An optimal energy storage mode determination method based on an entropy weight method

Abstract

This invention discloses a method for determining the optimal energy storage mode based on the entropy weight method, relating to the field of energy storage technology. The method includes the following steps: collecting historical data from wind turbines and the power grid; preprocessing this data using a sliding window to obtain a structured dataset; constructing a candidate feature set by combining battery energy storage characteristics; and using mutual information to filter out the feature set containing wind curtailment rate; establishing a wind speed prediction model; inputting wind speed data to obtain a prediction sequence; and then calculating the wind power output prediction sequence; constructing simulation models for lithium batteries, vanadium redox flow batteries, and flywheel energy storage using the backward Eulerian method, ion concentration field calculation method, and electromechanical state equation, respectively; and calculating each selected feature set based on the wind power output prediction results; improving the entropy weight method based on the weighted average vector optimization algorithm; and combining it with the analytic hierarchy process (AHP) to calculate the comprehensive score of each energy storage mode, comparing and determining the optimal method. This invention, based on the entropy weight method, can solve the problem of energy storage scheme adaptation deviation caused by the static solidification of weights in wind-storage synergy.

Description

Technical Field

[0001] This invention relates to the field of method determination engineering, specifically to a method for determining the optimal energy storage method based on the entropy weight method. Background Technology

[0002] As a core facility for renewable energy consumption and grid stability, energy storage devices directly impact energy efficiency, power supply reliability, and economic benefits. Industry data shows that my country's installed wind power capacity has exceeded 400GW, with a wind energy penetration rate of 11.2%, but the average annual wind curtailment rate remains as high as 8.5%, meaning a single fluctuation event can cause economic losses in the tens of millions of yuan. With the continuous expansion of wind power grid connection and the intensification of peak-valley load fluctuations, determining the optimal energy storage method has become an urgent requirement for the national energy security strategy.

[0003] Current methods for determining energy storage methods primarily rely on manually preset fixed weight parameters to apply a uniform standard to linearly weighted scoring of different types of solutions, such as lithium batteries, flow batteries, and flywheel energy storage. However, the main problem with these methods is that they ignore the differentiated characteristics of energy storage technologies. Core advantages such as the high energy density of lithium batteries, the long lifespan of flow batteries, and the fast response of flywheel energy storage are not quantitatively modeled, leading to an evaluation system that deviates from engineering realities and poses a serious threat to grid stability and efficient energy utilization. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method for determining the optimal energy storage method based on the entropy weight method, which solves the problem that the differentiated characteristics of existing energy storage technologies are ignored.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for determining the optimal energy storage method based on the entropy weight method, comprising the following steps: Step S1: Collect historical data of wind turbines and historical data of the power grid. Preprocess the historical data of wind turbines and historical data of the power grid using the sliding window method to obtain a structured dataset. Based on the characteristics of battery energy storage, construct a candidate feature set. Filter the candidate feature set using the mutual information feature filtering method to obtain a filtered feature set. The filtered feature set includes wind curtailment rate, power supply loss rate, and wind-storage complementarity rate. Step S2: Based on the autoregressive integral moving average model, construct a wind speed prediction model, input the wind speed data in the structured dataset into the wind speed prediction model to obtain a wind speed prediction sequence, and calculate the wind power output prediction sequence based on the wind speed prediction sequence. Step S3: Discretely model the lithium battery energy storage element using the backward Euler method and the constant admittance discrete adjoint model method to obtain the lithium battery energy storage simulation model; based on the lithium battery energy storage simulation model and the wind power output prediction sequence, calculate the screening feature set of lithium battery energy storage. Step S4: Model the vanadium redox flow battery using the ion concentration field calculation method to obtain the vanadium redox flow battery energy storage simulation model; based on the vanadium redox flow battery energy storage simulation model and the wind power output prediction sequence, calculate the screening feature set of vanadium redox flow battery energy storage. Step S5: Model the flywheel motor in flywheel energy storage using electromechanical state equations to obtain a flywheel energy storage simulation model; based on the flywheel energy storage simulation model and wind power output prediction values, calculate the selection feature set for flywheel energy storage; Step S6: Based on the vector weighting mechanism of the weighted average vector optimization algorithm, the static defects of the entropy weight method are improved by dynamically adjusting the weights, resulting in the improved entropy weight method; Step S7: The improved entropy weight method combined with the analytic hierarchy process (AHP) is used to calculate the screening feature sets for lithium battery energy storage, vanadium redox flow battery energy storage, and flywheel energy storage, respectively, to obtain the comprehensive scores for lithium battery energy storage, vanadium redox flow battery energy storage, and flywheel energy storage. By comparing the comprehensive scores of lithium battery energy storage, vanadium redox flow battery energy storage, and flywheel energy storage, the optimal energy storage method is determined.

[0006] Preferably, historical data of wind turbines and historical data of the power grid are collected, and the historical data of wind turbines and historical data of the power grid are preprocessed using the sliding window method to obtain a structured dataset including: The power grid dispatch center retrieves historical data from wind turbines and the power grid through a dedicated data interface. The historical data from wind turbines includes wind speed, real-time power, and start / stop status; the historical data from the power grid includes load demand, voltage frequency, and grid-connected power. The data is encrypted and transmitted to the local server, where it is stored after being sorted by timestamp. The data preprocessing process is as follows: Sliding window filtering of neighboring points: In the temporal alignment stage of data preprocessing, the sliding window method is used to provide a neighboring point filtering framework for Lagrange interpolation; Missing value imputation: For missing points in historical data, Lagrange interpolation is used for imputation. An interpolation function is constructed using four known data points near the missing time point. Based on the spatiotemporal relationship and observation values ​​of the four neighboring points, the imputation value is calculated through a weighted combination. The generated imputation value will completely restore the historical data value at that moment, resulting in the imputed historical data. Standardization process:

[0007] in, is the standardized eigenvalue, a dimensionless variable whose function is to eliminate dimensional differences; x is the original eigenvalue. It is the feature mean, representing the historical mean of a certain feature; It is the standard deviation of the feature, reflecting the range of fluctuation of the feature; Output structured dataset .

[0008] Preferably, the candidate feature set is filtered using a mutual information feature filtering method to obtain a filtered feature set including: Based on the characteristics of battery energy storage, namely, comprehensively considering the core performance parameters of the energy storage system during charging and discharging, such as power response speed, energy density, and cycle life, and combining its dynamic adaptation characteristics when operating in conjunction with wind farms, such as the ability to smooth wind power fluctuations and the effect of regulating load peaks and valleys, a candidate feature set C={ is constructed to comprehensively reflect the system-level response and collaborative performance of energy storage devices under actual operating conditions. , , }; Structured datasets column vectors in Through feature transformation function () Generate candidate features:

[0009] in, These are candidate features; h is the total number of parameters; It is a structured dataset The column vectors in the model; m is the total number of column vectors; the candidate features include the original parameters: wind speed fluctuation. Change rate, etc.; composite indicators: wind curtailment rate, power supply loss rate, wind-storage complementarity rate; The feature selection process described above is as follows: enter The candidate feature variables in the data are analyzed using mutual information measurement technology to quantify the statistical correlation between battery energy storage parameters and power supply quality indicators, thereby enabling the extraction of high-contribution features. Discrete variables:

[0010] Wherein, the independent variable X is Candidate feature variables; the target variable Y is the power supply stability index; It represents the correlation between discrete variable X and target variable Y, and its function is to quantify the statistical correlation between battery energy storage parameters and power supply quality indicators. It represents the discrete values ​​of the characteristic variable X, which are actually the set of wind turbine states; It represents the discrete values ​​of the target variable Y, which are actually the power grid frequency state; The joint probability is the joint probability of the grid frequency being at various frequencies when the wind turbine is in various states; It is the marginal probability, representing the probability of the wind turbine being in various states; It is the marginal probability, representing the probability of the power grid operating at various frequencies; Continuous variables:

[0011] in, It represents the correlation between the continuous variable X and the target variable Y; p(x) is the probability density function; Feature selection, based on mutual information values, involves classifying feature relevance levels as follows: MI is obtained as described above. and When MI is less than 0.4, the feature correlation level is weak and it is directly removed; when MI is greater than 0.4 and less than 0.7, the feature correlation level is moderate and it can be retained for auxiliary analysis; when MI is greater than 0.7, the feature correlation level is strong and it can be used as a core decision feature. With the stability of power supply from wind-storage power generation as the optimization objective, the candidate feature set C={ , , Mutual information values ​​with key indicators: Calculate the MI value of each parameter in relation to the power supply quality index; By setting a threshold τ=0.6, features with I(X,Y)≥τ are filtered to obtain wind curtailment rate, power loss rate, and wind-storage complementarity rate. Output the filtered feature set F:

[0012] in, It refers to the rate of curtailment of new energy sources, specifically the wind curtailment rate. = Curtailed wind power / Theoretical power generation × 100%; It is the power supply loss rate. =Power shortage / Total load demand × 100%; It refers to the wind-storage complementarity ratio, or simply wind-storage complementarity ratio. =Energy storage power supply / Total power supply × 100%.

