the hunting position is:

[0033] (4)

[0034] In the formula, During the daytime hunting phase, the hunting position of the i-th fox in the (k+1)th iteration is determined by its ears not rotating. During the daytime hunting phase, the Lüper fox, in the kth iteration, randomly selects the optimal hunting location by hunting without rotating its ears. During the daytime hunting phase, the hunting position of the i-th fox in the k-th iteration is determined by hunting without rotating its ears.

[0035] During the daytime hunting phase, when the Lüper fox's ears rotate, the formula for updating the hunting position is:

[0036] (5)

[0037] In the formula, During the daytime hunting phase, the i-th fox determines its hunting position by rotating its ears in the (k+1)th iteration. During the daytime hunting phase, the hunting position of the i-th fox after its ears have rotated.

[0038] Optionally, during the night hunting phase, when the Lüper fox's eyes do not rotate, the formula for updating the hunting position is:

[0039] (6)

[0040] In the formula, During the nighttime hunting phase, the i-th fox, in the (k+1)-th iteration, finds its hunting position by not rotating its eyes. During the nighttime hunting phase, the Lüper fox, in the kth iteration, randomly selects the optimal hunting position by hunting without rotating its eyes. During the nighttime hunting phase, the hunting position of the i-th fox in the k-th iteration is determined by hunting without rotating its eyes.

[0041] During the night hunting phase, when the Lüper fox's eyes rotate, the formula for updating the hunting position is:

[0042] (7)

[0043] In the formula, During the nighttime hunting phase, the i-th fox determines its hunting position by rotating its eyes during the (k+1)-th iteration. During the nighttime hunting phase, the hunting position of the i-th fox after its eyes have rotated. To limit the step size of the random walk, The angle of eye rotation;

[0044] During the night hunting phase, assuming the Lüper fox's ears do not rotate, the formula for updating the hunting position is as follows:

[0045] (8)

[0046] In the formula, During the nighttime hunting phase, the hunting position of the i-th fox in the (k+1)-th iteration is determined by its ears not rotating. During the nighttime hunting phase, the Lüper fox, in the kth iteration, randomly selects the optimal hunting position by hunting with its ears not rotating. During the nighttime hunting phase, the hunting position of the i-th fox in the k-th iteration is determined by hunting without rotating its ears.

[0047] During the night hunting phase, when the ears of the Lüper fox rotate, the formula for updating the hunting position is:

[0048] (9)

[0049] In the formula, During the nighttime hunting phase, the i-th fox determines its hunting position by rotating its ears during the (k+1)-th iteration. During the nighttime hunting phase, the hunting position of the i-th fox after its ears have rotated.

[0050] Optionally, in the Lüper fox optimization algorithm, if the hunting position cannot be updated iteratively through hearing and vision, the hunting position can be updated iteratively through the Lüper fox's sense of smell.

[0051] The formula for updating hunting locations is:

[0052] ,

[0053] In the formula, For either the daytime or nighttime hunting phase, the hunting location of the i-th fox during the (k+1)th iteration, determined by its sense of smell. , and For a random value in the range [0,1], The known optimal position vector is randomly selected, and smell is a function that varies with the number of iterations k to simulate the fox's sense of smell. Equation (10-1) indicates that the fox tracks prey based on scent, and Equation (10-2) indicates that when scent tracking fails, the fox will randomly search for nearby hunting locations.

[0054] Optionally, in the Lüper fox optimization algorithm, if the hunting position cannot be updated iteratively through hearing, vision, and smell, the hunting position is updated iteratively based on the principle of optimal individual movement; whereby the principle of optimal individual movement is that when Lüper foxes are foraging, they will move towards the individual in the group with the best hunting position.

[0055] The formula for updating hunting locations is:

[0056] ,

[0057] In the formula, For either the daytime or nighttime hunting phase, the i-th fox, in the (k+1)-th iteration, moves to the hunting position of the fox with the best hunting position in the group. and For positive integers, Let i be the hunting position of the i-th fox after the update relative to the prey position in the k-th iteration; The hunting position of the i-th fox after the k-th iteration relative to the prey position is expressed as follows:

[0058] (12)

[0059] In the formula, and It is a positive number.

[0060] Formulas (11-1) and (12) mimic the collective behavior of fox packs, expanding the possibilities for further exploration and utilization. In this invention, model selection based on formulas (11-1) and (12) can overcome local optima and consider the global picture.

