Floating wind power fluctuation stabilizing system and method

Abstract

The invention discloses a floating wind power fluctuation stabilizing system and method. The system comprises a wind power prediction module which predicts power based on an LSTM network and processes data by adopting a regression filling method; the model prediction control module is used for constructing a multi-objective optimization problem to coordinate grid-connected power smoothness, energy storage SOC maintenance and output minimization; the fluctuation decomposition module is used for decomposing power fluctuation into high-frequency and low-frequency components by using VMD; the hybrid energy storage module comprises flywheel energy storage and super-capacitor energy storage which are used for processing low-frequency and high-frequency components respectively; and the coordination controller dynamically distributes hybrid energy storage output. The floating wind power complex power fluctuation can be effectively coped with, the impact on a power grid is remarkably reduced, the multi-time scale out-of-limit times and the energy storage demand quantity are introduced to serve as evaluation indexes, a quantitative basis is provided for control strategy optimization and energy storage economic configuration, and the smooth effect, robustness and economical efficiency of the system are improved.

Description

Technical Field

[0001] This invention relates to the field of wind power generation and energy storage technology, and in particular to a floating wind power power fluctuation smoothing system and method. Background Technology

[0002] With the rapid development of wind power installed capacity, wind power equipment technology and operation and maintenance technology are constantly advancing. As an emerging technology, floating wind power elevates the power generation system to an altitude of 500 to 1000 meters using tethered cables, generating electricity from stable and strong high-altitude winds. It offers advantages such as material savings, low cost per kilowatt-hour, and rapid deployment. However, floating wind power systems face multiple disturbances, including wind shear, platform movement, and pitch vibration, resulting in complex and frequent power fluctuations that traditional control methods struggle to effectively mitigate.

[0003] Existing wind power mitigation methods often employ filtering, frequency division, or fitting strategies, but these suffer from issues such as sensitivity to control parameters, overcompensation, and large energy storage requirements, and fail to consider multi-timescale fluctuation characteristics. Furthermore, current technologies lack systematic evaluation indicators for wind power mitigation strategies. Specifically, existing methods lack a comprehensive assessment of "number of times exceeding limits across multiple timescales" and "energy storage demand," making it impossible to scientifically evaluate the mitigation effect and the economics of energy storage configuration. Therefore, there is an urgent need for a comprehensive control method capable of accurately predicting, dynamically decomposing, and coordinating energy storage, and possessing quantitative evaluation capabilities, to address the complex, multi-timescale power fluctuations caused by wind shear and platform movement in floating wind power. Summary of the Invention

[0004] The purpose of this invention is to provide a floating wind power fluctuation mitigation system and method, which aims to solve the technical problems in the existing system, such as poor mitigation effect due to the complex and variable fluctuations of floating wind power, sensitivity of control parameters, excessive energy storage capacity and lack of effective quantitative evaluation indicators.

[0005] To address the aforementioned problems, according to one aspect of this application, an embodiment of the present invention provides a floating wind power fluctuation mitigation system, comprising:

[0006] Wind power prediction module based on long short-term memory network The power output of floating wind power is predicted, and regression imputation is used to handle missing data and outliers;

[0007] The model prediction control module receives the output of the prediction module and constructs a multi-objective optimization problem to coordinate the smoothness of grid-connected power, the maintenance of the state of charge of energy storage, and the minimization of energy storage output.

[0008] The fluctuation decomposition module utilizes variational mode decomposition. Power fluctuations are decomposed into high-frequency and low-frequency components.

[0009] The hybrid energy storage module includes flywheel energy storage and supercapacitor energy storage, which respectively process the low-frequency and high-frequency components;

[0010] The coordination controller dynamically allocates the output of the hybrid energy storage module.

[0011] Furthermore, the system introduces the number of times the limit is exceeded at multiple time scales and the energy storage demand as evaluation indicators for the smoothing strategy; the number of times the limit is exceeded at multiple time scales is used to count the number of times the grid-connected power exceeds the allowable fluctuation threshold at different time scales (seconds, minutes); the energy storage demand is used to evaluate the minimum energy storage capacity and power required to achieve the target smoothing effect, and guide the economic configuration of the energy storage system.

[0012] In some embodiments, the wind power prediction module, The network input includes historical wind power data, and time-series memory is achieved through input gates, forget gates, and output gates. Its gating mechanism is defined by the following formula:

[0013] ;

[0014] ;

[0015] ;

[0016] ;

[0017] ;

[0018] ;

[0019] In the formula, and These are the forget gate, input gate, input node, output gate, intermediate output, and state unit state, respectively. and These are the corresponding gates and inputs. Intermediate output Multiplication matrix weights; These are the corresponding gate bias terms; This indicates element-wise multiplication; , They are respectively and Function changes.