[0013] Preferably, the calculated wind power output prediction sequence includes: Calculate wind power output, model wind turbine power characteristics, and input wind speed prediction sequence; Segmented output model :

[0014] in, It is the maximum power output that the wind turbine can continuously and stably generate at rated wind speed; v is the wind power forecast value, derived from... ; It is the wind turbine start-up threshold; It is the rated wind speed; u is the aerodynamic characteristic coefficient of the fan; This is the mechanical limit of the wind turbine; Output wind power prediction sequence :

[0015] Wind power output prediction sequence It serves as the unified input for S3-S5 energy storage simulations.

[0016] Preferably, based on the lithium battery energy storage simulation model and the wind power output prediction sequence, the selected feature set for lithium battery energy storage includes: Based on S2 output And lithium battery energy storage simulation model, calculate and screen feature set { , , } Lithium battery energy storage wind curtailment rate :

[0017] in, It is the wind curtailment rate of lithium battery energy storage, which is the ratio of the difference between the theoretical power generation of the wind turbine and the actual grid-connected power. Its function is to reflect the ability of the lithium battery energy storage device to absorb fluctuating wind power. The lower the value, the better the regulation effect of lithium battery energy storage. It represents the actual grid-connected wind power output and the simulated output of lithium battery energy storage. It is the theoretical maximum wind power, calculated from the wind speed-power curve; Lithium battery energy storage power supply loss rate :

[0018] in, It is the lithium battery energy storage power supply loss rate, which is the percentage difference between the load demand power and the actual power supplied. It is the power demand of the power grid, which is estimated from historical data; It is the grid-connected output power of the lithium battery energy storage device, and the simulated output of the lithium battery energy storage. Lithium-ion battery energy storage wind-storage complementarity :

[0019] in, It refers to the complementarity ratio of lithium battery energy storage to wind and energy storage, which is the proportion of power supplied by wind power and lithium battery energy storage in a coordinated manner, maximizing... To improve the utilization rate of renewable energy; It is the theoretical output of the wind turbine, derived from the wind power output prediction sequence; It is the total load, which comes from the real-time metering system of the power grid dispatch center.

[0020] Preferably, based on the vanadium redox flow battery energy storage simulation model and the wind power output prediction sequence, the selection feature set for vanadium redox flow battery energy storage is calculated to include: Based on S2 output And a simulation model of vanadium redox flow battery energy storage, calculating characteristic variables { , , }; Vanadium redox flow battery energy storage wind curtailment rate :

[0021] in, It is the wind curtailment rate of vanadium redox flow battery energy storage, which is the ratio of the difference between the theoretical power generation of the wind turbine and the actual grid-connected power. Its function is to reflect the ability of the vanadium redox flow battery energy storage device to absorb fluctuating wind power. The lower the value, the better the energy storage regulation effect. It is the theoretical maximum wind power, calculated from the wind speed-power curve; It represents the actual grid-connected wind power output and the simulated output of vanadium redox flow battery energy storage. Vanadium redox flow battery energy storage power supply loss rate :

[0022] in, It is the power supply loss rate, which is the percentage difference between the load demand power and the actual power supplied. It is the power demand of the power grid, which is estimated from historical data; It is the grid-connected output power of the vanadium redox flow battery energy storage device, and the simulated output of the vanadium redox flow battery energy storage. Vanadium redox flow battery energy storage wind-storage complementarity :

[0023] in, It refers to the wind-storage complementarity ratio of vanadium redox flow battery energy storage, which is the proportion of power supplied by wind power and vanadium redox flow battery energy storage in a coordinated manner, maximizing... To improve the utilization rate of renewable energy; It is the theoretical output of the wind turbine, derived from the wind power output prediction sequence; It is the total load, which comes from the real-time metering system of the power grid dispatch center.

[0024] Preferably, based on the flywheel energy storage simulation model and the predicted wind power output, the selected feature set for flywheel energy storage includes: Based on S2 output And flywheel energy storage simulation model, calculate characteristic variables { , , }; Flywheel energy storage wind curtailment rate :

[0025] in, It is the flywheel energy storage wind curtailment rate, which is the ratio of the difference between the theoretical power generation of the wind turbine and the actual grid-connected power. Its function is to reflect the ability of the flywheel energy storage device to absorb fluctuating wind power. The lower the value, the better the energy storage regulation effect. It is the theoretical maximum wind power, calculated from the wind speed-power curve; It represents the actual grid-connected wind power output and the simulated output of flywheel energy storage. Flywheel energy storage power supply loss rate :

[0026] in, It is the flywheel energy storage power supply loss rate, which is the percentage of the difference between the load demand power and the actual power supplied. It is the power demand of the power grid, which is estimated from historical data; It is the grid-connected output power of the flywheel energy storage device, and the simulated output of the flywheel energy storage. Flywheel energy storage wind-storage complementarity :

[0027] in, It refers to the flywheel energy storage-wind-storage complementarity ratio, which is the proportion of power supplied by wind power and flywheel energy storage in a coordinated manner, maximizing... To improve the utilization rate of renewable energy; It is the theoretical output of the wind turbine, derived from the wind power output prediction sequence; It is the total load, which comes from the real-time metering system of the power grid dispatch center.

[0028] Preferably, the vector weighting mechanism based on the weighted average vector optimization algorithm improves the static defects of the entropy weight method by dynamically adjusting the weights, resulting in an improved entropy weight method including: Standard entropy weighting :

[0029] Where n is the total number of data points; b is a temporary index used to traverse all features, and the weights are from the traditional entropy weight method. A fixed approach cannot adapt to the dynamic changes in the importance of indicators caused by the cyclical fluctuations in wind power, such as the wind curtailment rate being more important during peak wind power periods. The vector weighting mechanism of the INFO algorithm: Vector weighted mean definition:

[0030] in, It is a random weighting factor, and its function is to dynamically balance the two mean structures; and These are mean vector 1 and mean vector 2; Improved implementation process of the entropy weight method: Using an improved INFO algorithm to generate time-sensitive weights :

[0031]

[0032] in, m is the input vector; Quantity; The input weights are α, which is obtained from the above improvements; t corresponds to the wind power fluctuation period; the INFO operator dynamically adjusts the weights using a vector weighted average. It is the standard entropy weight method weight; It is the feature difference penalty weight; Divide the time period according to the wind power fluctuation cycle and match the weights. :

[0033] in, It is a fluctuation cycle. =4h; each time period Internally adopts independent ; Combine static entropy weights and dynamic weights to calculate the combined weights. :

[0034]

[0035] Where T is the total number of time periods; It is a static weighting percentage. ∈[0.4, 0.6], take =0.5; It is a static weight; It is the dynamic weight for the current time period.

