A method for lightweight interaction between an offshore wind farm and a power grid

Abstract

The present application relates to a kind of far sea wind farm and power grid light weight interaction method, including the pretreatment of power regulation target curve, obtain the trend component, periodic component, random component, shape complexity index, shape descriptor vector and inflection point parameter set corresponding to power regulation target curve;The Fourier coding vector corresponding to the periodic component of power regulation target curve is obtained;Based on the trend component of power regulation target curve, shape descriptor vector and Fourier coding vector and the actual power output curve of real-time acquisition far sea wind farm, obtain comprehensive shape matching degree;Based on the regulation ability dimension reduction coding vector of far sea wind farm and comprehensive shape matching degree, obtain the optimal state estimation of far sea wind farm regulation ability state;Based on the inflection point parameter set of the actual power output curve of far sea wind farm, Fourier coding vector, and optimal state estimation and comprehensive shape matching degree, hierarchical light weight data processing is carried out, and dynamic optimization data package is generated.

Description

Technical Field

[0001] This invention relates to the fields of offshore wind power control and smart grid technology, and in particular to a lightweight interaction method between offshore wind farms and power grids. Background Technology

[0002] After large-scale grid connection of offshore wind power, the interaction between offshore wind farms and the power grid has become a key factor affecting the safe and stable operation of the power system. Currently, the interaction between offshore wind farms and the power grid mainly adopts the traditional power-voltage control mode. This involves real-time transmission of detailed operating parameters such as active power, reactive power, voltage, and frequency from the offshore wind farm. The power grid dispatch center then formulates control strategies based on this information and issues specific power regulation commands to the offshore wind farm. This interaction method requires the offshore wind farm to continuously upload a large amount of operating data to the power grid dispatch center, including multi-dimensional information such as single-unit power output, wind speed changes, and equipment status. Simultaneously, the power grid dispatch center needs to process this massive amount of data in real time and frequently issue regulation commands. However, due to the distance between offshore wind farms and load centers, and the complex offshore communication environment, traditional detailed information exchange methods face numerous challenges such as communication bandwidth limitations, data transmission delays, and information security risks, seriously affecting the real-time performance and reliability of the interaction between offshore wind farms and the power grid. Existing methods for interaction between offshore wind farms and the power grid have many shortcomings, mainly reflected in the following aspects.

[0003] First, existing interaction methods lack a unified consideration of power regulation targets, often focusing only on the numerical matching of power output (whether the output at the current moment is equal to the power target value at the current moment), while ignoring the dynamic characteristics and shape features of the target curve during the grid regulation process. They cannot identify whether the target curve is "rising sharply" or "falling slowly", nor do they care about the trend, making it difficult for the interaction results to efficiently achieve the power system regulation targets. Secondly, traditional power interaction mechanisms are too complex, requiring the transmission and processing of massive amounts of data, resulting in low communication efficiency. They are ill-suited to the inherent volatility and rapid changes in wind power generation. In daily operation, the complex interaction process can lead to insufficient adjustment response speed, making it difficult to track rapid fluctuations in wind power. This problem becomes even more prominent in emergencies such as sudden changes in wind speed or grid failures, seriously affecting the power system's rapid adjustment capabilities and safe and stable operation.

[0004] Furthermore, existing methods lack effective feature extraction and encoding mechanisms for the regulation capabilities of offshore wind farms. This results in the inability to accurately quantify and compactly represent the multi-dimensional regulation capabilities of offshore wind farms at different time scales, such as upward and downward adjustments. Consequently, the flexible regulation potential of wind farms cannot be fully explored and utilized. At the same time, in the cloud (grid dispatch center)-edge (offshore wind farm control center) collaborative environment, there is a lack of a protocol system specifically designed for lightweight communication needs. Even if the aforementioned feature information is extracted, it is difficult to transmit it efficiently under limited bandwidth, further limiting the improvement of the interaction efficiency between offshore wind farms and the power grid. Summary of the Invention

[0005] Based on the above analysis, the embodiments of the present invention aim to provide a lightweight interaction method between offshore wind farms and the power grid, in order to solve the technical problems of low efficiency and slow dynamic response of existing methods for offshore wind power grid connection interaction.

[0006] The objective of this invention is mainly achieved through the following technical solutions: This invention provides a lightweight interaction method between offshore wind farms and the power grid, comprising the following steps: The power regulation target curve sent from the power grid dispatch center to the offshore wind farm control center is preprocessed to obtain the trend component, periodic component, random component, shape complexity index, shape descriptor vector and inflection point parameter set corresponding to the power regulation target curve; The Fourier encoded vector corresponding to the periodic component of the power regulation target curve is obtained; based on the trend component, shape descriptor vector, and Fourier encoded vector of the power regulation target curve, as well as the actual power output curve of the offshore wind farm collected in real time, the comprehensive shape matching degree is obtained. Based on the dimensionality reduction encoding vector of the regulation capability of the offshore wind farm and the comprehensive shape matching degree, the optimal state estimate of the regulation capability state of the offshore wind farm is obtained. Based on the set of inflection point parameters and Fourier encoded vectors of the actual power output curve of the offshore wind farm, as well as the optimal state estimation and comprehensive shape matching degree, hierarchical lightweight data processing is performed to generate dynamic optimization data packages for the offshore wind farm.

[0007] Furthermore, the process of obtaining the trend component, periodic component, random component, shape complexity index, shape descriptor vector, and inflection point parameter set corresponding to the power adjustment target curve includes: The trend component of the power adjustment target curve is extracted by using a moving average filter to obtain the trend component of the power adjustment target curve; The power adjustment target curve is detrended to obtain the detrended signal; Fourier spectrum analysis is performed on the detrended signal to extract the main periodic components and obtain the periodic components. Subtracting the trend component and the periodic component from the power adjustment target curve yields the random component of the power adjustment target curve. The shape complexity index is obtained by calculating the ratio of the mean of the absolute values ​​of the second-order differences of the power regulation target curve to the power standard deviation. Extract the peak position ratio, rising slope, falling slope, average rate of change, peak-to-valley ratio, and zero crossover rate of the power regulation target curve to form a vector, and obtain the shape descriptor vector of the power regulation target curve; The power regulation target curve is compressed using a piecewise linear approximation. The time interval is divided into a preset number of sub-intervals. The optimal slope and optimal intercept of each linear function are determined using the least squares method. The key inflection points between each sub-interval are extracted to obtain the inflection point parameter set.

[0008] Further, a Fourier series expansion is performed on the periodic components of the power adjustment target curve to obtain the Fourier encoded vectors corresponding to the periodic components of the power adjustment target curve, including: The periodic components of the power adjustment target curve are standardized with zero mean and unit variance to obtain the standardized periodic components of the power adjustment target curve. Based on the energy conservation criterion, select the option that satisfies the preceding conditions. The proportion of the cumulative energy of each frequency component to the total energy is not less than the minimum value of the preset energy retention rate threshold, thus obtaining the truncation order of the Fourier series. ; Cutoff order based on Fourier series The standardized periodic components of the power regulation target curve are expanded as follows: The Fourier series of each frequency component is used to obtain the Fourier coefficients of the power adjustment target curve; the Fourier coefficients of the power adjustment target curve include the DC component coefficients, the cosine coefficients of each harmonic, and the sine coefficients. The Fourier encoding vector of the power adjustment target curve is constructed based on the Fourier coefficients of the power adjustment target curve; wherein the Fourier encoding vector of the power adjustment target curve includes the DC component coefficients and the cosine and sine coefficients of each harmonic.

[0009] Furthermore, obtaining the comprehensive shape matching degree includes: The trend components of the actual power output curve of the offshore wind farm and the trend components of the power regulation target curve are time-aligned and amplitude-normalized. The trend component of the actual power output curve of the offshore wind farm is extracted by using a moving average filter to obtain the trend component of the actual power output curve of the offshore wind farm; the correlation coefficient between the first derivative sequence of the trend component of the actual power output curve of the offshore wind farm and the trend component of the power adjustment target curve is calculated as the trend similarity. Extract the peak position ratio, rising slope, falling slope, average rate of change, peak-to-valley ratio, and zero crossover rate of the actual power output curve of the offshore wind farm to form a vector, and obtain the shape descriptor vector of the actual power output curve of the offshore wind farm; calculate the shape correlation based on the dynamic time bending distance between the shape descriptor vectors of the actual power output curve of the offshore wind farm and the power regulation target curve. Fourier series expansion is performed on the periodic components of the actual power output curve of the offshore wind farm to obtain the Fourier encoded vector of the actual power output curve of the offshore wind farm corresponding to the periodic components; the energy distribution similarity of the Fourier encoded vector of the actual power output curve of the offshore wind farm and the Fourier encoded vector of the power adjustment target curve are compared in each frequency band. The trend similarity, shape correlation, and energy distribution similarity are weighted and summed to obtain the comprehensive shape matching degree between the power regulation target curve and the actual power output curve of the offshore wind farm.

[0010] Furthermore, the weighting coefficients for trend similarity, shape correlation, and energy distribution similarity are obtained as follows: Based on the relative importance of power grid dispatch in three dimensions—trend similarity, shape correlation, and energy distribution similarity—a judgment matrix is ​​constructed. Calculate the largest eigenvalue of the judgment matrix and its corresponding eigenvector, normalize the eigenvector, and obtain the weight coefficients of the trend similarity, the shape correlation and the energy distribution similarity; A consistency check is performed on the judgment matrix. If the consistency ratio is less than a preset consistency ratio threshold, the weight coefficients of the trend similarity, the shape correlation, and the energy distribution similarity are determined to be valid; otherwise, the judgment matrix is ​​reconstructed and the weights are recalculated until the consistency check is passed.

[0011] Furthermore, an instantaneous regulation capability matrix of offshore wind farms is constructed, and the regulation capability confidence level is calculated; based on the instantaneous regulation capability matrix and regulation capability confidence level of offshore wind farms, a dimension-reduced encoding vector of the regulation capability of offshore wind farms is obtained. This includes constructing the instantaneous regulation capability matrix of offshore wind farms and calculating the regulation capability confidence level, including: Based on the rated power of the offshore wind farm, the actual output power at the current moment, and the measured wind speed at the current moment, the rapid upward adjustment capability and rapid downward adjustment capability of the offshore wind farm under the first adjustment time scale, and the slow upward adjustment capability and slow downward adjustment capability under the second adjustment time scale are calculated respectively; wherein, the first adjustment time scale is shorter than the second adjustment time scale. Based on the rapid upward and downward adjustment capabilities of the offshore wind farm under the first adjustment time scale, and the slow upward and slow downward adjustment capabilities under the second adjustment time scale, an instantaneous adjustment capability matrix of the offshore wind farm is constructed. The confidence level of the regulation capability of offshore wind farms is calculated based on the standard deviation of wind speed forecasts, which reflects the reliability of the regulation capability of offshore wind farms under the current uncertainty of wind speed forecasts.

[0012] Furthermore, based on the instantaneous regulation capability matrix and regulation capability confidence of the offshore wind farm, a dimension-reduced encoding vector of the regulation capability of the offshore wind farm is obtained, including: Based on the rapid upward and downward adjustment capabilities of the offshore wind farm under the first adjustment time scale, the slow upward and slow downward adjustment capabilities of the offshore wind farm under the second adjustment time scale, the adjustment capability confidence of the offshore wind farm, the measured wind speed at the current moment, and the adjustable power range of the offshore wind farm at the current moment, the adjustment capability feature vector of the offshore wind farm is obtained; wherein, the adjustable power range of the offshore wind farm at the current moment is the sum of the rapid upward and rapid downward adjustment capabilities. Based on the historical data of the regulation capacity of offshore wind farms within a preset period, a historical operation data feature matrix is ​​constructed, and the covariance matrix corresponding to the historical data feature matrix is ​​calculated. Using the principle of spectral decomposition, the covariance matrix is ​​decomposed into eigenvalues ​​to obtain the corresponding orthogonal eigenvector matrix and eigenvalue matrix; Before choosing The orthogonal eigenvectors corresponding to the largest eigenvalues ​​form the projection matrix; The regulation capability feature vector of the offshore wind farm is projected onto the projection matrix to obtain the regulation capability dimensionality reduction encoding vector of the offshore wind farm.

[0013] Furthermore, based on the dimensionality reduction encoding vector of the offshore wind farm's regulation capability and the comprehensive shape matching degree, the optimal state estimate of the offshore wind farm's regulation capability state is obtained, including: The regulation capability state space model is established by using the dimensionality reduction encoding vector of the offshore wind farm as the state vector. The regulation capability state space model includes a state equation and an observation equation. The state equation describes the time evolution of the regulation capability dimensionality reduction encoding vector, and the observation equation describes the mapping relationship from the state vector to the observation vector. The comprehensive shape matching degree is used as observation feedback and is a component of the observation vector. Based on the optimal state estimate and control input at the current moment, perform Kalman filtering prediction to predict the state vector and prediction error covariance matrix at the next moment; Obtain the observation vector at the next time step, and calculate the Kalman gain based on the prediction error covariance matrix and the observation noise covariance matrix; Based on the Kalman gain fusion prediction and observation values, the optimal state estimate for the next time step is obtained, and the error covariance matrix is ​​updated.