[0020] In some implementations, the multi-objective optimization function of the model prediction control module is:

[0021] ;

[0022] In the formula, The weighting factors are dynamic and obtained through fuzzy rules. The median state of charge of the energy storage system is given by the energy storage system. The closer the energy storage system is to the median value, the better it can mitigate sudden changes in wind power output. yes Step size; The output power of the energy storage system; It refers to the state of charge of the energy storage system; This is for The state of charge of the energy storage unit at any given time.

[0023] In some implementations, the constraints of the optimization problem include:

[0024] ;

[0025] in, This is the permissible grid-connected power fluctuation. This refers to the rated power of the energy storage system. The quadratic programming method is used to solve the problem, which yields the energy storage output domain.

[0026] In some implementations, the wave decomposition module employs The algorithm decomposes power fluctuations into multiple intrinsic mode functions. It is then reconstructed into high-frequency and low-frequency components, with the high-frequency components ranging from milliseconds to seconds and the low-frequency components ranging from seconds to minutes.

[0027] In some implementations, the flywheel energy storage is used to absorb and release low-frequency fluctuations on the order of seconds to minutes, and the supercapacitor energy storage is used to compensate for high-frequency pulse power on the order of milliseconds to seconds.

[0028] This invention also provides a method for smoothing power fluctuations in floating wind power, comprising the following steps:

[0029] Historical wind power data was collected, and missing and outlier values ​​were handled using regression imputation.

[0030] use Networks predict future wind power output;

[0031] Model-based predictive control Construct a multi-objective optimization problem to generate energy storage control commands;

[0032] Variational mode decomposition Power fluctuations are decomposed into high-frequency and low-frequency components.

[0033] The low-frequency and high-frequency components are processed by flywheel energy storage and supercapacitor energy storage, respectively.

[0034] Furthermore, the method also includes a power mitigation effect evaluation step: calculating the number of power outages and energy storage demand at multiple time scales to quantitatively evaluate the power mitigation effect and the economics of energy storage system configuration.

[0035] In some embodiments, the The objective function of the optimization problem is:

[0036] ;

[0037] The solution is obtained using a quadratic programming method.

[0038] In some implementations, the The number of decomposed modes is set to 5-10, and the penalty factor is 2000.

[0039] This invention also provides a non-transitory storage medium storing a computer program, which, when executed by a processor, implements any of the methods described above.

[0040] Compared with the prior art, the floating wind power fluctuation suppression system and method of the present invention have at least the following beneficial effects:

[0041] pass Prediction and Optimization significantly reduces the impact of power fluctuations on the power grid.

[0042] use Decompose fluctuations to achieve precise allocation and efficient utilization of energy storage resources.

[0043] Introducing the number of time-scale limit exceedances and energy storage demand as evaluation indicators provides a quantitative basis for optimizing control strategies and capacity planning of energy storage systems, thereby improving system robustness.

[0044] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings. Attached Figure Description

[0045] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 This is a schematic diagram of a wind power energy storage system.

[0047] Figure 2 for Diagram of the network's internal input gates, output gates, and forget gates;

[0048] Figure 3 This is a diagram of a flywheel energy storage topology.

[0049] Figure 4 This is a diagram of a supercapacitor energy storage structure. Detailed Implementation

[0050] To further illustrate the technical means and effects adopted by the present invention to achieve the intended purpose, the specific embodiments, structures, features, and effects according to the present invention will be described in detail below with reference to the accompanying drawings and preferred embodiments. In the following description, different "an embodiment" or "an embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0051] In the description of this invention, it should be clearly stated that the terms "first," "second," etc., in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence; the terms "vertical," "lateral," "longitudinal," "front," "rear," "left," "right," "up," "down," "horizontal," etc., indicate orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, and are merely for the convenience of describing this invention, and do not mean that the device or element referred to must have a specific orientation or position, and therefore should not be construed as a limitation of this invention.

[0052] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0053] like Figures 1-4 As shown, this embodiment of the invention provides a floating wind power fluctuation mitigation system, comprising:

[0054] Wind power prediction module based on long short-term memory network The power output of floating wind power is predicted, and regression imputation is used to handle missing data and outliers;

[0055] The model prediction control module receives the output of the prediction module and constructs a multi-objective optimization problem to coordinate the smoothness of grid-connected power, the maintenance of the state of charge of energy storage, and the minimization of energy storage output.