[0036] Preferably, the comprehensive scores for lithium battery energy storage, vanadium redox flow battery energy storage, and flywheel energy storage include: Subjective weight Solution:

[0037]

[0038]

[0039] Where A is the judgment matrix; It is an eigenvector, the vector corresponding to the largest eigenvalue; It is the largest eigenvalue; n is the matrix order; It determines the elements of a matrix; It is the geometric mean; Combined weights synthesis:

[0040] in, It is a combined weight; verified through sensitivity analysis, when When the ratio is 0.5, the combined weights exhibit the best adaptability to wind power fluctuation cycles. Overall score of the plan : Quantitative indicators are converted into standardized scores using cost-based and benefit-based membership functions. ∈[0,1], together with the qualitative index transformation value, participates in the comprehensive score. calculate:

[0041] in, It is the overall score of scheme k; is the membership degree of index j to scheme k; n is the number of features in the selected feature set; j is the feature index; It is a comprehensive score for lithium battery energy storage; It is a comprehensive score for all-vanadium redox flow battery energy storage; It is a comprehensive score for flywheel battery energy storage.

[0042] Preferably, the optimal energy storage method is determined by comparing the comprehensive scores of lithium battery energy storage, vanadium redox flow battery energy storage, and flywheel energy storage, including: Optimal decision rule: Select the energy storage solution with the highest overall score. Scheme ranking:

[0043] Where k is the energy storage scheme type index; It is the comprehensive score of the k-th scheme; It is to take The largest k value; It is the index of the optimal solution that is ultimately selected.

[0044] This invention provides a method for determining the optimal energy storage mode based on the entropy weight method, involving machine learning and deep learning technologies, and has the following beneficial effects: (1) The optimal energy storage method based on the entropy weight method overcomes the bottleneck of adaptation accuracy in energy storage selection through the feature-weight-decision triple coupling mechanism, namely, the elimination of redundant interference from the strongly correlated feature set, so that the decision focuses on the differentiated characteristics of energy storage technology, and quantifies and models the core advantages of lithium battery high energy density, flow battery long life and flywheel energy storage fast response, ensuring fair competition among the three energy storage technologies.

[0045] (2) The optimal energy storage method based on the entropy weight method uses a high-precision wind power prediction sequence with a fitting error of ≤3.2% to drive a multi-energy storage collaborative model. It adopts minute-level sliding window joint simulation technology to output the optimized output curves of lithium battery, flow battery and flywheel energy storage. Combined with dynamic combination weights, it achieves three objectives: reducing wind curtailment rate, improving power supply reliability and achieving millimeter-level balance with optimal investment cost.

[0046] (3) The optimal energy storage method based on the entropy weight method achieves a leap in the spatiotemporal adaptability of energy storage scheme evaluation through dynamic weight adjustment and multi-dimensional feature fusion. The improved entropy weight method combines the dynamic adjustment of index weights with the wind power fluctuation cycle, automatically increasing the weight of wind curtailment rate during peak wind power periods and strengthening the wind-storage complementarity ratio during off-peak periods, which greatly shortens the weight response delay and significantly improves the accuracy of adaptation to transient wind power fluctuations. Attached Figure Description

[0047] Figure 1 This is a flowchart of a method for determining the optimal energy storage mode based on the entropy weight method proposed in this invention.

[0048] Figure 2 This invention provides a hierarchical diagram of the selection feature sets for three energy storage methods based on the entropy weight method, which is proposed as an optimal energy storage method in this invention.

[0049] Figure 3 This is a hierarchical diagram of the optimal energy storage method based on the entropy weight method proposed in this invention for determining the optimal energy storage mode. Detailed Implementation

[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] Please see Figure 1 This invention provides a technical solution: a method for determining the optimal energy storage method based on the entropy weight method. Specifically, the method for determining the optimal energy storage method based on the entropy weight method is provided below. Figure 1 The method includes the following steps: Step S1: Collect historical data of wind turbines and historical data of the power grid. Preprocess the historical data of wind turbines and historical data of the power grid using the sliding window method to obtain a structured dataset. Based on the characteristics of battery energy storage, construct a candidate feature set. Filter the candidate feature set using the mutual information feature filtering method to obtain a filtered feature set. The filtered feature set includes wind curtailment rate, power supply loss rate, and wind-storage complementarity rate.

[0052] The power grid dispatch center retrieves historical data of wind turbines (including wind speed, real-time power, start-stop status, etc.) and historical data of the power grid (including load demand, voltage frequency, grid-connected power, etc.) through a dedicated data interface; the data is encrypted and transmitted to the local server, and stored after being sorted by timestamp; the data integrity is automatically verified, and missing or abnormal data is marked, providing the raw dataset for subsequent sliding window preprocessing.

[0053] The data preprocessing process is as follows: Sliding window filtering of neighboring points: In the temporal alignment stage of data preprocessing, the sliding window method is used to provide a neighboring point filtering framework for Lagrange interpolation. Define a time window centered on the missing point; search for the four nearest known data points within the window; if there are fewer than four points in the window, extend it to 120 seconds; if there are more than four adjacent points, select the four points with the smallest time interval.

[0054] Missing value imputation: For missing points in historical data, Lagrange interpolation is used for imputation. An interpolation function is constructed using four known data points near the missing time point. The missing value is calculated by weighted combination based on the spatiotemporal relationship of the four neighboring points and the observed values.

[0055] The generated filler values ​​will completely restore the historical data values ​​at that moment, resulting in the filled historical data.

[0056] Standardization process:

[0057] in, is the standardized eigenvalue, a dimensionless variable whose function is to eliminate dimensional differences; x is the original eigenvalue. It is the feature mean, representing the historical mean of a certain feature; It is the characteristic standard deviation, which reflects the range of fluctuation of the characteristic.

[0058] Output structured dataset .

[0059] Based on the characteristics of battery energy storage, namely, comprehensively considering the core performance parameters of the energy storage system during charging and discharging, such as power response speed, energy density, and cycle life, and combining its dynamic adaptation characteristics when operating in conjunction with wind farms, such as the ability to smooth wind power fluctuations and the effect of regulating load peaks and valleys, a candidate feature set C={ is constructed to comprehensively reflect the system-level response and collaborative performance of energy storage devices under actual operating conditions. , , }

[0060] Structured datasets column vectors in Through feature transformation function () Generate candidate features:

[0061] in, These are candidate features; h is the total number of parameters; It is a structured dataset The column vectors in the model; m is the total number of column vectors. Candidate features include the original parameters: wind speed fluctuation. Wind curtailment rate, power supply loss rate, wind-storage complementarity rate, etc.

[0062] The feature selection process described above is as follows: enter Candidate feature variables in.

[0063] Using mutual information measurement technology, the statistical correlation between battery energy storage parameters and power supply quality indicators is quantified to achieve high-contribution feature extraction. Discrete variables (such as fan status flags, protection action signals, etc.):

[0064] Wherein, the independent variable X is Candidate feature variables; the target variable Y is the power supply stability index; It represents the correlation between discrete variable X and target variable Y, and its function is to quantify the statistical correlation between battery energy storage parameters and power supply quality indicators. It represents the discrete values ​​of the characteristic variable X, which are actually the set of wind turbine states; It represents the discrete values ​​of the target variable Y, which are actually the power grid frequency state; The joint probability is the joint probability of the grid frequency being at various frequencies when the wind turbine is in various states; It is the marginal probability, representing the probability of the wind turbine being in various states; It is the marginal probability, representing the probability that the power grid is at various frequencies.

[0065] Continuous variables (such as wind speed fluctuations, SOC change rate, wind curtailment rate, power supply loss rate, wind-storage complementarity rate, etc.):

[0066] in, It represents the correlation between the continuous variable X and the target variable Y; p(x) is the probability density function.

[0067] Feature selection, based on mutual information values, involves classifying feature relevance levels as follows: MI is obtained as described above. and When MI is less than 0.4, the feature correlation level is weak and it is directly removed; when MI is greater than 0.4 and less than 0.7, the feature correlation level is moderate and it can be retained for auxiliary analysis; when MI is greater than 0.7, the feature correlation level is strong and it can be used as a core decision feature.

