Offshore wind power-storage-vehicle grid collaborative optimization method based on wind speed probability scenario

Abstract

The application provides an offshore wind power-energy storage-vehicle network collaborative optimization method based on a wind speed probability scenario, and relates to the technical field of offshore wind power energy storage, wherein the method steps include obtaining wind speed data of an offshore wind farm and preprocessing; the wind speed data is decomposed into a trend component, a positive fluctuation component and a negative fluctuation component to construct a three-dimensional feature vector; the three-dimensional feature vector is subjected to probability clustering to generate multiple wind speed scenario parameters and scenario occurrence probabilities; an upper-layer multi-objective optimization model is constructed to obtain a power distribution strategy of the energy storage system between an electricity market and a frequency modulation auxiliary service market; a lower-layer two-stage joint clearing regulation model is constructed to perform joint clearing of the electricity market and the frequency modulation auxiliary service market, and the energy storage system is used for scenario-based adaptive frequency regulation; the market clearing and collaborative regulation results are fed back to the upper-layer multi-objective optimization model for evaluation to generate an optimal strategy, and a probability scenario-strategy mapping library is formed for calling.

Description

Technical Field

[0001] This invention relates to the field of offshore wind power energy storage technology, and in particular to a collaborative optimization method for offshore wind power-energy storage-vehicle-grid based on wind speed probability scenarios. Background Technology

[0002] With the large-scale development of nearshore and deep-sea wind power, offshore wind power has become an important power source for new power systems. However, offshore wind speeds exhibit significant randomness, abrupt changes, and multi-frequency fluctuations, causing wind turbine active power output to rise or fall sharply in a short period. When offshore wind power has a high penetration rate in regional power grids, such power fluctuations amplify the risk of system frequency deviation, placing enormous operational pressure on conventional thermal power units, interconnection lines, and frequency regulation reserve resources.

[0003] Currently, for the optimized operation of wind-storage integrated systems participating in the electricity market, existing technical solutions mostly adopt a single scenario for power planning optimization, modeling the electricity market and frequency regulation ancillary service market separately, configuring energy storage power separately, or focusing solely on economic efficiency. Furthermore, there is a lack of a collaborative mechanism based on wind speed scenario recognition between stationary energy storage and electric vehicle V2G resources, making it difficult to achieve dynamic adaptive optimization and control that coordinates the three major goals of economic benefits, frequency stability, and low-carbon environmental protection when facing uncertain wind speed changes. Summary of the Invention

[0004] The purpose of this invention is to provide a collaborative optimization method for offshore wind power-energy storage-vehicle-grid based on wind speed probability scenarios, in order to solve the problem mentioned in the background art that existing optimization operation schemes for wind-storage integrated systems and electricity markets are difficult to achieve dynamic adaptive optimization and control that coordinates the three major goals of economic benefits, frequency stability and low-carbon environmental protection when facing uncertain wind speed changes.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a collaborative optimization method for offshore wind power-energy storage-vehicle-grid based on wind speed probability scenarios, comprising the following steps: acquiring wind speed data from an offshore wind farm and preprocessing it; using variational mode decomposition to decompose the wind speed data into trend components and fluctuation components, and further splitting the fluctuation components into positive and negative fluctuation components according to their signs, calculating the statistics of each component to construct a three-dimensional feature vector; using a Gaussian mixture model to perform probabilistic clustering on the three-dimensional feature vector to generate multiple wind speed scenario parameters and scenario occurrence probabilities; based on the wind speed scenario parameters and scenario occurrence probabilities, constructing an upper-level multi-objective optimization model to maximize the total revenue of the energy storage system and frequency... With the objectives of minimizing rate stability and carbon emissions, a power allocation strategy for the energy storage system between the electricity market and the frequency regulation ancillary service market is obtained. Based on the power allocation strategy, a lower-level two-stage joint clearing and control model is constructed. In the first stage, the electricity market and the frequency regulation ancillary service market are jointly cleared. In the second stage, the energy storage system is used for scenario-based adaptive frequency regulation. When the energy storage system's regulation capacity is insufficient, V2G resources of electric vehicles are triggered for coordinated frequency regulation. The results of market clearing and coordinated regulation are fed back to the upper-level multi-objective optimization model for evaluation, generating the optimal strategy corresponding to each wind speed scenario, and forming a probabilistic scenario-strategy mapping library for use.

[0006] Optionally, the step of decomposing the wind speed data into trend components and fluctuation components using the variational mode decomposition method specifically includes: using the wind speed data as the original signal to be decomposed and setting the number of decomposed modes; constructing a variational constraint problem with the objective of minimizing the sum of the bandwidths of each mode, thereby transforming the decomposition of the wind speed data into an optimization solution process; introducing a Lagrange penalty operator to transform the variational constraint problem into an unconstrained optimization problem in augmented Lagrange form; using the alternating direction multiplier method to iteratively solve the unconstrained optimization problem, alternately updating the frequency domain estimate, center frequency, and Lagrange multipliers of each mode component in each iteration until the decomposition result converges; and separating the low-frequency trend component representing macroscopic changes in wind resources and the high-frequency fluctuation component representing instantaneous fluctuations from the converged mode components based on the level of the center frequency.

[0007] Optionally, the step of splitting the fluctuation component into positive and negative fluctuation components according to its sign and calculating the statistics of each component specifically includes: splitting the high-frequency fluctuation component into positive and negative fluctuation components according to the numerical sign corresponding to each time point; calculating the average value of the low-frequency trend component as a trend feature quantity characterizing the overall level of wind resources; calculating the average value of the positive fluctuation component as a positive fluctuation feature quantity characterizing the average intensity of positive disturbances; calculating the average value of the negative fluctuation component as a negative fluctuation feature quantity characterizing the average intensity of negative disturbances; and combining the trend feature quantity, the positive fluctuation feature quantity, and the negative fluctuation feature quantity to form the three-dimensional feature vector.

[0008] Optionally, the step of using a Gaussian mixture model to perform probabilistic clustering on the three-dimensional feature vector to generate multiple wind speed scene parameters and scene occurrence probabilities specifically includes: performing rare / abnormal sample management on the three-dimensional feature vector to remove or reduce the weight of samples whose occurrence frequency is lower than a frequency threshold or deviates from the main distribution by more than a deviation threshold; performing probabilistic clustering analysis on the managed three-dimensional feature vector using a Gaussian mixture model, and adaptively determining the optimal number of scene clusters through the Bayesian information criterion or the Akaike information criterion; fitting the Gaussian mixture model to obtain the Gaussian distribution parameters and scene occurrence probability for each wind speed scene, wherein the Gaussian distribution parameters include the feature mean vector, covariance matrix, and mixture weights reflecting the probability of occurrence of the wind speed scene, and the mixture weights characterize the prior occurrence probability of each wind speed scene; and selecting a representative sample set from the clustered three-dimensional feature vectors within each scene based on the principle of maximizing the posterior probability or minimizing the Mahalanobis distance, as the input to the lower-level two-stage joint clearing and control model.

