Wind storage collaborative control method, device, equipment, storage medium and program product

Abstract

This application relates to the field of wind power generation technology and discloses a wind-storage coordinated control method, device, equipment, storage medium, and program product. The method includes: acquiring operational data of a target wind farm and performing turbulent multi-timescale analysis on the operational data to obtain turbulent characteristic parameters characterizing fluctuations at different timescales; based on the turbulent characteristic parameters, solving a multi-objective optimization problem with wind turbine equivalent fatigue load, energy storage device lifespan loss, and grid frequency regulation requirements as optimization objectives using a multi-objective optimization algorithm to generate a coordinated control strategy; and executing the coordinated control strategy to coordinate the control of the wind turbine and the energy storage system. This application, by first performing turbulent multi-timescale analysis on the wind farm operational data to obtain turbulent characteristic parameters and then using a multi-objective optimization algorithm for coordinated decision-making, effectively solves the problem that related technologies cannot simultaneously guarantee wind turbine fatigue life, energy storage device lifespan, and grid frequency stability under complex turbulent conditions.

Description

Technical Field

[0001] This application relates to the field of wind power generation technology, specifically to wind-storage coordinated control methods, devices, equipment, storage media, and program products. Background Technology

[0002] In complex terrain conditions, wind farms often face operating environments with high turbulence intensity and frequent wind speed fluctuations. This turbulence has multi-timescale characteristics, including instantaneous gusts on the order of seconds as well as continuous pulsations on the order of minutes, resulting in complex fluctuations in the output power of wind turbines. Power fluctuations directly affect the frequency stability of the power grid, while the mechanical load impacts caused by turbulence can significantly accelerate fatigue damage to critical components of the wind turbine.

[0003] When faced with such multi-timescale coupled problems, the control methods of related technologies often struggle to coordinate multiple objectives such as power smoothing, load suppression, and energy storage lifetime maintenance, resulting in limited overall system performance. Summary of the Invention

[0004] This application provides a wind-storage coordinated control method, device, equipment, storage medium, and program product to solve the problem that related technologies cannot simultaneously guarantee the fatigue life of wind turbines, the lifespan of energy storage equipment, and the stability of grid frequency under complex turbulent conditions.

[0005] In a first aspect, this application provides a wind-storage coordinated control method, comprising: acquiring operational data of a target wind farm and performing turbulent multi-timescale analysis on the operational data to obtain turbulent characteristic parameters characterizing fluctuations at different timescales; based on the turbulent characteristic parameters, solving a multi-objective optimization problem with wind turbine equivalent fatigue load, energy storage equipment lifespan loss, and grid frequency regulation requirements as optimization objectives using a multi-objective optimization algorithm to generate a coordinated control strategy; and executing the coordinated control strategy to coordinately control the wind turbine and the energy storage system.

[0006] Beneficial Effects: This application first performs multi-timescale turbulence analysis on wind farm operation data to obtain in-depth turbulence characteristic parameters, and then uses a multi-objective optimization algorithm for collaborative decision-making based on these characteristic parameters. This effectively solves the problem that related technologies cannot simultaneously guarantee the fatigue life of wind turbines, the lifespan of energy storage devices, and the frequency stability of the power grid under complex turbulent conditions. This method transforms single-objective or sequential optimization into global collaborative optimization based on precise perception, enabling the generated collaborative control strategy to balance the three conflicting objectives of wind turbines, energy storage, and the power grid, thereby achieving optimal comprehensive performance of the wind-storage integrated system under complex and variable wind conditions.

[0007] In one optional implementation, the step of acquiring the operational data of the target wind farm and performing turbulence multi-timescale analysis on the operational data includes: acquiring a wind speed sequence sampled at a preset frequency as the operational data; processing the wind speed sequence using a signal decomposition algorithm to obtain turbulence components at different timescales; and determining the dominant turbulence type and turbulence level based on the energy distribution of the turbulence components.

[0008] Beneficial effects: By acquiring wind speed sequences sampled at a preset frequency (high frequency) and performing signal decomposition to obtain turbulence components at different time scales, and then determining the turbulence type and level based on energy distribution, this technique enables more refined and quantifiable turbulence perception. This not only distinguishes between different turbulence origins (such as mechanical turbulence and thermal turbulence) but also provides richer and more accurate input information for subsequent collaborative control strategies. This allows collaborative optimization decisions to be more targeted and adaptable to actual wind conditions, improving the accuracy and adaptability of control.

[0009] In one alternative implementation, the signal decomposition algorithm includes a wavelet transform algorithm or an empirical mode decomposition algorithm.

[0010] Beneficial effects: Both algorithms are particularly suitable for processing non-stationary wind speed signals and can effectively extract local features of turbulence at different time scales. This approach ensures the reliability and effectiveness of the multi-time-scale turbulence analysis process, laying the foundation for the stability of the entire control system and avoiding the risk of feature extraction distortion due to inappropriate analysis method selection, which in turn affects the optimization effect.

[0011] In one alternative implementation, determining the dominant turbulence type and turbulence level based on the energy distribution of the turbulence components includes: inputting the energy distribution of the turbulence components into a trained LSTM model to predict the future turbulence level.

[0012] Beneficial effects: By introducing an LSTM model to predict future turbulence levels, the collaborative optimization strategy is no longer a passive response based solely on the current state, but can predict wind conditions in advance, thereby generating predictive control commands. This can significantly reduce control lag caused by sudden turbulence changes, further improve the effects of power smoothing and load suppression, and enhance the system's robustness in dealing with complex wind conditions.

[0013] In one optional implementation, the multi-objective optimization problem solved using a multi-objective particle swarm optimization algorithm is employed. The process of generating a cooperative control strategy by solving the multi-objective optimization problem using this algorithm includes: initializing the particle swarm by randomly generating an initial position and velocity for each particle to represent the cooperative control strategy; entering an iterative optimization loop, and in each iteration, performing the following steps: dynamically adjusting the inertia weight and learning factor using an adaptive mechanism based on the ratio of the current iteration count to the maximum iteration count; updating the velocity and position of each particle based on the adaptively adjusted parameters, the individual historical optimal solution of each particle, and the global optimal solution selected from an external archive set, thereby generating a new candidate cooperative control strategy; and applying the new candidate... A cooperative control strategy is selected for multi-objective function evaluation and constraint satisfaction judgment. Based on the evaluation and judgment results, the individual historical optimal solution of each particle is updated, and the external archive set is updated. The non-dominated solutions generated in the current iteration are compared with the existing solutions in the archive set, and all non-dominated solutions are retained. After updating the external archive set, the non-dominated solutions in the archive set are screened based on the crowding distance to maintain the diversity of solution distribution in the objective space. In the comparison of the quality of solutions, feasible solutions that satisfy all constraints are selected first. For infeasible solutions that do not satisfy the constraints, their constraint violation degree is calculated, and a dynamic penalty coefficient that increases with the number of iterations is used for punishment. The iterative optimization loop is repeated until the pre-termination condition is met, and finally the optimal cooperative control strategy is selected from the external archive set.

[0014] Beneficial Effects: The adaptive adjustment mechanism enables the algorithm to perform a strong global search in the early stages of iteration, avoiding getting trapped in local optima; in the later stages of iteration, it can perform fine-grained local development, accelerating convergence. This mechanism effectively balances exploration and development, improves optimization efficiency, and ensures that high-quality cooperative control strategies can be quickly obtained within the short time required for real-time control of wind farms, meeting the real-time requirements of engineering applications. In the iterative optimization loop, by maintaining an external archive set and using congestion distance for filtering, the Pareto solution set obtained by the optimization algorithm is guaranteed to have good distribution diversity. This allows for a set of candidate strategies evenly distributed among multiple optimization objectives, rather than concentrated in a certain extreme direction, thus providing the possibility of flexibly selecting the most suitable cooperative control strategy according to real-time grid needs, enhancing the system's flexibility and practicality. In addition, by prioritizing feasible solutions and applying dynamic penalties to infeasible solutions, the optimization process is ensured to always move in the direction of satisfying all system hard constraints (such as wind turbine speed safety limits and energy storage SOC range). This greatly improves the convergence and practicality of the algorithm under complex constraints, effectively avoids the generation of invalid strategies that cannot be executed on actual devices, and ensures the engineering feasibility and security of the cooperative control strategy.

[0015] In an optional implementation, the step of dynamically adjusting the inertia weight and learning factor using an adaptive mechanism includes: dynamically calculating the inertia weight based on the ratio of the current iteration count to the total iteration count, such that the inertia weight linearly decreases from a preset maximum value to a preset minimum value as the ratio increases; adjusting the cognitive learning factor and social learning factor inversely based on the ratio of the current iteration count to the total iteration count, such that the cognitive learning factor decreases as the ratio increases, and the social learning factor c2 increases as the ratio increases; and introducing a sinusoidal perturbation term in the adjustment of the inertia weight, wherein the amplitude of the sinusoidal perturbation term decreases as the iteration count increases.

[0016] Beneficial Effects: In the early stages of iteration, the high inertia weight and the learning strategy biased towards individual experience synergistically promote the algorithm's extensive exploration of the solution space, effectively avoiding premature convergence. As the iteration progresses, the decreasing inertia weight and the shift of the learning strategy towards social experience smoothly transition to the fine-grained search stage, significantly accelerating the convergence speed towards a high-quality Pareto optimal solution. The introduction of the sinusoidal perturbation term provides the algorithm with the crucial ability to escape local extrema in the complex search space, and its decay characteristics ensure the stability of later convergence. Ultimately, this adaptive mechanism ensures that the optimization algorithm can efficiently and robustly generate uniformly distributed and high-performance collaborative control strategies in the complex problem of wind-storage collaborative control, which is characterized by high dimensionality, nonlinearity, and multiple conflicting objectives. This directly improves the overall control performance of wind farms in dealing with complex turbulent conditions, including power smoothing, fatigue load suppression, and grid frequency regulation response.

[0017] In one optional implementation, before generating a cooperative control strategy based on the turbulence characteristic parameters, the method further includes: statistically analyzing load cycles using the rainflow counting method based on the operating data; and calculating the equivalent fatigue load value according to the linear cumulative damage criterion.

[0018] Beneficial effects: By employing the rainflow counting method and the linear cumulative damage criterion to calculate the equivalent fatigue load, the abstract goal of "extending wind turbine life" is transformed into precisely quantifiable engineering parameters. This enables the optimization algorithm to make decisions based on an accurate fatigue damage accumulation model, rather than just empirical estimations, thereby significantly improving the accuracy and reliability of life prediction for key components. This allows the optimization goal of "extending wind turbine life" to be achieved scientifically and accurately.