[0056] The fluctuation decomposition module utilizes variational mode decomposition. Power fluctuations are decomposed into high-frequency and low-frequency components.

[0057] The hybrid energy storage module includes flywheel energy storage and supercapacitor energy storage, which respectively process the low-frequency and high-frequency components;

[0058] The coordination controller dynamically allocates the output of the hybrid energy storage module.

[0059] In this embodiment, the floating wind power system is raised to an altitude of 500 to 1000 meters via tethered cables to generate electricity. While this offers advantages such as material savings and low cost per kilowatt-hour, power fluctuations are complex and frequent due to wind shear, platform movement, and pitch vibration. This system integrates five modules to mitigate these fluctuations: the wind power prediction module first uses regression imputation to repair missing and outlier values ​​in historical wind power data, and then uses a long short-term memory network... Predict future power; the model prediction and control module receives the prediction results, constructs a multi-objective optimization problem, and coordinates the smoothness of grid-connected power, the maintenance of the energy storage state of charge, and the minimization of energy storage output; the fluctuation decomposition module uses variational mode decomposition... Power fluctuations are broken down into high-frequency (millisecond to second-level) and low-frequency (second to minute-level) components. In the hybrid energy storage module, flywheel energy storage handles the low-frequency component, while supercapacitor energy storage addresses the high-frequency component. A coordinating controller dynamically allocates the output of both. The modules work together to reduce the impact of power fluctuations on the power grid, improve energy storage utilization efficiency, enhance the system's adaptability to high-altitude wind farms, and lay the foundation for the large-scale application of floating wind power.

[0060] Figure 1 middle, , and These respectively represent the joint system in The grid power, energy storage unit power, and wind turbine output power at any given time. The control cycle of the energy storage unit; This represents the total capacity of the energy storage unit.

[0061] In some embodiments, the wind power prediction module, The network input includes historical wind power data, and time-series memory is achieved through input gates, forget gates, and output gates. Its gating mechanism is defined by the following formula:

[0062] ;

[0063] ;

[0064] ;

[0065] ;

[0066] ;

[0067] ;

[0068] In the formula, and These are the forget gate, input gate, input node, output gate, intermediate output, and state unit state, respectively. and These are the corresponding gates and inputs. Intermediate output Multiplication matrix weights; These are the corresponding gate bias terms; This indicates element-wise multiplication; , They are respectively and Function changes.

[0069] In this embodiment, the wind power prediction module The network uses historical wind power data as input, and its gating mechanism ensures prediction accuracy. Historical data is prone to missing or outlier values ​​due to abnormal collection and transmission. It is first imputed by regression, which builds a regression equation based on data correlation and adds a random term to repair the data, thus providing the network with high-quality data. The forget gate combines the current input with the intermediate output from the previous time step, through The function filters historical state information, removing redundancy; the input gate is also based on the intermediate output between the current input and the previous time step, borrowing... The function determines the update ratio. The function generates candidate information and updates the state unit; the output gate controls the output of the state unit information, which is then processed... and The function processes the intermediate output. This mechanism solves the gradient vanishing problem in traditional recurrent neural networks, accurately captures power timing characteristics, improves prediction accuracy, provides a reliable basis for subsequent optimization control, and avoids over-compensation or under-compensation of energy storage.

[0070] In some implementations, the multi-objective optimization function of the model prediction control module is:

[0071] ;

[0072] In the formula, The weighting factors are dynamic and obtained through fuzzy rules. The median state of charge of the energy storage system is given by the energy storage system. The closer the energy storage system is to the median value, the better it can mitigate sudden changes in wind power output. yes Step size; The output power of the energy storage system; The state of charge of the energy storage system ; This is for The state of charge of the energy storage unit at any given time.

[0073] In this embodiment, the wind power prediction module The network uses historical wind power data as input, and its gating mechanism ensures prediction accuracy. Historical data is prone to missing or outlier values ​​due to abnormal collection and transmission. It is first imputed by regression, which builds a regression equation based on data correlation and adds a random term to repair the data, thus providing the network with high-quality data. The forget gate combines the current input with the intermediate output from the previous time step, through The function filters historical state information, removing redundancy; the input gate is also based on the intermediate output between the current input and the previous time step, borrowing... The function determines the update ratio. The function generates candidate information and updates the state unit; the output gate controls the output of the state unit information, which is then processed... and The function processes the intermediate output. This mechanism solves the gradient vanishing problem in traditional recurrent neural networks, accurately captures power timing characteristics, improves prediction accuracy, provides a reliable basis for subsequent optimization control, and avoids over-compensation or under-compensation of energy storage.