[0014] Furthermore, a dynamic switching coding scheme based on the rate of change of the actual power output curve of offshore wind farms, and dynamic adjustment of the redundancy of forward error correction coding based on channel quality indicators, includes: Calculate the rate of change of the actual power output curve of the offshore wind farm to be developed; When the rate of change is less than a preset rate of change threshold, a high compression ratio coding scheme of differential coding plus Huffman compression is adopted; otherwise, a low compression ratio and high reliability coding scheme of direct coding plus forward error correction is adopted. The channel quality index is calculated based on the ratio of effective bandwidth to nominal bandwidth and the bit error rate. When the channel quality index is greater than the first preset threshold, the first redundancy is adopted; When the channel quality index is less than or equal to the first preset threshold and greater than the second preset threshold, the second redundancy is adopted. When the channel quality index is less than or equal to the second preset threshold, a third redundancy is used; Wherein, the first preset threshold is greater than the second preset threshold; the first redundancy is less than the second redundancy, and the second redundancy is less than the third redundancy.

[0015] Furthermore, based on the dynamically switched coding scheme and dynamically adjusted redundancy, as well as the inflection point parameter set, Fourier coding vector, optimal state estimation, and comprehensive shape matching degree of the actual power output curve of the offshore wind farm, hierarchical lightweight data processing is performed to generate a dynamically optimized data package for the offshore wind farm, including: The layering includes the application layer, coding layer, transport layer, and physical layer; At the application layer, the raw time-series data to be transmitted is classified into real-time control data, status monitoring data, and historical statistical data. At the coding layer, based on the dynamically switched coding scheme and the dynamically adjusted redundancy, the inflection point parameter set, Fourier coding vector, and corresponding optimal state estimation and comprehensive shape matching degree of the actual power output curve of the offshore wind farm are coded, compressed and forward-corrected. At the transport layer, an adaptive packet splitting mechanism is adopted to dynamically adjust the packet size based on the size of the original time-series data packets to be transmitted, the available bandwidth of the current channel, the inherent network latency, and the inherent protocol overhead of the data packets. At the physical layer, channel quality is monitored, providing parameters for the calculation of the channel quality metrics; Based on the results of the encoding compression and forward error correction coding, and the dynamically adjusted data packet size, the dynamically optimized data packet for the offshore wind farm is obtained.

[0016] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects: 1. This invention significantly improves the matching accuracy and response effect of regulation targets. By constructing a lightweight interactive model based on the shape characteristics of the regulation power target curve, this invention fundamentally changes the traditional interactive method that only focuses on power numerical matching. By introducing a shape complexity index, Fourier series encoding, and a multi-dimensional shape matching degree evaluation model, the system can accurately capture the dynamic characteristics and shape features of the target curve during grid regulation, including key information such as trend changes, periodic fluctuations, and transient responses. In particular, through a comprehensive evaluation of trend similarity, shape correlation, and energy distribution similarity, it ensures that the wind farm output not only approaches the target numerically but, more importantly, achieves a high degree of matching in the dynamic characteristics of the regulation process, thereby significantly improving the control effect in key scenarios such as grid frequency regulation and voltage support. Compared to the traditional interactive method based on real-time power numerical matching (i.e., the dispatch center issues specific power value commands, and the wind farm adjusts its output according to the command value), this invention can improve target tracking accuracy by more than 30%, especially performing better in dynamic scenarios such as grid faults or load changes. Here, target tracking accuracy is defined as the comprehensive shape matching degree in step S2. The evaluation metric is that, under the same communication bandwidth conditions, the method of this invention improves performance by more than 30% compared to the traditional method (e.g., (From 0.65 to over 0.90); 2. This invention significantly reduces communication data volume and improves transmission efficiency. The multi-level data compression and coding technology employed in this invention can significantly reduce the communication burden. First, Fourier series coding compresses the high-dimensional power regulation target curve into a compact vector containing the main frequency components, typically compressing the original power target regulation curve data to 20-30% of its original size. Second, principal component analysis-based regulation capability feature dimensionality reduction coding further compresses the multi-dimensional regulation capability information into a low-dimensional regulation capability coding vector, reducing the data dimensionality by 60-70% while maintaining over 95% of the information content. Finally, a layered lightweight data transmission protocol, through differential coding, Huffman compression, and adaptive packet segmentation, obtains dynamically optimized data packets for offshore wind farms, further improving transmission efficiency. Combining the effects of various compression technologies, this invention can reduce the overall communication data volume by more than 70%, significantly improving the real-time performance and reliability of information transmission in complex offshore communication environments, and effectively alleviating bandwidth limitations and transmission delays in offshore environments. 3. This invention fully explores and utilizes the flexible regulation potential of offshore wind farms. The multi-dimensional regulation capacity modeling and quantification method for offshore wind farms established in this invention can comprehensively identify and evaluate the regulation resources of wind farms. By constructing an instantaneous regulation capacity matrix for offshore wind farms that includes rapid regulation capacity, slow regulation capacity, and confidence level assessment, the regulation potential of wind farms at different time scales can be accurately quantified, avoiding the problem of conservative or inaccurate regulation capacity assessment in traditional methods. In particular, by considering multiple factors such as wind speed changes, equipment status, and operational constraints, the system can dynamically optimize the output allocation of each wind turbine, maximizing the utilization rate of the entire wind farm's regulation capacity. Practice shows that this refined regulation capacity modeling method can increase the actual adjustable capacity of wind farms by more than 25%, providing more flexible resources for the power grid and effectively supporting the safe and stable operation of power systems with a high proportion of new energy sources. 4. This invention enhances the rapid adaptation and response capabilities to wind power fluctuations. The dynamic regulation capability prediction and update mechanism designed in this invention significantly improves the system's adaptability to the inherent fluctuations of wind power. By establishing a dynamic state estimation model based on Kalman filtering, it can track the changing trend of wind farm regulation capability in real time and proactively update the regulation capability assessment based on information such as wind speed prediction and equipment status. When wind speed changes drastically, there is no need to wait for feedback from actual power output; instead, the control strategy is adjusted in advance based on the prediction model, thereby shortening the response time by more than 50%. Simultaneously, by introducing a regulation capability confidence assessment mechanism, the impact of wind speed prediction uncertainty on regulation capability can be quantified, allowing for full consideration of uncertainty factors when formulating control strategies and improving the robustness of the control scheme. This proactive adaptive mechanism enables offshore wind farms to better cope with rapid changes in offshore wind conditions, improving overall operational stability. 5. This invention improves communication reliability and cluster coordination capabilities in complex marine environments. Addressing the challenges of complex communication environments faced by offshore wind farms, the adaptive compression and redundancy control strategy designed in this invention significantly enhances the reliability of the communication system. By monitoring channel quality indicators in real time, the system can dynamically adjust the redundancy of forward error correction coding. Redundancy is reduced to increase transmission rate when channel quality is good, and redundancy is increased to ensure transmission reliability when channel quality is poor. Simultaneously, the hierarchical data transmission protocol supports flexible network topology, enabling multiple offshore wind farms to coordinate efficiently with the power grid dispatch center, avoiding the problem of excessive communication burden hindering large-scale cluster coordination in traditional methods. This intelligent adaptive mechanism allows the system to maintain stable communication performance even in complex environments such as harsh sea conditions and electromagnetic interference, improving communication success rate by more than 20%, providing crucial technical support for the coordinated and optimized operation of large-scale offshore wind farm clusters and the construction of new power systems.

[0017] In this invention, the above-described technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of this invention will be set forth in the following description, and some advantages may become apparent from the description or be learned by practicing the invention. The objects and other advantages of this invention can be realized and obtained from what is particularly pointed out in the description and drawings. Attached Figure Description

[0018] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts.

[0019] Figure 1 This is a flowchart of a lightweight interaction method between an offshore wind farm and the power grid in an embodiment of the present invention; Figure 2 This is a schematic diagram of the lightweight interactive architecture between offshore wind farms and power grids in an embodiment of the present invention; Figure 3 This is a schematic diagram of the functional modules of a lightweight interaction system between an offshore wind farm and the power grid in an embodiment of the present invention. Detailed Implementation

[0020] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0021] To address the aforementioned deficiencies and shortcomings of existing technologies in the interaction between offshore wind power and the power grid, this invention discloses a lightweight method for interaction between offshore wind farms and the power grid. This method significantly reduces the amount of communication data, improves interaction efficiency, and enhances the power system's adaptability to the volatility of offshore wind power while ensuring regulation accuracy. Furthermore, it outlines a dynamic regulation capability prediction and adaptive update mechanism to enable the power system to proactively respond to rapid fluctuations in offshore wind power output, thereby further enhancing the power system's adaptability to the volatility of offshore wind power.

[0022] Specifically, this includes: how to construct a lightweight interaction mechanism based on the shape characteristics of the power regulation target curve to achieve an efficient interaction mechanism with shape matching as the core; how to establish a feature extraction and encoding method for the regulation capacity of offshore wind farms to fully explore and characterize the flexible regulation potential of offshore wind farms; designing a lightweight communication protocol suitable for the cloud (power grid dispatch center)-edge (offshore wind farm centralized control center) collaborative environment to ensure the real-time and reliable transmission of interactive information in the offshore environment; and establishing a forward-looking update mechanism for dynamic regulation capacity based on state prediction, enabling the power system to predict changes in regulation capacity before sudden changes in wind speed and shorten response time.

[0023] This invention specifically relates to a lightweight modeling method for communication protocols between large-scale offshore wind farms and the power grid, particularly suitable for scenarios where large-scale offshore wind power clusters are connected to the grid via flexible DC transmission systems. Here, "large-scale" refers to scenarios where a single wind farm has an installed capacity of no less than 200MW, or where the total installed capacity of the wind farm cluster is no less than 500MW, such as an offshore wind power cluster with a total installed capacity of 900MW composed of three 300MW wind farms.

[0024] Example 1: A specific embodiment of the present invention discloses a lightweight interaction method between offshore wind farms and the power grid, such as... Figure 1 As shown, it includes the following steps: Step S1: Preprocess the power regulation target curve sent from the power grid dispatch center to the offshore wind farm control center to obtain the trend component, periodic component, random component, shape complexity index, shape descriptor vector and inflection point parameter set corresponding to the power regulation target curve; obtain the Fourier encoding vector corresponding to the periodic component of the power regulation target curve. Step S2: Based on the trend component, shape descriptor vector, and Fourier encoding vector of the power adjustment target curve, as well as the real-time acquired actual power output curve of the offshore wind farm, the comprehensive shape matching degree is obtained. Step S3: Based on the dimensionality reduction encoding vector of the regulation capability of the offshore wind farm and the comprehensive shape matching degree, obtain the optimal state estimate of the regulation capability state of the offshore wind farm. Step S4: Based on the set of inflection point parameters and Fourier encoding vector of the actual power output curve of the offshore wind farm, as well as the optimal state estimation and comprehensive shape matching degree, hierarchical lightweight data processing is performed to generate dynamic optimization data packages for the offshore wind farm.

[0025] This invention employs a technical approach combining power regulation target curve shape feature extraction, multi-dimensional encoding of offshore wind farm regulation capabilities, and a cloud-edge adaptive communication protocol. This method transforms traditional detailed power data interaction into compact information interaction based on shape features and regulation capabilities. By combining advanced algorithms such as Fourier series encoding, principal component analysis dimensionality reduction, and Kalman filter prediction, it can achieve regulation accuracy (based on comprehensive shape matching degree) while maintaining regulation precision. Under the premise of not less than 0.90, the amount of communication data will be reduced by more than 70%, which will significantly improve the interaction efficiency and response speed between offshore wind farms and the power grid.

[0026] The core innovations of the entire technical solution include: establishing a parameterized representation and similarity evaluation model for target power curves based on shape matching; realizing a feature extraction and dimensionality reduction encoding mechanism for the multi-dimensional adjustment capabilities of offshore wind farms; and designing a hierarchical lightweight data transmission protocol system adapted to the complex communication environment at sea. This technical solution effectively overcomes the technical bottlenecks of traditional interactive methods, such as heavy communication burden, high data redundancy, and difficulty in adapting to the rapid fluctuations in wind power, providing key technical support for the safe and reliable grid connection and intelligent operation control of large-scale offshore wind power.

[0027] like Figure 2 As shown, the power grid dispatch center, acting as the cloud, formulates a power regulation target curve based on load forecasting and sends it to the offshore wind farm control center (edge) via a lightweight communication link. The offshore wind farm control center, based on the received power regulation target curve and the real-time collected operating status of the offshore wind turbine cluster (terminal), generates control commands and sends them to each turbine in the offshore wind turbine cluster (terminal) for execution. Simultaneously, it uploads the actual power output curve of the offshore wind farm to the offshore wind farm control center. The offshore wind farm control center generates dynamically optimized data packets over time and, using the lightweight communication link, transmits them back to the power grid dispatch center within the effective dispatch cycle to achieve grid connection of a large offshore wind turbine cluster, realizing a lightweight interactive closed loop of cloud-edge-device collaboration.

[0028] Step S1 includes steps S11-S12.

[0029] Step S11: Obtain the power regulation target curve sent from the power grid dispatch center to the offshore wind farm control center.