[0061] Optionally, in the Lüper fox optimization algorithm, if the hunting position cannot be updated through auditory, visual, olfactory iterations and the principle of optimal individual movement, the hunting position is iteratively updated based on the worst-case animal behavior; where the worst-case animal behavior is that the Lüper fox cannot find prey in the nearby area, so the Lüper fox will move away from the nearby area.

[0062] The formula for updating hunting locations is:

[0063] (13)

[0064] In the formula, Let represent the fox's hunting position after the (k+1)th iteration in the worst-case scenario, during either the daytime or nighttime hunting phase. For either the daytime or nighttime hunting phase, the worst-case hunting position after the k-th iteration is given. This is a step size constraint for random walks.

[0065] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0066] The proposed electromechanical-electromagnetic adaptive collaborative modeling method for wind turbines establishes a multi-precision model library for electromechanical and electromagnetic systems. This allows for the selection of appropriate models / model combinations based on the actual simulation objectives, overcoming the shortcomings of traditional wind turbine modeling methods that rely solely on simple, fixed-precision models for full-condition simulations, leading to wasted computational resources or missed detection of critical physical phenomena. Furthermore, the introduced Lüpertz fox optimization algorithm iteratively updates the selected initial model / model combination, automatically selecting the most suitable model / model combination for the simulation objective based on real-time operating conditions. Attached Figure Description

[0067] Figure 1 The diagram shown is a flowchart of the electromechanical-electromagnetic adaptive collaborative modeling method for wind turbine generators in one embodiment of the present invention.

[0068] Figure 2 The diagram shown is a flowchart of the Lüperfox optimization algorithm in one embodiment of the present invention. Detailed Implementation

[0069] The following description, in conjunction with the accompanying drawings and specific embodiments, provides further details. In this description, it should be understood that the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature.

[0070] Example 1

[0071] This invention introduces an electromechanical-electromagnetic adaptive collaborative modeling method for wind turbine generators, including the following:

[0072] Based on the simulation objectives, an initial model / model combination is obtained from a pre-built multi-precision model library; the multi-precision model library includes electromechanical models, electromagnetic models, and hybrid models of electromechanical and electromagnetic models;

[0073] Based on the initial model / model combination, the Lüper Fox optimization algorithm is used to obtain the model / model combination that best matches the simulation target from a pre-built multi-precision model library.

[0074] In practical applications, traditional modeling methods for wind turbines suffer from computational burdens or insufficient accuracy. This invention overcomes these shortcomings by fusing electromechanical and electromagnetic models. Specifically, it first establishes a multi-precision model library covering system and component levels based on model complexity and relevant parameters. Then, it selects operating conditions based on the simulation objectives, derives specific dominant physical factors from these conditions, and selects appropriate models / model combinations based on these dominant physical factors. Finally, it optimizes this selection using the Lüpertz algorithm. If accuracy or computational burden issues remain, optimization continues until the computational cost is minimized while meeting accuracy requirements.

[0075] Example 2

[0076] Based on Example 1, this example introduces the specific implementation process of a wind turbine electromechanical-electromagnetic adaptive collaborative modeling method, such as... Figure 1 As shown, it specifically includes the following:

[0077] In one specific embodiment of the present invention, establishing a multi-precision model library includes establishing a system-level multi-precision model library covering the system level based on model complexity, and establishing a component-level multi-precision model library covering the component level based on generator-related parameters and converter-related parameters; wherein, the system level analyzes the wind power system, and the component level analyzes the components in the wind power system;

[0078] The system-level multi-precision model library includes a simplified electromechanical model L0, a simplified electromechanical-electromagnetic model L1, and a high-fidelity electromechanical-electromagnetic model L2; the component-level multi-precision model library includes generator models and converter models. Table 1 below illustrates the model precision, key features, applicable scenarios, and computational cost of each model. In other embodiments different from this invention, the model precision, key features, applicable scenarios, and computational cost may include other content. In Table 1, the model precision of the simplified electromechanical model L0, the simplified electromechanical-electromagnetic model L1, and the high-fidelity electromechanical-electromagnetic model L2 increases progressively. The generator model and the converter model have the same precision.