[0009] Optionally, the steps for constructing the upper-level multi-objective optimization model specifically include: based on the total revenue of the electricity market and the frequency regulation ancillary service market, subtracting the operating cost of thermal power, and weighting by scenario probability to obtain the maximum total system revenue; based on the root mean square value of the system frequency deviation, weighting by scenario probability to obtain the minimum frequency stability index; based on the sum of the products of thermal power unit output and carbon emission coefficient, weighting by scenario probability to obtain the minimum carbon emissions; and introducing energy storage power allocation constraints, energy storage state of charge constraints, and energy storage capacity limitation constraints between the electricity market and the frequency regulation market to obtain the upper-level multi-objective optimization model.

[0010] Optionally, the steps of constructing the lower-level two-stage joint clearing and control model specifically include: establishing a stochastic unit combination model, aiming to minimize the start-up and shutdown costs, no-load costs, and expected operating costs of all power sources under the wind speed scenario, determining the start-up and shutdown plan of the thermal power units, and introducing constraints such as system load balance, network power flow security, minimum start-up and shutdown time of units, and ramp rate; establishing a scenario-based security-constrained economic dispatch model, for each wind speed scenario, aiming to minimize real-time operating costs, performing refined allocation of power generation, and outputting the nodal marginal electricity price, line power flow, and energy storage charging and discharging plan under each wind speed scenario.

[0011] Optionally, the steps of the first stage of joint clearing of the electricity market and the frequency regulation ancillary service market specifically include: taking the capacity demand and mileage demand of the frequency regulation ancillary service market as optimization variables or constraints, and solving them synchronously with the clearing process of the electricity market in the lower-level two-stage joint clearing control model, so as to determine the energy storage system's output plan in the electricity market, the capacity reservation and mileage call plan in the frequency regulation market, and the corresponding market clearing price.

[0012] Optionally, the second stage utilizes the energy storage system for scenario-based adaptive frequency regulation, and triggers V2G resources of electric vehicles for coordinated frequency regulation when the energy storage system's regulation capacity is insufficient. Specifically, this includes: setting an adaptive droop control coefficient for the energy storage system that is associated with the current wind speed scenario; real-time monitoring of system frequency deviation, and when the frequency deviation exceeds a preset dead zone range, calculating and issuing a frequency regulation power command for the energy storage system based on the adaptive droop control coefficient; calculating the posterior probability of the current wind speed data belonging to each wind speed scenario based on the Gaussian mixture model, evaluating the state of charge of the energy storage system and its over-limit risk probability during future rolling scheduling periods, and triggering the electric vehicle cluster to participate in frequency regulation when the risk probability exceeds a preset risk threshold; and controlling the electric vehicle cluster to operate in a vehicle-to-grid discharge mode or a grid-to-vehicle charging mode according to the direction of system power deficit or surplus, coordinating with the energy storage system to suppress frequency deviation.

[0013] Optionally, the step of feeding back the results of market clearing and coordinated regulation to the upper-level multi-objective optimization model for evaluation and generating the optimal strategy corresponding to each wind speed scenario specifically includes: using the nodal marginal electricity price obtained from the lower-level joint clearing, the actual energy storage power, and the system frequency stability index and carbon emissions calculated after frequency regulation as input parameters and constraints of the upper-level multi-objective optimization model; and in the upper-level multi-objective optimization model, re-evaluating and optimizing the power allocation strategy and scenario-based control parameters based on the feedback information.

[0014] Optionally, the step of forming a probability scenario-policy mapping library for invocation specifically includes: for each wind speed scenario, storing its corresponding optimal power allocation policy, adaptive droop control coefficient, and electric vehicle trigger threshold as a complete policy combination to jointly constitute the probability scenario-policy mapping library; calculating the posterior probability of each wind speed scenario based on real-time wind speed data, and directly calling the policy combination corresponding to the wind speed scenario with the highest posterior probability, or weighting and fusing the policy combinations of each wind speed scenario according to the posterior probability to generate the final executed hybrid policy.

[0015] Compared with the prior art, the beneficial effects of the present invention are:

[0016] This application constructs a complete scheme encompassing scenario perception, decision optimization, collaborative regulation, and closed-loop evaluation. Through variational mode decomposition and probabilistic clustering using Gaussian mixture models, it generates typical wind speed scenarios with clear physical meaning and probability of occurrence, providing precise probabilistic input for optimization decisions and significantly enhancing the adaptability and robustness of the strategy to actual wind conditions. By establishing an upper-level multi-objective optimization model, it achieves unified decision-making on power allocation between energy storage and the frequency regulation market, thus explicitly considering frequency stability and low-carbon objectives while pursuing the maximization of total system revenue, solving the problems of fragmented multi-markets and difficulty in coordinating multiple objectives. Through joint clearing of the lower-level market and scenario-based adaptive frequency regulation, combined with a risk probability triggering mechanism to mobilize V2G resources from electric vehicles, it forms a dynamic complementarity between fixed and mobile energy storage resources, significantly improving the system's ability to suppress frequency fluctuations and its regulatory flexibility. Finally, through closed-loop feedback, a probabilistic scenario-policy mapping library is formed, enabling the entire method to possess online self-learning and rolling optimization capabilities, achieving a leap from static planning to dynamic adaptive operation, and comprehensively improving the economic, safety, and environmental benefits of offshore wind power clusters. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the method steps of the present invention. Detailed Implementation

[0018] The present invention will now be clearly and completely described in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0019] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be used interchangeably where appropriate for the embodiments of this application described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0020] Those skilled in the art will understand that, unless explicitly stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in the specification of this application means the presence of features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or wireless coupling. The term “and / or” as used herein includes all or any units and all combinations of one or more associated listed items.

[0021] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0022] It should be understood that the sequence number and size of each step in this embodiment do not imply the order of execution. The execution order of each process is determined by its function and internal logic, and should not constitute any limitation on the implementation process of this application embodiment.