[0030] The power regulation target curve that varies over time, issued by the power grid dispatch center, is shown below: , Formula (1) in, To adjust the effective duration of the command, the unit is seconds or minutes. This indicates that the power grid dispatch center hopes that offshore wind farms will develop over time. Follows the ideal power output.

[0031] The power adjustment target curve is distributed in the form of discrete sampling points, as shown below: Formula (2) in, The sampling point number is from 1 to... , This indicates that the power regulation target curve issued by the power grid dispatch center to the offshore wind farm control center is in the [number]th [year]. The power target value at each discrete sampling point, i.e., the power target value of the offshore wind farm at the [number]th ... The time corresponding to each discrete sampling point The ideal power output should be; The sampling interval; Simply put, That is, the first one in the scheduling instruction The required power value at each time point; The value depends on the sampling frequency and the total time length. , This represents the total number of sampling points. Typically, if the scheduling instruction duration is 10 minutes and the sampling period is 1 second, then... .

[0032] This power regulation target curve represents the power grid dispatch center's expectation, based on load forecasting, that offshore wind farms will perform well in the future. The ideal power output trajectory that is followed is the sequence of target values ​​for adjusting the output of offshore wind farms.

[0033] Step S11 obtains the power regulation target curve represented by discrete sampling points issued by the power grid dispatch center based on load forecast, which serves as the input benchmark for subsequent feature extraction, shape matching and lightweight interaction.

[0034] Step S12 involves preprocessing the power adjustment target curve to obtain its corresponding structured features. These structured features include the trend component, periodic component, random component, shape complexity index, shape descriptor vector, and inflection point parameter set corresponding to the power adjustment target curve.

[0035] The obtained set of trend components, periodic components, random components, shape complexity index, shape descriptor vector, and inflection point parameters corresponding to the power adjustment target curve includes: The trend component of the power adjustment target curve is extracted by using a moving average filter to obtain the trend component of the power adjustment target curve; The power adjustment target curve is detrended to obtain the detrended signal; Fourier spectrum analysis is performed on the detrended signal to extract the main periodic components and obtain the periodic components. Subtracting the trend component and the periodic component from the power adjustment target curve yields the random component of the power adjustment target curve. The shape complexity index is obtained by calculating the ratio of the mean of the absolute values ​​of the second-order differences of the power regulation target curve to the power standard deviation. Extract the peak position ratio, rising slope, falling slope, average rate of change, peak-to-valley ratio, and zero crossover rate of the power regulation target curve to form a vector, and obtain the shape descriptor vector of the power regulation target curve; The power regulation target curve is compressed using a piecewise linear approximation. The time interval is divided into a preset number of sub-intervals. The optimal slope and optimal intercept of each linear function are determined using the least squares method. The key inflection points between each sub-interval are extracted to obtain the inflection point parameter set.

[0036] The first layer involves signal decomposition, extracting multi-resolution components from the target power adjustment target curve. This includes extracting trend components, periodic components, and random components.

[0037] Objective: To decompose the original power regulation target curve into three mutually orthogonal components according to frequency and scale of change, so that each component can serve as the data basis for subsequent steps.

[0038] The power regulation target curve is decomposed into trend component, periodic component and random component by a shape feature extraction algorithm based on differential sequence statistical analysis and piecewise linear fitting.

[0039] Power regulation target curve The decomposition process is as follows: (1) Extracting trend components : Trend components are extracted using moving average filtering, a time series analysis method suitable for extracting low-frequency trend components.

[0040] For discrete sampling sequences Using a window length of Moving average filter extraction ,as follows: Formula (3) in, The length of the sliding window is selected based on the adjustment cycle; for example, it is taken as... A smooth window corresponding to 15 sampling points; This is a temporary loop variable inside the sliding window; it is used to represent the relative position index of each sampling point within the window, and its scope is limited to the summation symbol ∑, used to traverse each sampling point within the sliding window. The power adjustment target curve is in the first... The target power value at each sampling point.

[0041] Trend Components It reflects the direction of change of the power regulation target curve over the overall time scale, such as gradually increasing from 100MW to 150MW, and is used to determine the macro trend of power regulation.

[0042] The value is determined based on the ratio of a typical power grid regulation cycle (approximately 15 to 30 seconds) to the sampling frequency (usually 1 Hz). In practical engineering, it can be determined based on the characteristics of the dispatch command and the wind turbine response speed. Adjustments can be made within the specified range.

[0043] (2) Extracting periodic components : First, the time series of the original power regulation target curve is detrended to obtain the detrended signal, as shown below: Formula (4) Signals after detrending Fourier spectrum analysis was performed to extract the periodic components corresponding to the main frequency domain peaks, and then reconstructed into periodic components. .

[0044] The periodic component reflects the regular fluctuation characteristics in the power regulation target curve, such as the power regulation fluctuation caused by the periodic changes in grid load, and is used to guide offshore wind farms to perform periodic power regulation actions.

[0045] (3) Calculate the random components As shown below: Formula (5) The random component is the residual term after subtracting the trend component and the periodic component from the original power regulation target curve. It reflects the unpredictable irregular disturbances in the power regulation target curve and is used to subsequently assess the difficulty and uncertainty of the regulation task.

[0046] The logical relationship between the power regulation target curve and the trend component, periodic component, and random component is shown below: Formula (6) The second layer: complexity quantification, calculating the shape complexity index of the power regulation target curve. .

[0047] Objective: To use a scalar indicator Quantitatively characterizing the drastic changes in the shape of the power adjustment target curve provides an objective basis for subsequent adaptive compression strategies.

[0048] The difference sequence is calculated as follows: The first-order difference (instantaneous rate of change) sequence of the power regulation target curve is calculated as follows: Formula (7) in, For the first-order difference, for the th... Instantaneous rate of change at each sampling point; For the first Each sampling point corresponds to The target power value at any given time.

[0049] The first-order difference sequence represents the change in the target power value between two adjacent sampling points, i.e., the instantaneous rate of change; ① A positive value indicates an increase in power; ② A negative value indicates a decrease in power.

[0050] The second-order difference (derivative of the rate of change) sequence is calculated as follows: Formula (8) in, For the second-order difference, is the change in the first-order difference; For the first Each sampling point corresponds to The target power value at any given time.

[0051] This invention extracts the shape features of the power regulation target curve by analyzing the statistical characteristics of the difference sequence.

[0052] Second-order difference sequence Represents a first-order difference sequence The change in the value reflects the curvature and inflection point information of the power adjustment target curve. The larger the absolute value, the steeper or more abrupt the local curve of the power adjustment target curve.

[0053] Calculate the shape complexity index: Shape complexity index of power regulation target curve As shown below: Formula (9) in, Power sequence for target power adjustment curve standard deviation Power sequence for target power adjustment curve The mean of the target power values ​​at each discrete sampling point.

[0054] The physical meaning of formula (9) is: the shape complexity index SC measures the ratio of the mean of the absolute values ​​of the second difference of the power regulation target curve to the standard deviation of the power, reflects the severity of the shape change of the power regulation target curve, and measures the degree of shape tortuosity of the power regulation target curve under the unit power change amplitude.

[0055] ①The larger the SC value, the more inflection points and abrupt changes there are in the power regulation target curve, and the more difficult the regulation is; ②The smaller the SC value, the flatter the power regulation target curve, and the simpler the regulation task.

[0056] The shape complexity index SC provides a quantitative basis for selecting the encoding scheme in the subsequent adaptive compression strategy.

[0057] Engineering criteria: ① The power adjustment target curve is flat, the adjustment task is simple, and a high compression ratio encoding scheme can be selected; ② The power adjustment target curve has a large number of inflection points and abrupt changes, making adjustment difficult and requiring a low compression ratio and high-fidelity coding scheme. ③ When the value is between 0.1 and 0.2, a medium compression strategy is adopted.

[0058] For example, a smooth, sloping ascending curve A curve containing multiple power reversal fluctuations. It can reach 0.3~0.5.

[0059] The third layer involves geometric feature extraction, constructing a shape descriptor vector for the power adjustment target curve. .

[0060] Objective: To construct a low-dimensional vector using six scalar parameters with explicit geometric meanings. It compactly depicts the overall shape of the power adjustment target curve, which is used to calculate trend similarity in subsequent shape matching evaluation.

[0061] Construct the shape descriptor vector of the power regulation target curve as follows: Formula (10) in, Key characteristic parameters of the power regulation target curve: peak position ratio (Ratio of peak occurrence time to total duration T), rising slope descent slope Average rate of change (Mean of absolute values ​​of first-order difference sequences), peak-to-valley ratio (The ratio of the difference between the maximum and minimum values ​​to the mean) Zero crossover rate (the frequency of sign changes in a first-order difference sequence).

[0062] The shape descriptor lower vector of the power regulation curve in this invention includes Key feature parameters are shown below: Formula (11) The physical meaning of formula (11) is that the shape descriptor vector The overall geometric shape of the power regulation target curve is abstracted into six key parameters and extracted into a low-dimensional vector representation. Each scalar parameter compactly characterizes the overall shape of the power adjustment target curve, with dimension m=6. This allows subsequent shape comparisons to be performed in a 6-dimensional feature space. This vector serves as the input for trend similarity calculation in subsequent shape matching evaluation, enabling shape comparisons to be conducted in a 6-dimensional feature space, rather than the original... Performed in a 3D sampling space, the computational complexity is reduced from Down to This significantly reduces computational complexity.

[0063] ① Peak position ratio This indicates the time of peak occurrence and the effective duration of the adjustment command. The ratios are shown below: Formula (12) in, The target power sequence; The peak time is within the entire time interval. Within, the absolute time corresponding to when the power adjustment target curve reaches its maximum value (peak), in seconds or minutes.

[0064] Indicates all Among the sampling points, find the point with the largest power value, i.e., the peak value; "Argument of the Maximum" is an abbreviation for the independent variable at which the maximum power value is obtained; in this case, it is the time point. In a discrete power series, the maximum power value is first found, and then the time point corresponding to this maximum power value is extracted and assigned to... .

[0065] The timing and location of the peak value reflect whether the center of gravity of the power regulation target curve is shifted forward or backward.

[0066] The timing of the peak appearance on the power grid regulation target curve: ①If This indicates that the peak appears in the first half of the process, requiring offshore wind farms to rise sharply first and then fall slowly. ②If This indicates that the peak occurs in the second half of the process, requiring the wind farm to rise slowly first and then rise sharply. ③If This indicates that the power regulation target curve is roughly symmetrical.

[0067] For example, assume the total duration of the scheduling instructions =600 seconds (10 minutes); Within these 10 minutes, the target power reaches its peak at the 240th second, i.e. =240 seconds.

[0068] Then the calculation shows that, =240 / 600=0.4; that is, the peak appears at 40% of the total duration (slightly earlier), and the power grid requires the offshore wind farm to respond quickly and exert its power early.

[0069] ② Slope of ascent This represents the average rate of ascent from the starting point to the peak value, reflecting how quickly the power system needs to climb, as shown below: Formula (13) in, Peak power; The starting power is the power value corresponding to the start time of the power adjustment target curve, which is the initial power target at the time the scheduling instruction is issued. This is the starting point.

[0070] ③ Descent slope This represents the average rate of decline from peak to end, reflecting how quickly the power system needs to recover, as shown below: Formula (14) in, For the final power, The endpoint.

[0071] The physical meaning of is the average decrease in power per unit time when the power regulation target curve falls from the peak point back to the end point.

[0072] ④ Average rate of change , representing the mean of the absolute values ​​of the first-order difference sequence, as shown below: Formula (15) The physical meaning of is to represent the overall average level of power change amplitude among all adjacent sampling points of the power regulation target curve.

[0073] ⑤ Peak-to-valley ratio This represents the ratio of the difference between the maximum and minimum values ​​to the mean, as shown below: Formula (16) in, Valley power; Average power; This represents the absolute peak-to-trough difference in the target power adjustment curve, indicating the power range that needs to be traversed from the lowest point to the highest point. A larger value indicates a greater adjustment range required by the target curve.

[0074] This represents the magnitude of the peak-to-valley difference in the power regulation target curve relative to the average level, reflecting the relative strength of the power regulation amplitude; dividing the absolute peak-to-valley difference by the average power yields a dimensionless relative index. This facilitates comparisons between wind farms of different power levels.

[0075] ⑥ Zero crossover rate indicates the frequency of sign changes in the first-order difference sequence, reflecting the frequency of reversal of the power regulation target curve.

[0076] All features are calculated on a normalized time axis to eliminate the influence of absolute time length; They collectively reflect the asymmetry of the power regulation target curve, with a rapid rise and a slow fall, or vice versa; The combined effect reflects the degree of fluctuation in the power regulation target curve.

[0077] The fourth layer involves lossy compression and phased approximate parameterization to obtain the inflection point parameter set.

[0078] Objective: To achieve, at the cost of controllable loss of accuracy, Each sampling point is compressed into a compact representation of a small number of line segment parameters, providing a set of inflection point parameters that can be directly output for lightweight communication link transmission in step S4.

[0079] Non-stage approximation: The power regulation target curve is approximated by piecewise linear approximation. A compact parameterized representation.

[0080] The preset number of segments is ( ), time interval Divided into The time intervals are divided into several sub-intervals, as shown below: Formula (17) in, For segmentation nodes, the boundary time between each sub-interval satisfies ; This is the first sub-interval; For the first Sub-intervals.