[0079] Table 1 Multi-precision model library

[0080]

[0081] In one specific embodiment of the present invention, the generator-related parameters include frequency, active power, reactive power, stator voltage, stator current, rotor voltage, rotor current, power factor, and generator temperature; the converter-related parameters include rated power, voltage range, conversion efficiency, protection level, and cooling method.

[0082] Selecting a model / model combination from the multi-precision model library based on the simulation objective involves selecting an operating condition based on the simulation objective, obtaining the dominant physical factors based on the operating condition, and then selecting a model / model combination from the multi-precision model library based on the dominant physical factors. In one specific embodiment of the present invention, the operating condition includes wind speed type, power grid status, and control mode; wherein, wind speed type includes steady-state wind, turbulent wind, and extreme gusts; power grid status includes normal, voltage sag, frequency fluctuation, and fault; and control mode includes maximum power point tracking, power limiting, and fault ride-through. In other embodiments different from those of the present invention, the operating condition may include other parameters in conjunction with specific application scenarios.

[0083] In practical applications, such as Figure 1 As shown, the dominant physical factors are obtained based on the specific simulation objectives and operating parameters. These factors are then input into a pre-built multi-precision model library. Using the Lüpertzfox optimization algorithm, the model combination with the current optimal precision is selected. Since the system-level model precision L2>L1>L0, when the precision requirements are not met, the model is upgraded; when the computational cost exceeds the limit, the model is downgraded.

[0084] In one specific embodiment of this invention, the selection of models / model combinations from the multi-precision model library based on the simulation objective is shown in Table 2 below. In Table 2, the key output quantities represent the simulation objective. For example, if the simulation objective is annual power generation estimation, and the dominant physical factors are identified as the power curve and wind speed distribution, combined with the key features of each model in Table 1 and the cost requirements of the calculation, the ultra-simplified electromechanical model L0 is recommended.

[0085] Table 2. Model / Model Combination Selection Based on Simulation Objectives

[0086]

[0087] In one specific embodiment of the present invention, the selection of the above-mentioned model / model combination is optimized according to the Lüper Fox optimization algorithm, including an initialization phase, a daytime hunting phase, and a nighttime hunting phase. The flowchart of the Lüper Fox optimization algorithm is as follows: Figure 2 As shown, the appropriate formula is selected to update the hunting location of the Lüper fox by combining the relationship between visual power, auditory power, and random numbers.

[0088] Based on the initial model / model combination, the Lüperfox optimization algorithm is used to obtain the model / model combination that best matches the simulation target from a pre-built multi-precision model library, including:

[0089] If the initial model is obtained, the initial hunting position in the initialization phase of the Lüper Fox optimization algorithm is used to simulate the initial model. Depending on the type of hunting position, during the daytime or nighttime hunting phase, the hunting position is iteratively updated by checking whether the Lüper Fox's eyes and ears rotate, until the updated hunting position meets the preset requirements. Among these steps, optimizing the hunting position involves reselecting a model from the multi-precision model library.

[0090] The system-level multi-precision model library corresponds to the daytime hunting stage, and the component-level multi-precision model library corresponds to the nighttime hunting stage.

[0091] If the initial model combination is obtained, for each model in the model combination, the initial hunting position of the model combination is simulated by the Lüper Fox optimization algorithm in the initialization stage. Depending on the type of hunting position, during the daytime hunting stage or the nighttime hunting stage, the hunting position is iteratively updated by checking whether the Lüper Fox's eyes and ears rotate, until the updated hunting position meets the preset requirements.

[0092] The formula for generating the initial hunting position in the initialization phase of the Lüper Fox optimization algorithm is as follows:

[0093] (1)

[0094] In the formula, i represents the index of the Lüper fox, j represents the search dimension, and rand represents a random number in the range [0,1]. Let represent the initial hunting position of the i-th Lüper fox in dimension j. This indicates the lower bound of the search domain in search dimension j. This indicates the upper bound of the search domain in search dimension j.

[0095] In one specific embodiment of the present invention, during the daytime hunting phase, the updating of the hunting position of the Lüper fox includes visual hunting position updating without eye rotation, visual hunting position updating with eye rotation, auditory hunting position updating without ear rotation, and auditory hunting position updating with ear rotation.