[0019] In one optional implementation, the grid frequency regulation requirement is defined as a set of constraints for the multi-objective optimization problem, the set of constraints including: frequency regulation response time not greater than 10 seconds, and the energy storage capacity reserved for frequency regulation not less than 20% of the rated capacity of the energy storage system.

[0020] Beneficial effects: This scheme provides some measurable and verifiable hard standards, which not only provide clear constraints for the optimization algorithm, but also enable the final generated collaborative control strategy to strictly meet the grid compliance requirements, significantly improving the reliability and credibility of wind farms participating in grid frequency regulation.

[0021] In one optional implementation, the execution of the coordinated control strategy includes: generating instructions to control the supercapacitor to perform rapid charging and discharging for power fluctuations on a second-level time scale; and generating instructions to control the energy storage battery to perform charging and discharging and the wind turbine to perform pitch regulation for power fluctuations on a minute-level or higher time scale.

[0022] Beneficial Effects: By assigning differentiated control methods—combining supercapacitors and energy storage batteries with wind turbine pitch control—to power fluctuations on different time scales (seconds, minutes, and above), this scheme achieves optimal allocation of control resources. It fully leverages the respective advantages of supercapacitors' rapid response and energy storage batteries' high energy density, avoiding waste of control resources or excessive operation of individual devices. This effectively reduces equipment wear and improves overall economic efficiency while efficiently smoothing power fluctuations.

[0023] Secondly, this application provides a wind-storage coordinated control device, the device comprising: a turbulence sensing module, used to acquire operating data of a target wind farm and perform turbulence multi-timescale analysis on the operating data to obtain turbulence characteristic parameters characterizing fluctuations at different timescales; a coordinated optimization module, used to solve a multi-objective optimization problem based on the turbulence characteristic parameters, using a multi-objective optimization algorithm to optimize the equivalent fatigue load of the wind turbine, the life loss of the energy storage equipment, and the grid frequency regulation requirements, and generate a coordinated control strategy; and a control execution module, used to execute the coordinated control strategy to coordinate the control of the wind turbine and the energy storage system.

[0024] Thirdly, this application provides an electronic device, including: a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to perform the wind-storage coordinated control method of the first aspect or any corresponding embodiment described above.

[0025] Fourthly, this application provides a computer-readable storage medium storing computer instructions for causing a computer to execute the wind-storage coordinated control method of the first aspect or any corresponding embodiment described above.

[0026] Fifthly, this application provides a computer program product, including computer instructions, which are used to cause a computer to execute the wind-storage coordinated control method of the first aspect or any corresponding embodiment described above. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0028] Figure 1 This is a schematic diagram illustrating an application scenario according to an embodiment of this application;

[0029] Figure 2 This is a schematic flowchart of the first type of wind-storage coordinated control method according to an embodiment of this application;

[0030] Figure 3 This is a second flowchart illustrating the wind-storage coordinated control method according to an embodiment of this application;

[0031] Figure 4 This is a schematic diagram of the third process of the wind-storage coordinated control method according to the embodiments of this application;

[0032] Figure 5 This is a schematic diagram of the fourth process of the wind-storage coordinated control method according to the embodiments of this application;

[0033] Figure 6 This is a schematic diagram of the relationship between generator rotor speed and maximum power output during the wind turbine control process according to an embodiment of this application;

[0034] Figure 7 This is a structural block diagram of a wind-storage coordinated control device according to an embodiment of this application;

[0035] Figure 8 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0037] It is understood that before using the technical solutions disclosed in the various embodiments of this application, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this application in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.

[0038] In the field of new energy power generation control technology, wind farm operation faces challenges posed by complex environments such as mountainous terrain. Mountainous terrain (such as valleys and ridges) intensifies turbulence, causing wind speeds to fluctuate dramatically on a scale of seconds to minutes. This fluctuation is directly transmitted to the generator, causing high-frequency oscillations in output power, which in turn leads to grid frequency deviations and threatens system stability. At the same time, turbulence-induced aerodynamic loads subject blades, towers, and transmission chains to periodic alternating stresses. For example, the bending moment at the blade root may change by more than 20% within one second due to turbulence, leading to a 3-5 times increase in the rate of fatigue damage accumulation and significantly shortening component lifespan.

[0039] While wind turbine energy storage control technology has made some progress in related technologies, it still has the following limitations:

[0040] The charging and discharging strategy does not take into account multiple objectives: it focuses too much on a single objective (such as power smoothing or load suppression) and does not fully consider the dynamic coupling relationship between equipment fatigue life, battery life and grid frequency stability, resulting in local optimization problems.

[0041] Single control parameter: Some schemes rely solely on wind speed prediction (such as 5-minute average wind speed), ignoring instantaneous changes in turbulence, resulting in energy storage charging and discharging response lagging behind power fluctuations (delay of 10-15 seconds).

[0042] Static threshold setting: Using a fixed charge and discharge power threshold (such as ±10% of theoretical power) cannot adapt to changes in turbulence intensity. It over-adjusts in low turbulence and under-adjusts in high turbulence.

[0043] Lack of dynamic adjustment mechanism: The collaborative control strategy cannot be adjusted in real time according to the multi-timescale characteristics of turbulence (second-level, minute-level, hour-level), resulting in poor control effect under complex wind conditions.

[0044] Insufficient quantification of hazards: The quantitative relationship between power deviation and grid stability and unit life has not been established, and the coordinated control strategy lacks specificity.

[0045] As one optional application scenario in the embodiments of this application, such as Figure 1 As shown, the wind-storage coordinated control method of this application can be deployed on one or more terminal devices 103 at the wind farm. The terminal device 103 is connected to field devices such as wind turbine 101 and energy storage system 102 through the wind farm's internal control network 110, and is used to acquire their operating data and execute the control algorithm of this application, thereby generating and issuing coordinated control commands.

[0046] In this embodiment, terminal device 103 mainly refers to a dedicated computing device deployed in an industrial field, capable of data acquisition and real-time computing. Specific examples include, but are not limited to, industrial computers, edge computing gateways, or embedded industrial control computers. Field devices include wind turbine main controllers, energy storage converters, and related sensors. Network 110 can be a wired industrial network, examples of which include, but are not limited to, industrial Ethernet, fiber optic ring networks, etc.

[0047] It should be noted that, Figure 1 This is merely an example of one application scenario. This application can also be deployed on a high-performance computing unit integrated within a wind turbine controller or energy storage converter, and is not limited to the specific hardware configuration shown in the figure. The embodiments of this application will be described below with reference to the accompanying drawings.

[0048] According to an embodiment of this application, a wind-storage coordinated control method embodiment is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0049] This embodiment provides a wind and energy storage coordinated control method, which can be used in the aforementioned terminal devices, such as industrial computers, edge computing gateways, or embedded industrial control computers (the executing entity is described in conjunction with the actual situation). Figure 2 This is a flowchart of the wind-storage coordinated control method according to an embodiment of this application, such as... Figure 2 As shown, the process includes the following steps:

[0050] Step S201: Obtain the operation data of the target wind farm and perform turbulence multi-timescale analysis on the operation data to obtain turbulence characteristic parameters that characterize the fluctuation characteristics at different time scales.

[0051] It should be noted that in this step S201, the operating data refers to the physical quantity data generated by the wind farm during operation, reflecting its state and external environment, including the wind speed sequence collected by the anemometer, the blade root strain collected by the wind turbine sensor, the generator speed and power, and the battery state of charge (SOC) and temperature provided by the energy storage system.

[0052] Turbulence multi-timescale analysis is a signal processing technique used to decompose complex turbulent wind speed signals into wave components at different time scales for study. Turbulence characteristic parameters are indicators used to quantitatively describe turbulence characteristics, including turbulence intensity reflecting wave intensity, turbulence integral scale reflecting the average size of large eddies, and the energy proportion of each scale component obtained through analysis.

[0053] Step S201 transforms the raw, mixed operational data into structured information that accurately describes the current and future short-term wind conditions. The terminal device, acting as the execution entity, first acquires the raw operational data stream from the field network; subsequently, its built-in analysis algorithm performs in-depth processing on this data—namely, turbulent multi-timescale analysis—decomposing the signal and calculating key turbulent characteristic parameters.

[0054] Step S202: Based on turbulence characteristic parameters, a multi-objective optimization problem with wind turbine equivalent fatigue load, energy storage equipment life loss and power grid frequency regulation requirements as optimization objectives is solved by a multi-objective optimization algorithm to generate a cooperative control strategy.

[0055] It should be noted that multi-objective optimization algorithms are mathematical methods used to find the optimal balance point among multiple conflicting objectives; while cooperative control strategies refer to a set of instructions that coordinate the actions of wind turbines and energy storage systems. For example, a strategy may include a combination of specific actions such as "instructing the supercapacitor to absorb a 200kW power spike," "instructing the wind turbine pitch angle to increase by 0.5 degrees," and "instructing the battery to discharge at 50kW power for 3 minutes."

[0056] In step S202, the terminal device, based on the turbulence characteristic parameters provided in step S201, initiates a multi-objective optimization algorithm to solve a complex mathematical problem. The goal of this problem is to simultaneously minimize wind turbine fatigue, energy storage loss, and power fluctuations while meeting grid requirements. The algorithm ultimately generates an optimal coordinated control strategy. Essentially, this process transforms a complex engineering trade-off problem into a computable mathematical model, and the powerful computing capabilities of the terminal device automatically identify the action plan with the highest overall benefit under the current wind conditions.

[0057] Step S203: Execute the coordinated control strategy to coordinate the control of the wind turbine and the energy storage system.

[0058] It should be noted that wind turbine units refer to wind power generation equipment including wind turbine blades, gearboxes, generators, pitch systems, etc., while energy storage systems are hybrid energy storage power stations composed of lithium battery packs, supercapacitors, energy storage converters, etc.

[0059] In step S203, the terminal device translates the collaborative control strategy generated in step S202 into specific instructions that can be understood by the underlying hardware, and then executes and issues these instructions. These instructions act simultaneously on the wind turbine and the energy storage system, achieving collaborative control of both.