[0023] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0024] Please refer to Figure 1This invention discloses a collaborative optimization method for offshore wind power, energy storage, and vehicle-to-grid systems based on wind speed probability scenarios. The method includes the following steps: acquiring wind speed data from an offshore wind farm and preprocessing it; using variational mode decomposition to decompose the wind speed data into trend and fluctuation components, further separating the fluctuation components into positive and negative fluctuation components based on their signs, and calculating the statistics of each component to construct a three-dimensional feature vector; using a Gaussian mixture model to perform probabilistic clustering on the three-dimensional feature vector to generate multiple wind speed scenario parameters and scenario occurrence probabilities; and constructing an upper-level multi-objective optimization model based on the wind speed scenario parameters and scenario occurrence probabilities to maximize the total revenue of the energy storage system and maintain frequency stability. With the objectives of minimizing performance indicators and carbon emissions, a power allocation strategy for the energy storage system between the electricity market and the frequency regulation ancillary service market is obtained. Based on the power allocation strategy, a lower-level two-stage joint clearing and control model is constructed to jointly clear the electricity market and the frequency regulation ancillary service market. The energy storage system is used for scenario-based adaptive frequency regulation, and when the energy storage system's regulation capacity is insufficient, V2G resources of electric vehicles are triggered for coordinated frequency regulation. The results of market clearing and coordinated regulation are fed back to the upper-level multi-objective optimization model for evaluation, generating the optimal strategy corresponding to each wind speed scenario, and forming a probabilistic scenario-strategy mapping library for use.

[0025] Specifically, wind speed sequence data of the target offshore wind farm is acquired, and missing value imputation, anomaly handling, standardization, and time scale unification are performed. Variational mode decomposition (VMD) is used to decompose the wind speed sequence into trend and fluctuation components. The fluctuation components are further divided into positive and negative fluctuations based on their signs. Statistics are calculated, and a three-dimensional feature vector is constructed as input for scene recognition. Rare / anomaly sample processing is performed on the three-dimensional feature vector, and a Gaussian mixture model is used for probabilistic clustering of the processed three-dimensional feature vector. Wind speed scene parameters and scene occurrence probabilities are output, and a representative sample set is extracted. A higher-level multi-objective optimization model is established with the optimization objectives of maximizing total system revenue, minimizing frequency stability indicators, and minimizing carbon emissions. Expectation weighting is applied based on scene occurrence probabilities, and the cross-market power allocation ratio of energy storage and scene-specific control parameters are output. A two-stage joint clearing and control model is established at the lower level. The first stage runs a stochastic safety-constrained unit combination model to determine the start-up and shutdown status and operational commitment plans of conventional generator units. The second stage runs a safety-constrained economic dispatch model for each typical wind speed probability scenario, obtaining the active power output plan of each generator unit, the line power flow distribution in the power network, and the marginal electricity price of each node. These results are then aggregated according to scenario probabilities for evaluation by the upper-level multi-objective optimization model and frequency control input. Based on the clearing results of the safety-constrained economic dispatch and the upper-level decision, an adaptive droop control coefficient and control dead zone associated with the wind speed scenario are used to regulate the frequency power response of the energy storage system. When the adjustable capacity margin of the fixed energy storage is insufficient, and the probability of its state of charge exceeding the limit exceeds a preset threshold, a collaborative control strategy for the electric vehicle cluster is triggered. Under operating conditions of system power surplus or power shortage, the grid charges the vehicles or the vehicles discharge to the grid to suppress system frequency deviation, and the system frequency stability index is calculated for closed-loop evaluation. The price, output, and frequency regulation effects from the lower layer, along with carbon emission calculations, are fed back into the upper-layer multi-objective optimization model for evaluation and trade-offs, forming a probabilistic scenario-policy mapping library. This library supports weighted fusion and invocation based on scenario-specific posterior probabilities. In actual power allocation, wind speed characteristics and load forecasts are updated according to a rolling time window, allowing for online fine-tuning of the strategy. When the log-likelihood of the scenario model decreases significantly or the distribution drift exceeds a threshold, scenario model re-evaluation and strategy retraining are triggered to ensure the long-term effectiveness and robustness of the strategy.

[0026] This application constructs a complete scheme encompassing scenario perception, decision optimization, collaborative regulation, and closed-loop evaluation. Through variational mode decomposition and probabilistic clustering using Gaussian mixture models, it generates typical wind speed scenarios with clear physical meaning and probability of occurrence, providing precise probabilistic input for optimization decisions and significantly enhancing the adaptability and robustness of the strategy to actual wind conditions. By establishing an upper-level multi-objective optimization model, it achieves unified decision-making on power allocation between energy storage and the frequency regulation market, thus explicitly considering frequency stability and low-carbon objectives while pursuing the maximization of total system revenue, solving the problems of fragmented multi-markets and difficulty in coordinating multiple objectives. Through joint clearing of the lower-level market and scenario-based adaptive frequency regulation, combined with a risk probability triggering mechanism to mobilize V2G resources from electric vehicles, it forms a dynamic complementarity between fixed and mobile energy storage resources, significantly improving the system's ability to suppress frequency fluctuations and its regulatory flexibility. Finally, through closed-loop feedback, a probabilistic scenario-policy mapping library is formed, enabling the entire method to possess online self-learning and rolling optimization capabilities, achieving a leap from static planning to dynamic adaptive operation, and comprehensively improving the economic, safety, and environmental benefits of offshore wind power clusters.

[0027] In some embodiments, the step of decomposing the wind speed data into trend components and fluctuation components using the variational mode decomposition method specifically includes: using the wind speed data as the original signal to be decomposed and setting the number of decomposed modes; constructing a variational constraint problem with the objective of minimizing the sum of the bandwidths of each mode, thereby transforming the decomposition of the wind speed data into an optimization solution process; introducing a Lagrange penalty operator to transform the variational constraint problem into an unconstrained optimization problem in augmented Lagrange form; using the alternating direction multiplier method to iteratively solve the unconstrained optimization problem, alternately updating the frequency domain estimate, center frequency, and Lagrange multipliers of each mode component in each iteration until the decomposition result converges; and separating the low-frequency trend component representing macroscopic changes in wind resources and the high-frequency fluctuation component representing instantaneous fluctuations from the converged mode components based on the level of the center frequency.