[0081] By using linear functions, the original sampling points of the fitted power adjustment target curve are obtained in each sub-interval, and the connection points of adjacent line segments are extracted as key inflection points to achieve data compression.

[0082] The linear functions within the subinterval are shown below: Formula (18) in, The power value is the result of fitting a linear function. , The first The slope and intercept of the segment.

[0083] The slope of each segment of the linear function is determined using the least squares method. and intercept That is, for the first The solution for the segment is as follows: Formula (19) in, They are respectively the first The optimal slope and optimal intercept of the segment.

[0084] Inflection point extraction: The connection points of adjacent line segments are the key inflection points. The set of inflection point parameters is shown below: Formula (20) in, For the first The time coordinates of each inflection point; For the first The power coordinates of each inflection point at time [time] The power value is calculated from a linear function at that location; These are the coordinate pairs of the inflection point.

[0085] In this way, the original The power regulation target curve for each sampling point is compressed into a curve containing... The coordinates of each inflection point and A compact representation of the group slope and intercept parameters.

[0086] Compression efficiency: Raw data volume of the power regulation target curve: Power values ​​at each sampling point (each value requires a 32-bit floating-point number); Compressed data size: Coordinates of each inflection point (each containing both time and power values) + Group slope, intercept (two values ​​per group), total One parameter; Compression ratio is approximately ;when At that time, the compression ratio is approximately .

[0087] The set of inflection point parameters is as follows: Formula (21) Step S1 preprocesses the power regulation target curve issued by the power grid dispatch center, extracting features and performing four-layer processing on the high-dimensional original power regulation target curve: signal decomposition, complexity quantification, ensemble feature extraction, and piecewise linear compression. This transforms the curve into a curve containing trend components, periodic components, random components, and shape complexity indices. Shape descriptor vector Structured low-dimensional features, including piecewise linear inflection point parameters, provide a unified data entry point and feature support for subsequent shape matching evaluation, frequency domain coding, dynamic prediction, and lightweight transmission.

[0088] Step S2 includes steps S21-S22.

[0089] Step S21: Perform Fourier series expansion based on the periodic components to obtain the Fourier encoded vector corresponding to the periodic components of the power adjustment target curve.

[0090] Perform a Fourier series expansion on the periodic components of the power adjustment target curve to obtain the Fourier encoded vectors corresponding to the periodic components of the power adjustment target curve, including: The periodic components of the power adjustment target curve are standardized with zero mean and unit variance to obtain the standardized periodic components of the power adjustment target curve. Based on the energy conservation criterion, select the option that satisfies the preceding conditions. The proportion of the cumulative energy of each frequency component to the total energy is not less than the minimum value of the preset energy retention rate threshold, thus obtaining the truncation order of the Fourier series. ; Cutoff order based on Fourier series The standardized periodic components of the power regulation target curve are expanded as follows: The Fourier series of each frequency component is used to obtain the Fourier coefficients of the power adjustment target curve; the Fourier coefficients of the power adjustment target curve include the DC component coefficients, the cosine coefficients of each harmonic, and the sine coefficients. The Fourier encoding vector of the power adjustment target curve is constructed based on the Fourier coefficients of the power adjustment target curve; wherein the Fourier encoding vector of the power adjustment target curve includes the DC component coefficients and the cosine and sine coefficients of each harmonic.

[0091] This step is a frequency domain compression layer.

[0092] To further reduce the representation dimension of the power adjustment target curve, this invention employs Fourier series expansion technology to perform frequency domain encoding on the shape of the power adjustment target curve.

[0093] In step S12, the power adjustment target curve has been decomposed into a trend component, a periodic component, and a random component; the trend component reflects the macro trend, the periodic component reflects regular fluctuations, and the random component is the unpredictable residual.

[0094] Fourier series are best suited for processing signal components with periodicity and regularity. Therefore, this invention only performs Fourier encoding on the periodic components, rather than encoding the entire original power adjustment target curve. The advantages of this approach are: ① Higher signal-to-noise ratio: After removing trend and random components, the frequency domain characteristics of the periodic components are purer and the energy of the Fourier coefficients is more concentrated. ② Higher compression efficiency: With increased energy concentration, the same energy retention rate can be achieved with fewer coefficients; ③ The physical meaning is clearer: the encoding result directly corresponds to the periodic adjustment needs, rather than a mixed signal that is mixed with trends and noise.

[0095] The high-dimensional sampling sequence in the time domain is transformed into a low-dimensional feature space in the frequency domain, generating a compact Fourier encoded vector. .

[0096] The reason for choosing Fourier series is: ① Signal matching: The target curve for wind power regulation usually has obvious periodic and quasi-periodic characteristics, and Fourier series can naturally capture these frequency domain characteristics. ② Orthogonality and independence: Fourier series are orthogonal, and the frequency components are independent of each other, which makes it easy to sort them by importance and truncate low-energy components, so as to achieve lossy compression with controllable accuracy. ③ Low computational complexity: The calculation process for Fourier coefficients is mature and has low computational complexity, making it suitable for real-time operation in environments with limited computing resources on the edge.

[0097] Core task: While preserving the main characteristics of the power regulation objective function (energy retention rate ≥ 95%), expand the data dimensions from... Each sampling point is compressed to The Fourier coefficients provide frequency domain feature inputs for subsequent shape matching evaluation and lightweight communication transmission.

[0098] The first step is to prepare the input by obtaining the standardized periodic components.

[0099] Standardization process, as follows: Before performing the Fourier series expansion, the periodic components extracted in step S12 are first processed. Standardization is performed. The standardization method is zero-mean unit variance normalization, which yields the standardized periodic components, as shown below: Formula (22) in, For standardized periodic components, dimensionless, with a mean of 0 and a variance of 1; The mean of the periodic component sequence represents the center level of the periodic component. The standard deviation of the periodic component represents the degree of dispersion of the periodic component.

[0100] The purpose of standardization: ① Eliminate the influence of dimensions: The regulation tasks of different power levels have different amplitude levels (such as 100MW level vs 1000MW level). After standardization, they are unified into dimensionless signals, so that the Fourier coefficients only reflect the shape characteristics of the curve rather than the magnitude of the amplitude. ② Unified threshold standard: After standardization, the threshold in the energy preservation criterion (formula (25)) is... It has the same physical meaning for all regulation tasks; ③ Eliminate amplitude differences and maintain numerical stability: Avoid numerical overflow or precision loss in Fourier coefficient calculation due to excessive amplitude, so that the coding result only reflects the shape characteristics of the power adjustment target curve.

[0101] The second step is to expand the standardized periodic components into Fourier series to obtain the Fourier coefficients.

[0102] Standardized periodic components In the interval Expanding this into a Fourier series, as shown below: Formula (23) in, These are the DC component coefficients of the Fourier series. After standardization, the average value of the signal is 0. Theoretically, ; Reflects the average level of the standardized target curve; This represents the index of the frequency component. ; is the truncation order of the Fourier series, indicating the number of principal frequency components retained. It is a positive integer, usually between 5 and 15; For the first Cosine coefficients of each frequency component; For the first The sine coefficients of each frequency component; Let be the angular frequency, the th The angular frequency of a first harmonic, expressed in rad / s.

[0103] The fundamental component has a period of . (Overall fluctuations in the signal); Second harmonic, period is ; : The second harmonic, with a period of .

[0104] For wind power regulation targets, low-order harmonics ( Normally, this corresponds to slow adjustment needs at the minute level, while higher-order harmonics correspond to rapid fluctuations at the second level.

[0105] The third step is to calculate the Fourier coefficients.

[0106] The Fourier coefficients are calculated as follows: Formula (24) Considering the time characteristics of offshore wind power regulation, the cutoff order in formula (23) The specific value is determined by the energy conservation criterion, that is, selecting the minimum value that satisfies the following conditions. value: Formula (25) in, The energy retention rate typically ranges from 0.90 to 0.99; for example, a preset threshold is used. The value is 0.95, meaning that the energy of the first K frequency components is retained, which accounts for more than 95% of the total energy.

[0107] In practical applications, adjustments can be made based on the required adjustment accuracy and communication bandwidth limitations. Under these conditions, for a typical wind power regulation target curve, The value is usually between 5 and 15. The energy of the nth harmonic; For the front Accumulated energy at each stage; The total energy is the sum of the energies of all frequency components.

[0108] Selection logic: ① Starting from n=1, accumulate the energy of each frequency component; ② When the proportion of accumulated energy to total energy first reaches or exceeds the threshold. Stop accumulating.

[0109] The value of n at this point is the truncation order. .

[0110] By arranging all Fourier coefficients (DC component + sine and cosine coefficients of each harmonic) in an ordered manner, the original high-dimensional power regulation target curve is obtained using this encoding method, including... A vector of Fourier coefficients is shown below: Formula (26) in, For Fourier encoded vectors, it is a compact representation of the shape of the power adjustment target curve in the frequency domain; This is the DC component; It is the fundamental cosine function; It is the fundamental sine function; It is the second harmonic cosine function; It is the second harmonic sine function; For the first Cosine coefficients of each frequency component; For the first The sine coefficients of each frequency component; It is the set of real numbers; Represents the Fourier encoded vector It is A real vector of dimension 1, meaning that all elements of the vector (each Fourier coefficient) are real numbers.

[0111] Fourier encoded vector The energy distribution similarity will be used as part of the shape matching evaluation in step S22. The calculation input is also one of the core data transmitted in the lightweight communication protocol in step S4.

[0112] Compared to the original Each sampling point, Fourier encoded vector Significantly reduces data transmission volume.

[0113] ① Original periodic component data volume: One sampling point, typical value This corresponds to a 10-minute instruction and a 1-second sampling interval. ② Data volume after encoding: One coefficient, typical value , corresponding to 21 coefficients; ③ Compression ratio: .

[0114] Step S21 converts the periodic components from a high-dimensional sampling sequence in the time domain into a low-dimensional Fourier encoded vector in the frequency domain. While retaining over 95% of the energy information of the curve shape, the data dimension was expanded from... Each sampling point is compressed to With a Fourier coefficient, the compression ratio can typically reach over 10:1.

[0115] Step S22, based on the trend component of the power adjustment target curve Shape description Predicate vector and Fourier code vector And the actual power output curves of offshore wind farms collected in real time. The overall shape matching degree is obtained. .

[0116] The obtained comprehensive shape matching degree includes: The trend components of the actual power output curve of the offshore wind farm and the trend components of the power regulation target curve are time-aligned and amplitude-normalized. The trend component of the actual power output curve of the offshore wind farm is extracted by using a moving average filter to obtain the trend component of the actual power output curve of the offshore wind farm; the correlation coefficient between the first derivative sequence of the trend component of the actual power output curve of the offshore wind farm and the trend component of the power adjustment target curve is calculated as the trend similarity. Extract the peak position ratio, rising slope, falling slope, average rate of change, peak-to-valley ratio, and zero crossover rate of the actual power output curve of the offshore wind farm to form a vector, and obtain the shape descriptor vector of the actual power output curve of the offshore wind farm; calculate the shape correlation based on the dynamic time bending distance between the shape descriptor vectors of the actual power output curve of the offshore wind farm and the power regulation target curve. Fourier series expansion is performed on the periodic components of the actual power output curve of the offshore wind farm to obtain the Fourier encoded vector of the actual power output curve of the offshore wind farm corresponding to the periodic components; the energy distribution similarity of the Fourier encoded vector of the actual power output curve of the offshore wind farm and the Fourier encoded vector of the power adjustment target curve are compared in each frequency band. The trend similarity, shape correlation, and energy distribution similarity are weighted and summed to obtain the comprehensive shape matching degree between the power regulation target curve and the actual power output curve of the offshore wind farm.

[0117] Step S22 is the evaluation criterion layer in the overall technical solution of this invention. An actual power output curve for an offshore wind farm was established. Power regulation target curve issued by the power grid dispatch center Quantitative assessment of the shape matching degree between them.

[0118] Core task: To construct a comprehensive shape matching index by integrating three dimensions: trend similarity, shape correlation, and energy distribution similarity. This enables a comprehensive and quantitative evaluation of the regulation effect of offshore wind farms.

[0119] To quantitatively assess the shape matching degree between the actual power output curve and the power regulation target curve of an offshore wind farm, this invention innovatively constructs a multi-dimensional shape matching degree assessment model that integrates trend similarity, shape correlation, and energy distribution similarity.

[0120] The actual power output curve of an offshore wind farm is as follows The power adjustment target curve is Power adjustment target curve Issued by the power grid dispatch center; the actual power output curve of the offshore wind farm is as follows: Data is collected in real time through the SCADA (Supervisory Control And Data Acquisition) system of the offshore wind farm control center.

[0121] First, the two curves are aligned in time and their amplitudes are normalized.

[0122] Constructing a comprehensive shape matching degree As shown below: Formula (27) in, Trend similarity measures the degree of consistency in the changing trends of two curves; Shape correlation measures the degree to which the overall shapes of two curves match. Energy distribution similarity measures whether the energy distribution of two curves is consistent across frequency bands. , and Weighting coefficients for trend similarity, shape correlation, and energy distribution similarity, respectively; The trend similarity is obtained by calculating the correlation coefficient of the first derivatives of the two curves. As shown below: Formula (28) in, This is the first derivative (i.e., the rate of change of power) of the actual power output curve of the wind farm. The first derivative of the power regulation target curve; The standard deviation of the first derivative sequence of the actual power output curve; The standard deviation of the first derivative sequence of the power regulation target curve; It represents the covariance between two sequences.