[0096] The eye-non-rotation visual hunting position update is calculated using the following formula:

[0097] (2),

[0098] In the formula, k is the number of iterations. During the daytime hunting phase, the hunting position of the i-th fox in the (k+1)th iteration is determined by its eyes not rotating. During the daytime hunting phase, the Lüper fox, in the kth iteration, randomly selects the optimal hunting position by hunting without rotating its eyes. During the daytime hunting phase, the hunting position of the i-th fox in the k-th iteration is determined by its eyes not rotating. For random numbers different from rand, is the global optimal position vector, p is a random value in the range [0,1], s is the visual power, and h is the auditory power;

[0099] The eye rotation visual hunting position update is calculated using the following formula:

[0100] (3)

[0101] In the formula, During the daytime hunting phase, the i-th fox determines its hunting position by rotating its eyes in the (k+1)th iteration. During the daytime hunting phase, the hunting position of the i-th fox after its eyes have rotated. To limit the step size of the random walk, The angle of eye rotation;

[0102] The ear-non-rotating auditory hunting position update is calculated using the following formula:

[0103] (4)

[0104] In the formula, During the daytime hunting phase, the hunting position of the i-th fox in the (k+1)th iteration is determined by its ears not rotating. During the daytime hunting phase, the Lüper fox, in the kth iteration, randomly selects the optimal hunting location by hunting without rotating its ears. During the daytime hunting phase, the hunting position of the i-th fox in the k-th iteration is determined by hunting without rotating its ears.

[0105] Ear rotation auditory hunting position update is calculated using the following formula:

[0106] (5)

[0107] In the formula, During the daytime hunting phase, the i-th fox determines its hunting position by rotating its ears in the (k+1)th iteration. During the daytime hunting phase, the hunting position of the i-th fox after its ears have rotated.

[0108] In one specific embodiment of the present invention, during the night hunting phase, the updating of the hunting position of the Lüper fox includes visual hunting position updating without eye rotation, visual hunting position updating with eye rotation, auditory hunting position updating without ear rotation, and auditory hunting position updating with ear rotation.

[0109] The eye-non-rotation visual hunting position update is calculated using the following formula:

[0110] (6)

[0111] In the formula, During the nighttime hunting phase, the i-th fox, in the (k+1)-th iteration, finds its hunting position by not rotating its eyes. During the nighttime hunting phase, the Lüper fox, in the kth iteration, randomly selects the optimal hunting position by hunting without rotating its eyes. During the nighttime hunting phase, the hunting position of the i-th fox in the k-th iteration is determined by hunting without rotating its eyes.

[0112] The eye rotation visual hunting position update is calculated using the following formula:

[0113] (7)

[0114] In the formula, During the nighttime hunting phase, the i-th fox determines its hunting position by rotating its eyes during the (k+1)-th iteration. During the nighttime hunting phase, the hunting position of the i-th fox after its eyes have rotated. To limit the step size of the random walk, The angle of eye rotation;

[0115] The ear-non-rotating auditory hunting position update is calculated using the following formula:

[0116] (8)

[0117] In the formula, During the nighttime hunting phase, the hunting position of the i-th fox in the (k+1)-th iteration is determined by its ears not rotating. During the nighttime hunting phase, the Lüper fox, in the kth iteration, randomly selects the optimal hunting position by hunting with its ears not rotating. During the nighttime hunting phase, the hunting position of the i-th fox in the k-th iteration is determined by hunting without rotating its ears.

[0118] Ear rotation auditory hunting position update is calculated using the following formula:

[0119] (9)

[0120] In the formula, During the nighttime hunting phase, the i-th fox determines its hunting position by rotating its ears during the (k+1)-th iteration. During the nighttime hunting phase, the hunting position of the i-th fox after its ears have rotated.

[0121] In one specific embodiment of the present invention, in the Lüper fox optimization algorithm, if the hunting position cannot be updated iteratively through hearing and vision, the hunting position is updated iteratively through the olfactory sense of the Lüper fox.

[0122] The formula for updating hunting locations is:

[0123] ,

[0124] In the formula, For either the daytime or nighttime hunting phase, the hunting location of the i-th fox during the (k+1)th iteration, determined by its sense of smell. , and For a random value in the range [0,1], The known optimal position vector is randomly selected, and smell is a function that varies with the number of iterations k to simulate the fox's sense of smell. Equation (10-1) indicates that the fox tracks prey based on scent, and Equation (10-2) indicates that when scent tracking fails, the fox will randomly search for nearby hunting locations.