[0068] With the stability of power supply from wind-storage power generation as the optimization objective, the candidate feature set C={ , , Mutual information values ​​with key indicators: Calculate the MI value of each parameter in relation to the power supply quality index; A threshold τ = 0.6 is set to filter features where I(X, Y) ≥ τ. The resulting parameters are wind curtailment rate, power loss rate, and wind-storage complementarity rate.

[0069] Output the filtered feature set F:

[0070] in, It refers to the rate of curtailment of new energy sources, specifically the wind curtailment rate. = Curtailed wind power / Theoretical power generation × 100%; It is the power supply loss rate. =Power shortage / Total load demand × 100%; It refers to the wind-storage complementarity ratio, or simply wind-storage complementarity ratio. =Energy storage power supply / Total power supply × 100%.

[0071] This step completed the preprocessing and feature engineering of the wind-storage system data. Time series were aligned using sliding window technology, missing historical data were filled using Lagrange interpolation, and Z-score standardization was used to eliminate dimensional differences, thus constructing a structured dataset. Based on mutual information, three strongly correlated features—wind curtailment rate, power loss rate, and wind-storage complementarity rate—were selected to form a feature set. This step provides high-quality input for subsequent wind speed prediction (S2), establishes core evaluation dimensions for energy storage simulation (S3-S5), and lays the index foundation for comprehensive evaluation (S7), serving as the data cornerstone for determining energy storage methods.

[0072] Step S2: Based on the autoregressive integral moving average model, construct a wind speed prediction model, input the wind speed data in the structured dataset into the wind speed prediction model to obtain a wind speed prediction sequence, and calculate the wind power output prediction sequence based on the wind speed prediction sequence.

[0073] The wind speed prediction model uses an autoregressive integral moving average (ARIMA) model to predict time-series wind speed data. Input wind speed history sequence V, from a structured dataset Extracted and obtained:

[0074] Where T is the number of parameters; It is the speed of historical winds.

[0075] First-order difference transform, performing difference operations based on d=1:

[0076] in, It is the value after first-order difference, which is a stable wind speed sequence that eliminates trends; The original wind speed at time t; T is the total number of time points, i.e., the total amount of data; thus, the stationary sequence Y={ , ,..., }

[0077] ARIMA(2,1,1) model fitting, difference equation as follows:

[0078] in, It is an autoregressive coefficient of order 2, which is used to quantify the correlation between the current wind speed and the previous moment. It is a weighted value with a lag of 2 periods and an order of 2, which quantifies the correlation between the current wind speed and the previous two time points. It was the wind speed at the previous moment; These are the wind speeds at the first two moments; It is the moving average coefficient, with an order of 1, which corrects for the prediction bias of the previous time step. It is a white noise sequence, consisting of unpredictable transient fluctuations. ; It is the intensity of white noise.

[0079] Parameter estimation is performed by solving for the coefficients using maximum likelihood estimation.

[0080] in, It was the wind speed at the previous moment; The wind speeds are the wind speeds at the first two moments; f() is the normal probability density function. It is a white noise sequence.

[0081] Verify model quality and calculate AIC value:

[0082] in, These are the optimal parameters, and their function is to determine the minimum optimal parameters. The model with the minimum AIC is the model that should be selected. It is the total number of parameters. =p+q=3; It is the maximum value of the model likelihood function, which measures the degree of data fit; It is the logarithm of the maximum likelihood value, and its function is to quantify the explanatory power of the model.

[0083] The output is a wind speed prediction model:

[0084] Input the real-time wind speed data collected by the sensor.

[0085] Output: Generate a wind speed forecast sequence for the next 24 hours.

[0086] in, This is a wind speed prediction sequence.

[0087] Wind power output calculation, wind turbine power characteristic modeling, input wind speed prediction value v, from .

[0088] Segmented output model :

[0089] in, It is the maximum power output that the wind turbine can continuously and stably generate at rated wind speed; v is the wind power forecast value, derived from... ; It is the wind turbine start-up threshold. =3m / s; That is the rated wind speed. =12m / s; u is the aerodynamic characteristic coefficient of the wind turbine, the value of which is determined by the tip speed ratio and the blade pitch angle, with a typical range of 1.8–2.3, and is adjusted according to the aerodynamic characteristics; It is the mechanical limit of the wind turbine. =25m / s.

[0090] Output wind power prediction sequence :

[0091] Wind power output prediction sequence It serves as the unified input for S3-S5 energy storage simulations.

[0092] This step utilizes the ARIMA model to predict wind speed and calculate wind power output. First, the historical wind speed sequence is processed using first-order differencing to eliminate trends and construct a stationary sequence. The autoregressive coefficients and moving average coefficients are then solved using maximum likelihood estimation, and the model quality is verified. The fitted model outputs a predicted wind speed sequence for the next 24 hours, which is then combined with the wind turbine power characteristic curve to calculate the theoretical wind power output sequence. This prediction result provides a unified input benchmark for the subsequent three energy storage simulations (S3-S5), directly impacting the simulation accuracy of core indicators such as wind curtailment rate and power supply reliability.

[0093] Step S3: Discretely model the lithium battery energy storage element using the backward Euler method and the constant admittance discrete adjoint model method to obtain the lithium battery energy storage simulation model; based on the lithium battery energy storage simulation model and the wind power output prediction sequence, calculate the screening feature set of lithium battery energy storage.

[0094] The backward Euler method is used as a numerical integration method to discretize and model dynamic components in a circuit.

[0095] The modeling process for capacitor components is as follows: Differential equation of a capacitor element:

[0096] Solve for the voltage value at time t using the backward Euler method:

[0097] After combining the two formulas above and rearranging them, we can obtain:

[0098] in, It is the current value at time t; It is a simulated step size; yes The voltage value at that moment; is the voltage value at time t; C is the capacitance value. This formula discretizes the capacitor element into values ​​of . electrical conductivity With value Discrete equivalent model of the historical current source J(t) in parallel.

[0099] The basic passive components include resistors R, inductors L, and capacitors C. To reduce the number of nodes in the network, series and parallel combinations of basic units such as RL, RC, and LC are treated as a single branch unit. After discretization using the backward Euler method, the general form of its equivalent model can be obtained:

[0100]

[0101] in, It is the current value at time t; It is the current source component in the equivalent circuit that is determined by the historical state; for The coefficient of the voltage value at time, for The coefficient of the current value at any given time. , and The value depends on the admittances of R, L, and C, and how they are combined. For example, for element R, =1 / R, =0, =0; for the combined circuit RC of R and C, = , = , = , = .

[0102] The modeling method for switching elements, using a discrete adjoint model combined with the backward Euler method, can also represent all switching elements in the circuit as equivalent to a parallel historical current source with conductance: The equivalent equations for the capacitor and inductor after discretization are:

[0103] in, The equivalent conductance of the switch; The voltage across the switch; The historical current source of the switch; This represents the current flowing through the switch.

[0104] For capacitors:

[0105] in, It is the equivalent conductance of a capacitor, and its function is to convert the capacitance into equivalent resistance characteristics; It is the switching equivalent capacitance, typically 10nF; It is the equivalent current source of the capacitor, and its function is to replace the dynamic current characteristics of the capacitor. It represents the capacitor voltage at a historical moment, and its function is to provide a memory of the voltage at the previous time step.

[0106] For inductors:

[0107] in, It is the equivalent conductance of an inductor, and its function is to convert the inductance into the equivalent resistance characteristic; It is the switching equivalent inductance (typical value 1μH); It is an inductor equivalent current source, and its function is to replace the dynamic current characteristics of an inductor. It is the inductor current at a historical moment, providing current memory of the previous time step.