[0123] Trend similarity Essentially, it is the Pearson correlation coefficient between two rate of change sequences.

[0124] ① The actual power output curve of the offshore wind farm is completely consistent with the power regulation target curve in terms of change direction and time point, rising and falling at the same time, and the rate of change is proportional. ② The actual power output curve of an offshore wind farm is not linearly correlated with the direction of change of the power regulation target curve; they are completely out of sync. ③ The actual power output curve of offshore wind farms changes in the opposite direction to the power regulation target curve; the target rises while the actual output falls.

[0125] Shape correlation The calculation is performed using a dynamic time-bending algorithm, as shown below: Formula (29) in, This represents the dynamic time curvature distance between the actual power output curve and the power regulation target curve of an offshore wind farm. The number of sampling points for the two curves (i.e., the common sequence length N after time alignment) is used to normalize the distance. The standard deviation of the merged sequence of the two curves is used to eliminate the influence of power magnitude. The range of values ​​is The closer the value is to 1, the more closely the two curves match in shape.

[0126] ① The actual curve closely matches the target curve under the condition of allowable time offset, and the shapes are almost identical, with only a slight phase difference possible. ② The match is moderate, but there are some morphological differences. ③ Their shapes vary greatly, and some are even completely unrelated.

[0127] Energy distribution similarity The results were obtained by comparing the energy distribution of the two curves in different frequency bands.

[0128] Specifically, let the Fourier encoding vector of the actual power output curve of an offshore wind farm be... The Fourier encoded vector of the power adjustment target curve is The energies of each frequency band are as follows: Formula (30) Energy distribution similarity As shown below: Formula (31) in, The total energy of the actual power output curve of an offshore wind farm. The total energy for the power adjustment target curve. The energy of all frequency components. The value range is [0, 1]. The closer the value is to 1, the more consistent the frequency domain energy distribution of the two curves are.

[0129] ① The energy distribution of the two curves across all frequency bands is almost identical, and their frequency domain characteristics are highly consistent. ② The discrepancy indicates that the actual power output curve of the offshore wind farm has too much or too little power in certain frequency components. ③ The energy distribution is completely different.

[0130] The weighting coefficients for trend similarity, shape correlation, and energy distribution similarity are obtained as follows: Based on the relative importance of power grid dispatch in three dimensions—trend similarity, shape correlation, and energy distribution similarity—a judgment matrix is ​​constructed. Calculate the largest eigenvalue of the judgment matrix and its corresponding eigenvector, normalize the eigenvector, and obtain the weight coefficients of the trend similarity, the shape correlation and the energy distribution similarity; A consistency check is performed on the judgment matrix. If the consistency ratio is less than a preset consistency ratio threshold, the weight coefficients of the trend similarity, the shape correlation, and the energy distribution similarity are determined to be valid; otherwise, the judgment matrix is ​​reconstructed and the weights are recalculated until the consistency check is passed.

[0131] Weighting coefficient Adjustments should be made according to specific application scenarios to meet the requirements. Its value is determined using the Analytic Hierarchy Process (AHP) based on the application scenario.

[0132] Specifically, a judgment matrix is ​​constructed based on the relative importance of the three dimensions of trend tracking, shape matching, and energy allocation by the power grid dispatch center. The weights of each dimension are obtained by calculating the eigenvector corresponding to the largest eigenvalue of the judgment matrix. For example, in a scenario where frequency regulation is the primary objective, the judgment matrix is ​​constructed as shown in Table 1.

[0133] Table 1: Example of the judgment matrix for frequency adjustment target scenarios

[0134] Matrix elements Indicates the first The first criterion is relative to the first The importance of these criteria: ① Trend similarity is slightly more important than shape correlation; ② Trend similarity is slightly more important than energy distribution similarity; ③ Shape correlation is slightly more important than energy distribution similarity.

[0135] (1) Calculate the largest eigenvalue

[0136] Judgment matrix in frequency adjustment scenario (Judgment Matrix) as an example, as shown below: Formula (32) Normalize each column of the judgment matrix as follows: First column: Formula (33) Similarly: Second column: Third column:

[0137] The weight vector is obtained by averaging the rows, as follows: Formula (34) Right now Retaining two digits, the normalized eigenvector is [0.54, 0.30, 0.16]. Rounding down, we get... It focuses on the ability to track and adjust trends.

[0138] Calculate the consistency vector ,as follows: Formula (35) The largest eigenvalue is calculated as follows: Formula (36) Theoretically, when the judgment matrices are completely identical, , where n is the order of the judgment matrix, and the closer it is to 3, the better the consistency.

[0139] Calculate the consistency ratio : The consistency index is calculated as follows: Formula (37) for The table of average random consistency indices proposed by Saaty, random consistency indices ; Formula (38) Calculate the largest eigenvalue Consistency ratio If the consistency test is passed; If the judgment matrix contains a logical contradiction, the judgment matrix is ​​reconstructed until... .

[0140] In scenarios where voltage support is the primary objective, They pay more attention to the consistency of energy distribution.

[0141] Overall shape matching The comprehensive index, with a value range of [0,1], is used to comprehensively evaluate the tracking effect of offshore wind farms on the power regulation target curve from three orthogonal dimensions.

[0142] Step S22 establishes a quantitative evaluation mechanism for the shape matching degree between the actual power output curve and the target power regulation curve of the wind farm. By comprehensively evaluating three dimensions—trend similarity, shape correlation, and energy distribution similarity—the comprehensive shape matching degree is obtained. As a core indicator for measuring the regulation effect of wind farms, this indicator is used for the observation feedback of Kalman filtering in step S4, and as uplink feedback information transmitted to the dispatch center in the communication protocol of step S4, providing a quantitative basis for dispatch decisions. When the value is below a preset threshold (e.g., 0.85), the adaptive optimization of the control strategy is triggered.

[0143] Step S2 involves Fourier series frequency domain encoding of the periodic components of the power regulation target curve, and combining its trend components, shape descriptor vector, and the real-time acquired actual power output curve of the offshore wind farm. Through weighted fusion of trend similarity, shape correlation, and energy distribution similarity, a comprehensive shape matching degree is obtained, providing a quantitative evaluation basis for subsequent dynamic prediction and adaptive transmission.

[0144] Step S3 includes steps S31-S32.

[0145] Step S31: Construct the instantaneous regulation capability matrix of offshore wind farms and calculate the regulation capability confidence level of offshore wind farms.

[0146] Construct an instantaneous regulation capability matrix for offshore wind farms and calculate the regulation capability confidence level; based on the instantaneous regulation capability matrix and regulation capability confidence level of offshore wind farms, obtain the dimension-reduced encoding vector of the regulation capability of offshore wind farms. This includes constructing the instantaneous regulation capability matrix of offshore wind farms and calculating the regulation capability confidence level, including: Based on the rated power of the offshore wind farm, the actual output power at the current moment, and the measured wind speed at the current moment, the rapid upward adjustment capability and rapid downward adjustment capability of the offshore wind farm under the first adjustment time scale, and the slow upward adjustment capability and slow downward adjustment capability under the second adjustment time scale are calculated respectively; wherein, the first adjustment time scale is shorter than the second adjustment time scale. Based on the rapid upward and downward adjustment capabilities of the offshore wind farm under the first adjustment time scale, and the slow upward and slow downward adjustment capabilities under the second adjustment time scale, an instantaneous adjustment capability matrix of the offshore wind farm is constructed. The confidence level of the regulation capability of offshore wind farms is calculated based on the standard deviation of wind speed forecasts, which reflects the reliability of the regulation capability of offshore wind farms under the current uncertainty of wind speed forecasts.

[0147] Modeling the regulation capability of offshore wind farms is a core component of lightweight interactive design. This invention constructs a mathematical model of the multidimensional regulation capability of offshore wind farms, considering the regulation characteristics of wind farms at different time scales.

[0148] Establish a quantitative model for the upregulation and downregulation capabilities on two time scales: a fast (second-level) first adjustment time scale and a slow (minute-level) second adjustment time scale. A confidence level assessment is introduced to quantify the impact of wind speed forecast uncertainty on capacity estimation. Why is it necessary to establish a multi-dimensional regulation capability model for offshore wind farms? Traditional methods for characterizing the regulation capability of wind farms are extremely crude—usually only providing a general "adjustable capacity" or "ramp rate," which has the following fundamental problems: ① Time scale aliasing: Failure to distinguish between pitch / converter (second-level response) and unit switching (minute-level response); ② Lack of directionality: The upward and downward adjustment capabilities are subject to different constraints (the upward adjustment is limited by the rated power margin, while the downward adjustment is limited by the minimum output limit), and combining them into a single value cannot reflect the true flexibility. ③ Uncertainty ignored: The impact of wind speed prediction error on capacity estimation is not quantified, which may lead the dispatch center to formulate infeasible control strategies based on overly optimistic capacity assessments.

[0149] Constructing an instantaneous regulation capability matrix for offshore wind farms As shown below: Formula (39) in, and These represent the rapid upregulation and downregulation capabilities at the first adjustment time scale, respectively. and This indicates the ability to slowly increase and decrease the rate of adjustment at the second adjustment time scale.

[0150] Each regulation capacity component is calculated as follows: Formula (40) in, Rated power for offshore wind farms This represents the actual output power of the offshore wind farm at the current moment. This is for power margin; This is the wind speed-to-power conversion coefficient, reflecting the ability of wind speed changes to be converted into power regulation. Its value is determined based on the power curve characteristics of the wind turbine. For example, it is taken as... ; This represents the measured wind speed at the current moment. The first adjustment time scale is the time constant for rapid adjustment (corresponding to the response time scale of wind turbine pitch control and converter power regulation). For example, the first adjustment time scale is set to a value of... ; This represents the rapid response capability value. This is the minimum allowable output power for an offshore wind farm (usually 10%-15% of the rated power). The power reduction rate coefficient reflects the wind farm's ability to actively reduce power output. For example, it is taken as... ; This is the slow power increase rate coefficient, reflecting the slow power increase capability achieved by the wind farm through methods such as gradually increasing the number of wind turbines put into operation. An example is taken as... ; The second adjustment time scale is the time constant for slow adjustment. For example, the second adjustment time scale is set to a value of [value missing]. ; For example, the slow down-adjustment rate coefficient is taken as... Slow-speed adjustment is mainly achieved by adjusting the number of wind turbines in operation and by making large-scale adjustments to the pitch angle. The time scale is generally on the order of minutes.

[0151] The first adjustment time scale, for rapid up-adjustment capability, is subject to two constraints, and the smaller of the two values ​​is taken: ① Static constraints (power margin): Even with favorable wind conditions, the output of a wind farm cannot exceed its rated power. Therefore, the upper limit of the adjustable amount is the difference between the current power and the rated power. ; ② Dynamic Constraints (Wind Speed ​​Response Capability): Within a 10-second fast response window, the maximum power increase of a wind farm is limited by the current wind speed conditions and the response capability of the wind turbine's pitch / converter. This dynamic constraint is characterized.

[0152] The deeper meaning of the coefficient: It is not a simple proportional constant; it implicitly incorporates the following factors: ① The sensitivity of the wind turbine power coefficient to the blade pitch angle; ② Upper limit of converter power modulation speed; ③ The inertial energy release rate of the unit: some units can achieve short-term power boost through rotor kinetic energy release.

[0153] For example, Meaning: Within 10 seconds, each unit of wind speed can be converted into approximately The power increment. For example, when When, the dynamic limit is .

[0154] This means that within the first adjustment timescale, the wind farm can decrease at a rate of 15% / s of its current output (i.e., a reduction of 15% of its current power per second), multiplied by the time constant. The maximum reduction is 150% of the current output, but this is subject to static constraints. Due to limitations, the actual reduction cannot exceed the difference between the current output and the minimum output.

[0155] The confidence assessment of the regulation capacity of offshore wind farms, considering the impact of wind speed forecast uncertainty on regulation capacity, is shown below: Formula (41) in, The confidence level of the adjustment capability is a scalar with a value range of (0, 1], reflecting the reliability of the adjustment capability assessment results under the current uncertainty of wind speed prediction. Let be the standard deviation of the wind speed forecast. When the wind speed is relatively low (i.e., the wind speed prediction is relatively accurate), Approaching 1; when When it is large, A value close to 0 indicates low reliability of the assessment results of the regulatory capacity. The reference standard deviation is used for normalization, for example, .

[0156] The purpose of step S31 is to establish a mathematical quantitative model of the wind farm’s upward and downward adjustment capabilities under the first and second adjustment time scales, and to introduce a confidence assessment to reflect the impact of wind speed prediction uncertainty.

[0157] The instantaneous regulation capability matrix of offshore wind farms obtained in step S31 and the confidence level of adjustment ability The input data used as the principal component analysis dimensionality reduction encoding in step S32 is encoded together with the confidence level and the instantaneous adjustment capability matrix, so that the principal components after dimensionality reduction retain the "reliability" information. At the same time, the confidence information is used to adjust the confidence weight of the prediction in step S4 Kalman filtering. When the confidence level is high, the observation noise is small (more trust in the actual measurement), and when the confidence level is low, the observation noise is large (more trust in the model prediction).