[0028] Specifically, the raw wind speed data is processed using variational mode decomposition (VMD), which divides the wind speed signal into fluctuation and trend components. VMD achieves signal decomposition through the principle of minimizing variation, concentrating the various modes of signal as much as possible in the frequency domain. This yields sub-signals reflecting different frequency characteristics, allowing the extraction of different frequency components of the wind speed signal to support subsequent probabilistic scenario classification and business model selection. The specific process of VMD is as follows:

[0029] Construct a variational problem: given a signal, the goal of variational mode decomposition is to decompose it into K modes. The frequency characteristics of each mode are concentrated, and the frequencies of these modes are as non-overlapping as possible. The objective function of variational mode decomposition can be expressed as:

[0030] ;

[0031] In the formula: This is wind speed data for offshore wind farms, where K represents the number of modes. It's a Dirac function. It is a convolution operator. It is modal. It is the center frequency. It is the k-th eigenmode component obtained after decomposition. It is a complex exponential function used to shift the spectrum of a mode to the baseband.

[0032] Introducing the Lagrange penalty operator λ transforms the constrained variational problem into an unconstrained variational problem, resulting in the augmented expression:

[0033] ;

[0034] In the formula: λ is the augmented Lagrangian function, and λ is the Lagrange penalty operator. It is the regularization parameter.

[0035] Iterative solution, first initialize the parameters , , Set up a loop to update the solution;

[0036] ;

[0037] ;

[0038] ;

[0039] ;

[0040] In the formula: It is the spectrum of the k-th modal component in the (n+1)-th iteration. It is the spectrum of the original input signal. It is the sum of the spectra of all modes except the k-th mode after the nth iteration. It is the spectrum of the Lagrange multipliers after the (n+1)th iteration. It is the center frequency of the k-th mode after the (n+1)-th iteration. It is noise margin. , and They are , and Fourier transform, It is the convergence tolerance, the threshold for determining whether the algorithm has stopped.

[0041] To accurately segment wind speed scenarios, this application employs a variational mode decomposition algorithm to decouple the raw time-series wind speed data, precisely separating the trend component and the fluctuation component. The low-frequency trend component characterizes the total distribution characteristics of wind resources, while the high-frequency fluctuation component reflects the transient disturbance characteristics of wind power. By calculating the average values ​​of the trend and fluctuation components, a feature vector with clear physical meaning can be constructed.

[0042] This application employs variational mode decomposition, which adaptively and accurately decomposes the original wind speed signal into trend components and fluctuation components with clear physical meaning. This provides a high-quality source signal with pure frequency components for subsequent feature extraction and scene classification, fundamentally ensuring the accuracy and reliability of scene recognition.

[0043] In some embodiments, the step of splitting the fluctuation component into positive and negative fluctuation components according to its sign and calculating the statistics of each component specifically includes: splitting the high-frequency fluctuation component into positive and negative fluctuation components according to the numerical sign corresponding to each time point; calculating the average value of the low-frequency trend component as a trend feature quantity characterizing the overall level of wind resources; calculating the average value of the positive fluctuation component as a positive fluctuation feature quantity characterizing the average intensity of positive disturbances; calculating the average value of the negative fluctuation component as a negative fluctuation feature quantity characterizing the average intensity of negative disturbances; and combining the trend feature quantity, the positive fluctuation feature quantity, and the negative fluctuation feature quantity to form the three-dimensional feature vector.

[0044] Specifically, since the fluctuation component contains both positive and negative values, directly averaging the fluctuation component may result in a value close to zero, thus losing effective information about the fluctuation intensity. To address this issue, this application decomposes the fluctuation component into positive and negative fluctuations when calculating the average value of the trend component. Then, the positive and negative fluctuation components are averaged separately. This method effectively preserves the characteristic information of the fluctuation component and, together with the trend component, constructs a three-dimensional feature vector. This provides a more accurate basis for subsequent scene classification.

[0045] ;

[0046] ;

[0047] ;

[0048] In the formula, The total time step, and These are the high-frequency and low-frequency mode components after decomposing the original wind speed data. , and These are the mean of the trend component and the mean of the positive and negative fluctuation components, respectively.

[0049] This application splits the wave components into positive and negative waves according to their signs and calculates their mean values ​​separately. This allows the constructed three-dimensional feature vector to fully retain the steady-state trend of wind speed, the intensity of positive disturbances, and the intensity of negative disturbances. It effectively avoids the loss of wave information due to mutual cancellation during the averaging process, and provides input features with higher discriminative power and more comprehensive representation for probabilistic clustering.

[0050] In some embodiments, the step of using a Gaussian mixture model to perform probabilistic clustering on the three-dimensional feature vector to generate multiple wind speed scene parameters and scene occurrence probabilities specifically includes: performing rare / abnormal sample management on the three-dimensional feature vector to remove or reduce the weight of samples whose occurrence frequency is lower than a frequency threshold or deviates from the main distribution by more than a deviation threshold; performing probabilistic clustering analysis on the managed three-dimensional feature vector using a Gaussian mixture model, and adaptively determining the optimal number of scene clusters through the Bayesian information criterion or the Akaike information criterion; fitting the Gaussian mixture model to obtain the Gaussian distribution parameters and scene occurrence probability for each wind speed scene, wherein the Gaussian distribution parameters include the feature mean vector, covariance matrix, and mixture weights reflecting the probability of occurrence of the wind speed scene, and the mixture weights characterize the prior occurrence probability of each wind speed scene; and selecting a representative sample set from the clustered three-dimensional feature vectors within each scene based on the principle of maximizing the posterior probability or minimizing the Mahalanobis distance, as the input to the lower-level two-stage joint clearing and control model.

[0051] Specifically, after completing the variational mode decomposition and feature extraction of the wind speed signal, a Gaussian mixture model is used for probabilistic clustering to classify the scene based on the three-dimensional feature vector sample set composed of the trend component and the fluctuation component. The probabilistic clustering includes: processing the samples by removing or reducing the weight of samples with extremely low occurrence frequency and significant anomalies to avoid the bias effect of abnormal data on scene parameter estimation; and fitting a Gaussian mixture model to the processed samples to obtain the parameters and occurrence probability of each scene. For any sample... Belongs to the The posterior probability of a scenario can be calculated using the following formula:

[0052] ;

[0053] In the formula, For posterior probability, For the first The mixed weights of each scenario, i.e., the probability of each scenario occurring. , These are the scene mean and covariance, respectively. This represents the Gaussian distribution density. Number of scenes. Information criteria BIC / AIC can be used for selection, and a representative sample set can be selected in each scenario based on the principle of maximizing posterior probability or minimizing Mahalanobis distance, which will be used as input for subsequent power allocation and joint clearing.

[0054] This application uses Gaussian mixture model probabilistic clustering combined with sample governance and information criteria to adaptively segment representative typical wind speed probability scenarios and accurately quantify the occurrence probability and internal sample distribution of each scenario. This enables a refined and probabilistic description of complex sea wind conditions, providing robust and interpretable probabilistic inputs for subsequent optimization decisions.