[0060] The wind-storage coordinated control method provided in this embodiment first performs turbulent multi-timescale analysis on wind farm operation data to obtain in-depth turbulent characteristic parameters. Then, based on these characteristic parameters, a multi-objective optimization algorithm is used for coordinated decision-making. This effectively solves the problem that related technologies cannot simultaneously guarantee the fatigue life of wind turbines, the lifespan of energy storage devices, and the frequency stability of the power grid under complex turbulent conditions. This method transforms single-objective or sequential optimization into global coordinated optimization based on precise perception. This allows the generated coordinated control strategy to balance the three conflicting objectives of wind turbines, energy storage, and the power grid, thereby achieving optimal overall performance of the wind-storage integrated system under complex and variable wind conditions.

[0061] This embodiment provides a wind and energy storage coordinated control method, which can be used in the aforementioned terminal devices, such as industrial computers, edge computing gateways, or embedded industrial control computers. Figure 3 This is a flowchart of the wind-storage coordinated control method according to an embodiment of this application, such as... Figure 3 As shown, the process includes the following steps:

[0062] Step S301: Obtain the operation data of the target wind farm and perform turbulence multi-timescale analysis on the operation data to obtain turbulence characteristic parameters that characterize the fluctuation characteristics at different time scales.

[0063] Specifically, step S301 includes:

[0064] Step S3011: Obtain the wind speed sequence sampled at a preset frequency as operating data.

[0065] It should be noted that conventional meteorological observations usually collect data at a frequency of 1 to 10 minutes. In this step S3011, the preset frequency is a high frequency, that is, a sampling rate that can capture wind speed fluctuations on a time scale of seconds or even shorter.

[0066] Optionally, in some specific embodiments, the sampling frequency can be 50Hz, that is, 50 wind speed data points are collected per second, to ensure that the second-level vortex information contained in the turbulence can be accurately captured.

[0067] Step S3012: The wind speed sequence is processed using a signal decomposition algorithm to obtain turbulence components at different time scales.

[0068] It should be noted that the signal decomposition algorithm is used to process the "raw" wind speed sequence, which is a mixture of fluctuations at various scales, to obtain a series of clear, time-scale separated turbulence components.

[0069] Optionally, in some embodiments of this application, step S3012 includes:

[0070] Step a1: The wind speed sequence is processed using wavelet transform algorithm or empirical mode decomposition algorithm to obtain turbulence components at different time scales.

[0071] It should be noted that the wavelet transform algorithm is a time-frequency analysis method based on preset basis functions (wavelet mother functions). It achieves the localization and analysis of different frequency components of a signal by scaling and shifting the basis functions. The empirical mode decomposition algorithm, on the other hand, is a completely data-driven adaptive signal decomposition method that can decompose a signal into a series of eigenmode functions from high frequency to low frequency.

[0072] In step a1, both algorithms are particularly suitable for processing non-stationary wind speed signals and can effectively extract local features of turbulence at different time scales. This approach ensures the reliability and effectiveness of the multi-time-scale turbulence analysis process, laying the foundation for the stability of the entire control system and avoiding the risk of feature extraction distortion due to inappropriate selection of analysis methods, which in turn affects the optimization effect.

[0073] Step S3013: Based on the energy distribution of the turbulent components, determine the dominant turbulence type and turbulence level.

[0074] It should be noted that energy component refers to the proportion of energy carried by turbulent components at different time scales in the total turbulent energy; turbulence dominant type refers to determining whether the turbulence is mechanical turbulence caused by fixed obstacles such as terrain or thermal turbulence caused by uneven heating of the earth's surface, based on the scale range in which the energy is mainly concentrated; and turbulence grade is a classification of the intensity of turbulence based on comprehensive parameters such as turbulence intensity.

[0075] Optionally, in some embodiments of this application, step S3013 above includes:

[0076] Step b1 involves inputting the energy distribution of the turbulence components into a trained LSTM model to predict the future turbulence level.

[0077] It should be noted that the trained LSTM model refers to a recurrent neural network model that has been trained using a large amount of historical wind field data and possesses time series prediction capabilities. Because the evolution of turbulence is highly nonlinear, the LSTM model can automatically learn this complex pattern from historical data, and its prediction accuracy is far superior to simple linear extrapolation methods, thus ensuring the effectiveness of look-ahead control.

[0078] In step b1, by introducing an LSTM model to predict future turbulence levels, the collaborative optimization strategy is no longer a passive response based solely on the current state, but can predict wind conditions in advance, thereby generating predictive control commands. This can significantly reduce control lag caused by sudden turbulence changes, further improve the effects of power smoothing and load suppression, and enhance the system's robustness in dealing with complex wind conditions.

[0079] In steps S3011 to S3013 above, turbulence components at different time scales are obtained by acquiring wind speed sequences sampled at a preset frequency (high frequency) and performing signal decomposition. Then, the turbulence type and level are determined based on energy distribution. This technique makes turbulence perception more refined and quantifiable. This not only distinguishes between different turbulence origins (such as mechanical turbulence and thermal turbulence) but also provides richer and more accurate input information for subsequent collaborative control strategies. This allows collaborative optimization decisions to be more targeted and adaptable to actual wind conditions, improving the accuracy and adaptability of control.

[0080] Step S302: Based on turbulence characteristic parameters, a multi-objective optimization problem is solved using a multi-objective optimization algorithm, with the optimization objectives being the equivalent fatigue load of the wind turbine, the lifespan loss of the energy storage equipment, and the grid frequency regulation requirements, to generate a cooperative control strategy. For details, please refer to [link to relevant documentation]. Figure 2 Step S202 of the illustrated embodiment will not be described again here.

[0081] Step S303: Execute the coordinated control strategy to perform coordinated control of the wind turbine and energy storage system. For details, please refer to [link to relevant documentation]. Figure 2 Step S203 of the illustrated embodiment will not be described again here.

[0082] This embodiment provides a wind and energy storage coordinated control method, which can be used in the aforementioned terminal devices, such as industrial computers, edge computing gateways, or embedded industrial control computers. Figure 4 This is a flowchart of the wind-storage coordinated control method according to an embodiment of this application, such as... Figure 4 As shown, the process includes the following steps:

[0083] Step S401: Obtain the operational data of the target wind farm and perform multi-timescale turbulence analysis on the operational data to obtain turbulence characteristic parameters characterizing the fluctuation characteristics at different time scales. For details, please refer to... Figure 2 Step S201 of the illustrated embodiment will not be described again here.

[0084] Step S402: Based on turbulence characteristic parameters, a multi-objective optimization problem with wind turbine equivalent fatigue load, energy storage equipment life loss and power grid frequency regulation requirements as optimization objectives is solved by a multi-objective optimization algorithm to generate a cooperative control strategy.

[0085] Optionally, in some embodiments of this application, a multi-objective optimization problem is solved using a multi-objective particle swarm optimization algorithm. The algorithm used is multi-objective particle swarm optimization. The process of solving the multi-objective optimization problem using multi-objective particle swarm optimization to generate a cooperative control strategy includes:

[0086] Step S4021: Initialize the particle swarm by randomly generating an initial position and velocity for each particle that represents the cooperative control strategy.

[0087] Step S4022: Enter the iterative optimization loop. Specifically, in each iteration, the following steps are performed:

[0088] Step c1: Based on the ratio of the current iteration number to the maximum iteration number, dynamically adjust the inertia weight and learning factor using an adaptive mechanism.

[0089] It's important to note that inertia weight is a core parameter in particle swarm optimization (PSO) algorithms, directly influencing the tendency of particles (i.e., candidate solutions) to maintain their previous velocity during the search process. A larger inertia weight allows particles to explore the global search space more effectively, reducing the risk of getting trapped in local optima, but it also slows down convergence. Conversely, a smaller inertia weight allows particles to perform a more refined local search within the current region, accelerating convergence, but it can also lead to premature entrapment in local optima, causing them to miss the opportunity to find a globally better solution.

[0090] Therefore, an adaptive mechanism is used to dynamically adjust the inertia weight, so that the inertia weight is no longer fixed, but changes automatically and dynamically according to preset rules as the algorithm iterates.

[0091] In step c1, this adaptive mechanism enables the algorithm to perform a strong global search in the early stages of iteration, avoiding getting trapped in local optima; while in the later stages of iteration, it can perform fine-grained local exploration, accelerating convergence. This mechanism effectively balances exploration and development, improves optimization efficiency, and ensures that a high-quality collaborative control strategy can be quickly obtained within the short time required for real-time control of wind farms, meeting the real-time requirements of engineering applications.

[0092] Step c2 updates the velocity and position of each particle based on the adaptively adjusted parameters, the individual historical best solution of each particle, and the global best solution selected from the external archive set, thereby generating a new candidate cooperative control strategy.

[0093] It should be noted that the external archive set is an elite solution set independent of the current iteration particle swarm, used to store and display the best candidate solutions found in previous iterations.

[0094] Step c3: Evaluate the new candidate cooperative control strategy using a multi-objective function and determine the satisfaction of constraints. Based on the evaluation and determination results, update the individual historical optimal solution of each particle and update the external archive set. Compare the non-dominated solution generated in the current iteration with the existing solutions in the archive set and retain all non-dominated solutions.

[0095] It's important to note that non-dominated solutions are those for all optimization objectives, where no other solution is strictly better. In other words, solutions that cannot improve one objective by sacrificing another represent the optimal set of trade-offs, i.e., Pareto optimal solutions.

[0096] Specifically, the steps for updating the external archive set include: in each iteration of the algorithm, comparing all non-dominated solutions generated by the current particle swarm with existing solutions in the archive set. If the new solution dominates some old solutions in the archive set, these dominated old solutions are deleted, and the new solution is added to the archive set. If the new solution does not dominate any solutions in the archive set, the new solution is directly added to the archive set.

[0097] Step c4: After updating the external archive set, the non-dominated solutions in the archive set are filtered based on the crowding distance in order to maintain the diversity of solution distribution in the target space.

[0098] It should be noted that in step c4, the crowding distance is an indicator used to quantify the density of a solution among its neighboring solutions in the target space. The larger the crowding distance, the more "open" the solution is, and the more unique the solution it represents.

[0099] When the size of the external archive set exceeds a preset limit, filtering must be performed to prevent infinite expansion. The filtering principle is to prioritize retaining solutions with large crowding distances. More specifically, in step d2, the crowding distance of each solution in the archive set needs to be calculated, and the solution with the smallest crowding distance is deleted until the size of the archive set returns to within the limit.