[0158] Step S32: Based on the instantaneous regulation capability matrix and regulation capability confidence of the offshore wind farm, obtain the dimensionality reduction encoding vector of the regulation capability of the offshore wind farm.

[0159] Based on the instantaneous regulation capability matrix and regulation capability confidence of the offshore wind farm, a dimension-reduced encoding vector of the regulation capability of the offshore wind farm is obtained, including: Based on the rapid upward and downward adjustment capabilities of the offshore wind farm under the first adjustment time scale, the slow upward and slow downward adjustment capabilities of the offshore wind farm under the second adjustment time scale, the adjustment capability confidence of the offshore wind farm, the measured wind speed at the current moment, and the adjustable power range of the offshore wind farm at the current moment, the adjustment capability feature vector of the offshore wind farm is obtained; wherein, the adjustable power range of the offshore wind farm at the current moment is the sum of the rapid upward and rapid downward adjustment capabilities. Based on the historical data of the regulation capacity of offshore wind farms within a preset period, a historical operation data feature matrix is ​​constructed, and the covariance matrix corresponding to the historical data feature matrix is ​​calculated. Using the principle of spectral decomposition, the covariance matrix is ​​decomposed into eigenvalues ​​to obtain the corresponding orthogonal eigenvector matrix and eigenvalue matrix; Before choosing The orthogonal eigenvectors corresponding to the largest eigenvalues ​​form the projection matrix; The regulation capability feature vector of the offshore wind farm is projected onto the projection matrix to obtain the regulation capability dimensionality reduction encoding vector of the offshore wind farm.

[0160] Dimensionality reduction encoding of moderating ability features based on principal component analysis: To achieve lightweight transmission of regulation capability information, this invention employs Principal Component Analysis (PCA) to reduce the multidimensional regulation capability characteristics of wind farms to a lower-dimensional principal component space for dimensionality reduction encoding, generating compact variable vectors. .

[0161] ① Construct a 7-dimensional regulation capability feature vector that includes regulation capability, confidence level, wind speed, and adjustable power range. ; ② Calculate the covariance matrix based on historical operating data and perform eigenvalue decomposition; ③ Based on the criterion of cumulative contribution rate ≥ 95%, the top performers are selected. The principal components constitute the projection matrix. ; ④ Project the original 7-dimensional features onto By using the principal component space, we obtain the dimensionality reduction encoding. .

[0162] Constructing the regulation capability feature vector As shown below: Formula (42) in, The current adjustable power range of the wind farm is defined as follows: This refers to the sum of the ability to quickly increase and decrease power, reflecting the range of power that a wind farm can adjust in a short period of time.

[0163] A feature matrix is ​​constructed by collecting historical operational data. As shown below: Formula (43) in, The feature matrix is ​​composed of historical operating data, with each column representing a feature vector of the adjustment capability at a given moment. This represents the total number of sampling times for historical data. Calculate the covariance matrix : Formula (44) in, Characteristic matrix The covariance matrix reflects the correlation between features in each dimension.

[0164] Eigenvalue decomposition of the covariance matrix is ​​performed because... Since C is a real symmetric positive semi-definite matrix, according to the spectral decomposition theorem, C can be uniquely decomposed into: Formula (45) Wherein, the orthogonal eigenvector matrix , For the first eigenvectors; eigenvalue matrix , The eigenvalues ​​are arranged in descending order; For feature vectors The dimension (d=7 in this invention, corresponding to the 7 components in formula (42)).

[0165] Before choosing The eigenvectors corresponding to the largest eigenvalues ​​constitute the projection matrix. As shown below: Formula (46) in, The selection is based on the cumulative contribution rate criterion, i.e., the previous... The cumulative variance contribution rate of each principal component is not less than 95%.

[0166] Formula (47) Adjustment capability dimensionality reduction encoding As shown below: Formula (48) This encoding method compresses the original high-dimensional adjustment capability vector into a low-dimensional principal component vector, significantly reducing the amount of data transmitted while preserving the main information.

[0167] Principal component analysis (PCA) was used to determine the 7-dimensional moderating ability eigenvectors. Based on the criterion of cumulative contribution rate ≥95%, dimensionality reduction and compression are performed to a 3-4 dimensional principal component encoding vector, i.e., dimensionality reduction encoding of adjustment capability. This significantly reduces data dimensionality while preserving key information, providing a compact capability representation for subsequent Kalman filter prediction and lightweight communication transmission.

[0168] Step S33: Based on the dimensionality reduction encoding vector of the regulation capability of the offshore wind farm and the comprehensive shape matching degree, the optimal state estimate of the regulation capability state of the offshore wind farm is obtained.

[0169] Based on the dimensionality reduction encoding vector of the offshore wind farm's regulation capability and the comprehensive shape matching degree, the optimal state estimate of the offshore wind farm's regulation capability state is obtained, including: The regulation capability state space model is established by using the dimensionality reduction encoding vector of the offshore wind farm as the state vector. The regulation capability state space model includes a state equation and an observation equation. The state equation describes the time evolution of the regulation capability dimensionality reduction encoding vector, and the observation equation describes the mapping relationship from the state vector to the observation vector. The comprehensive shape matching degree is used as observation feedback and is a component of the observation vector. Based on the optimal state estimate and control input at the current moment, perform Kalman filtering prediction to predict the state vector and prediction error covariance matrix at the next moment; Obtain the observation vector at the next time step, and calculate the Kalman gain based on the prediction error covariance matrix and the observation noise covariance matrix; Based on the Kalman gain fusion prediction and observation values, the optimal state estimate for the next time step is obtained, and the error covariance matrix is ​​updated.

[0170] A dynamic adjustment capability prediction and update mechanism is established in this invention, taking into account the time-varying nature of wind farm adjustment capability.

[0171] ① Establish a linear dynamic system model of the state evolution of regulatory capacity; ② Proactively estimate the regulatory capacity at the next moment through the predictive step; ③ Correct prediction bias by integrating real-time observation data through update steps; ④ Provide optimal estimates and uncertainty measures for lightweight communication transmission and adaptive compression in step S4.

[0172] The evolution of regulatory capacity is modeled as a linear dynamic system, with state equations (prediction model) and observation equations (measurement model), as shown below: Formula (49) in, Let t+1 be the state vector of the adjustment capability at the next time step, that is, the state value of the dimension-reduced adjustment capability in the principal component space after one time step. This is the state vector (adjustment capability after dimensionality reduction). To control inputs (such as changes in wind speed). This is the observation vector. and These represent process noise and observation noise, respectively. State transition matrix. The observation matrix was obtained through a systematic identification method. Determined based on sensor configuration; To control the input matrix.

[0173] The prediction steps of Kalman filtering are as follows: Formula (50) in, Based on the current moment Information for the next moment A one-step prediction of the state of regulatory capacity; This is the optimal state estimate at time t (i.e., the adjustment capability estimate after incorporating current observation information). This is the error covariance matrix of the current state estimate, reflecting the degree of uncertainty in the current estimate; This is the error covariance matrix for one-step prediction, reflecting the degree of uncertainty in the predicted value. For process noise The covariance matrix reflects the uncertainty of the state evolution model of the regulatory capacity. State transition matrix. It is obtained through a system identification method, specifically by utilizing the state vector sequence of adjacent time points in historical operational data. The linear dynamic relationship was fitted using the least squares method, and the solution is as follows: Formula (51) Control input matrix Reflecting external inputs (such as changes in wind speed) The impact of changes in scheduling instructions, etc., on the state of regulation capacity, and Synchronization is obtained through system identification. Observation matrix. Determined based on sensor configuration, this represents the mapping relationship from the state vector to observables. In this invention, if the dimensionality-reduced state vector... If it can be directly observed, then Take the identity matrix.

[0174] The update steps are as follows: Formula (52) in, To observe noise The covariance matrix reflects the uncertainty of sensor measurements; Kalman gain; For a moment The Kalman gain matrix determines the optimal weighting ratio between the predicted and observed values; To integrate The optimal state estimate after receiving time-series observation information. When the Kalman gain... When the value is larger, it indicates that the system trusts the observations more; when When the value is smaller, it indicates that the system trusts the model's predictions more. It is an identity matrix. This is the updated error covariance matrix after fusing observation information; its diagonal elements reflect the residual uncertainty of the estimation of each state component. This matrix is ​​used to determine the uncertainty of the adaptive compression strategy in step S4.

[0175] This step aims to establish a dynamic prediction and real-time update mechanism for wind farm regulation capacity using Kalman filtering, addressing the time-varying nature of regulation capacity. By using state prediction, it enables a forward-looking estimate of the regulation capacity at the next moment, allowing for adjustments to the control strategy in advance without waiting for actual measurement feedback, significantly shortening the system response time. Simultaneously, by continuously correcting the predicted values ​​through the fusion of actual observation data, the accuracy of the regulation capacity assessment is ensured.

[0176] The output of this step is the optimal state estimate. This will serve as the core adjustment capability data for uplink transmission in the lightweight communication protocol of step S4, while also predicting the error covariance. Used for uncertainty judgment of the adaptive compression strategy in step S4.

[0177] Step S4, specifically.

[0178] Based on the dynamically switched coding scheme and dynamically adjusted redundancy, and using the inflection point parameter set, Fourier coding vector, optimal state estimation, and comprehensive shape matching degree of the actual power output curve of the offshore wind farm, hierarchical lightweight data processing is performed to generate a dynamically optimized data package for the offshore wind farm, including: The layering includes the application layer, coding layer, transport layer, and physical layer; At the application layer, the raw time-series data to be transmitted is classified into real-time control data, status monitoring data, and historical statistical data. At the coding layer, based on the dynamically switched coding scheme and the dynamically adjusted redundancy, the inflection point parameter set, Fourier coding vector, and corresponding optimal state estimation and comprehensive shape matching degree of the actual power output curve of the offshore wind farm are coded, compressed and forward-corrected. At the transport layer, an adaptive packet splitting mechanism is adopted to dynamically adjust the packet size based on the size of the original time-series data packets to be transmitted, the available bandwidth of the current channel, the inherent network latency, and the inherent protocol overhead of the data packets. At the physical layer, channel quality is monitored, providing parameters for the calculation of the channel quality metrics; Based on the results of the encoding compression and forward error correction coding, and the dynamically adjusted data packet size, the dynamically optimized data packet for the offshore wind farm is obtained.

[0179] like Figure 2As shown, the power grid dispatch center serves as the cloud, the offshore wind farm control center serves as the edge, and the offshore wind turbine cluster serves as the terminal.

[0180] This invention presents a layered, lightweight data transmission protocol suitable for cloud-edge collaborative environments. The protocol architecture consists of four layers: application layer, encoding layer, transport layer, and physical layer.

[0181] At the application layer, the transmitted data is divided into three categories based on the importance and timeliness requirements of the data: real-time control data, status monitoring data, and historical statistical data.

[0182] The coding layer employs a compression strategy combining differential coding and entropy coding for time series data. First, calculate the difference sequence. Then, Huffman coding is used to compress the difference values.

[0183] Compression ratio Defined as: Formula (53) in, and These represent the lengths of the original data and the compressed data, respectively.

[0184] The transport layer employs an adaptive packet splitting mechanism, dynamically adjusting the size of data packets based on network conditions.

[0185] Let the network latency be... The channel bandwidth is Data packet size The optimization objective is: Formula (54) in, The size of the data packet to be optimized (in bits); This refers to the packet loss rate. The available bandwidth of the current channel (unit: bit / s); The inherent latency of the network (unit: seconds). For normal transmission time, This represents the expected additional delay due to packet loss and retransmission. The constraints are: ,in For the minimum overhead of the protocol header and checksum, an example is taken. ; For the channel's maximum transmission unit (MTU) limitation, exemplarily take . For each data packet, a fixed protocol overhead (header, checksum, etc.) is taken as an example. At this point, there exists a non-trivial optimal solution. Value. By estimating network parameters online, the packet size is dynamically optimized to minimize the total transmission time.

[0186] The purpose of this step is to design a layered protocol architecture suitable for complex communication environments in the open sea. Through the coordinated work of four layers—data classification at the application layer, differential and Huffman compression at the coding layer, adaptive packetization at the transport layer, and channel monitoring at the physical layer—efficient transmission of communication data can be achieved.

[0187] The inputs for this step are the inflection point parameter set from step S12 and the Fourier encoded vector from step S2. And the adjustment capability dimensionality reduction encoding vector output from step S4 After being encoded, compressed, and packetized in this step, this compact data enables efficient information exchange between the dispatch center and the wind farm control center.

[0188] Adaptive compression and redundancy control strategy: A dynamic switching coding scheme based on the rate of change of the actual power output curve of offshore wind farms, dynamically adjusting the redundancy of forward error correction coding based on channel quality indicators, includes: Calculate the rate of change of the actual power output curve of the offshore wind farm to be developed; When the rate of change is less than a preset rate of change threshold, a high compression ratio coding scheme of differential coding plus Huffman compression is adopted; otherwise, a low compression ratio and high reliability coding scheme of direct coding plus forward error correction is adopted. The channel quality index is calculated based on the ratio of effective bandwidth to nominal bandwidth and the bit error rate. When the channel quality index is greater than the first preset threshold, the first redundancy is adopted; When the channel quality index is less than or equal to the first preset threshold and greater than the second preset threshold, the second redundancy is adopted. When the channel quality index is less than or equal to the second preset threshold, a third redundancy is used; Wherein, the first preset threshold is greater than the second preset threshold; the first redundancy is less than the second redundancy, and the second redundancy is less than the third redundancy.