[0125] In one specific embodiment of the present invention, in the Lüper fox optimization algorithm, if the hunting position cannot be updated iteratively through hearing, vision, and smell, the hunting position is updated iteratively based on the optimal individual movement principle; wherein, the optimal individual movement principle is that when Lüper foxes are foraging, they will move towards the individual in the group with the best hunting position;

[0126] The formula for updating hunting locations is:

[0127] ,

[0128] In the formula, For either the daytime or nighttime hunting phase, the i-th fox, in the (k+1)-th iteration, moves to the hunting position of the fox with the best hunting position in the group. and For positive integers, Let i be the hunting position of the i-th fox after the update relative to the prey position in the k-th iteration; The hunting position of the i-th fox after the k-th iteration relative to the prey position is expressed as follows:

[0129] (12)

[0130] In the formula, and It is a positive number.

[0131] Formulas (11-1) and (12) mimic the collective behavior of fox packs, expanding the possibilities for further exploration and utilization. In this invention, model selection based on formulas (11-1) and (12) can overcome local optima and consider the global picture.

[0132] In one specific embodiment of the present invention, in the Lüper fox optimization algorithm, if the hunting position cannot be updated through auditory, visual, olfactory iteration and the optimal individual movement principle, the hunting position is iteratively updated based on the worst-case animal behavior; wherein, the worst-case animal behavior is that the Lüper fox cannot find prey in the nearby area, and the Lüper fox will move away from the nearby area.

[0133] (13)

[0134] In the formula, Let represent the fox's hunting position after the (k+1)th iteration in the worst-case scenario, during either the daytime or nighttime hunting phase. For either the daytime or nighttime hunting phase, the worst-case hunting position after the k-th iteration is given. This is a step size constraint for random walks.

[0135] This embodiment updates the hunting position for the next iteration by updating the hunting position of the fox in the current iteration under different conditions. This is equivalent to obtaining the optimal choice of the model or model combination for the next moment based on the dominant physical factors at the current moment.

[0136] In the initialization phase, the Lüper fox optimization algorithm starts the optimization process by randomly creating a set of initial solutions, and then divides it into two phases: daytime and nighttime, based on the system-level model and the component-level model. If the system-level model is selected, the fox's vision is greater than its hearing, and rand ≥ 0.25, then it enters the daytime hunting phase, using the non-rotating visual hunting position update, i.e., formula (2). At this time, the initial solution is used as... The optimal solution of the current model can be obtained by iterating using formula (2) until the required accuracy and cost are met.

[0137] When a fox approaches the fox with the best hunting position in the group, it is easy to generate a local optimum. This is equivalent to only considering one of the dominant physical factors in actual application. Formula (12) can overcome the defect of finding the local optimum and consider the global situation.

[0138] For example, when a fox is in the worst-case scenario and cannot find prey in the vicinity, it is equivalent to a complex simulation target with many dominant physical factors in actual applications, making it impossible to find the optimal solution for the model. Formula (13) can address this deficiency by attempting to move to a promising area and away from the worst-case scenario. The corresponding practical approach is to simplify the complex simulation target or reduce the dominant factors.