[0055] In some embodiments, the steps of constructing the upper-level multi-objective optimization model specifically include: based on the total revenue of the electricity market and the frequency regulation ancillary service market, subtracting the thermal power operating cost and weighting by scenario probability to obtain the maximum total system revenue; based on the root mean square value of the system frequency deviation, weighting by scenario probability to obtain the minimum frequency stability index; based on the sum of the products of thermal power unit output and carbon emission coefficient, weighting by scenario probability to obtain the minimum carbon emissions; and introducing energy storage power allocation constraints, energy storage state of charge constraints, and energy storage capacity limitation constraints between the electricity market and the frequency regulation market to obtain the upper-level multi-objective optimization model.

[0056] Specifically, system frequency stability is a crucial indicator for assessing grid security and the balance between power generation and consumption. The inertia of a power system is a key factor in resisting frequency changes caused by external disturbances, effectively mitigating the rate of frequency change (ROCOF) and thus ensuring system frequency stability. System inertia H and the rate of frequency change (ROCOF) are calculated using the following formulas:

[0057] ;

[0058] ;

[0059] In the formula, It is the moment of inertia. It is the rotor speed. It is the rated power. It is the total number of generators. It is the change in active power. It is the nominal frequency.

[0060] The system frequency at a certain moment and the deviation of the system from the rated frequency can be calculated by the following formulas:

[0061] ;

[0062] ;

[0063] The formula for calculating the System Stability Index (SFSI) is as follows:

[0064] ;

[0065] In the formula: It is a frequency control dead zone. and These are their upper and lower limits, respectively. and The maximum power that frequency-regulated charging and discharging can provide for energy storage is when or If the system is in an unstable state, it is considered to need to store energy for charging and discharging to maintain system stability.

[0066] To maximize the benefits of configuring energy storage systems in wind farms within both the electricity market and ancillary services market, while also considering grid frequency stability and carbon emission reduction, a multi-objective, two-layer collaborative framework model is established. The upper layer, with the comprehensive objectives of maximizing total system revenue, minimizing the system frequency stability index (SFSI), and minimizing carbon emissions, determines the power allocation ratio of energy storage between the electricity market and the ancillary services market, as well as the corresponding business model. The lower layer, under the conditions given by the upper layer's decisions, jointly clears the electricity market and the frequency regulation market, outputting market node prices, power output plans for each generation and energy storage unit, and frequency regulation response results. These clearing results are then fed back to evaluate and constrain the upper-layer objectives, thus forming a tightly coupled, scenario-oriented, and coordinated allocation framework between the upper and lower layers.

[0067] The upper-level model aims to determine the power allocation strategy of the energy storage system in the electricity market and ancillary services market for different wind speed probability scenarios. The model uses multiple objective functions to maximize the total system revenue (electricity market revenue, frequency regulation market revenue), minimize the system frequency stability index (SFSI), and minimize system carbon emissions. It employs expected value weighting based on the probability of each scenario and can introduce a risk term to improve robustness in tail scenarios. The upper-level objective function can be expressed as:

[0068] ;

[0069] ;

[0070] ;

[0071] In the formula, For the first The probability of each scenario occurring is given by the mixture weights of the Gaussian mixture model. For the scene The marginal price at the bottom clearing node. , , These represent the contributions of thermal power, wind power, and energy storage to the electricity market, respectively. For time intervals; For energy storage frequency regulation capacity, For frequency-controlled mileage; and These are capacity-based electricity pricing and mileage-based electricity pricing, respectively. Carbon emission factor This is the cost function for thermal power.

[0072] The upper-level constraints are as follows:

[0073] ;

[0074] ;

[0075] In the formula, , For energy storage charging and discharging efficiency, For energy storage capacity; , These are the charging and discharging power components, respectively.

[0076] The multi-objective optimization model constructed in this application is the first to solve the total system revenue, frequency stability index and carbon emissions in a unified way under the framework of probabilistic scenario weighting. This enables the energy storage power allocation strategy to transparently weigh economic, safety and environmental goals, thereby achieving synergistic optimization of multiple objectives at the strategy level and avoiding the one-sidedness of a single objective orientation.

[0077] In some embodiments, the step of constructing the lower-level two-stage joint clearing and control model specifically includes: establishing a stochastic unit combination model to determine the start-up and shutdown plan of thermal power units with the goal of minimizing the start-up and shutdown costs, no-load costs, and expected operating costs of all power sources under the wind speed scenario, and introducing constraints such as system load balance, network power flow security, minimum start-up and shutdown time of units, and ramp rate; establishing a scenario-based security-constrained economic dispatch model to perform refined allocation of power generation for each wind speed scenario with the goal of minimizing real-time operating costs, and outputting the nodal marginal electricity price, line power flow, and energy storage charging and discharging plan under each wind speed scenario.

[0078] Specifically, the lower-level model is divided into an energy market clearing model and a lower-level ancillary services market frequency regulation model. The energy market clearing model aims to complete the generation planning and nodal tariff calculation under different wind speed probability scenarios based on the energy storage allocation ratio determined by the upper level and the adjustable resource characteristics of wind, thermal, and energy storage. The model adopts a two-stage stochastic SCUC–SCED solution method.

[0079] The first stage involves constructing a stochastic SCUC clearing model. The objective function for this stage is to minimize the system operating cost, and its calculation formula is as follows:

[0080] ;

[0081] ;

[0082] ;

[0083] ;

[0084] In the formula, For the operating costs of the system's thermal power units, The cost of generating electricity from offshore wind turbines, The cost of charging and discharging energy storage power stations. and Separate quotes and outputs for thermal power units. , Costs associated with starting and stopping thermal power units. and The offshore wind turbines recently submitted their on-grid electricity prices and power outputs. , The prices are for charging and discharging energy storage devices. , These represent the charge / discharge capacity declared by the energy storage equipment as of the previous day; T represents the system operating time. This is the start-up / shutdown state of the thermal power plant. For start / stop action variables; For the no-load cost of thermal power, This represents start-up and shutdown costs; the meanings of the other symbols are consistent with those of the upper layer.