[0189] To further improve communication efficiency, this invention proposes an adaptive compression and redundancy control strategy. The compression algorithm and redundancy are dynamically adjusted based on the data's variation characteristics and network conditions. The rate of change of the original time series data is defined. As shown below: Formula (55) in, The first time series data to be transmitted Each sampled value (which can be the adjustment capability dimensionality reduction encoding vector in step S2 or the time-series update value of the Fourier coefficients in step S2); The length of the sampling sequence; To prevent small constants from being divided by zero, for example, take... .

[0190] The range of values ​​is , The larger the value, the more drastic the data change.

[0191] when When it is smaller ( The data changes gradually, so a high compression ratio encoding scheme (such as differential coding + Huffman compression) is used. when When it is larger ( Due to the drastic changes in data, a coding scheme with a low compression ratio but higher transmission reliability (such as direct coding + forward error correction) is adopted.

[0192] The threshold of 0.1 can be adjusted according to the actual application scenario.

[0193] Redundancy control strategies are based on adaptive adjustment of channel quality, and channel quality indicators... As shown below: Formula (56) in, The value range of is (0, 1]. The closer to 1, the better the channel quality; For bit error rate, and These are the effective bandwidth and the nominal bandwidth, respectively.

[0194] according to The redundancy of forward error correction coding is dynamically adjusted. To maximize transmission efficiency while ensuring data transmission reliability, among which Redundancy is defined as the proportion of error-correcting redundant data to the total amount of transmitted data. When the channel quality is good, redundancy is reduced to increase the transmission rate; when the channel quality is poor, redundancy is increased to ensure transmission reliability.

[0195] For example, the first preset threshold is 0.8, and the first redundancy is 0.05. hour, ; The second preset threshold is 0.5, and the second redundancy is 0.15; when hour, ; The third redundancy is 0.30; when hour, .

[0196] Step S4 involves classifying the data to be transmitted into real-time control data, status monitoring data, and historical statistical data at the application layer; encoding and compressing inflection point parameters, Fourier coding vectors, optimal state estimates, and comprehensive shape matching degrees at the coding layer based on the dynamically switched coding scheme and dynamically adjusted redundancy; dynamically adjusting the data packet size at the transport layer using an adaptive packet splitting mechanism; and monitoring channel quality at the physical layer to provide parameters for calculating channel quality indicators and generating dynamically optimized data packets, thereby achieving efficient and reliable two-way information exchange between offshore wind farms and the power grid.

[0197] Example 2: A specific embodiment of the present invention discloses a lightweight interaction system between an offshore wind farm and the power grid, thereby realizing the lightweight interaction method between an offshore wind farm and the power grid described in Embodiment 1. The specific implementation of each module is as described in the corresponding section of Embodiment 1.

[0198] like Figure 3 As shown, the system includes a multi-dimensional feature extraction module M1, a shape matching degree evaluation module M2, an adjustment capability prediction module M3, and an adaptive hierarchical lightweight communication module M4. The multi-dimensional feature extraction module M1 is used to preprocess the power regulation target curve sent from the power grid dispatch center to the offshore wind farm control center to obtain the trend component, periodic component, random component, shape complexity index, shape descriptor vector and inflection point parameter set corresponding to the power regulation target curve. The shape matching degree evaluation module M2 is used to obtain the Fourier encoding vector corresponding to the periodic component of the power adjustment target curve; based on the trend component, shape descriptor vector and Fourier encoding vector of the power adjustment target curve and the actual power output curve of the offshore wind farm collected in real time, the comprehensive shape matching degree is obtained. The regulation capacity prediction module M3 is used to obtain the optimal state estimate of the regulation capacity state of the offshore wind farm based on the regulation capacity dimensionality reduction encoding vector of the offshore wind farm and the comprehensive shape matching degree. The adaptive hierarchical lightweight communication module M4 is used to perform hierarchical lightweight data processing based on the inflection point parameter set, Fourier encoding vector, optimal state estimation and comprehensive shape matching degree of the actual power output curve of the offshore wind farm, to generate dynamic optimization data packets for the offshore wind farm.

[0199] Since the system in this embodiment and the method in Embodiment 1 are related and can be referenced from each other, this description is redundant and will not be repeated here. Because this system embodiment shares the same principle as the above method embodiment, it also possesses the corresponding technical effects of the above method embodiment.

[0200] This method is implemented as follows: The basic idea of ​​this invention is to construct a complete lightweight interaction method and system between offshore wind farms and the power grid. This is achieved by deploying a multi-dimensional, multi-scale feature extraction module M1 and a frequency domain coding and shape matching evaluation module M2 in the power grid dispatch center; and by deploying a multi-dimensional regulation capability modeling and dimensionality reduction coding module M3, a dynamic prediction and optimal estimation module M4, and an adaptive hierarchical lightweight communication module M5 in the wind farm control center. A lightweight data transmission protocol is implemented in the communication link to achieve efficient and reliable information interaction between offshore wind farms and the power grid.

[0201] This system primarily addresses key issues in traditional interactive methods, such as large communication data volume, low matching degree of adjustment targets, and difficulty in adapting to wind power fluctuations. Through techniques such as shape feature encoding, adjustment capability dimensionality reduction, and adaptive communication, it reduces the communication burden by more than 70% while ensuring adjustment accuracy, significantly improving the grid connection efficiency and grid friendliness of offshore wind farms.

[0202] The cloud-edge-device communication protocol stack is configured, deploying a layered, lightweight data transmission protocol stack in the power grid dispatch center, offshore wind farm control center, and key wind turbine nodes. This includes a data classification mechanism at the application layer, a differential compression algorithm at the coding layer, an adaptive packet splitting strategy at the transport layer, and channel quality monitoring functions at the physical layer. An adaptive compression and redundancy control module is configured to dynamically adjust the coding scheme and error correction redundancy based on data change rate and channel quality. Real-time channel quality monitoring equipment is deployed in the offshore communication link to provide decision-making support for the adaptive communication strategy.

[0203] System integration and joint commissioning testing integrates the power grid dispatch center and the offshore wind farm control center, establishing an end-to-end lightweight interactive link from the dispatch center to the offshore wind farm. System functional testing verifies the accuracy of power regulation target curve encoding and decoding, the accuracy of regulation capacity prediction, and the reliability of the communication protocol. Stress testing is conducted under different wind conditions and network conditions to ensure stable system operation in harsh marine environments. A system performance monitoring mechanism is established to track key indicators such as communication data volume, regulation accuracy, and response time in real time.

[0204] On-site deployment and optimization: The system is deployed at the target offshore wind farm, and parameters are optimized and adjusted based on actual sea conditions and communication environment. Customized configuration and adjustment capability modeling parameters and communication protocol parameters are developed for the specific wind farm's turbine configuration and cable network topology.

[0205] A trial run of no less than three months will be conducted to collect operational data and evaluate system performance. Based on the evaluation results, the algorithm parameters and control strategies will be fine-tuned to ensure the system meets the design performance targets. The main parameters to be optimized include: Fourier series cutoff order (Adjustment range 5-15), Energy retention threshold (Adjustment range 0.90-0.99), Shape matching degree weighting coefficient Kalman filtering process noise covariance matrix and observation noise covariance matrix Diagonal elements, data change rate in adaptive compression strategy The handover threshold (adjustable range 0.05~0.20) and channel quality indicators Redundancy switching thresholds, etc.

[0206] Operation and maintenance: After the system is put into formal operation, daily operation and maintenance are carried out. Dispatchers set power regulation targets for offshore wind farms through a lightweight interactive interface at the power grid dispatch center. The system automatically extracts and encodes the shape features of the target curve and sends it to the wind farm's central control center via a lightweight communication protocol. The central control center receives the encoded target curve and, combined with real-time regulation capability assessment results, automatically generates power allocation instructions for each wind turbine. The wind turbines execute corresponding power regulation actions according to the received instructions and feed back the execution results to the central control center. Throughout the process, the system continuously monitors the shape matching degree index, and automatically triggers control strategy optimization when the matching degree falls below a preset threshold. Operators need to regularly check key system performance indicators, including communication data volume, regulation accuracy, and response time, to ensure the system operates in an optimal state.

[0207] Data management and maintenance: Establish a comprehensive data management system to classify, store, and back up target curve data, regulation capability assessment data, and communication performance data generated during system operation. Implement a data quality monitoring mechanism to automatically identify and handle abnormal data, ensuring data integrity and accuracy. Regularly upgrade system software and optimize algorithms, continuously improving Fourier coding parameters, principal component analysis models, and Kalman filter parameters based on operational experience and technological advancements. Establish a fault diagnosis and recovery mechanism so that the system can automatically switch to a backup communication link or degraded operation mode when communication is interrupted or equipment fails. Conduct monthly system performance evaluations to analyze trends in communication efficiency, regulation accuracy, and reliability, providing data support for system optimization. Perform comprehensive system maintenance annually, including hardware inspection, software upgrades, and parameter calibration, to ensure long-term stable system operation.

[0208] In summary, the lightweight interaction method and system between offshore wind farms and power grids according to embodiments of the present invention have the following beneficial effects: 1. This invention significantly improves the matching accuracy and response effect of regulation targets. By constructing a lightweight interactive model based on the shape characteristics of the regulation power target curve, this invention fundamentally changes the traditional interactive method that only focuses on power numerical matching. By introducing a shape complexity index, Fourier series encoding, and a multi-dimensional shape matching degree evaluation model, the system can accurately capture the dynamic characteristics and shape features of the target curve during grid regulation, including key information such as trend changes, periodic fluctuations, and transient responses. In particular, through a comprehensive evaluation of trend similarity, shape correlation, and energy distribution similarity, it ensures that the wind farm output not only approaches the target numerically but, more importantly, achieves a high degree of matching in the dynamic characteristics of the regulation process, thereby significantly improving the control effect in key scenarios such as grid frequency regulation and voltage support. Compared to the traditional interactive method based on real-time power numerical matching (i.e., the dispatch center issues specific power value commands, and the wind farm adjusts its output according to the command value), this invention can improve target tracking accuracy by more than 30%, especially performing better in dynamic scenarios such as grid faults or load changes. Here, target tracking accuracy is defined as the comprehensive shape matching degree in step S2. The evaluation metric is that, under the same communication bandwidth conditions, the method of this invention improves performance by more than 30% compared to the traditional method (e.g., (From 0.65 to over 0.90); 2. This invention significantly reduces communication data volume and improves transmission efficiency. The multi-level data compression and coding technology employed in this invention can significantly reduce the communication burden. First, Fourier series coding compresses the high-dimensional power regulation target curve into a compact vector containing the main frequency components, typically compressing the original power target regulation curve data to 20-30% of its original size. Second, principal component analysis-based regulation capability feature dimensionality reduction coding further compresses the multi-dimensional regulation capability information into a low-dimensional regulation capability coding vector, reducing the data dimensionality by 60-70% while maintaining over 95% of the information content. Finally, a layered lightweight data transmission protocol, through differential coding, Huffman compression, and adaptive packet segmentation, obtains dynamically optimized data packets for offshore wind farms, further improving transmission efficiency. Combining the effects of various compression technologies, this invention can reduce the overall communication data volume by more than 70%, significantly improving the real-time performance and reliability of information transmission in complex offshore communication environments, and effectively alleviating bandwidth limitations and transmission delays in offshore environments. 3. This invention fully explores and utilizes the flexible regulation potential of offshore wind farms. The multi-dimensional regulation capacity modeling and quantification method for offshore wind farms established in this invention can comprehensively identify and evaluate the regulation resources of wind farms. By constructing an instantaneous regulation capacity matrix for offshore wind farms that includes rapid regulation capacity, slow regulation capacity, and confidence level assessment, the regulation potential of wind farms at different time scales can be accurately quantified, avoiding the problem of conservative or inaccurate regulation capacity assessment in traditional methods. In particular, by considering multiple factors such as wind speed changes, equipment status, and operational constraints, the system can dynamically optimize the output allocation of each wind turbine, maximizing the utilization rate of the entire wind farm's regulation capacity. Practice shows that this refined regulation capacity modeling method can increase the actual adjustable capacity of wind farms by more than 25%, providing more flexible resources for the power grid and effectively supporting the safe and stable operation of power systems with a high proportion of new energy sources. 4. This invention enhances the rapid adaptation and response capabilities to wind power fluctuations. The dynamic regulation capability prediction and update mechanism designed in this invention significantly improves the system's adaptability to the inherent fluctuations of wind power. By establishing a dynamic state estimation model based on Kalman filtering, it can track the changing trend of wind farm regulation capability in real time and proactively update the regulation capability assessment based on information such as wind speed prediction and equipment status. When wind speed changes drastically, there is no need to wait for feedback from actual power output; instead, the control strategy is adjusted in advance based on the prediction model, thereby shortening the response time by more than 50%. Simultaneously, by introducing a regulation capability confidence assessment mechanism, the impact of wind speed prediction uncertainty on regulation capability can be quantified, allowing for full consideration of uncertainty factors when formulating control strategies and improving the robustness of the control scheme. This proactive adaptive mechanism enables offshore wind farms to better cope with rapid changes in offshore wind conditions, improving overall operational stability. 5. This invention improves communication reliability and cluster coordination capabilities in complex marine environments. Addressing the challenges of complex communication environments faced by offshore wind farms, the adaptive compression and redundancy control strategy designed in this invention significantly enhances the reliability of the communication system. By monitoring channel quality indicators in real time, the system can dynamically adjust the redundancy of forward error correction coding. Redundancy is reduced to increase transmission rate when channel quality is good, and redundancy is increased to ensure transmission reliability when channel quality is poor. Simultaneously, the hierarchical data transmission protocol supports flexible network topology, enabling multiple offshore wind farms to coordinate efficiently with the power grid dispatch center, avoiding the problem of excessive communication burden hindering large-scale cluster coordination in traditional methods. This intelligent adaptive mechanism allows the system to maintain stable communication performance even in complex environments such as harsh sea conditions and electromagnetic interference, improving communication success rate by more than 20%, providing crucial technical support for the coordinated and optimized operation of large-scale offshore wind farm clusters and the construction of new power systems.