[0074] In some implementations, the constraints of the optimization problem include:

[0075] ;

[0076] in, This is the permissible grid-connected power fluctuation. This refers to the rated power of the energy storage system. The quadratic programming method is used to solve the problem, which yields the energy storage output domain.

[0077] In this embodiment, the constraints of the model predictive control module define the boundaries for safe system operation. Firstly, it limits the energy storage output to within the rated power to protect the equipment from overload damage. Secondly, it sets a grid-connected power fluctuation threshold, requiring that the difference in grid-connected power between adjacent times does not exceed the limit, meeting grid stability requirements and reducing the impact on grid frequency and voltage. Thirdly, it limits the energy storage state of charge to a safe range to avoid overcharging and over-discharging, ensuring energy storage safety and retaining a regulatory margin to cope with sudden fluctuations. These constraints, in conjunction with the multi-objective optimization function, in... The quadratic programming method is used to quickly obtain the energy storage output domain that satisfies the constraints. The constraints ensure that the system optimization does not exceed the safety threshold, achieving a balance between performance and safety, and improving system reliability.

[0078] In some implementations, the wave decomposition module employs The algorithm decomposes power fluctuations into multiple intrinsic mode functions. It is then reconstructed into high-frequency and low-frequency components, with the high-frequency components ranging from milliseconds to seconds and the low-frequency components ranging from seconds to minutes.

[0079] In this embodiment, the wave decomposition module's The algorithm accurately breaks down power fluctuations. It is particularly effective for handling the complex fluctuations inherent in floating wind power. First, set 5-10 modes (to balance decomposition accuracy and efficiency) and a penalty factor of 2000 (to avoid mode aliasing). Decompose power fluctuations into multiple intrinsic mode functions (EMFs), then reconstruct them by frequency into high-frequency (millisecond to second level, corresponding to turbulence and high-frequency platform vibration) and low-frequency (second to minute level, corresponding to slow wind speed changes and long-period platform motion) components. This algorithm is more adaptable and accurate than traditional filtering and frequency division methods. Precise fluctuation decomposition allows for differentiated processing of hybrid energy storage modules, avoiding waste of energy storage resources, improving smoothing efficiency, and providing a clear basis for subsequent energy storage scheduling.

[0080] In some implementations, the flywheel energy storage is used to absorb and release low-frequency fluctuations on the order of seconds to minutes, and the supercapacitor energy storage is used to compensate for high-frequency pulse power on the order of milliseconds to seconds.

[0081] In this embodiment, the flywheel energy storage and supercapacitor energy storage of the hybrid energy storage module are divided according to their characteristics. The flywheel energy storage includes components such as radial magnetic levitation bearings, a sealed shell, and a flywheel hub. The magnetic levitation bearings reduce friction, and the sealed shell protects the operating environment, giving it the advantages of large capacity and long-term energy storage. It is specifically designed to absorb and release low-frequency fluctuations in the range of seconds to minutes, and to cope with power fluctuations caused by slow wind speed changes and long-cycle platform movements. The supercapacitor energy storage consists of electrolyte, separator plates, and electrodes. It has a fast response and high charging and discharging efficiency, and is focused on compensating for high-frequency pulse power in the range of milliseconds to seconds, smoothing out the instantaneous impact caused by turbulence and high-frequency platform vibrations. The two work together to cover fluctuations of different frequencies, leveraging their respective advantages, reducing the load on a single energy storage device, extending its service life, reducing the cost of energy storage configuration, and ensuring stable grid-connected power.

[0082] This invention also provides a method for smoothing power fluctuations in floating wind power, comprising the following steps:

[0083] Historical wind power data was collected, and missing and outlier values ​​were handled using regression imputation.

[0084] use Networks predict future wind power output;

[0085] Model-based predictive control Construct a multi-objective optimization problem to generate energy storage control commands;

[0086] Variational mode decomposition Power fluctuations are decomposed into high-frequency and low-frequency components.

[0087] The low-frequency and high-frequency components are processed by flywheel energy storage and supercapacitor energy storage, respectively.