[0085] The constraints are:

[0086] Load balance constraints:

[0087] ;

[0088] Network trend constraints:

[0089] ;

[0090] ;

[0091] Thermal power unit start-up and shutdown constraints:

[0092] ;

[0093] ;

[0094] ;

[0095] ;

[0096] Thermal power unit ramping constraints:

[0097] ;

[0098] ;

[0099] Thermal power unit output constraints:

[0100] ;

[0101] Wind turbine output constraints:

[0102] ;

[0103] Energy storage device charge and discharge constraints:

[0104] ;

[0105] ;

[0106] ;

[0107] Energy storage capacity constraints:

[0108] ;

[0109] ;

[0110] In the formula, , , , These represent the number of thermal power units, offshore wind turbines, energy storage devices, and load nodes, respectively. denoted as node load. H represents the power transfer distribution factor. Ag, Aw, and Ae represent the network topologies of thermal power units, offshore wind turbines, and energy storage devices, respectively. This represents the maximum branch power of the network. , This is a counter for the start-up and shutdown time of the thermal power unit. , This refers to the start-up and shutdown duration of the thermal power unit. This indicates the start-up and shutdown status of the thermal power unit. , Costs associated with starting and stopping thermal power units. , Cost of a single start-up and shutdown of a thermal power unit. , This refers to the ramp rate of the thermal power unit. , These represent the minimum and maximum generating capacities of the thermal power unit, respectively. This refers to the maximum charging and discharging power of the energy storage device.

[0111] This application establishes a two-stage clearing model of random unit combination and scenario-based economic dispatch, which can formulate a safe, reliable and economically efficient day-ahead power generation plan and real-time dispatch scheme under the premise of considering wind speed uncertainty, providing accurate market price signals and system operation status inputs for upper-level decision-making.

[0112] In some embodiments, the step of jointly clearing the electricity market and the frequency regulation ancillary service market specifically includes: taking the capacity demand and mileage demand of the frequency regulation ancillary service market as optimization variables or constraints, and solving them synchronously with the clearing process of the electricity market in the lower-level two-stage joint clearing control model, so as to determine the energy storage system's output plan in the electricity market, the capacity reservation and mileage call plan in the frequency regulation market, and the corresponding market clearing price.

[0113] Specifically, the second phase involves constructing a scenario-based SCED clearing model. After obtaining the thermal power unit output plan and energy storage charging and discharging arrangements determined based on the SCUC model, wind farm operators will further run the SCED clearing model to achieve refined control of wind power output. This phase aims to minimize the operating costs of wind farms, and the specific solution model is shown in the following equation:

[0114] ;

[0115] ;

[0116] ;

[0117] ;

[0118] The constraints are as follows:

[0119] ;

[0120] ;

[0121] ;

[0122] In the formula, For marginal energy cost, This refers to the marginal cost of congestion.

[0123] After the SCED model is completed, its clearing results will be fed back to the upper-level decision-making model; the system frequency deviation will be transmitted to the frequency control module of the ancillary services market for subsequent frequency control strategy formulation and execution.

[0124] This application achieves unified and optimized allocation of resources in the electricity market and the frequency regulation ancillary services market by jointly clearing out the electricity market and the frequency regulation ancillary services market. This avoids resource allocation conflicts and overall efficiency losses caused by market fragmentation, thereby maximizing the comprehensive benefits of flexible resources such as energy storage participating in multiple markets.

[0125] In some embodiments, the step of utilizing the energy storage system for scenario-based adaptive frequency regulation and triggering V2G resources of electric vehicles for coordinated frequency regulation when the energy storage system's regulation capacity is insufficient specifically includes: setting an adaptive droop control coefficient associated with the current wind speed scenario for the energy storage system; monitoring the system frequency deviation in real time, and when the frequency deviation exceeds a preset dead zone range, calculating and issuing a frequency regulation power command for the energy storage based on the adaptive droop control coefficient; calculating the posterior probability of the current wind speed data belonging to each wind speed scenario based on the Gaussian mixture model, evaluating the state of charge of the energy storage system and its over-limit risk probability in future rolling scheduling periods, and triggering the electric vehicle cluster to participate in frequency regulation when the risk probability exceeds a preset risk threshold; and controlling the electric vehicle cluster to operate in a vehicle-to-grid discharge mode or a grid-to-vehicle charging mode according to the direction of system power deficit or surplus, and cooperating with the energy storage system to suppress frequency deviation.

[0126] Specifically, electric vehicles (EVs) participate in the electricity market as a demand-side controllable resource, and their stochastic uplink / downlink electricity capacity is determined by the following formula:

[0127] ;

[0128] In the formula, The number of available EVs. The system sets the rated charging / discharging power for each vehicle. Based on system operating parameters, including wind power output, system load level, and remaining charging / discharging capacity of stationary energy storage, an EV charging / discharging coordination strategy is designed: when wind power output is high and the remaining charging capacity of stationary energy storage is insufficient, the EV is controlled to absorb excess energy in G2V mode; when the system load is high and stationary energy storage cannot meet the grid demand, the EV is controlled to release energy to the grid in V2G mode, thereby improving system regulation flexibility and providing stable backup support for the electricity market.

[0129] To align with probabilistic scenarios, this application employs a risk probability triggering mechanism: within a rolling window, the risk probability of the fixed energy storage SOC reaching the upper / lower limit is assessed based on the scenario's posterior probability. When this probability exceeds a threshold, EV coordination is triggered in advance, thereby avoiding the lag problem of passively adjusting EVs only after the frequency offset has significantly deteriorated.

[0130] This application achieves a leap from fixed parameters and passive response to dynamic adaptation and forward-looking decision-making in frequency regulation by setting a scenario-adaptive droop control coefficient and combining it with a risk warning mechanism based on posterior probability to trigger V2G collaboration. This significantly improves the system's rapid response capability, regulation flexibility, and reliability in the face of power disturbances.

[0131] In some embodiments, the step of feeding back the results of market clearing and coordinated regulation to the upper-level multi-objective optimization model for evaluation and generating the optimal strategy corresponding to each wind speed scenario specifically includes: using the nodal marginal electricity price obtained from the lower-level joint clearing, the actual energy storage power, and the system frequency stability index and carbon emissions calculated after frequency regulation as input parameters and constraints of the upper-level multi-objective optimization model; and in the upper-level multi-objective optimization model, re-evaluating and optimizing the power allocation strategy and scenario-based control parameters based on the feedback information.

[0132] Specifically, after each rolling optimization cycle, the lower-level two-stage joint clearing and control model and frequency control module output a series of key actual operating data. These data include: the nodal marginal electricity price reflecting market supply and demand, the actual power utilized by the energy storage system, the system frequency stability index calculated after frequency control, and carbon emissions calculated based on the actual output of thermal power units. This data is fed back in real time and used as key input parameters and boundary constraints, re-injected into the upper-level multi-objective optimization model. Within the model, based on this feedback information from the actual system, the actual achieved values ​​of the three objectives—total system revenue, frequency stability index, and carbon emissions—are recalculated. Subsequently, multi-objective decision theory and optimization algorithms are used to compare and analyze the feedback results with the expected objectives, thereby re-evaluating and optimizing the power allocation strategy decided by the upper-level model and the adaptive control parameters corresponding to each scenario. This process can be iterated multiple times by setting convergence criteria until the adjustment amount of the power allocation strategy tends to stabilize, thereby ensuring that the strategy generated for each wind speed probability scenario is a Pareto optimal solution or satisfactory solution that has been verified by actual systems and can achieve the best balance between economy, safety and environmental protection.