[0108] If the selected switch parameters , and time step satisfy:

[0109] That is, only the value of the current source is related to the switching state S, while the conductance value is independent of the switching state S, then the switching equation... for:

[0110] in, It is an equivalent constant admittance; It is the inductor current at a historical moment, activated when S=1, representing the energy memory of the inductor in the closed state; It is the capacitor voltage at a historical moment. It is activated when S=0, representing the energy memory of the capacitor in the open state.

[0111] Due to the limitations of the above formula, the admittance matrix H becomes a constant, allowing for rapid matrix calculations. Otherwise, especially in more complex circuits, computational efficiency would be significantly reduced. Furthermore, considering the essential characteristics of the switching states of switching devices, smaller L and C values ​​result in smaller modeling errors. Therefore, a discrete adjoint model is chosen to simulate the switching element.

[0112] A three-phase voltage source full-bridge inverter with an LCL filter is used to achieve AC-DC conversion, and a current control strategy is used to control the inverter.

[0113] To study lithium-ion battery energy storage in more detail, we first decoupled the capacitor, which has relatively weak electrical characteristics, dividing the entire energy storage device into a DC side and an AC side, and then conducted modeling and simulation studies on each side. The DC side includes the lithium-ion battery and the Buck / Boost converter, while the AC side includes the inverter, LCL filter, load, and AC voltage source.

[0114] Buck / Boost converter: A Buck / Boost converter increases or decreases the input voltage based on a given duty cycle (D): when D is greater than 0.5, it enters Boost mode, resulting in a voltage increase in the circuit; when D is less than 0.5, it enters Buck mode, resulting in a voltage decrease in the circuit. The mathematical relationship between output voltage and input voltage is as follows: When in Boost mode:

[0115] When in Buck mode:

[0116] in, Input voltage; This refers to the output voltage. For switching transistors Duty cycle; For switching transistors Duty cycle; .

[0117] Closed-loop control equations for a Buck / Boost converter: PI regulation of the voltage outer loop and current inner loop:

[0118] in, It is the voltage error value, reflecting the deviation between the actual output and the target; This is the reference voltage, representing the target voltage value that is expected to be achieved. It is the actual output voltage, the output voltage value detected in real time by the sensor; It is the voltage loop proportional coefficient, and its function is to quickly respond to voltage deviations; It is the voltage loop integral coefficient, and its function is to eliminate steady-state error; It is the voltage error integral term, and its function is to accumulate historical deviations over time. It is a current reference command, which is the desired current value output to the inner loop; It is the actual inductor current, and its function is to detect the value in real time through a current sensor. It is the current error value, which reflects the deviation between the actual current and the command. It is the proportional coefficient of the current loop, and its function is to quickly respond to current deviation; It is the integral coefficient of the current loop, and its function is to eliminate the steady-state error of the current. It is the integral term of current error, and its function is to accumulate historical current deviations.

[0119] Voltage loop parameters and Determined by frequency domain scanning:

[0120]

[0121] in, It is the crossover frequency (taken as 1 / 10 of the switching frequency), which is the frequency at which the open-loop gain drops to 1; It is the bus capacitor, representing the capacitance value of the DC bus capacitor; It is the integration time constant, take =0.1s, its function is to eliminate the speed adjustment factor of steady-state error.

[0122] DC / AC converter (inverter): In the current control strategy of three-phase inverters, PI regulation, dq coordinate transformation, and pulse width modulation (PWM) techniques are used. PI regulation is relatively mature, simple in structure, robust, and accurate, thus becoming the mainstream regulation technique in control systems. To simplify the control scheme, a dynamic average model of the three-phase inverter is obtained using a dq reference frame. This transformation is very beneficial because decoupling control reduces the dimension from three to two. Among pulse width modulation techniques, sinusoidal pulse width modulation (SPWM) is the most common and popular PWM technique due to its ease of implementation and good performance.

[0123] To improve stability during grid faults, a low-voltage ride-through (LVRT) control strategy is specifically configured:

[0124] in, It is an active current command; It is a reactive current command; This is the rated current of the energy storage converter; It is the proportionality coefficient; It is the integral coefficient; It is the d-axis current error; It is the q-axis current error; It is the rated voltage; That is the rated current.

[0125] The control logic detects the grid voltage in real time. Implement mode switching: when At the same time, dual closed-loop PI control is used to ensure accurate power transmission; when Automatic execution at that time: set up Provides reactive power support, forcibly increasing the reactive current to 110% of the rated value.

[0126] according to Active current limiting and automatic active current limiting protection equipment.

[0127] After the fault is cleared, there is a 500ms delay before returning to normal mode.

[0128] In a VSI control system, the three-phase voltage in the circuit , , It can be represented as:

[0129] Where E is the peak value of the grid phase voltage, which is 311V. The angular frequency of the power grid is 100π.

[0130] The first transformation is to convert the three-phase voltage into two decoupled phase voltages. and Relationship:

[0131] The second transformation is to decouple the two phase voltages. and Converted into two rotations and This transformation is called the αβ-to-dq transformation, and its relation is:

[0132] in, It is a functional component; It is the reactive component.

[0133] For open-loop control of inverters, the inverter output can be directly controlled using PWM technology. PWM is achieved by adjusting the ON and OFF cycles of the inverter's switching devices. Among all PWM techniques, SPWM is the most reliable because it directly determines the frequency of the inverter output waveform based on the waveform function of the control signal.

[0134] The SPWM modulation process is as follows: The modulated wave is compared with the carrier wave to generate a switching signal:

[0135] in, It is a triangular carrier wave. It is a modulated wave.

[0136] In this control method, the pulse amplitude is constant, while the duty cycle varies over time. By modulating the pulse width, the inverter's output voltage is controlled, and the total harmonic distortion (THD) can be reduced. The switching device is ON when the reference sine wave is greater than the carrier triangular wave, and OFF when the reference sine wave is less than the carrier triangular wave. In a three-phase voltage source inverter, three sine waves are required as reference signals, and their phase shifted by 120° according to the desired output voltage frequency.

[0137] filter: Because the output voltage and current of the inverter may contain high-order harmonics, the low-pass filter between the grid and the inverter has high power quality requirements. Therefore, an LCL filter is selected. The LCL filter transfer function... :

[0138] in, It is the inverter-side inductor; It is the grid-side inductance; It is a filter capacitor; It is the Laplace operator.

[0139] Filter state equation:

[0140] in, It is the inverter-side inductor current, which characterizes the inverter's output energy. It is the grid-side inductor current, which reflects the quality of grid-connected power. This is the capacitor voltage, and its function is to buffer voltage fluctuations; It is the inverter bridge arm output voltage, and it is the equivalent output voltage of the PWM switch; It is the grid voltage and the system synchronization reference.

[0141] Lithium battery equivalent model:

[0142]

[0143] in, Open circuit voltage; K is the battery polarization parameter; This is the operating current of the lithium battery; It is the equivalent internal resistance; This is the polarization curve steepness coefficient, reflecting the electrode reaction kinetics, with a typical value range of 0.5–2.0. ; It is in a charged state.

[0144] Lithium battery discharge power :

[0145] in, It is the actual terminal voltage of the lithium battery, obtained from the discharge mode in the equivalent model of the lithium battery; This is the actual terminal current of the lithium battery.

[0146] SOC dynamics (dynamically reflecting changes in the state of charge of a lithium battery by integrating the cumulative effect of charging and discharging current over time):

[0147] in, It is the initial state of charge; This refers to the rated capacity of the lithium battery.

[0148] Grid-connected power synthesis:

[0149]

[0150] in, It is the net output power of the lithium battery actually injected into the power grid at time t; It's wind power; It is the energy storage and discharge of lithium batteries; It refers to lithium battery energy storage charging, and the charging power of the lithium battery (a negative value indicates power absorbed from the grid). It is the maximum power that the power grid can accept, which is determined by the substation capacity; It is the theoretical maximum wind power, calculated from the wind speed-power curve.