[0088] In this embodiment, the smoothing method operates according to the process of "data preprocessing - power prediction - optimized control - fluctuation decomposition - energy storage smoothing". First, historical wind power data is collected, and missing and outlier values ​​are repaired using regression imputation; then... The network predicts future power and captures time-series characteristics; then, based on model predictive control, a multi-objective optimization problem is constructed, using dynamic weighting factors to coordinate objectives, and constrained quadratic programming is used to generate energy storage commands; subsequently, ... The algorithm (modal number 5-10, penalty factor 2000) decomposes power fluctuations into high- and low-frequency components; finally, flywheel energy storage handles the low-frequency components, and supercapacitor energy storage addresses the high-frequency components. This method solves the problems of parameter sensitivity and overcompensation in traditional wind power smoothing methods, takes into account fluctuations across multiple time scales, improves smoothing effect and energy storage efficiency, and provides a feasible path for grid connection of floating wind power.

[0089] In some implementations, the The objective function of the optimization problem is:

[0090] ;

[0091] The solution is obtained using a quadratic programming method.

[0092] In this embodiment, The objective function of the optimization problem focuses on grid-connected power smoothness, energy storage state of charge maintenance, and minimizing energy storage output. Dynamic weighting factors (obtained through fuzzy rules) coordinate priorities, and the median state of charge is set to 0.5. The function transforms multi-objective optimization into a mathematical model. The solution is obtained using quadratic programming. Quadratic programming is highly efficient and accurate. During the solution process, it strictly adheres to constraints such as the rated power of the energy storage, the grid-connected power fluctuation threshold, and the energy storage state of charge range, quickly converging to the optimal solution and outputting energy storage charging and discharging commands. This solution method meets the real-time control requirements of the system, ensuring command feasibility. While guaranteeing grid connection stability, it improves energy storage utilization efficiency, extends energy storage life, and reduces operating costs.

[0093] In some implementations, the The number of decomposed modes is set to 5-10, and the penalty factor is 2000.

[0094] In this embodiment, The decomposition parameters (5-10 modes, penalty factor 2000) were experimentally determined to suit the power fluctuation characteristics of floating wind power. Too few modes lead to incomplete decomposition, failing to separate all frequency components; too many modes increase computational load and reduce real-time performance. A mode number of 5-10 balances accuracy and efficiency. A penalty factor that is too small easily leads to mode aliasing, distorting the decomposition results; a penalty factor that is too large over-constrains the mode bandwidth, failing to reflect the true fluctuations. A penalty factor of 2000 ensures the independence of each intrinsic mode function. This parameter combination allows... The system efficiently and accurately separates power fluctuations into high- and low-frequency components, providing a basis for differentiated processing of hybrid energy storage, avoiding waste of energy storage resources, improving the smoothing effect, and ensuring the system's rapid response to wind field changes.

[0095] This invention also provides a non-transitory storage medium storing a computer program, which, when executed by a processor, implements any of the methods described above.

[0096] In this embodiment, the computer program stored on the non-transitory storage medium can fully implement the method for smoothing fluctuations in floating wind power. The non-transitory nature of the storage medium ensures long-term stable storage of the program, is compatible with various processors and control systems, and facilitates widespread application. During program execution, historical data is first collected and repaired using regression imputation, and then... Predict power, then generate energy storage commands based on model predictive control, and then use... The system decomposes fluctuations and ultimately controls and smooths them using hybrid energy storage. The various modules of the program are seamlessly integrated, allowing for flexible parameter adjustments to adapt to different wind farms, reducing operation and maintenance costs, and enabling the smoothing algorithm to be implemented in engineering applications, thus promoting the large-scale development of floating wind power in the renewable energy sector.

[0097] This invention proposes a combination Wind power prediction models and as well as A flywheel-supercapacitor hybrid energy storage system is proposed to mitigate power output fluctuations in floating wind power. This technical solution fully utilizes several advanced technologies, including wind power output prediction, fluctuation decomposition, and energy storage system scheduling and control. Through precise prediction and dynamic adjustment, it effectively mitigates wind power fluctuations. The following will describe in detail the various components of the technical solution and their working principles.