[0100] In step c4, by maintaining an external archive set and using congestion distance for filtering, the Pareto solution set obtained by the optimization algorithm is ensured to have good distribution diversity. This allows for the acquisition of a set of candidate strategies that are evenly distributed across multiple optimization objectives, rather than being concentrated in a single extreme direction. This makes it possible to flexibly select the most suitable cooperative control strategy based on real-time power grid needs, enhancing the system's flexibility and practicality.

[0101] In step c5, during the comparison of the quality of solutions, feasible solutions that satisfy all constraints are selected first. For infeasible solutions that do not satisfy the constraints, the constraint violation degree is calculated, and a dynamic penalty coefficient that increases with the number of iterations is used for penalty.

[0102] The multi-objective particle swarm optimization algorithm employs a hierarchical constraint handling strategy. It's important to note that constraints in the wind-storage co-optimization problem refer to the physical or safety limitations that must be met, such as the upper limit of wind turbine speed, the state of charge limit of the energy storage system, and the rated output power. Constraint violation is an indicator used to quantify the "infeasibility" of an infeasible solution, typically calculated as the sum of deviations from all violated constraints. The dynamic penalty coefficient is a penalty weight that gradually increases with the number of iterations, used to transform the constraint violation into a penalty on the objective function value. As iterations proceed, the penalty for infeasible solutions becomes increasingly severe, weakening their competitiveness in the population and leading to their gradual elimination.

[0103] In step c5, by prioritizing feasible solutions and imposing dynamic penalties on infeasible solutions, the optimization process is ensured to always proceed in the direction of satisfying all system hard constraints (such as wind turbine speed safety limits and energy storage SOC range). This greatly improves the convergence and practicality of the algorithm under complex constraints, effectively avoids generating invalid strategies that cannot be executed on actual equipment, and guarantees the engineering feasibility and security of the cooperative control strategy.

[0104] Step S4023: Repeat the iterative optimization loop until the pre-termination condition is met, and finally select the optimal cooperative control strategy from the external archive set.

[0105] Optionally, in some embodiments of this application, before generating the cooperative control strategy based on turbulence characteristic parameters, the following steps are also included:

[0106] Step d1: Based on the running data, the load cycle is counted using the rainflow counting method;

[0107] It should be noted that load cycles are the basic unit of fatigue analysis, referring to the process by which the stress or strain of a component fluctuates from the starting point and returns to the starting point. Rainflow counting is a standard algorithm for processing random load sequences. It can decompose complex fluctuations into a series of complete stress cycles (full cycles) and semi-cycles by identifying the peaks and troughs in the load-time history, and statistically analyze the amplitude (stress range) and mean of each cycle.

[0108] Step d2: Calculate the equivalent fatigue load value according to the linear cumulative damage criterion.

[0109] It should be noted that the equivalent fatigue load value is a constant amplitude stress level, and the fatigue damage caused by it is equal to the damage caused by the actual complex variable amplitude load spectrum. This value can be used to condense the complex load history into a comparable parameter with clear physical meaning.

[0110] In these embodiments, by employing rainflow counting and linear cumulative damage criteria to calculate equivalent fatigue loads, the abstract goal of "extending wind turbine life" is transformed into precisely quantifiable engineering parameters. This enables the optimization algorithm to make decisions based on accurate fatigue damage accumulation models, rather than just empirical estimates, thereby significantly improving the accuracy and reliability of life prediction for key components. This allows the optimization goal of "extending wind turbine life" to be achieved scientifically and accurately.

[0111] Optionally, in some embodiments of this application, the power grid frequency regulation requirement is defined as a set of constraints for a multi-objective optimization problem. These constraints include: the frequency regulation response time is no more than 10 seconds, and the energy storage capacity reserved for frequency regulation is no less than 20% of the rated capacity of the energy storage system.

[0112] It should be noted that frequency regulation response time refers to the time required from the point when the grid frequency deviation exceeds the dead zone until the energy storage system's output power reaches the specified value; it measures the system's rapid response capability. Frequency regulation reserved capacity, on the other hand, refers to the available capacity of the energy storage system specifically reserved for responding to grid frequency changes and not used for other purposes. It measures the system's continuous support capability.

[0113] This scheme provides some measurable and verifiable hard standards, which not only provide clear constraints for the optimization algorithm, but also ensure that the final generated collaborative control strategy can strictly meet the grid compliance requirements, significantly improving the reliability and credibility of wind farms participating in grid frequency regulation.

[0114] Step S403: Execute the coordinated control strategy to perform coordinated control of the wind turbine and energy storage system. For details, please refer to [link to relevant documentation]. Figure 2 Step S203 of the illustrated embodiment will not be described again here.

[0115] This embodiment provides a wind and energy storage coordinated control method, which can be used in the aforementioned terminal devices, such as industrial computers, edge computing gateways, or embedded industrial control computers. Figure 5 This is a flowchart of the wind-storage coordinated control method according to an embodiment of this application, such as... Figure 5 As shown, the process includes the following steps:

[0116] Step S501: Obtain the operational data of the target wind farm and perform multi-timescale turbulence analysis on the operational data to obtain turbulence characteristic parameters characterizing the fluctuation characteristics at different time scales. For details, please refer to... Figure 2 Step S201 of the illustrated embodiment will not be described again here.

[0117] Step S502: Based on turbulence characteristic parameters, a multi-objective optimization problem is solved using a multi-objective optimization algorithm, with the optimization objectives being the equivalent fatigue load of the wind turbine, the lifespan loss of the energy storage equipment, and the grid frequency regulation requirements, to generate a cooperative control strategy. For details, please refer to step S402 of the aforementioned embodiment or... Figure 2 Step S202 of the illustrated embodiment will not be described again here.

[0118] Step S503: Execute the coordinated control strategy to coordinate the control of the wind turbine and the energy storage system.

[0119] Specifically, in step S503 above, the steps for implementing the coordinated control strategy include:

[0120] Step S5031: For power fluctuations on a second-level time scale, generate instructions to control the supercapacitor to perform rapid charging and discharging.

[0121] It should be noted that power fluctuations on a second-scale timescale refer to rapid, high-frequency power changes with a period of less than a few seconds. These fluctuations are usually caused by small-scale turbulent eddies, resulting in dramatic changes but short durations. Supercapacitors, on the other hand, are energy storage devices with extremely high power density and extremely fast charging and discharging speeds, with response times reaching the millisecond level, but relatively low energy density.

[0122] When the system detects or predicts second-level fluctuations, it immediately generates targeted instructions to drive the supercapacitor to charge and discharge rapidly: quickly charging to absorb energy when there is excess power and quickly discharging to replenish energy when there is insufficient power. This effectively filters out high-frequency "glitch" in the power waveform, thereby significantly improving power quality, while preventing these rapid fluctuations from being transmitted to the wind turbine drive chain and reducing mechanical load impact.

[0123] Step S5032: For power fluctuations on a time scale of minutes or above, generate commands to control the charging and discharging of the energy storage battery and the pitch adjustment of the wind turbine.

[0124] It should be noted that power fluctuations on time scales of minutes and above refer to power changes with periods ranging from several minutes to tens of minutes or even longer. These fluctuations are usually caused by weather systems, large-scale topographic effects, or continuous wind speed changes, and while they involve large amounts of energy, the changes are relatively slow.

[0125] Because of the enormous energy fluctuations occurring on a minute-by-minute scale, supercapacitors, with their limited capacity, are insufficient to handle this, while energy storage batteries are well-suited for energy dispatching on a longer scale. Simultaneously, wind turbines, through pitch regulation, limit or increase power capture at the source, working in synergy with the energy storage system. By generating coordinated commands, they can share power deficits or jointly absorb power surpluses. In specific examples, as wind speed increases, the energy storage battery can be prioritized for charging to absorb excess energy; if power continues to increase, the turbine pitch is activated, increasing the pitch angle to reduce wind energy capture. This synergy not only smooths grid-connected power but also significantly reduces the frequency and amplitude of the turbine pitch system's operations by having the energy storage system undertake some regulation tasks, thereby directly reducing low-cycle fatigue damage to key components and extending turbine lifespan.

[0126] It should also be noted that in practical applications, when turbulence occurs before the rated wind speed, doubly fed wind turbines mainly achieve load reduction operation through two methods: one is to reduce the power coefficient of the wind turbine by increasing the pitch angle, and the other is to change the operating point of the wind turbine by controlling the generator speed, thereby achieving load reduction operation of the unit.

[0127] Reference Figure 6 The figure shows the power-speed curve of a wind turbine generator, used to illustrate the power output pattern under different control methods. The meanings of each parameter are as follows:

[0128] 1) Coordinate axis parameters, vertical axis: unit output power, P max The horizontal axis represents the rated (maximum) output power of the unit; the horizontal axis represents the mechanical speed of the unit (usually referring to the rotor speed). ω opt It is the optimal rotational speed (corresponding to the best operating condition for the wind turbine to capture wind energy).

[0129] 2) Curves and control parameters, β (Pitch angle): The pitch angle of the wind turbine blades (the angle between the blades and the airflow). β 2> β 1 indicates that the larger the pitch angle, the lower the power output at the same speed (the lower the curve). Acceleration control (point B): By increasing the rotor speed, the unit is brought towards the optimal speed. ω opt Proximity control improves wind energy capture efficiency. Deceleration control (point C): Reduces rotor speed to avoid exceeding the unit's safe speed or rated power. Pitch angle control (point A): Adjusts the pitch angle (e.g., increasing...) to improve wind energy capture efficiency. β By altering the power-speed curve, power limiting can be achieved (often used when wind speed is too high to prevent power from exceeding limits). P maxThis curve illustrates the "power variation with rotational speed under different pitch angles." By combining acceleration, deceleration, and pitch angle control, the unit can maintain efficient or safe operation under different wind speeds.

[0130] Pitch angle control controls the active power output of a wind turbine to be lower than its rated value by increasing the pitch angle. When using pitch angle control, the wind turbine operates stably at the optimal speed for the current wind speed. Under constant wind speed and constant speed conditions, the pitch angle controller increases the pitch angle from a certain value to a certain value, thereby limiting wind energy capture and reducing active power output. However, this method has a relatively slow response time, and the pitch actuator is a mechanical component. Frequent pitch angle adjustments inevitably increase the fatigue load on the turbine and shorten its service life.