[0151] Predict the selected feature variables: Based on S2 output And lithium battery energy storage simulation model, calculate and screen feature set { , , }; Lithium battery energy storage wind curtailment rate :

[0152] in, It is the wind curtailment rate of lithium battery energy storage, which is the ratio of the difference between the theoretical power generation of the wind turbine and the actual grid-connected power. Its function is to reflect the ability of the lithium battery energy storage device to absorb fluctuating wind power. The lower the value, the better the regulation effect of lithium battery energy storage. It represents the actual grid-connected wind power output and the simulated output of lithium battery energy storage. It is the theoretical maximum wind power, calculated from the wind speed-power curve.

[0153] Lithium battery energy storage power supply loss rate :

[0154] in, It is the lithium battery energy storage power supply loss rate, which is the percentage difference between the load demand power and the actual power supplied. It is the power demand of the power grid, which is estimated from historical data; It is the grid-connected output power of the lithium battery energy storage device, and the simulated output of the lithium battery energy storage.

[0155] Lithium-ion battery energy storage wind-storage complementarity :

[0156] in, It refers to the complementarity ratio of lithium battery energy storage to wind and energy storage, which is the proportion of power supplied by wind power and lithium battery energy storage in a coordinated manner, maximizing... To improve the utilization rate of renewable energy; It is the theoretical output of the wind turbine, derived from the wind power output prediction sequence; It is the total load, which comes from the real-time metering system of the power grid dispatch center.

[0157] Output filtering feature set ={ , , Proceed to step S7.

[0158] This step completes the refined modeling and characteristic simulation of lithium battery energy storage. The backward Euler method is used to discretize the components, and the equivalent model of the switching devices is constructed using the constant admittance discrete adjoint model method. Buck / Boost converter voltage regulation is achieved through dual closed-loop control; and an LVRT strategy is configured. Based on the wind power output prediction sequence of S2, the wind curtailment rate of lithium battery energy storage is simulated and calculated. Lithium battery energy storage power supply loss rate and the complementarity rate of lithium battery energy storage and wind energy storage Output a unique feature set for lithium battery energy storage This provides key input for the S7 comprehensive evaluation.

[0159] Step S4: Model the vanadium redox flow battery using the ion concentration field calculation method to obtain the vanadium redox flow battery energy storage simulation model; based on the vanadium redox flow battery energy storage simulation model and the wind power output prediction sequence, calculate the screening feature set of vanadium redox flow battery energy storage.

[0160] The ion-exchange membrane used in vanadium redox flow batteries is a proton exchange membrane, which theoretically allows only hydrogen ions to pass through while hindering vanadium ions. However, due to the existence of ion diffusion forces, this hindrance is not absolute. Vanadium ions of different valence states will diffuse through the ion-exchange membrane due to the concentration difference between the positive and negative electrodes, and will undergo corresponding self-discharge reactions with the ions at the opposite electrode. Currently, ion-exchange membranes still face a common problem: while allowing hydrogen ions to pass through efficiently during operation, other ions will also diffuse through the ion-exchange membrane under the influence of ion diffusion forces.

[0161] Chemical reactions of self-discharge at positive and negative electrodes caused by vanadium ion diffusion: In the positive electrode electrolyte:

[0162]

[0163] In the negative electrode electrolyte:

[0164]

[0165] In the above vanadium ion diffusion process, the vanadium ion diffusion flux can be calculated using Fick's law:

[0166] in, This represents the diffusion flux of vanadium ions in various valence states across the ion-difference membrane. This indicates the concentration of vanadium ions in each valence state; This represents the diffusion coefficient of vanadium ions in different valence states; This indicates the thickness of the ion exchange membrane.

[0167] Based on the stoichiometric relationships of the charge-discharge chemical reactions in a vanadium redox flow battery, the dynamic differential equations for the concentrations of vanadium ions in different valence states in the electrolytes at both the positive and negative electrodes can be obtained:

[0168]

[0169]

[0170]

[0171] in, It is the volume of the electrolyte at the positive and negative electrodes; It is the area of ​​the ion separator in a single cell; This is the charging and discharging current of the vanadium redox flow battery. A current of 0 indicates that the battery is in an open circuit state. * This indicates the change in vanadium ion concentration caused by the chemical reactions during battery charging and discharging. It represents the number of individual cells; F is the Faraday constant.

[0172] Substituting the diffusion flux equations for vanadium ions in each valence state obtained from Fick's law into the dynamic differential equation for vanadium ion solubility and integrating it, the real-time instantaneous ion solubility of vanadium ions in each valence state can be calculated based on the known initial solubility of vanadium ions.

[0173] in, It is the concentration of vanadium ions in the a-th valence state at time t; This is the initial vanadium ion concentration; It is the rate of change of concentration; It refers to the charging and discharging time.

[0174] By incorporating the vanadium ion diffusion effect of the vanadium redox flow battery into the battery's equivalent circuit through a controlled voltage source controlled by vanadium ion concentration, the model achieves both high-precision simulation calculations and simplification requirements. Based on the equivalent circuit model, and using current as the excitation for charging and discharging the battery, the equivalent model of the vanadium redox flow battery is as follows:

[0175] in, It is the battery terminal voltage; It is the open-circuit voltage; It is the electrode voltage; It is the rate of change of the electric double layer voltage; It is the resistance of the solution; It is the operating current; It is the equivalent capacitance; It is a reactive resistor; This is the total input current; It is a fixed parallel resistor.

[0176] The specific implementation process of modeling is as follows: First, based on the charging and discharging current, parameters, and initial concentrations of ions in each valence state of the vanadium redox flow battery, the instantaneous concentrations of vanadium ions in each valence state are calculated using the dynamic differential equation of vanadium ion concentration. Then, by directly controlling the controlled voltage source using ion concentration, the open-circuit voltage of a single vanadium redox flow cell was calculated using the Nernst equation for the electrochemical reaction. :

[0177] in, It is the standard electrode potential; It is the gas constant; It is the absolute temperature; F is the Faraday constant; It is the concentration of oxidation state; It is the reduced concentration.

[0178] Then, the open-circuit voltage of the vanadium redox flow battery stack is derived from the stack voltage equation. ; Finally, by inputting the open-circuit voltage of the fuel cell stack into the controlled voltage source and setting the various parameter values ​​in the equivalent circuit, the equivalent operating terminal voltage of the vanadium redox flow battery can be measured. .

[0179] Using the above method, the changes in ion concentration caused by ion diffusion can be tracked in real time and reflected in the changes in the operating characteristics of the vanadium redox flow battery, thereby realizing the equivalent circuit modeling of the vanadium redox flow battery considering ion diffusion.

[0180] The vanadium redox flow battery is connected to the power distribution network through a power conversion system (PCS). The PCS is the core component of the energy storage device, and the vanadium redox flow battery and the PCS together form a complete energy storage device. Based on the characteristics of the vanadium redox flow battery, a two-stage DC / DC and DC / AC structure is adopted for the PCS.

[0181] DC / DC converter (dual closed loop): A complementary PWM control method is adopted.

[0182] Voltage outer loop (regulation):

[0183] in, This is the inductor current reference value; It is the DC bus voltage deviation; It is the voltage loop proportionality coefficient; It is the voltage loop integral coefficient.

[0184] Inner current loop (fast tracking):

[0185] in, It is the PWM duty cycle, which directly controls the proportion of the switching device's on-time within one cycle; It is the real-time current value flowing through the inductor in the circuit; It is current deviation; It is the current loop proportionality coefficient; It is the integral coefficient of the current loop.

[0186] DC / AC converter control (PQ decoupling): Power loop control (outer loop):

[0187] The power loop is used to generate d-axis and q-axis current reference values. and Based on active power ( ) and reactive power ( The error is adjusted using PI.

[0188] in, This is the d-axis current reference value; This is the q-axis current reference value; and These are the reference active power and reactive power, respectively. and This is the actual measured power; and These are the PI parameters of the d-axis power loop; and These are the PI parameters of the q-axis power loop; It is the Laplace operator.

[0189] Current loop control (inner loop):

[0190] The current loop calculates the inverter output voltage command based on the output of the power loop. and ), employing feedforward decoupling techniques to counteract coupling effects.