[0133] This application feeds back the actual effects of lower-level market clearing and frequency regulation to the upper-level model, forming a closed-loop evaluation and optimization circuit. This allows the power allocation strategy to be dynamically adjusted and self-improved based on actual operating results, ensuring the long-term effectiveness and adaptability of the strategy.

[0134] In some embodiments, the step of forming a probability scenario-policy mapping library for invocation specifically includes: for each wind speed scenario, storing its corresponding optimal power allocation policy, adaptive droop control coefficient, and electric vehicle trigger threshold as a complete policy combination to jointly constitute the probability scenario-policy mapping library; calculating the posterior probability of each wind speed scenario based on real-time wind speed data, and directly calling the policy combination corresponding to the wind speed scenario with the highest posterior probability, or weighting and fusing the policy combinations of each wind speed scenario according to the posterior probability to generate the final executed hybrid policy.

[0135] Specifically, in the offline phase, for each typical wind speed probability scenario generated by Gaussian mixture model clustering, the optimal strategy set obtained after closed-loop evaluation and optimization, including the optimal power allocation ratio of the energy storage system between the electricity and frequency regulation markets, the adaptive droop control coefficient of energy storage matching the characteristics of the scenario, and the risk probability threshold for triggering V2G collaboration of electric vehicles, is stored as a complete and indivisible strategy combination. All strategy combinations for all scenarios together constitute a quickly searchable knowledge base, namely the probability scenario-strategy mapping library. During online operation, the system acquires the latest wind speed data in real time and uses the pre-trained Gaussian mixture model to quickly calculate the posterior probability that the current wind speed feature belongs to each typical scenario in the library. Based on this posterior probability, the system adopts two intelligent invocation strategies: one is the dominant scenario invocation strategy, which directly invokes the complete strategy combination corresponding to the typical scenario with the highest posterior probability, suitable for scenarios where the current wind characteristics are very clear; the other is the hybrid strategy generation strategy, which, when the posterior probability distribution is relatively dispersed, uses a weighted approach based on posterior probability to weight and fuse strategy combinations of multiple related scenarios, generating a more robust and smoother hybrid strategy for execution. In addition, to ensure the long-term effectiveness of the mapping library, the system will continuously monitor the drift of wind speed feature distribution. When the model's log-likelihood value is found to be continuously lower than the threshold, the system will automatically trigger the scene re-clustering and the overall update process of the policy library, so that the entire system has the ability to continuously learn and self-evolve.

[0136] In actual power allocation, wind speed characteristics and load forecasts are updated according to rolling time windows, and the strategy is fine-tuned online. When the log likelihood of the Gaussian mixture model decreases significantly or the distribution drift exceeds the threshold, the scenario model is re-evaluated and the strategy is retrained to ensure the long-term effectiveness and robustness of the strategy.

[0137] This application constructs a probabilistic scenario-policy mapping library and supports calling or weighted fusion policies based on posterior probability, enabling optimization results to be directly and quickly applied to online operation. This achieves a seamless transition from offline optimization to online adaptive decision-making, ultimately forming an intelligent operation system with continuous learning and evolution capabilities.

[0138] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an electronic device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0139] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, database, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0140] The above are merely embodiments of the present invention and do not limit the patent scope of the present invention. Any equivalent modifications made based on the content of the present invention's specification and drawings, or direct or indirect applications in related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for coordinated optimization of offshore wind power, energy storage, and vehicle-grid based on wind speed probability scenarios, characterized by the following steps: include: Acquire wind speed data from offshore wind farms and perform preprocessing; The wind speed data is decomposed into trend components and fluctuation components using the variational mode decomposition method. The fluctuation components are then split into positive fluctuation components and negative fluctuation components according to their signs. The statistics of each component are calculated to construct a three-dimensional feature vector. A Gaussian mixture model is used to perform probabilistic clustering on the three-dimensional feature vectors to generate multiple wind speed scene parameters and scene occurrence probabilities. Based on the wind speed scenario parameters and the probability of scenario occurrence, an upper-level multi-objective optimization model is constructed. With the objectives of maximizing the total revenue of the energy storage system, minimizing the frequency stability index, and minimizing carbon emissions, the power allocation strategy of the energy storage system between the electric energy market and the frequency regulation ancillary service market is obtained. Based on the power allocation strategy, a lower-level two-stage joint clearing and control model is constructed. In the first stage, the electric energy market and the frequency regulation ancillary service market are jointly cleared. In the second stage, the energy storage system is used to carry out scenario-based adaptive frequency regulation. When the energy storage system's regulation capacity is insufficient, the electric vehicle V2G resources are triggered to carry out coordinated frequency regulation. The results of market clearing and coordinated regulation are fed back to the upper-level multi-objective optimization model for evaluation, generating the optimal strategy corresponding to each wind speed scenario, and forming a probability scenario-strategy mapping library for use.

2. The offshore wind power-energy storage-vehicle-grid collaborative optimization method based on wind speed probability scenarios as described in claim 1, characterized in that, The step of decomposing the wind speed data into trend components and fluctuation components using the variational mode decomposition method specifically includes: The wind speed data is used as the original signal to be decomposed, and the number of modes after decomposition is set. A variational constraint problem is constructed with the objective of minimizing the sum of the bandwidths of each mode, thereby transforming the decomposition of the wind speed data into an optimization solution process; By introducing the Lagrange penalty operator, the variational constraint problem is transformed into an unconstrained optimization problem in augmented Lagrange form; The unconstrained optimization problem is solved iteratively using the alternating direction multiplier method. In each iteration, the frequency domain estimate, center frequency, and Lagrange multiplier of each modal component are updated alternately until the decomposition result converges. From the converged modal components, based on the center frequency, low-frequency trend components that characterize macroscopic changes in wind resources and high-frequency fluctuation components that characterize instantaneous fluctuations are separated.