[0131] Variable speed load shedding control adjusts the generator rotor speed by increasing or decreasing it to a new operating point below the maximum power output point. Variable speed control for load shedding in doubly-fed wind turbines can be further divided into acceleration control and deceleration control, as described in [reference]. Figure 6 Points B and C are shown in the diagram. Since deceleration control may cause stability issues such as small-signal instability, acceleration control is generally used to achieve variable speed load reduction. Acceleration control for load reduction of doubly-fed induction generator (DFIG) wind turbines can be achieved by adjusting the active power reference value of the wind turbine to control the generator speed, thereby reducing the active power output of the wind turbine. However, this control method is not suitable at high wind speeds, and the generator speed after acceleration should not exceed its upper limit.

[0132] The frequency response decay caused by rotor deceleration means that the wind turbine cannot achieve the expected power-frequency static characteristics. The steady-state frequency response of the unit is nonlinear and difficult to predict, which may lead to uneven power distribution among multiple generator units in the system.

[0133] Wind turbine control phase: This is divided into TC (torque control) and PC (pitch angle control). When the unit operates below the rated wind speed (e.g., below 1 m / s), encountering gusts, the generator speed of the wind turbine is prone to exceeding 1.1 times the protection limit, leading to overspeed shutdown. To prevent power loss due to pitch control before reaching rated power, the transition from TC to PC is often designed so that the generator speed exceeds the rated speed and the unit's power reaches the rated value. Therefore, under conditions of rapidly increasing wind speed, the generator speed rises sharply, but the power does not reach the rated value, and pitch control does not activate. However, when the power reaches the rated value, the speed is already very high, and initiating pitch control at this point is no longer sufficient to effectively and promptly suppress generator overspeed.

[0134] When the unit operates above the rated wind speed, the wind turbine operates in the PC phase. This is due to two reasons: first, the huge inertia of the wind turbine causes a lag of about 1 second between the occurrence of gusts and the change in generator speed; second, due to the nonlinear characteristics of blade aerodynamics, the larger the blade angle is when designing the PC, the smaller the gain of the controller. This results in a relatively slow blade back-off rate when the wind speed rises sharply, which cannot effectively suppress wind turbine overspeed.

[0135] The wind-storage coordinated control method provided in this embodiment specifies differentiated control by supercapacitors and energy storage batteries in conjunction with wind turbine pitch control for power fluctuations on different time scales, such as second-level and minute-level and above. This scheme achieves optimal allocation of control resources. It fully leverages the respective advantages of supercapacitors' rapid response and energy storage batteries' high energy density, avoiding waste of control resources or excessive operation of individual devices. Thus, it effectively reduces equipment losses and improves overall economic efficiency while efficiently smoothing power fluctuations.

[0136] This embodiment also provides a wind-storage coordinated control device, which is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0137] In conjunction with the above embodiments, in order to illustrate the implementation details of the technical solution in more detail in some embodiments of this application, the specific implementation process of the wind-storage coordinated control system is now supplemented as follows.

[0138] Step S201: Obtain the operation data of the target wind farm and perform turbulence multi-timescale analysis on the operation data to obtain turbulence characteristic parameters that characterize the fluctuation characteristics at different time scales.

[0139] In step S201, the multi-timescale turbulence analysis is based on the collected three-dimensional wind speed data to calculate core turbulence parameters (using a sliding window process with a window size of 10 minutes and a sliding step of 1 minute, consistent with the statistical characteristics of turbulence). The calculated turbulence characteristic parameters include turbulence intensity reflecting wave intensity, turbulence integral scale reflecting the average size of large eddies, and the energy proportion of each scale component obtained through analysis.

[0140] Among them, turbulence intensity ( T 1) is a quantitative index of turbulence fluctuation intensity, and the formula is:

[0141]

[0142] in, u , v ,w σ represents the three-dimensional wind speed components in the downwind, lateral, and vertical directions, respectively. u , σ v , σ w Then they respectively represent u , v , w The standard deviation of wind speed in each direction reflects the fluctuation range of wind speed in the three directions.

[0143] T The solution process for 1 includes:

[0144] (1) Calculate the average wind speed within the 10-minute window:

[0145]

[0146] in, N Let N be the number of data points within the window. For a 10-minute window with a sampling frequency of 50Hz, then N = 50 × 60 × 10 = 30000. u i , v i , w i The first in each of the three directions i The instantaneous flow velocity at a given moment.

[0147] (2) Calculate the fluctuating wind speed in the three directions, i.e., the deviation between the instantaneous wind speed and the average value:

[0148] , , ;

[0149] , , ;

[0150] in, u i , v i , w i For the first i The pulsating wind speed in each direction at any given moment. The average flow velocity in each direction within the window (the result of averaging the instantaneous flow velocities in all directions within a 10-minute window).

[0151] (3) Calculate the standard deviation of pulsations in the three directions:

[0152] ,

[0153] ,

[0154] ,

[0155] (4) Substitution T Formula 1 yields the turbulence intensity within the window.

[0156] In addition, the turbulence integral scale ( L The integral scale reflects the characteristic size of large turbulent eddies and is calculated using the autocorrelation method (along the downwind direction). u (Components):

[0157]

[0158] Then calculate the fluctuating wind speed. u autocorrelation function :

[0159]

[0160] in τ The time lag is denoted by E (in seconds), and E is the expected value.

[0161] turn up = 1 / e The lag time is approximately the upper limit of the integral. τ 0 (truncation time).

[0162] Trapezoidal numerical integration Multiply by the average wind speed to obtain the turbulence integral scale. L (Unit: m).

[0163] More specifically, step S201 includes:

[0164] In step S2011, wind speed sequences sampled at a preset frequency are obtained as operating data.

[0165] In step S2012, the wind speed sequence is processed using a wavelet transform algorithm or an empirical mode decomposition algorithm to obtain turbulence components at different time scales.

[0166] The process of distinguishing between mechanical turbulence and thermal turbulence by decomposing the frequency characteristics of turbulence signals using wavelet transform is as follows:

[0167] (1) Selection of wavelet basis and decomposition scale

[0168] The db4 wavelet (Daubechies4) is selected, which is suitable for non-stationary signals (turbulence is a non-stationary random process).

[0169] Decomposition scale: At a sampling rate of 50Hz, the decomposition is performed in 8 layers (corresponding frequency ranges are as follows), covering the main energy frequency bands of turbulence.

[0170] (2) Feature calculation

[0171] For each sliding window u′ i The pulsating signal is decomposed into 8 levels of wavelet to obtain the detail coefficients (high frequency) and approximation coefficients (low frequency) of each level.

[0172] Calculate the energy percentage of each layer, and the energy of a certain layer. ( c k For the first k Layer coefficient, E k For the first k Layer wavelet energy (representing the energy of that frequency component), total energy ;No. k Layer energy ratio This is used to distinguish between mechanical turbulence and thermal turbulence. (Energy percentage in mechanical turbulence) (Low frequency band), proportion of thermal turbulence energy (High frequency band).

[0173] Step S2013: Based on the energy distribution of the turbulent components, determine the dominant turbulence type and turbulence level. This step S2013 includes: inputting the energy distribution of the turbulent components into a trained LSTM model to predict the future turbulence level.

[0174] The steps for predicting the turbulence level (low / medium / high) for the next 10 minutes using LSTM based on historical turbulence parameters and spectral characteristics include:

[0175] (1) Dataset construction, which includes:

[0176] Input features (statistics for each sample within a 10-minute window);

[0177] Turbulence parameters: T 1. Turbulent Integral Scale L ;

[0178] Spectral characteristics: P mech , P therm ;

[0179] Historical time series: Extract features from the first 5 windows (50 minutes) to construct the input sequence. ,in F For the above four-dimensional features, t Indicates the start time of the prediction.

[0180] F t-5 Indicates the current time t The feature data from the previous 5th time window, F t-1 express t The feature data from the first time window. Each time window is 10 minutes long, therefore... X It actually includes t The historical feature sequence of the previous 50 consecutive minutes (5 windows) is used as input to the LSTM model for prediction. t Turbulence level (low / medium / high) for the next 10 minutes.

[0181] Output labels, categorized by turbulence level: Low: T 1 < 0.1; In this case: 0.1 ≤ T 1≤0.2; Height: T 1 > 0.2;

[0182] The data is divided into a 70% training set, a 20% validation set, and a 10% test set (covering different weather conditions, such as cloudy, sunny, rainy, snowy, and windy days).

[0183] Based on this, the dominant push flow type is determined according to the energy proportion of wavelet decomposition. Specifically, if... P mech > P therm +0.1 (significantly higher low-frequency energy) indicates that the dominant type is mechanical turbulence; if P therm > P mech If the value is +0.1 (significantly higher high-frequency energy), the dominant type is mechanical turbulence; otherwise, it is mixed turbulence.

[0184] Finally, the output results are integrated, and the final output format of the module is updated every minute.

[0185] Step S202: Based on turbulence characteristic parameters, a multi-objective optimization problem with wind turbine equivalent fatigue load, energy storage equipment life loss and power grid frequency regulation requirements as optimization objectives is solved by a multi-objective optimization algorithm to generate a cooperative control strategy.

[0186] This step specifically includes:

[0187] Step S2021: Initialize the particle swarm by randomly generating an initial position and velocity for each particle that represents the cooperative control strategy.

[0188] Step S2022: Enter the iterative optimization loop. Specifically, in each iteration, the following steps are performed:

[0189] Step c1: Based on the ratio of the current iteration number to the maximum iteration number, dynamically adjust the inertia weight and learning factor using an adaptive mechanism;

[0190] Step c2 updates the velocity and position of each particle based on the adaptively adjusted parameters, the individual historical best solution of each particle, and the global best solution selected from the external archive set, thereby generating a new candidate cooperative control strategy.

[0191] Step c3: Evaluate the new candidate cooperative control strategy using a multi-objective function and determine the satisfaction of constraints. Based on the evaluation and determination results, update the individual historical optimal solution of each particle and update the external archive set. Compare the non-dominated solution generated in the current iteration with the existing solutions in the archive set and retain all non-dominated solutions.

[0192] Step c4: After updating the external archive set, the non-dominated solutions in the archive set are filtered based on the crowding distance in order to maintain the diversity of solution distribution in the target space.

[0193] In step c5, during the comparison of the quality of solutions, feasible solutions that satisfy all constraints are selected first. For infeasible solutions that do not satisfy the constraints, the constraint violation degree is calculated, and a dynamic penalty coefficient that increases with the number of iterations is used for penalty.

[0194] Step S2023: Repeat the iterative optimization loop until the pre-termination condition is met, and finally select the optimal cooperative control strategy from the external archive set.