[0098] It is a special type of recurrent neural network It is suitable for processing time-series data and has the ability to remember long-term dependencies. Wind power, as time-series data, is affected by factors such as wind speed and weather conditions, and typically exhibits strong time-series characteristics. Therefore, Networks can effectively predict future wind power by learning from historical wind power data, thus providing accurate input for subsequent control strategies. Historical wind power data is a crucial factor in power prediction; however, due to system equipment issues, problems such as data acquisition anomalies, input anomalies, and transmission anomalies can easily occur, leading to data outliers and missing data in wind power data. Data outliers refer to a portion of the dataset whose numerical distribution differs significantly from other data. When conducting data analysis or model training, this outlier can severely interfere with the results. On a graph, this manifests as large spikes mixed in with a continuous data stream, similar to large noise in a signal acting at a single point in time. Numerically, outliers are typically two close time points interspersed with a value that differs significantly from them. Missing data, on the other hand, is random, represented on a graph as a discontinuity in a data stream, causing the data stream to lose its continuity. Numerically, discontinuities are typically empty or zero.

[0099] Here are a few specific examples:

[0100] Example 1

[0101] This invention first uses regression imputation to clean the dataset, detects and repairs outliers, and then performs a standardization transformation to improve the accuracy of the prediction model.

[0102] 101: Regression imputation refers to establishing a regression equation using the complete dataset based on the relationship between existing data and other attribute values. During the imputation process, a random term is added to the imputed values, and then the regression equation's predicted values ​​are applied to fill in the missing data. For objects containing null values, known wind power values ​​are substituted into the equation to estimate the missing wind power values, which are then filled into the original wind power data.

[0103] 102: It is an improvement. ,Depend on It was proposed in 1997. The internal structure of the network's input gate, output gate, and forget gate is as follows: Figure 1 As shown. Through its internal input gate Output gate Forgotten Gate This technology enables the memorization of temporal features and addresses issues such as gradient vanishing, and is currently widely used in popular fields such as natural language processing and computer vision. Meanwhile... Compared to It can remember and forget information over longer periods, and performs exceptionally well in predicting time-series data. Wind power data, in particular, is a type of time-series data that varies over time and exhibits distinct temporal characteristics. Networks have been widely used in the field of wind power prediction.

[0104] 103: In practical control systems, the behavior of the system is difficult to predict completely due to changes in the external environment, incomplete knowledge of system parameters, measurement errors, and other reasons. This unpredictability is called uncertainty. The power fluctuations of the wind power system involved in this invention are affected by uncertainties in the external environment such as wind speed and climate, making precise control strategies difficult. The energy storage and release process in a flywheel-supercapacitor hybrid energy storage system is often nonlinear. Due to the complex dynamic characteristics (such as nonlinearity, coupling, and uncertainty) and constraints of practical control systems, traditional control methods cannot provide accurate modeling and control effects when facing these challenges. Therefore, to address these problems, this invention applies... As an advanced optimization control technique, it has significant advantages in handling multivariable and multi-constraint problems involving complex control systems. Through model prediction, optimization solution, and constraint handling, It can effectively address problems that are difficult to solve with traditional control methods, and improve the stability, efficiency and security of the system.

[0105] 104: For example Figure 2 As shown, the power smoothing control of the wind power energy storage system is based on the premise that the power in the future period is predictable. Based on the output results of the models in 102 and 103, the control command of the energy storage system is generated according to Equation (11) within the energy storage dispatch control interval. In the actual control process, if the state of charge of the energy storage system is not considered and only the power smoothing and grid connection power requirements are considered, the energy storage system may experience capacity limitations, especially when the wind turbine is under extreme weather conditions. At this time, the rapid change in wind speed requires the state of charge of the energy storage system to be kept at an intermediate level in order to meet the purpose of power smoothing control. Therefore, in step 104, under the premise of ensuring the wind power smoothing requirements and safety, from the perspective of ensuring the charging and discharging capacity of the energy storage system under extreme weather conditions, an optimized control strategy for realizing the power smoothing of the wind power energy storage system is proposed. The objective function of the optimized control of the wind power energy storage system is defined in step 204. At the beginning of each control cycle, The controller will receive data from upstream. The predictive model provides a sequence of floating wind power forecasts for a future period (e.g., 60 seconds). The controller then constructs a multi-objective optimization problem within a finite time domain. The objective function of this problem aims to simultaneously coordinate three crucial objectives: optimal grid-connected power smoothness, maintaining the energy storage device's state of charge (SOC) at approximately 0.5, and minimizing the energy storage device's output.

[0106] 105: Building upon steps 101-104, this step focuses on... It can decompose complex signals into multiple local vibration modes, is suitable for processing signals with significant local frequency changes, and can effectively extract intrinsic mode functions. In this invention, a ground-deployed flywheel-supercapacitor hybrid energy storage system is configured for the floating power generation system and the floating wind power system, and managed by an intelligent coordination controller.