3. The offshore wind power-energy storage-vehicle-grid collaborative optimization method based on wind speed probability scenarios as described in claim 2, characterized in that, The step of splitting the fluctuation component into positive and negative fluctuation components according to its sign and calculating the statistics of each component specifically includes: The high-frequency fluctuation component is divided into positive fluctuation component and negative fluctuation component according to the numerical sign corresponding to each time point; Calculate the average value of the low-frequency trend components as a trend characteristic quantity characterizing the overall level of wind resources; Calculate the average value of the positive wave components as a positive wave characteristic quantity characterizing the average strength of the positive disturbance; Calculate the average value of the negative wave components as a negative wave characteristic quantity characterizing the average intensity of the negative disturbance; The trend feature, positive fluctuation feature, and negative fluctuation feature are combined to form the three-dimensional feature vector.

4. The offshore wind power-energy storage-vehicle-grid collaborative optimization method based on wind speed probability scenarios as described in claim 1, characterized in that, The step of using a Gaussian mixture model to perform probabilistic clustering on the three-dimensional feature vector to generate multiple wind speed scene parameters and scene occurrence probabilities specifically includes: Rare / abnormal sample management is performed on the three-dimensional feature vector to remove or reduce the weight of samples whose frequency is lower than the frequency threshold or whose deviation from the main distribution exceeds the deviation threshold; Gaussian mixture model is used to perform probabilistic clustering analysis on the three-dimensional feature vectors after treatment, and the optimal number of scene clusters is adaptively determined by Bayesian information criterion or Akaike information criterion. By fitting the Gaussian mixture model, the Gaussian distribution parameters and the probability of occurrence of each wind speed scenario are obtained. The Gaussian distribution parameters include the feature mean vector, covariance matrix, and mixture weights reflecting the probability of occurrence of each wind speed scenario. The mixture weights represent the prior probability of occurrence of each wind speed scenario. Within each scenario, representative sample sets are selected from the three-dimensional feature vectors after cluster analysis based on the principle of maximizing posterior probability or minimizing Mahalanobis distance, and used as input to the lower-level two-stage joint clearing control model.

5. The offshore wind power-energy storage-vehicle-grid collaborative optimization method based on wind speed probability scenarios according to claim 4, characterized in that, The steps for constructing the upper-level multi-objective optimization model specifically include: Based on the total revenue from the electricity market and the frequency regulation ancillary service market, after subtracting the thermal power operating costs, the system's total revenue is weighted according to scenario probability to obtain the maximum total system revenue. Based on the root mean square value of the system frequency deviation, a frequency stability index that is minimized is obtained by weighting the values ​​according to the scenario probability. Based on the sum of the products of thermal power unit output and carbon emission coefficient, and weighted according to scenario probability, the minimum carbon emission is obtained. By introducing constraints on energy storage power allocation, energy storage state of charge, and energy storage capacity limitation between the electricity market and the frequency regulation market, an upper-level multi-objective optimization model is obtained.

6. The offshore wind power-energy storage-vehicle-grid collaborative optimization method based on wind speed probability scenarios as described in claim 5, characterized in that, The specific steps for constructing the lower-level two-stage joint clearing control model include: A stochastic unit combination model is established to minimize the start-up and shutdown costs, no-load costs, and expected operating costs of all power sources under the wind speed scenario. The start-up and shutdown plan of the thermal power units is determined, and constraints such as system load balancing, network power flow security, minimum start-up and shutdown time of the units, and ramp rate are introduced. Establish a scenario-based safety-constrained economic dispatch model. For each wind speed scenario, with the goal of minimizing real-time operating costs, perform refined allocation of power generation and output the node marginal electricity price, line power flow, and energy storage charging and discharging plan for each wind speed scenario.

7. The offshore wind power-energy storage-vehicle-grid collaborative optimization method based on wind speed probability scenarios according to claim 6, characterized in that, The specific steps for the joint clearing of the electricity market and the frequency regulation ancillary services market in the first phase include: The capacity and mileage requirements of the frequency regulation ancillary service market are used as optimization variables or constraints, and solved synchronously with the clearing process of the electricity market in the lower-level two-stage joint clearing control model to determine the energy storage system's output plan in the electricity market, its capacity reservation and mileage call plan in the frequency regulation market, and the corresponding market clearing price.

8. The offshore wind power-energy storage-vehicle-grid collaborative optimization method based on wind speed probability scenarios according to claim 7, characterized in that, The second stage utilizes the energy storage system for scenario-based adaptive frequency regulation, and triggers coordinated frequency regulation of electric vehicle V2G resources when the energy storage system's regulation capacity is insufficient. The specific steps include: Set an adaptive droop control coefficient for the energy storage system that is relevant to the current wind speed scenario; The system monitors the frequency deviation in real time. When the frequency deviation exceeds the preset dead zone range, the system calculates and issues a frequency regulation power command for energy storage based on the adaptive droop control coefficient. Based on the Gaussian mixture model, the posterior probability of the current wind speed data belonging to each wind speed scenario is calculated, the state of charge of the energy storage system and its over-limit risk probability in the future rolling scheduling period are evaluated, and when the risk probability exceeds the preset risk threshold, the electric vehicle cluster is triggered to participate in frequency regulation. Based on the direction of system power deficit or surplus, the electric vehicle cluster is controlled to operate in either vehicle-to-grid discharge mode or grid-to-vehicle charging mode, working in conjunction with the energy storage system to suppress frequency deviation.

9. The offshore wind power-energy storage-vehicle-grid collaborative optimization method based on wind speed probability scenarios according to claim 1, characterized in that, The step of feeding back the results of market clearing and coordinated regulation to the upper-level multi-objective optimization model for evaluation, and generating the optimal strategy corresponding to each wind speed scenario, specifically includes: The marginal electricity price of the nodes obtained from the joint clearing at the lower level, the actual power of energy storage, and the system frequency stability index and carbon emissions calculated after frequency regulation are used as the input parameters and constraints of the upper-level multi-objective optimization model. In the upper-level multi-objective optimization model, the power allocation strategy and scenario-based control parameters are re-evaluated and optimized based on feedback information.

10. The offshore wind power-energy storage-vehicle-grid collaborative optimization method based on wind speed probability scenarios according to claim 9, characterized in that, The steps for creating a probabilistic scenario-policy mapping library for use specifically include: For each wind speed scenario, its corresponding optimal power allocation strategy, adaptive droop control coefficient, and electric vehicle trigger threshold are stored as a complete strategy combination, which together constitute the probability scenario-strategy mapping library. The posterior probability of each wind speed scenario is calculated based on real-time wind speed data. The strategy combination corresponding to the wind speed scenario with the highest posterior probability is directly called, or the strategy combination of each wind speed scenario is weighted and fused according to the posterior probability to generate the final execution hybrid strategy.

Images