[0191] in, It is the d-axis output voltage command; It is the q-axis output voltage command; and These are the actual d-axis and q-axis currents, which are measured values. and It is the component of the grid voltage in the dq coordinate system; It is the grid angular frequency; L is the filter inductance value; and These are the PI parameters of the current loop.

[0192] The discharge power of the vanadium redox flow battery is:

[0193] in, It is the actual terminal voltage of the vanadium redox flow battery, which is obtained from the discharge mode in the equivalent model of the vanadium redox flow battery. This is the actual terminal current of the vanadium redox flow battery; and The output is generated in real time from the equivalent circuit model.

[0194] SOC dynamics (the cumulative effect of charge and discharge current over time is calculated by integration, dynamically reflecting the change in the state of charge of a vanadium redox flow battery):

[0195] in, It is the initial state of charge; It is the rated capacity of a vanadium redox flow battery.

[0196] Grid-connected power synthesis:

[0197]

[0198] in, It is the net output power of the vanadium redox flow battery actually injected into the grid at time t; It's wind power; It is an all-vanadium redox flow battery for energy storage and discharge; It is the energy storage and charging of the vanadium redox flow battery, and the charging power of the vanadium redox flow battery (a negative value indicates the power absorbed from the grid). It is the maximum power that the power grid can accept, which is determined by the substation capacity; It is the theoretical maximum wind power, calculated from the wind speed-power curve.

[0199] Predict the selected feature variables: Based on S2 output And a simulation model of vanadium redox flow battery energy storage, calculating characteristic variables { , , }

[0200] Vanadium redox flow battery energy storage wind curtailment rate :

[0201] in, It is the wind curtailment rate of vanadium redox flow battery energy storage, which is the ratio of the difference between the theoretical power generation of the wind turbine and the actual grid-connected power. Its function is to reflect the ability of the vanadium redox flow battery energy storage device to absorb fluctuating wind power. The lower the value, the better the energy storage regulation effect. It is the theoretical maximum wind power, calculated from the wind speed-power curve; It represents the actual grid-connected wind power output and the simulated output of vanadium redox flow battery energy storage.

[0202] Vanadium redox flow battery energy storage power supply loss rate :

[0203] in, It is the power supply loss rate, which is the percentage difference between the load demand power and the actual power supplied. It is the power demand of the power grid, which is estimated from historical data; It is the grid-connected output power of the vanadium redox flow battery energy storage device, and the simulated output of the vanadium redox flow battery energy storage device.

[0204] Vanadium redox flow battery energy storage wind-storage complementarity :

[0205] in, It refers to the wind-storage complementarity ratio of vanadium redox flow battery energy storage, which is the proportion of power supplied by wind power and vanadium redox flow battery energy storage in a coordinated manner, maximizing... To improve the utilization rate of renewable energy; It is the theoretical output of the wind turbine, derived from the wind power output prediction sequence; It is the total load, which comes from the real-time metering system of the power grid dispatch center.

[0206] Output filtering feature set ={ , , Proceed to step S7.

[0207] This step involves modeling and simulating the ion diffusion dynamics of a vanadium redox flow battery. Based on Fick's law, the transmembrane diffusion flux of vanadium ions is quantified, and dynamic differential equations for the concentration of ions at various valence states are constructed. The ion concentration is converted into open-circuit voltage using the Nernst equation, and the output voltage is determined by combining this with an equivalent circuit model. A two-stage PCS control system is employed: a dual-closed-loop DC / DC converter and PQ decoupling in the DC / AC converter. The wind curtailment rate of the vanadium redox flow battery energy storage system is calculated based on the S2 wind power output prediction sequence. Vanadium redox flow battery energy storage power supply loss rate and the wind-storage complementarity of vanadium redox flow battery energy storage Output a feature set specific to vanadium redox flow batteries This provides key input for the S7 comprehensive evaluation.

[0208] Step S5: Model the flywheel motor in flywheel energy storage using electromechanical state equations to obtain a flywheel energy storage simulation model; based on the flywheel energy storage simulation model and wind power output prediction values, calculate the screening feature set for flywheel energy storage.

[0209] The flywheel's dynamic model is as follows: The energy storage equation for flywheel energy storage:

[0210] in, It is the energy that can be released, measured in joules (J). It is the total moment of inertia, in units of ; This is the maximum rotational speed of the flywheel, measured in rad / s. It is the minimum speed of the flywheel.

[0211] Equations of motion for a flywheel motor (from a dynamic perspective):

[0212] in, J is the electromagnetic torque; J is the total moment of inertia, which is the sum of the moment of inertia of the motor and the moment of inertia of the flywheel. B is the rotor's mechanical angular acceleration; B is the viscosity coefficient. It is the flywheel speed; This is the load torque. The total electromagnetic torque of the motor is the remaining flywheel torque after overcoming the damping torque and the load torque.

[0213] The flywheel motor model is as follows: Three-phase voltage equations:

[0214] in, , , It is the three-phase phase voltage of the stator winding; , , It is the three-phase current of the stator winding; , , It is the back electromotive force of the stator winding; Phase resistance; The self-inductance of each phase winding of the stator; It is mutual inductance between the stator windings.

[0215] Relationship between back electromotive force and rotational speed:

[0216] It is the back electromotive force; It is a motor phase identifier; is the electromotive force constant; a1, b1, c1 are the three-phase winding identifiers.

[0217] Electromagnetic torque equation (electromagnetic field angle):

[0218] in, It is electromagnetic torque; It is the number of pole pairs of the motor; It is mechanical angular velocity; and These are the stator flux linkage components in the dq coordinate system; and It is the stator current component in the dq coordinate system.

[0219] dynamics Define external mechanical effects and electromagnetic fields. Revealing the inherent generation logic, the two together constitute a complete control closed loop for flywheel energy storage.

[0220] The power conversion model is as follows: Boost chopper circuit:

[0221]

[0222] in, It is the output voltage; It is the input voltage; It is the duty cycle of the switching transistor; It is the maximum value of the input voltage; It is the target output voltage value.

[0223] SVPWM control equations, active power reactive power :

[0224] in, , These are the d and q components of the grid voltage, respectively. , These are the d and q components of the grid current.

[0225] Voltage control equation:

[0226] Where R and L are resistance and inductance, respectively; , , , These are the d and q components of the grid voltage and the voltage components controlled by the grid-side d-axis and q-axis, respectively. , These are the d and q components of the grid current.

[0227] Flywheel SOC definition:

[0228] in, This indicates the charging state of the flywheel at time t; It is the efficiency coefficient, taken as... =0.92.

[0229] Grid-connected power synthesis:

[0230]

[0231] in, It is the net output power of the flywheel energy storage actually injected into the grid at time t; It's wind power; It is flywheel energy storage and discharge; It is flywheel energy storage charging, and it is the charging power (a negative value indicates that the power is absorbed from the grid). It is the maximum power that the power grid can accept, which is determined by the substation capacity; It is the theoretical maximum wind power, calculated from the wind speed-power curve.

[0232] Predict the selected feature variables: Based on S2 output And flywheel energy storage simulation model, calculate characteristic variables { , , }

[0233] Flywheel energy storage wind curtailment rate :

[0234] in, It is the flywheel energy storage wind curtailment rate, which is the ratio of the difference between the theoretical power generation of the wind turbine and the actual grid-connected power. Its function is to reflect the ability of the flywheel energy storage device to absorb fluctuating wind power. The lower the value, the better the energy storage regulation effect. It is the theoretical maximum wind power, calculated from the wind speed-power curve; It represents the actual grid-connected wind power output and the simulated output of flywheel energy storage.

[0235] Flywheel energy storage power supply loss rate :

[0236] in, It is the flywheel energy storage power supply loss rate, which is the percentage of the difference between the load demand power and the actual power supplied. It is the power demand of the power grid, which is estimated from historical data; It is the grid-connected output power of the flywheel energy storage device, and the simulated output of the flywheel energy storage device.