[0139] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A wind turbine electro-mechanical-electromagnetic adaptive co-modeling method, characterized in that, The system comprises: According to the simulation target, an initial model / model combination is obtained from a pre-constructed multi-precision model library; wherein the multi-precision model library comprises electromechanical models, electromagnetic models, and mixed electromechanical and electromagnetic models; According to the initial model / model combination, a model / model combination most suitable for the simulation target is obtained from the pre-constructed multi-precision model library by using the Lyapunov fox optimization algorithm; The multi-precision model library is constructed, comprising: Based on the model complexity, a system-level multi-precision model library covering the system hierarchy is established; wherein the system-level model is used for analyzing the wind power system; the system-level model comprises electromechanical models and mixed electromechanical and electromagnetic models, the electromechanical models comprise a super-simple electromechanical model L0, and the mixed electromechanical and electromagnetic models comprise a simplified electromechanical-electromagnetic model L1 and a high-fidelity electromechanical-electromagnetic model L2; Based on the generator parameters and the converter parameters, a component-level multi-precision model library covering the component hierarchy is established; wherein the component-level model is used for analyzing the components in the wind power system; the component-level model comprises electromagnetic models, and the electromagnetic models comprise generator models and converter models; According to the initial model / model combination, a model / model combination most suitable for the simulation target is obtained from the pre-constructed multi-precision model library by using the Lyapunov fox optimization algorithm, comprising: If the initial model is obtained, the initial hunting position in the initialization stage of the Lyapunov fox optimization algorithm is used to simulate the initial model; according to the type of the hunting position, the hunting position is iteratively updated in the daytime hunting stage or the nighttime hunting stage through the rotation of the eyes and ears of the Lyapunov fox until the updated hunting position meets the preset requirements; wherein the optimized hunting position is a reselected model from the multi-precision model library; The system-level multi-precision model library corresponds to the daytime hunting stage, and the component-level multi-precision model library corresponds to the nighttime hunting stage; If the initial model combination is obtained, for each model in the model combination, the initial hunting position in the initialization stage of the Lyapunov fox optimization algorithm is used to simulate the model in the model combination; according to the type of the hunting position, the hunting position is iteratively updated in the daytime hunting stage or the nighttime hunting stage through the rotation of the eyes and ears of the Lyapunov fox until the updated hunting position meets the preset requirements; The formula for generating the initial hunting position in the initialization stage of the Lyapunov fox optimization algorithm is: ，（1）， where i denotes the index of the Lyubersky fox, j denotes the search dimension, and rand denotes a random number in [0, 1], denotes the initial hunting position of the i-th Lyubersky fox in dimension j, denotes the lower bound of the search domain in search dimension j, denotes the upper bound of the search domain in search dimension j.

2. The wind turbine generator electro-mechanical-electromagnetic adaptive co- modeling method according to claim 1, characterized in that, The generator parameters comprise frequency, active power, reactive power, stator voltage, stator current, rotor voltage, rotor current, power factor, and generator temperature; and the converter parameters comprise rated power, voltage range, conversion efficiency, protection level, and cooling method.

3. The wind turbine generator electro-mechanical- electromagnetic adaptive co- modeling method in accordance with claim 1, characterized by, According to the simulation target, an initial model / model combination is obtained from a pre-constructed multi-precision model library, comprising: According to the simulation target, the operating conditions of the wind power system are determined; wherein the operating conditions comprise wind speed types, power grid states, and control modes, the wind speed types comprise steady wind, turbulence, and extreme gust, the power grid states comprise normal, voltage sag, frequency fluctuation, and fault, and the control modes comprise maximum power tracking, power limitation, and fault ride-through; According to the operating conditions, the physical factors of the dominant operating conditions are determined; According to the physical factors of the dominant operating conditions, the initial model / model combination is determined.

4. The wind turbine generator electro-mechanical- electromagnetic adaptive co- modeling method in accordance with claim 1, characterized by, In the daytime hunting stage, the formula for updating the hunting position of the Lü Pèi fox when the eyes are not rotating is: ，（2）， where k is the iteration number, is the hunting position of the i-th fox at the (k+1)-th iteration by eye not rotation during the daytime hunting stage, is the random selection of the best hunting position of the Lyubersky fox at the k-th iteration by eye not rotation during the daytime hunting stage, is the hunting position of the i-th fox at the k-th iteration by eye not rotation during the daytime hunting stage, is a random number different from rand, is the global optimal position vector, p is a random value in the range of [0, 1], s is the visual strength, and h is the auditory strength. In the daytime hunting stage, the formula for updating the hunting position of the Lü Pèi fox when the eyes are rotating is: ，（3）， wherein is the hunting position of the i-th fox at the (k+1)-th iteration by eye rotation for the daytime hunting phase, is the hunting position of the i-th fox after eye rotation for the daytime hunting phase, is the step limit of the random walk, is the eye rotation angle; In the daytime hunting stage, the formula for updating the hunting position of the Lü Pèi fox when the ears are not rotating is: ，（4）， wherein is the hunting position of the i-th fox at the (k+1)-th iteration through the day hunting phase by not rotating the ears, is the random selection of the optimal hunting position of the Lüpold fox at the k-th iteration through the day hunting phase by not rotating the ears, is the hunting position of the i-th fox at the k-th iteration through the day hunting phase by not rotating the ears. In the daytime hunting stage, the formula for updating the hunting position of the Lü Pèi fox when the ears are rotating is: ，（5）， wherein is the hunting position of the i-th fox at the (k+1)-th iteration of the ear rotation for the daytime hunting phase, is the hunting position of the i-th fox after the ear rotation for the daytime hunting phase.