[0195] In steps S2021 to S2023, it should be noted that, for the multi-objective optimization requirements of the wind power-energy storage collaborative control system (conflicting objectives: fatigue mitigation, energy storage life, power smoothing, frequency regulation response), traditional single-objective optimization algorithms cannot simultaneously take into account multiple objectives. Multi-objective optimization algorithms need to solve two core problems: convergence of solutions (approaching the Pareto optimal front) and diversity (uniformity of solution distribution).

[0196] Because the Multi-Objective Particle Swarm Optimization (MOPSO) algorithm has advantages in real-time performance, global optimization capability, and multi-objective adaptability, it is chosen for multi-objective optimization and is specifically improved to meet the real-time performance and constraint complexity requirements of wind power systems.

[0197] To address the issues of slow convergence and insufficient solution diversity in traditional MOPSO under high-dimensional constraints, three improvement mechanisms are designed:

[0198] (1) Adaptive parameter adjustment mechanism

[0199] The core parameter of the particle swarm optimization algorithm (inertia weight) ω Learning factorsc 1, c 2) Directly impacts search performance; traditional fixed parameters struggle to balance "global exploration" and "local development." Improvements are as follows:

[0200] Inertia weight ω Dynamic adjustment: Increase in the initial stage (first 40% of iterations) ω To enhance global exploration (avoiding local optima) and reduce the impact in later stages. ω To accelerate convergence:

[0201]

[0202] in: ω max =0.9, ω min =0.4, iter This represents the current iteration number. iter max =80 (balancing real-time performance and accuracy). A sinusoidal perturbation term is introduced to avoid convergence stagnation in the later stages.

[0203] Adaptive allocation of learning factors: cognitive factors c 1. Individual experience and social factors c 2. (Group experience) is adjusted in reverse with iteration:

[0204]

[0205]

[0206] Early stage c 1> c 2 (emphasizing individual exploration), later stage c 1 < c 2 (Emphasis on group collaboration convergence).

[0207] (2) Elite archiving and diversity maintenance mechanism

[0208] An external archive is used to store non-dominated solutions (Pareto optimal solutions). It is necessary to guarantee the convergence (approaching the true optimal front) and diversity (solutions are uniformly distributed in the objective space). The improvement is as follows:

[0209] Archive update strategy: Calculate the objective function value of the particle in each iteration. J 1 J 4. Filter out non-dominated solutions (i.e., no other solution is better than all objectives); if the new solution dominates an existing solution in the archive, delete the dominated solution and add the new solution; if the new solution and the archived solution do not dominate each other, calculate the crowding distance between the new solution and all solutions in the archive (to measure the degree of dispersion of solutions), and only keep the solutions with larger crowding distances (to ensure diversity).

[0210] Crowded distance calculation: for each target dimension J k Unsave the archive J k Sort, number i The crowding distance for each solution is:

[0211]

[0212] The congestion distance of boundary solutions is set to infinity and is preferentially retained. When the archive size exceeds a threshold (e.g., 100), the solution with the smallest congestion distance (the redundant solution in the densest region) is deleted.

[0213] (3) Constraint handling mechanism

[0214] The system has multiple types of hard constraints (such as wind turbine speed limits and energy storage SOC range). Traditional penalty function methods are prone to convergence deviations. Therefore, a hierarchical constraint handling strategy is designed.

[0215] Feasibility priority: In determining the dominance relationship of solutions, feasible solutions (that satisfy all constraints) take precedence over infeasible solutions;

[0216] Constraint violation quantification: For infeasible solutions, calculate the degree of violation. CV (Deviation from all constraints):

[0217]

[0218] in The deviation of the i-th constraint (e.g.) (This indicates that the pitch angle exceeds the limit).

[0219] Dynamic penalty coefficient: The penalty coefficient is small in the early stages of iteration (allowing a small amount of exploration of the infeasible region), and increases in the later stages (forcing convergence to the feasible region):

[0220]

[0221] =10, ensuring that the objective function value of infeasible solutions is significantly penalized.

[0222] Based on these improvements, the detailed flow of the multi-objective optimization algorithm for steps S2021 to S2023 is as follows:

[0223] (1) Initialization: The dimensions of the decision variables are set to (The number of variables that need to be adjusted in the optimization problem is 5), particle swarm size N=60 (balancing computational cost and diversity); initial particle positions are randomly generated: u i U(Lower bound, Upper bound), Initial velocity: v i U ( v max , v max ), v max =0.2 × variable range (limiting velocity to prevent particles from deviating from feasible); initialize individual optimal solutions. pbest i =u i External archive set Archive= .

[0224] (2) Calculation of objective functions and constraints: For each particle, calculate four objective functions based on the current state variables x (wind speed, SOC, etc.). J 1 J 4. Calculate the degree of constraint violation. CV Mark feasible solutions / infeasible solutions.

[0225] (3) Update individual optimal and archive set: If the current particle Dominate ,but ;like For non-dominated solutions, add them to the archive set Archive and prune the archives according to the crowding distance (maintain size ≤ 100).

[0226] (4) Global optimal solution (gbest) selection: Randomly select one solution from the archive set as the global optimal solution (based on roulette, the solution with a larger crowding distance has a higher probability of being selected) to avoid the algorithm converging to a local area.

[0227] (5) Particle velocity and position update: The velocity update formula is as follows:

[0228]

[0229] , ~ The number is random, and the speed is truncated at... .

[0230] in, for t At the +1st iteration, the... i The velocity of each particle; The inertial weight (controls the degree to which the particle retains its previous velocity, and has been adaptively adjusted in the improved mechanism); For the first t During the nth iterationi The velocity of each particle; For learning factors; A random number, following the rules of... U Uniform distribution of (0,1); The optimal position of the i-th particle (the optimal solution position found during the iteration process of this particle). The position of the i-th particle in the t-th iteration (the particle position in the current iteration, corresponding to the value of the decision variable); The globally optimal position (the Pareto optimal solution position selected from the external archive set); This is the upper limit of velocity (limiting the maximum velocity of particles to prevent them from deviating from the feasible region).

[0231] The position update formula is as follows: When the search exceeds the boundary, it forces the search back to the feasible region to avoid invalid searches.

[0232] (6) Convergence judgment: If the number of iterations reaches If the standard deviation of the crowding distance changes by less than 0.5% over 10 consecutive generations of archive sets, stop iterating.

[0233] (7) Final solution decision: Select the final cooperative control strategy from the archive set. A comprehensive evaluation using fuzzy membership functions is employed.

[0234] For each objective, define the membership degree:

[0235]

[0236] The larger the value, the better; the final solution is... ,in For real-time weighting (e.g., when grid frequency regulation is prioritized) The rest are 0.2).

[0237] in, The fuzzy membership degree of the k-th target quantifies the "quality" of a single target (value range [0,1]), with a larger value indicating better performance of the target; The function value for the k-th objective; The minimum value of the k-th target in the archive set; Let be the maximum value of the i-th target in the archive set.

[0238] It should be noted that, through the above steps, a multi-objective decision-making framework for the coordinated optimization of energy storage charging and discharging strategies and wind turbine control can be constructed under the multi-timescale characteristics of turbulence. Then, by dynamically coupling multiple optimization objectives—namely, energy storage system lifetime constraints, blade and tower fatigue damage constraints, and grid frequency fluctuation constraints—the control strategy can be achieved.

[0239] The steps for calculating the equivalent fatigue load (blade and tower fatigue damage constraints) of a wind turbine include:

[0240] Step e1: Based on the running data, the load cycle is counted using the rainflow counting method;

[0241] Step e2: Calculate the equivalent fatigue load value according to the linear cumulative damage criterion.

[0242] It should be noted that, under turbulent conditions, the functional relationship between the equivalent fatigue load of wind turbine components and wind speed and turbulence intensity can be constructed by combining the equivalent fatigue load formula with the cyclic statistical results of the rainflow counting method. The core logic is to transform the complex load history caused by turbulence into an equivalent constant amplitude load in order to quantify fatigue damage. The specific process is as follows:

[0243] The equivalent fatigue load (Leq) is calculated based on the Miner linear cumulative damage criterion and the cyclic statistical results of the rainflow counting method. Its general formula is:

[0244]

[0245] In this formula, L i For the first i Level load amplitude (obtained by rainflow counting method). n i For the first i The number of cycles of the load level (obtained by rainflow counting method). m The slope of the material's SN curve is the negative reciprocal (reflecting the material's sensitivity to fatigue; for example, in blade composites, it is typically taken as...). m =10), N ref The reference number of cycles (usually the total number of cycles within the design life, such as 10⁷ or 10⁸).

[0246] Based on this, the nonlinear relationship between the equivalent load and the turbulence intensity is the equivalent fatigue load ( L eq The correlation between ) and turbulence intensity is nonlinearly positive. According to the formula, L eq For load amplitude ( L i ) and number of cycles ( n i Changes in ) are highly sensitive, especially when m When the value is large (e.g., in blade composite materials) m =10), the increase in load amplitude caused by turbulence will lead to L eq Significant increase.

[0247] Furthermore, it should be noted that turbulence also has a mechanism for influencing equivalent fatigue loads; that is, turbulence significantly increases fatigue damage to wind turbine components by increasing the amplitude and frequency of load fluctuations. Specifically, this manifests as follows:

[0248] (1) Load amplitude ( L i ) and turbulence intensity ( I T The relationship between )

[0249] Turbulence causes drastic changes in wind speed within a short period of time, increasing the stress amplitude on the components. L i The load increases. For example, in the Mann turbulence model, when the ratio of lateral to longitudinal turbulence intensity increases from 0.7 to 1.0, the loads at the blade root, rotor, and yaw position increase by 5% to 10%.

[0250] The relationship between load amplitude and turbulence intensity can be approximated as: , where V is the wind speed and k is an index related to the characteristics of the wind field (usually taken as 1 to 2).

[0251] (2) Number of cycles ( n i Relationship between ) and turbulence intensity

[0252] Turbulence causes more frequent fluctuations in the load history, and the number of cycles counted by the rainflow counting method ( n i The number of cycles increases significantly. For example, in the Kaimal turbulence model, for every 0.05 increase in turbulence intensity, the number of cycles may increase by 15% to 20%.