[0107] Using the total energy storage domain obtained from 104 as the target energy storage output, this invention adopts... The algorithm can decompose it into different components and then reconstruct it into high-frequency and low-frequency parts for allocation, such as... Figure 3 , 4 As shown, the flywheel energy storage is responsible for absorbing and releasing low-to-medium frequency, high-frequency energy fluctuations ranging from seconds to minutes (such as those caused by slow wind speed changes or long-cycle platform movements). The supercapacitor is responsible for compensating for high-frequency, instantaneous pulse power ranging from milliseconds to seconds (such as those caused by turbulence or high-frequency platform vibrations).

[0108] In summary, through steps 101-105 above, the energy storage system can maintain a relatively ideal state of charge level in terms of power regulation, and has a significant effect on smoothing out wind power.

[0109] Example 2

[0110] The following section, combining specific calculation formulas and principle analysis, further introduces the scheme in Example 1, as detailed in the description below:

[0111] 201: The estimated value of missing values ​​in wind power data can be expressed as:

[0112] ;

[0113] In the formula, For missing variables, Is with Complete variables that exhibit a linear regression relationship It is a constant. It is a random term added to the fill value during the filling process, thereby reducing prediction error and overcoming the defect of distorted sample distribution.

[0114] Regression imputation can make full use of the information in the data itself to solve the problem of collinearity of predictor variables. This method can handle situations with a large amount of missing data and the imputed values ​​are more accurate.

[0115] 202: Moment Input features include Input vector Neuron state vector and output vector The forget gate, input gate, and output gate perform logical operations and nonlinear transformations based on the input characteristics to obtain the output value. The intrinsic relationship is shown in equations (2) to (7):

[0116] ;

[0117] ;

[0118] ;

[0119] ;

[0120] ;

[0121] ;

[0122] In the formula, and These are the forget gate, input gate, input node, output gate, intermediate output, and state unit state, respectively. and These are the corresponding gates and inputs. Intermediate output Multiplication matrix weights; These are the corresponding gate bias terms; This indicates element-wise multiplication; , They are respectively and Function changes.

[0123] 203: To address fluctuations and prediction errors in wind power output, energy storage systems are first used to compensate for these errors. These systems can absorb excess electricity predicted during wind power forecasting (when actual wind power output exceeds the predicted value) or release energy when wind power output is insufficient, thus compensating for the difference between actual and actual wind power output. After prediction error compensation, [further steps are taken]. The method further mitigates the volatility of wind power. The goal is to optimize the charging and discharging processes of energy storage devices (flywheels, supercapacitors) at each time step based on future wind power output forecasts, in order to reduce the impact of wind power fluctuations on the power grid. This is achieved by solving the optimization problem. It will predict future wind power fluctuations at each time step, combined with the power constraints and state of charge of the energy storage system. Constraints are imposed, and the output of the energy storage device is adjusted in real time to smooth out power fluctuations. The ultimate control objective is to keep the error between the actual wind power output and the predicted value within an acceptable range, while taking into account the operating conditions of the energy storage device. This requires the energy storage system's charging and discharging strategy to not only compensate for prediction errors but also consider the energy storage device's performance. Limiting and setting maximum power output ensures system stability.

[0124] ;

[0125] ;

[0126] ;

[0127] In the formula, for The state of charge of the energy storage unit at any given time. , , These respectively represent the joint system in The grid power, energy storage unit power, and wind turbine output power at any given time. For the control cycle of the energy storage unit, This represents the total capacity of the energy storage unit.

[0128] Let the state variable be... Input variables Uncertain variables Then the discretized state-space equation of the wind power energy storage system can be expressed in the following form:

[0129] ;

[0130] In the formula, , , .

[0131] 204: Under the premise that the wind power fluctuation rate meets the allowable requirements, and taking into account the minimum output and charge-discharge balance of the overall hybrid energy storage, the following is obtained: Solve for the objective function, with the goal of minimizing the output of energy storage in a predicted time domain.

[0132] ;

[0133] In the formula, The weighting factors are dynamic and obtained through fuzzy rules. The median state of charge of the energy storage system is given by the energy storage system. The closer the energy storage system is to the median value, the better it can mitigate sudden changes in wind power output. yes The step size.

[0134] The constraints for the optimization process are:

[0135] ;

[0136] In the formula, This is the permissible grid-connected power fluctuation. This refers to the rated power of the energy storage system. The quadratic programming method is used to solve the problem, which yields the energy storage output domain.