5. The wind turbine generator electro-mechanical-electromagnetic adaptive co- modeling method in accordance with claim 4, characterized in that, in In the nighttime hunting stage, the formula for updating the hunting position of the Lü Pèi fox when the eyes are not rotating is: ，（6）， wherein is the hunting position of the i-th fox at the (k+1)-th iteration of the night hunting phase by eye not rotation, is the random selection of the optimal hunting position of the Lüpold fox at the k-th iteration of the night hunting phase by eye not rotation, is the hunting position of the i-th fox at the k-th iteration of the night hunting phase by eye not rotation, In the nighttime hunting stage, the formula for updating the hunting position of the Lü Pèi fox when the eyes are rotating is: ，（7）， wherein is the hunting position of the i-th fox at the (k+1)-th iteration of the hunting phase by eye rotation, is the hunting position of the i-th fox after eye rotation at the hunting phase, is the step limit of the random walk, is the eye rotation angle; In the nighttime hunting stage, the formula for updating the hunting position of the Lü Pèi fox when the ears are not rotating is: ，（8）， wherein is the hunting position of the i-th fox at the (k+1)-th iteration of the night hunting phase by ear not rotation, is the random selection of the optimal hunting position of the Lüpold fox at the k-th iteration of the night hunting phase by ear not rotation, is the hunting position of the i-th fox at the k-th iteration of the night hunting phase by ear not rotation, In the nighttime hunting stage, the formula for updating the hunting position of the Lü Pèi fox when the ears are rotating is: ，（9）， wherein is the hunting position of the i-th fox at the (k+1)-th iteration of the ear rotation for the night hunting phase, is the hunting position of the i-th fox after the ear rotation for the night hunting phase.

6. The wind turbine generator electro-mechanical-electromagnetic adaptive co- modeling method in accordance with claim 5, characterized in that, in In the Lü Pèi fox optimization algorithm, if the hunting position cannot be updated through auditory and visual iteration, the hunting position is updated through Lü Pèi fox olfactory iteration; The formula for updating the hunting position is: ， wherein, is the hunting position of the i-th fox at the (k+1)-th iteration by smell during the day hunting phase or the night hunting phase, , and is a random value in the range of [0, 1], is a randomly selected known optimal position vector, smell is a function that varies with the iteration number k, and is used to simulate the fox's olfactory ability; formula (10-1) represents that the fox tracks the prey according to the smell, and formula (10-2) represents that when the smell tracking fails, the fox will randomly search for a hunting position nearby.

7. The wind turbine electromechanical-magnetic adaptive co-modeling method according to claim 6, characterized in that, In the Lü Pèi fox optimization algorithm, if the hunting position cannot be updated through auditory, visual, and olfactory iteration, the hunting position is updated through the optimal individual movement principle, wherein the optimal individual movement principle is that the Lü Pèi fox will move towards the individual with the best hunting position in the group when foraging; The formula for updating the hunting position is: ， wherein is the hunting position of the i-th fox at the (k+1)-th iteration, and is a positive constant, is the hunting position of the i-th fox at the k-th iteration after updating with respect to the prey position; is the hunting position of the i-th fox at the k-th iteration after updating with respect to the prey position, expressed as follows: ，（12）， wherein and are normal numbers.

8. The wind turbine electromechanical-magnetic adaptive co-modeling method according to claim 7, characterized in that, In the Lü Pèi fox optimization algorithm, if the hunting position cannot be updated through auditory, visual, olfactory iteration, and the optimal individual movement principle, the hunting position is updated through the worst-case animal behavior, wherein the worst-case animal behavior is that the Lü Pèi fox cannot find prey in the adjacent area, and the Lü Pèi fox will move away from the adjacent area; The formula for updating the hunting position is: ，（13）， wherein is the hunting position after the worst case iteration of the k+1th day hunting phase or night hunting phase for the fox, is the hunting position after the worst case iteration of the kth day hunting phase or night hunting phase, is the step limit for the random walk.

Images