[0253] The relationship between the number of cycles and the turbulence intensity can be expressed as: ,in p and q The experience index (usually taken as) p =0.5, q =1)

[0254] (3) Functional relationship between equivalent fatigue load and turbulence intensity and wind speed

[0255] Substituting the expressions for load amplitude and number of cycles into the equivalent fatigue load formula, we get:

[0256]

[0257] The simplified relationship between the equivalent fatigue load and turbulence intensity and wind speed can be expressed as:

[0258]

[0259] in, CThese are constants related to the material and component geometry;

[0260] α and β The experience index (usually taken as) α =1.0~1.5, β =1.5~2.5).

[0261] (4) Mapping function between equivalent fatigue load and economic benefits

[0262] Based on the above, it is also necessary to clarify the equivalent fatigue load ( L eq The relationship between equivalent fatigue load and equipment maintenance or replacement costs. Equivalent fatigue load is an indicator that measures the fatigue damage that equipment experiences during operation. A higher equivalent fatigue load usually means that the equipment is more prone to fatigue damage, thus requiring more frequent maintenance or earlier replacement, which directly leads to increased maintenance or equipment replacement costs. Conversely, reducing the equivalent fatigue load can reduce these costs.

[0263] A mathematical model is established to map the reduction of equivalent fatigue load to economic benefits (i.e., reduction in operation and maintenance or replacement costs). Assumptions:

[0264] C op Current maintenance and equipment replacement costs (baseline costs).

[0265] C op，new Reduce the maintenance and equipment replacement costs after reducing the equivalent fatigue load.

[0266] △ L eq The reduction in equivalent fatigue load ( ).

[0267] △ C op The reduction in maintenance and equipment replacement costs ( ).

[0268] A nonlinear function can be found: .

[0269] This function can describe the relationship between the reduction of equivalent fatigue load and the reduction of operation and maintenance costs.

[0270] II. The calculation process for the lifespan loss of energy storage equipment (lifespan constraint of energy storage system) is as follows:

[0271] First, the constraints of charging and discharging energy storage devices need to be considered, as follows:

[0272] (1) Charging power constraints

[0273]

[0274] Among them, P bat，min P represents the minimum charging and discharging power of the battery. bat，max P represents the maximum charge / discharge power of the battery. bat ( t (Time) t The battery charging and discharging power.

[0275] (2) Battery temperature constraints

[0276]

[0277] in, T bat，min This is the lowest temperature at which the battery operates. T bat，max This is the highest operating temperature for the battery. T bat ( t (Time) t Battery temperature.

[0278] (3) Battery internal resistance constraint conditions

[0279]

[0280] in, R int，min This represents the minimum internal resistance of the battery. R int，max This represents the maximum value of the battery's internal resistance. R int ( t (Time) t The internal resistance of the battery.

[0281] (4) Constraint on the number of iterations

[0282]

[0283] in, N cycles，min This represents the minimum number of battery cycles. N cycles，max This represents the maximum number of battery cycles. N cycles ( t (Time) t The number of battery cycles.

[0284] (5) Battery life constraints

[0285]

[0286] in, L loss，min This represents the minimum value for battery life loss. L loss，max This represents the minimum value for battery life loss. L loss This is a function representing battery life loss.

[0287] Based on the above constraints, it should be noted that the lifespan loss of energy storage devices is related to the number of charge-discharge cycles, generally defined by the total amount of charge-discharge capacity in each cycle. Therefore, lifespan loss is directly proportional to the amount of energy stored during charge-discharge. Furthermore, discharging energy at a higher state of charge has less impact on its lifespan, while discharging at a lower state of charge accelerates its degradation. In this application, the energy storage devices mainly include lithium-ion batteries and supercapacitors.

[0288] (1) The following formula is the lithium-ion battery cycle aging with respect to the average state of charge of energy storage ( SOC ), depth of discharge ( DOD The stress expression for temperature can be used to calculate the lifespan degradation caused by each charge-discharge cycle.

[0289]

[0290] In the formula: This refers to the lifespan of a lithium-ion battery that decreases due to cycle aging per unit time, compared to... DOD , SOC It is related to temperature; For lithium-ion batteries DOD Stress coefficient. This model can effectively assess the degree of battery cycle aging under different operating conditions, where, DOD The stress coefficient expression is a complex exponential form.

[0291] The cost of lifespan loss per unit charge / discharge capacity of energy storage batteries is The cost of life loss is expressed as:

[0292] .

[0293] (2) The cost loss of each charge of a supercapacitor needs to be considered in conjunction with the total life-cycle cost of the equipment, the cycle life, and the lifespan consumption of each charge-discharge cycle. The initial purchase cost of the supercapacitor is allocated over all its effective charge-discharge cycles; the cost loss per charge is the cost allocation per cycle. Each charge-discharge cycle of a supercapacitor corresponds to one charge-discharge cycle (consuming one lifespan), therefore, the cost loss per charge is the "unit cost of the total life-cycle cost allocated to the actual cycle life," i.e.:

[0294]

[0295] in, The cost of each charge and discharge cycle. This represents the total cost over the entire lifespan of a supercapacitor. This represents the actual number of times it was used.

[0296] Based on this, the cycle life of supercapacitors decreases with increasing temperature, conforming to the Arrhenius aging law: the higher the temperature, the faster the electrochemical reaction rate, and the more significant the lifespan degradation. Considering temperature and depth of charge / discharge (…),… DOD The relationship between the actual cycle life (considering temperature and depth of charge / discharge corrections) and the rated life of a supercapacitor, considering factors such as temperature and depth of charge / discharge, is as follows:

[0297]

[0298] Substituting the actual cycle life into the formula, the final formula is:

[0299]

[0300] in, Cost of each charge (unit: yuan / charge). Total cost over the entire lifespan of the equipment; Rated lifespan of equipment (e.g., the rated lifespan of Maxwell supercapacitors is typically 500,000 to 1,000,000 cycles at 25°C). Fitting through accelerated aging tests The gas constant is 8.314 J / (mol·K). This refers to the actual operating temperature. Reference operating temperature, k The influence coefficient is the depth of charge / discharge (DOD), which is the depth of charge / discharge (dimensionless, 0~1).

[0301] III. The calculation process for the power grid frequency regulation requirements (power grid frequency fluctuation constraints) is as follows:

[0302] It should be noted that power grid frequency regulation needs to respond to system frequency deviation ( , =50Hz), the power deficit is compensated by rapid charging and discharging of energy storage to maintain frequency stability. More specifically, this includes the following constraints:

[0303] (1) Frequency response characteristic constraints

[0304] Response speed: Primary frequency regulation (seconds): Energy storage needs to respond to frequency deviation within 10 seconds (e.g., GB / T34120 requires primary frequency regulation response time ≤ 10s); Secondary frequency regulation (minutes): Response time ≤ 60 seconds, and needs to track AGC (Automatic Generation Control) commands.

[0305] Frequency dead zone and limiting: Dead zone, no action is taken when the frequency deviation is within ±0.05Hz (to avoid frequent adjustments), i.e., |Δ f When | < 0.05Hz, energy storage does not participate in frequency regulation; when the frequency deviation exceeds ±0.5Hz, emergency limiting is triggered (such as full-power charging and discharging of energy storage) to prevent system instability.

[0306] (2) Frequency modulation capacity and power constraints

[0307] Frequency regulation capacity: Energy storage needs to reserve sufficient capacity for frequency regulation, typically 20% to 30% of its rated capacity (to avoid depletion of capacity due to load shedding), that is:

[0308]

[0309] in, Reserve capacity for FM modulation. This is the rated capacity.

[0310] Charging and discharging power matching: The frequency regulation power of energy storage must be proportional to the frequency deviation (droop characteristic), that is:

[0311]

[0312] in, This is the frequency modulation factor (e.g., 100kW / Hz, meaning that the energy storage output is 100kW when the frequency deviation is 1Hz), and .

[0313] It should be noted that, based on the aforementioned constraints, after the model is deployed at the edge, it acquires state-space parameters of the current state of various devices by obtaining relevant parameters from turbulence sensing devices, the state of supercapacitors, the state of energy storage batteries, the current state of wind turbines (including operating status, cumulative equivalent fatigue load of components, etc.), and grid frequency regulation requirements. Through the optimization strategy model, it outputs corresponding collaborative control optimization strategies. After solving the problem, a model with multiple inputs such as wind speed, turbulence intensity, ambient temperature, and grid frequency is obtained, with the mapping space representing the collaborative control strategy for supercapacitor, energy storage battery charging and discharging, and wind turbine control.

[0314] Step S203: Execute the coordinated control strategy to coordinate the control of the wind turbine and the energy storage system.

[0315] Through the above methods, this application reduces the levelized cost of electricity (LCOE) by more than 15% by multi-objective collaborative optimization while ensuring the safety of the wind power-energy storage combined system throughout its entire life cycle. At the same time, it controls the fatigue life loss rate of the blade tower to within 80% of the design value and meets the stringent requirements of the power grid for frequency regulation of new energy power plants (such as frequency deviation ≤ ±0.05Hz and response time ≤ 200ms).

[0316] This embodiment provides a wind-storage coordinated control device, such as... Figure 7 As shown, it includes:

[0317] The turbulence sensing module 601 is used to acquire the operating data of the target wind farm and perform turbulence multi-timescale analysis on the operating data to obtain turbulence characteristic parameters that characterize the fluctuation characteristics at different time scales.

[0318] The collaborative optimization module 602 is used to solve a multi-objective optimization problem based on turbulence characteristic parameters, with the optimization objectives being the equivalent fatigue load of the wind turbine, the life loss of the energy storage equipment, and the frequency regulation requirements of the power grid, and to generate a collaborative control strategy.

[0319] The control execution module 603 is used to execute the coordinated control strategy to coordinate the control of the wind turbine and the energy storage system.

[0320] More specifically, in some embodiments of this application, the turbulence sensing module 601 can be implemented in hardware through a sensor system and an edge computing unit deployed at the wind farm site.

[0321] Specifically, a three-cup or ultrasonic turbulence anemometer is installed at the height of the wind turbine hub, sampling at a high frequency of 50Hz to simultaneously collect wind speed, wind direction, and turbulence intensity (TI) data. The data is transmitted via RS485 or fiber optic communication to an industrial computer or edge computing gateway, which serves as an edge computing unit (ECU). This ECU is responsible for running signal processing algorithms, such as using wavelet transform or empirical mode decomposition (EMD) to process the wind speed sequence, obtaining turbulence components at different time scales, and determining the dominant turbulence type and level based on energy distribution. Furthermore, a trained LSTM model can be integrated to predict future turbulence trends.