[0137] 205: Evaluation steps for mitigation effect: In order to quantitatively evaluate the mitigation effect of this method and the economics of energy storage configuration, two core evaluation indicators are introduced: the number of times the limit is exceeded at multiple time scales and the energy storage demand.

[0138] Statistics on the number of times limits are exceeded at multiple time scales: Allowable thresholds for grid-connected power fluctuations are set at two time scales: second-level and minute-level. Real-time monitoring of grid-connected power The statistics show that it surpasses the target within a rolling time window (e.g., 1 minute, 10 minutes). The fewer the number of times, the better the suppression effect.

[0139] Energy storage demand assessment: based on Optimization results and The decomposed power components are used to estimate the critical capacity of the energy storage system required to meet the smoothing requirements. With critical power This assessment comprehensively considers both historical extreme fluctuations and forecast uncertainties. Its calculation method is as follows: analyzing the target output sequence. The cumulative energy distribution (charge / discharge energy envelope) and power extremes over a typical operating cycle (e.g., 24 hours) are used to determine energy storage configuration parameters that meet a specific confidence level (e.g., 99%). This metric directly serves the capacity planning and cost control of energy storage systems.

[0140] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0141] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A floating wind power fluctuation smoothing system, characterized in that, include: Wind power prediction module based on long short-term memory network The power output of floating wind power is predicted, and regression imputation is used to handle missing data and outliers; The model prediction control module receives the output of the prediction module and constructs a multi-objective optimization problem to coordinate the smoothness of grid-connected power, the maintenance of the state of charge of energy storage, and the minimization of energy storage output. The fluctuation decomposition module utilizes variational mode decomposition. Power fluctuations are decomposed into high-frequency and low-frequency components. The hybrid energy storage module includes flywheel energy storage and supercapacitor energy storage, which respectively process the low-frequency and high-frequency components; The coordination controller dynamically allocates the output of the hybrid energy storage module.

2. The system according to claim 1, characterized in that, In the wind power prediction module The network input includes historical wind power data, and time-series memory is achieved through input gates, forget gates, and output gates. Its gating mechanism is defined by the following formula: ； ； ； ； ； ; In the formula, and These are the forget gate, input gate, input node, output gate, intermediate output, and state unit state, respectively. and These are the corresponding gates and inputs. Intermediate output Multiplication matrix weights; These are the corresponding gate bias terms; This indicates element-wise multiplication; , They are respectively and Function changes.

3. The system according to claim 1, characterized in that, The multi-objective optimization function of the model prediction control module is: ； In the formula, The weighting factors are dynamic and obtained through fuzzy rules. The median state of charge of the energy storage system is given by the energy storage system. The closer the energy storage system is to the median value, the better it can mitigate sudden changes in wind power output. yes Step size; The output power of the energy storage system; It refers to the state of charge of the energy storage system; This is for The state of charge of the energy storage unit at any given time.

4. The system according to claim 3, characterized in that, The constraints of the optimization problem include: ； in, This is the permissible grid-connected power fluctuation. This refers to the rated power of the energy storage system. The quadratic programming method is used to solve the problem, which yields the energy storage output domain.

5. The system according to claim 1, characterized in that, The wave decomposition module adopts The algorithm decomposes power fluctuations into multiple intrinsic mode functions. It is then reconstructed into high-frequency and low-frequency components, with the high-frequency components ranging from milliseconds to seconds and the low-frequency components ranging from seconds to minutes.

6. The system according to claim 1, characterized in that, The flywheel energy storage is used to absorb and release low-frequency fluctuations on the order of seconds to minutes, while the supercapacitor energy storage is used to compensate for high-frequency pulse power on the order of milliseconds to seconds.

7. A method for smoothing power fluctuations in floating wind power, characterized in that, Includes the following steps: Historical wind power data was collected, and missing and outlier values ​​were handled using regression imputation. use Networks predict future wind power output; Model-based predictive control Construct a multi-objective optimization problem to generate energy storage control commands; Variational mode decomposition Power fluctuations are decomposed into high-frequency and low-frequency components. The low-frequency and high-frequency components are processed by flywheel energy storage and supercapacitor energy storage, respectively.

8. The method according to claim 7, characterized in that, The The objective function of the optimization problem is: ； The solution is obtained using a quadratic programming method.

9. The method according to claim 7, characterized in that, The The number of decomposed modes is set to 5-10, and the penalty factor is 2000.

10. A non-transitory storage medium storing a computer program, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 7-9.

Images