[0209] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0210] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A lightweight interaction method between offshore wind farms and power grids, characterized in that, include: The power regulation target curve sent from the power grid dispatch center to the offshore wind farm control center is preprocessed to obtain the trend component, periodic component, random component, shape complexity index, shape descriptor vector and inflection point parameter set corresponding to the power regulation target curve; The Fourier encoded vector corresponding to the periodic component of the power regulation target curve is obtained; based on the trend component, shape descriptor vector, and Fourier encoded vector of the power regulation target curve, as well as the actual power output curve of the offshore wind farm collected in real time, the comprehensive shape matching degree is obtained. Based on the dimensionality reduction encoding vector of the regulation capability of the offshore wind farm and the comprehensive shape matching degree, the optimal state estimate of the regulation capability state of the offshore wind farm is obtained. Based on the set of inflection point parameters and Fourier encoded vectors of the actual power output curve of the offshore wind farm, as well as the optimal state estimation and comprehensive shape matching degree, hierarchical lightweight data processing is performed to generate dynamic optimization data packages for the offshore wind farm.

2. The lightweight interaction method between offshore wind farms and power grids according to claim 1, characterized in that, The obtained set of trend components, periodic components, random components, shape complexity index, shape descriptor vector, and inflection point parameters corresponding to the power adjustment target curve includes: The trend component of the power adjustment target curve is extracted by using a moving average filter to obtain the trend component of the power adjustment target curve; The power adjustment target curve is detrended to obtain the detrended signal; Fourier spectrum analysis is performed on the detrended signal to extract the main periodic components and obtain the periodic components. Subtracting the trend component and the periodic component from the power adjustment target curve yields the random component of the power adjustment target curve. The shape complexity index is obtained by calculating the ratio of the mean of the absolute values ​​of the second-order differences of the power regulation target curve to the power standard deviation. Extract the peak position ratio, rising slope, falling slope, average rate of change, peak-to-valley ratio, and zero crossover rate of the power regulation target curve to form a vector, and obtain the shape descriptor vector of the power regulation target curve; The power regulation target curve is compressed using a piecewise linear approximation. The time interval is divided into a preset number of sub-intervals. The optimal slope and optimal intercept of each linear function are determined using the least squares method. The key inflection points between each sub-interval are extracted to obtain the inflection point parameter set.

3. The lightweight interaction method between offshore wind farms and power grids according to claim 1, characterized in that, Perform a Fourier series expansion on the periodic components of the power adjustment target curve to obtain the Fourier encoded vectors corresponding to the periodic components of the power adjustment target curve, including: The periodic components of the power adjustment target curve are standardized with zero mean and unit variance to obtain the standardized periodic components of the power adjustment target curve. Based on the energy conservation criterion, select the one that satisfies the preceding... The proportion of the cumulative energy of each frequency component to the total energy is not less than the minimum value of the preset energy retention rate threshold, thus obtaining the truncation order of the Fourier series. ; Cutoff order based on Fourier series The standardized periodic components of the power regulation target curve are expanded as follows: The Fourier series of each frequency component is used to obtain the Fourier coefficients of the power adjustment target curve; the Fourier coefficients of the power adjustment target curve include the DC component coefficients, the cosine coefficients of each harmonic, and the sine coefficients. The Fourier encoding vector of the power adjustment target curve is constructed based on the Fourier coefficients of the power adjustment target curve; wherein the Fourier encoding vector of the power adjustment target curve includes the DC component coefficients and the cosine and sine coefficients of each harmonic.

4. The lightweight interaction method between offshore wind farms and power grids according to claim 3, characterized in that, The obtained comprehensive shape matching degree includes: The trend components of the actual power output curve of the offshore wind farm and the trend components of the power regulation target curve are time-aligned and amplitude-normalized. The trend component of the actual power output curve of the offshore wind farm is extracted by using a moving average filter to obtain the trend component of the actual power output curve of the offshore wind farm; the correlation coefficient between the first derivative sequence of the trend component of the actual power output curve of the offshore wind farm and the trend component of the power adjustment target curve is calculated as the trend similarity. Extract the peak position ratio, rising slope, falling slope, average rate of change, peak-to-valley ratio, and zero crossover rate of the actual power output curve of the offshore wind farm to form a vector, and obtain the shape descriptor vector of the actual power output curve of the offshore wind farm; calculate the shape correlation based on the dynamic time bending distance between the shape descriptor vectors of the actual power output curve of the offshore wind farm and the power regulation target curve. Fourier series expansion is performed on the periodic components of the actual power output curve of the offshore wind farm to obtain the Fourier encoded vector of the actual power output curve of the offshore wind farm corresponding to the periodic components; the energy distribution similarity of the Fourier encoded vector of the actual power output curve of the offshore wind farm and the Fourier encoded vector of the power adjustment target curve are compared in each frequency band. The trend similarity, shape correlation, and energy distribution similarity are weighted and summed to obtain the comprehensive shape matching degree between the power regulation target curve and the actual power output curve of the offshore wind farm.

5. The lightweight interaction method between offshore wind farms and power grids according to claim 4, characterized in that, The weighting coefficients for trend similarity, shape correlation, and energy distribution similarity are obtained as follows: Based on the relative importance of power grid dispatch in three dimensions—trend similarity, shape correlation, and energy distribution similarity—a judgment matrix is ​​constructed. Calculate the largest eigenvalue of the judgment matrix and its corresponding eigenvector, normalize the eigenvector, and obtain the weight coefficients of the trend similarity, the shape correlation, and the energy distribution similarity; A consistency check is performed on the judgment matrix. If the consistency ratio is less than a preset consistency ratio threshold, the weight coefficients of the trend similarity, the shape correlation, and the energy distribution similarity are determined to be valid; otherwise, the judgment matrix is ​​reconstructed and the weights are recalculated until the consistency check is passed.

6. The lightweight interaction method between offshore wind farms and power grids according to claim 5, characterized in that, Construct an instantaneous regulation capability matrix for offshore wind farms and calculate the regulation capability confidence level; based on the instantaneous regulation capability matrix and regulation capability confidence level of offshore wind farms, obtain the dimension-reduced encoding vector of the regulation capability of offshore wind farms. This includes constructing the instantaneous regulation capability matrix of offshore wind farms and calculating the regulation capability confidence level, including: Based on the rated power of the offshore wind farm, the actual output power at the current moment, and the measured wind speed at the current moment, the rapid upward adjustment capability and rapid downward adjustment capability of the offshore wind farm under the first adjustment time scale, and the slow upward adjustment capability and slow downward adjustment capability under the second adjustment time scale are calculated respectively; wherein, the first adjustment time scale is shorter than the second adjustment time scale. Based on the rapid upward and downward adjustment capabilities of the offshore wind farm under the first adjustment time scale, and the slow upward and slow downward adjustment capabilities under the second adjustment time scale, an instantaneous adjustment capability matrix of the offshore wind farm is constructed. The confidence level of the regulation capability of offshore wind farms is calculated based on the standard deviation of wind speed forecasts, which reflects the reliability of the regulation capability of offshore wind farms under the current uncertainty of wind speed forecasts.

7. The lightweight interaction method between offshore wind farms and power grids according to claim 6, characterized in that, Based on the instantaneous regulation capability matrix and regulation capability confidence of the offshore wind farm, a dimension-reduced encoding vector of the regulation capability of the offshore wind farm is obtained, including: Based on the rapid upward and downward adjustment capabilities of the offshore wind farm under the first adjustment time scale, the slow upward and slow downward adjustment capabilities of the offshore wind farm under the second adjustment time scale, the adjustment capability confidence of the offshore wind farm, the measured wind speed at the current moment, and the adjustable power range of the offshore wind farm at the current moment, the adjustment capability feature vector of the offshore wind farm is obtained; wherein, the adjustable power range of the offshore wind farm at the current moment is the sum of the rapid upward and rapid downward adjustment capabilities. Based on the historical data of the regulation capacity of offshore wind farms within a preset period, a historical operation data feature matrix is ​​constructed, and the covariance matrix corresponding to the historical data feature matrix is ​​calculated. Using the principle of spectral decomposition, the covariance matrix is ​​decomposed into eigenvalues ​​to obtain the corresponding orthogonal eigenvector matrix and eigenvalue matrix; Before choosing The orthogonal eigenvectors corresponding to the largest eigenvalues ​​form the projection matrix; The regulation capability feature vector of the offshore wind farm is projected onto the projection matrix to obtain the regulation capability dimensionality reduction encoding vector of the offshore wind farm.

8. The lightweight interaction method between offshore wind farms and power grids according to claim 7, characterized in that, Based on the dimensionality reduction encoding vector of the offshore wind farm's regulation capability and the comprehensive shape matching degree, the optimal state estimate of the offshore wind farm's regulation capability state is obtained, including: The regulation capability state space model is established by using the dimensionality reduction encoding vector of the offshore wind farm as the state vector. The regulation capability state space model includes a state equation and an observation equation. The state equation describes the time evolution of the regulation capability dimensionality reduction encoding vector, and the observation equation describes the mapping relationship from the state vector to the observation vector. The comprehensive shape matching degree is used as observation feedback and is a component of the observation vector. Based on the optimal state estimate and control input at the current moment, perform Kalman filtering prediction to predict the state vector and prediction error covariance matrix at the next moment; Obtain the observation vector at the next time step, and calculate the Kalman gain based on the prediction error covariance matrix and the observation noise covariance matrix; Based on the Kalman gain fusion prediction and observation values, the optimal state estimate for the next time step is obtained, and the error covariance matrix is ​​updated.

9. The lightweight interaction method between offshore wind farms and power grids according to claim 8, characterized in that, A dynamic switching coding scheme based on the rate of change of the actual power output curve of offshore wind farms, which dynamically adjusts the redundancy of forward error correction coding based on channel quality indicators, includes: Calculate the rate of change of the actual power output curve of the offshore wind farm to be developed; When the rate of change is less than a preset rate of change threshold, a high compression ratio coding scheme of differential coding plus Huffman compression is adopted; otherwise, a low compression ratio and high reliability coding scheme of direct coding plus forward error correction is adopted. The channel quality index is calculated based on the ratio of effective bandwidth to nominal bandwidth and the bit error rate. When the channel quality index is greater than the first preset threshold, the first redundancy is adopted; When the channel quality index is less than or equal to the first preset threshold and greater than the second preset threshold, the second redundancy is adopted; When the channel quality index is less than or equal to the second preset threshold, a third redundancy is used; Wherein, the first preset threshold is greater than the second preset threshold; the first redundancy is less than the second redundancy, and the second redundancy is less than the third redundancy.

10. The lightweight interaction method between offshore wind farms and power grids according to claim 9, characterized in that, Based on the dynamically switched coding scheme and dynamically adjusted redundancy, and using the inflection point parameter set, Fourier coding vector, optimal state estimation, and comprehensive shape matching degree of the actual power output curve of the offshore wind farm, hierarchical lightweight data processing is performed to generate a dynamically optimized data package for the offshore wind farm, including: The layering includes the application layer, coding layer, transport layer, and physical layer; At the application layer, the raw time-series data to be transmitted is classified into real-time control data, status monitoring data, and historical statistical data. At the coding layer, based on the dynamically switched coding scheme and the dynamically adjusted redundancy, the inflection point parameter set, Fourier coding vector, and corresponding optimal state estimation and comprehensive shape matching degree of the actual power output curve of the offshore wind farm are coded, compressed and forward-corrected. At the transport layer, an adaptive packet splitting mechanism is adopted to dynamically adjust the packet size based on the size of the original time-series data packets to be transmitted, the available bandwidth of the current channel, the inherent network latency, and the inherent protocol overhead of the data packets. At the physical layer, channel quality is monitored, providing parameters for the calculation of the channel quality metrics; Based on the results of the encoding compression and forward error correction coding, and the dynamically adjusted data packet size, the dynamically optimized data packet for the offshore wind farm is obtained.

Images