[0322] Building upon this foundation, the core of the collaborative optimization module 602 is an improved multi-objective particle swarm optimization (MOPSO) algorithm running on an edge computing unit (ECU). This ECU connects to the wind turbine main control system and energy storage system monitoring unit via an internal wind farm control network (such as an industrial Ethernet network) to acquire data including the wind turbine's equivalent fatigue load (calculated using rainflow counting and linear cumulative damage criteria), the energy storage battery's state of charge (SOC), temperature, internal resistance, cycle count, and real-time grid frequency, which serve as inputs to the optimization problem. The algorithm dynamically adjusts the inertia weights and learning factors through an adaptive mechanism, maintains an external archive set to store non-dominated solutions, and employs a hierarchical constraint processing strategy to ensure the generation of the optimal collaborative control strategy under hard constraints such as wind turbine speed safety limits, energy storage SOC range, and grid frequency regulation requirements (e.g., response time ≤ 10 seconds, reserved capacity ≥ 20%).

[0323] The control execution module 603 converts the generated collaborative control strategy into specific control commands through the ECU's communication interface (such as MODBUS TCP) and sends them to the actuators. For power fluctuations on a second-level timescale, it generates commands to control the supercapacitor bank (e.g., configured as 1MW / 1MWh) to perform rapid charging and discharging through its DC / DC converter; for power fluctuations on a minute-level or longer timescale, it generates collaborative commands to control the lithium battery energy storage system (e.g., configured as 5MW / 10MWh) to perform charging and discharging, and the wind turbine pitch control system to perform pitch angle adjustment. This module ensures the accurate and real-time execution of the collaborative control strategy, realizing the collaborative control of the wind and energy storage systems.

[0324] The wind-storage coordinated control device provided in this application can execute the wind-storage coordinated control method provided in any embodiment of this application, and has the corresponding functional modules and beneficial effects for executing the method. Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0325] Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0326] The following is a detailed reference. Figure 8 The diagram illustrates a structural schematic suitable for implementing the electronic device described in the embodiments of this application. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 701, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 702 or a program loaded from memory 708 into random access memory (RAM) 703. The RAM 703 also stores various programs and data required for the operation of the electronic device. The processor 701, ROM 702, and RAM 703 are interconnected via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0327] Typically, the following devices can be connected to I / O interface 705: input devices 706 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 707 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 708 including, for example, magnetic tapes, hard disks, etc.; and communication devices 709. Communication device 709 allows electronic devices to exchange data via wireless or wired communication with other devices. Although Figure 7 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0328] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication device 709, or installed from memory 708, or installed from ROM 702. When the computer program is executed by processor 701, it performs the functions defined in the wind-storage coordinated control method of embodiments of this application.

[0329] Figure 8 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0330] This application also provides a computer-readable storage medium. The methods described in this application can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the wind and energy storage coordinated control method shown in the above embodiments is implemented.

[0331] A portion of this application can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to this application through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0332] Although embodiments of this application have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of this application, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A wind storage collaborative control method, characterized in that, include: The operation data of the target wind farm is acquired, and turbulence multi-timescale analysis is performed on the operation data to obtain turbulence characteristic parameters that characterize the fluctuation characteristics at different time scales. Based on the turbulence characteristic parameters, a multi-objective optimization problem with wind turbine equivalent fatigue load, energy storage equipment life loss and power grid frequency regulation requirements as optimization objectives is solved by a multi-objective optimization algorithm to generate a collaborative control strategy. The aforementioned coordinated control strategy is implemented to coordinate the control of the wind turbine and the energy storage system. The multi-objective optimization problem is solved using a multi-objective particle swarm optimization algorithm. The process of generating a cooperative control strategy by solving the multi-objective optimization problem using the multi-objective particle swarm optimization algorithm includes: Initialize the particle swarm by randomly generating an initial position and velocity for each particle, representing the cooperative control strategy. Enter the iterative optimization loop, and in each iteration, perform the following steps: Based on the ratio of the current iteration count to the maximum iteration count, an adaptive mechanism is used to dynamically adjust the inertia weight and learning factor. Based on the adaptively adjusted parameters, the individual historical optimal solution of each particle, and the global optimal solution selected from the external archive set, the velocity and position of each particle are updated, thereby generating new candidate cooperative control strategies. The new candidate cooperative control strategy is evaluated using a multi-objective function and the constraint satisfaction is judged. Based on the evaluation and judgment results, the individual historical optimal solution of each particle is updated, and the external archive set is updated. The non-dominated solution generated in the current iteration is compared with the existing solutions in the archive set, and all non-dominated solutions are retained. After updating the external archive set, the non-dominated solutions within the archive set are filtered based on the crowding distance to maintain the diversity of solution distribution in the target space; In comparing the quality of solutions, feasible solutions that satisfy all constraints are selected first. For infeasible solutions that do not satisfy the constraints, the degree of constraint violation is calculated and a dynamic penalty coefficient that increases with the number of iterations is used for penalty. Repeat the iterative optimization loop until the pre-termination condition is met, and select the optimal cooperative control strategy from the external archive set; The steps of dynamically adjusting the inertia weights and learning factors using an adaptive mechanism include: Based on the ratio of the current iteration number to the total iteration number, the inertia weight is dynamically calculated, so that the inertia weight decreases linearly from a preset maximum value to a preset minimum value as the ratio increases. Based on the ratio of the current iteration number to the total iteration number, the cognitive learning factor and the social learning factor are adjusted in reverse, so that the cognitive learning factor decreases as the ratio increases, and the social learning factor c2 increases as the ratio increases. A sinusoidal perturbation term is introduced into the adjustment of the inertia weight, and the amplitude of the sinusoidal perturbation term decreases as the number of iterations increases.

2. The wind-storage coordinated control method according to claim 1, characterized in that, The steps of acquiring the operational data of the target wind farm and performing turbulence multi-timescale analysis on the operational data include: The wind speed sequence sampled at a preset frequency is obtained as the operational data; The wind speed sequence was processed using a signal decomposition algorithm to obtain turbulence components at different time scales; Based on the energy distribution of the turbulent components, the dominant turbulence type and turbulence level are determined.

3. The wind-storage coordinated control method according to claim 2, characterized in that, The determination of the dominant turbulence type and turbulence level based on the energy distribution of the turbulence components includes: The energy distribution of the turbulence components is input into a trained LSTM model to predict the future turbulence level.

4. The wind-storage coordinated control method according to claim 1, characterized in that, Before generating the cooperative control strategy based on the turbulence characteristic parameters, the following steps are also included: Based on the aforementioned operational data, load cycles are statistically analyzed using the rainflow counting method. The equivalent fatigue load value is calculated according to the linear cumulative damage criterion.

5. The wind-storage coordinated control method according to claim 1, characterized in that, The power grid frequency regulation requirement is defined as a set of constraints for the multi-objective optimization problem. The set of constraints includes: the frequency regulation response time is no more than 10 seconds, and the energy storage capacity reserved for frequency regulation is no less than 20% of the rated capacity of the energy storage system.

6. The wind-storage coordinated control method according to claim 1, characterized in that, The execution of the collaborative control strategy includes: For power fluctuations on a timescale of seconds, commands are generated to control the supercapacitor to charge and discharge rapidly. For power fluctuations on time scales of minutes and above, commands are generated to control the charging and discharging of energy storage batteries and the pitch adjustment of wind turbines.

7. A wind storage cooperative control device, characterized by, The device includes: The turbulence sensing module is used to acquire the operational data of the target wind farm and perform turbulence multi-timescale analysis on the operational data to obtain turbulence characteristic parameters that characterize the fluctuation characteristics at different time scales. The collaborative optimization module is used to solve a multi-objective optimization problem based on the turbulence characteristic parameters, with the optimization objectives being the equivalent fatigue load of the wind turbine, the life loss of the energy storage equipment, and the frequency regulation requirements of the power grid, and to generate a collaborative control strategy. The control execution module is used to execute the coordinated control strategy to coordinate the control of the wind turbine and the energy storage system. The multi-objective optimization problem is solved using a multi-objective particle swarm optimization algorithm. The process of generating a cooperative control strategy by solving the multi-objective optimization problem using the multi-objective particle swarm optimization algorithm includes: Initialize the particle swarm by randomly generating an initial position and velocity for each particle, representing the cooperative control strategy. Enter the iterative optimization loop, and in each iteration, perform the following steps: Based on the ratio of the current iteration count to the maximum iteration count, an adaptive mechanism is used to dynamically adjust the inertia weight and learning factor. Based on the adaptively adjusted parameters, the individual historical optimal solution of each particle, and the global optimal solution selected from the external archive set, the velocity and position of each particle are updated, thereby generating new candidate cooperative control strategies. The new candidate cooperative control strategy is evaluated using a multi-objective function and the constraint satisfaction is judged. Based on the evaluation and judgment results, the individual historical optimal solution of each particle is updated, and the external archive set is updated. The non-dominated solution generated in the current iteration is compared with the existing solutions in the archive set, and all non-dominated solutions are retained. After updating the external archive set, the non-dominated solutions within the archive set are filtered based on the crowding distance to maintain the diversity of solution distribution in the target space; In comparing the quality of solutions, feasible solutions that satisfy all constraints are selected first. For infeasible solutions that do not satisfy the constraints, the degree of constraint violation is calculated and a dynamic penalty coefficient that increases with the number of iterations is used for penalty. Repeat the iterative optimization loop until the pre-termination condition is met, and select the optimal cooperative control strategy from the external archive set; The steps of dynamically adjusting the inertia weights and learning factors using an adaptive mechanism include: Based on the ratio of the current iteration number to the total iteration number, the inertia weight is dynamically calculated, so that the inertia weight decreases linearly from a preset maximum value to a preset minimum value as the ratio increases. Based on the ratio of the current iteration number to the total iteration number, the cognitive learning factor and the social learning factor are adjusted in reverse, so that the cognitive learning factor decreases as the ratio increases, and the social learning factor c2 increases as the ratio increases. A sinusoidal perturbation term is introduced into the adjustment of the inertia weight, and the amplitude of the sinusoidal perturbation term decreases as the number of iterations increases.

8. An electronic device, characterized in that, include: A memory and a processor are interconnected, the memory stores computer instructions, and the processor executes the wind-storage coordinated control method according to any one of claims 1 to 6 by executing the computer instructions.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute the wind-storage coordinated control method according to any one of claims 1 to 6.

10. A computer program product, characterized in that, It includes computer instructions for causing a computer to execute the wind-storage coordinated control method according to any one of claims 1 to 6.

Images