[0237] Flywheel energy storage wind-storage complementarity :

[0238] in, It refers to the flywheel energy storage-wind-storage complementarity ratio, which is the proportion of power supplied by wind power and flywheel energy storage in a coordinated manner, maximizing... To improve the utilization rate of renewable energy; It is the theoretical output of the wind turbine, derived from the wind power output prediction sequence; It is the total load, which comes from the real-time metering system of the power grid dispatch center.

[0239] Output filtering feature set ={ , , Proceed to step S7.

[0240] This step implements electromechanical modeling and grid-connected simulation of flywheel energy storage. A flywheel energy model is constructed based on the moment of inertia equation; the electromagnetic characteristics of the motor are described using three-phase voltage equations, and electromechanical coupling is established by combining the motion equations; grid-connected power regulation is achieved using a Boost chopper circuit and SVPWM control. The flywheel SOC is defined as a function of rotational speed. The wind curtailment rate of flywheel energy storage is calculated based on the S2 wind power output prediction sequence simulation. Flywheel energy storage power supply loss rate and flywheel energy storage wind-storage complementarity Output flywheel energy storage-specific feature set This provides key input for the S7 comprehensive evaluation.

[0241] Step S6: Based on the vector weighting mechanism of the weighted average vector optimization algorithm, the static defects of the entropy weight method are improved by dynamically adjusting the weights, resulting in the improved entropy weight method.

[0242] Standard entropy weight method: Data standardization:

[0243]

[0244] in, These are the original data values, representing the value of sample i in index j; It is the minimum value of index j; It is the maximum value of index j; It is a standardized value.

[0245] The reason for distinguishing between benefit-oriented and cost-oriented indicators is that the wind-storage complementarity rate, an indicator of benefit-oriented indicators, is better the higher it is, while the wind curtailment rate, an indicator of cost-oriented indicators, is better the lower it is. Only after distinguishing and standardizing these indicators can they be weighted and summed.

[0246] Information entropy calculate:

[0247]

[0248] Where m is the total number of data; It is the information entropy value; the smaller the value, the more ordered the information. It is a probability distribution; It is the natural logarithm of the probability; It is the summation of probability entropy values; It is the normalization coefficient.

[0249] Standard entropy weighting :

[0250] Where n is the total number of data points; b is a temporary index used to iterate through all features. Traditional entropy weighting method weights. A fixed approach cannot adapt to the dynamic changes in the importance of indicators caused by wind power fluctuation cycles, such as the wind curtailment rate being more important during peak wind power periods.

[0251] The vector weighting mechanism of the INFO algorithm: Vector weighted mean definition:

[0252] in, It is a random weighting factor, and its function is to dynamically balance the two mean structures; and These are mean vector 1 and mean vector 2:

[0253]

[0254] in, Its function is to capture the differences within sample a; It is the best sample; It is a boundary sample; It is the worst sample; Its function is to measure the deviation between the current solution and the reference solution; To prevent division by zero of extremely small constants.

[0255] Feature difference penalty weight :

[0256] in, ( ) is the inverted cosine function; the larger the value, the smaller the sample difference. ( ) is the exponential decay term, which penalizes samples with large differences; It is a characteristic function. = ; It is a difference-sensitive factor, and the empirical value is taken. =0.5.

[0257] Dynamic weight adjustment factor: Original INFO algorithm scaling factor :

[0258]

[0259] in, This is the current iteration number; It is a random number uniformly distributed in the interval [0,1]. It is the maximum number of iterations; It is the original exponential decay factor; It is the original random scaling factor.

[0260] Improvements (introducing dynamic adjustment of the cosine function):

[0261] in, is the periodic adjustment factor, a dynamic parameter that matches the characteristics of wind fluctuations; g is the current iteration number; It represents the maximum number of iterations. The cosine function makes α vary periodically in [0, 0.02] to adapt to the wind power fluctuation cycle; the coefficient 0.2 is determined by the wind power fluctuation cycle spectrum analysis to match the 4-hour dominant fluctuation mode.

[0262] Improved implementation process of the entropy weight method: Using an improved INFO algorithm to generate time-sensitive weights :

[0263]

[0264] in, m is the input vector; Quantity; The input weights are α, which is obtained from the above improvements; t corresponds to the wind power fluctuation period. The INFO operator dynamically adjusts the weights using a vector-weighted average. It is the standard entropy weight method weight; It is the feature difference penalty weight.

[0265] Divide the time period according to the wind power fluctuation cycle and match the weights. :

[0266] in, It is a fluctuation cycle. =4h; each time period Internally adopts independent .

[0267] Combine static entropy weights and dynamic weights to calculate the combined weights. :

[0268]

[0269] Where T is the total number of time periods; It is a static weighting percentage. ∈[0.4, 0.6], take =0.5; It is a static weight; It is the dynamic weight for the current time period.

[0270] This improvement resulted in the improved entropy weight method.

[0271] This step proposes an improvement to address the static limitations of the traditional entropy weighting method. By introducing the vector weighting mechanism of the INFO algorithm and designing a dynamic adjustment factor based on the characteristics of wind power fluctuation cycles, the feature weights are corrected in real time. This mechanism synthesizes the static entropy weights and dynamic weights proportionally, enabling the weights to adaptively adjust with the wind power fluctuation cycle. This solves the decision-making lag problem caused by fixed weights and provides a dynamic weighting basis for the comprehensive evaluation of S7 multi-energy storage systems.

[0272] Step S7: The improved entropy weight method combined with the analytic hierarchy process (AHP) is used to calculate the screening feature sets for lithium battery energy storage, vanadium redox flow battery energy storage, and flywheel energy storage, respectively, to obtain the comprehensive scores for lithium battery energy storage, vanadium redox flow battery energy storage, and flywheel energy storage. By comparing the comprehensive scores of lithium battery energy storage, vanadium redox flow battery energy storage, and flywheel energy storage, the optimal energy storage method is determined.

[0273] AHP subjective weight calculation: Judgment Matrix The structure is as follows:

[0274]

[0275] Subjective weight Solution (Eigenvector method):

[0276]

[0277]

[0278] Where A is the judgment matrix; It is an eigenvector, the vector corresponding to the largest eigenvalue; It is the largest eigenvalue; n is the matrix order; It determines the elements of a matrix; It is the geometric mean.

[0279] Consistency check:

[0280]

[0281] Among them, CI is the consistency index; CR is the consistency ratio; and RI is the random consistency index.

[0282] Improved Entropy Weight Method Objective Weight : The objective weight is obtained through step S6. .

[0283] Combined weights synthesis:

[0284] in, It is a combined weight; verified through sensitivity analysis, when When the weight ratio is 0.5, the combined weights exhibit the best adaptability to wind power fluctuation cycles.

[0285] Membership matrix construction: Quantitative indicators of efficiency (the higher the value, the better):

[0286] Where 'a' is the lower limit of the indicator, and 'a = min' b is the upper limit of the indicator, b = max x is the actual indicator value.

[0287] Cost-related quantitative indicators (the smaller the value, the better):

[0288] Where 'a' is the lower limit of the indicator, and a = min b is the upper limit of the indicator, b = max x is the actual indicator value.

[0289] Qualitative indicators :

[0290] Overall score of the plan : Quantitative indicators (such as wind curtailment rate and power loss rate) are converted into standardized scores through cost-based membership functions and benefit-based membership functions. ∈[0,1], together with the qualitative index transformation value, participates in the comprehensive score. calculate:

[0291] in, It is the overall score of scheme k; is the membership degree of index j to scheme k; n is the number of features in the feature set to be screened (n=3); j is the feature index; It is a comprehensive score for lithium battery energy storage; It is a comprehensive score for all-vanadium redox flow battery energy storage; It is a comprehensive score for flywheel battery energy storage.

[0292] Optimal decision rule: Scheme ranking:

[0293] Where k is the energy storage scheme type index; It is the comprehensive score of the k-th scheme; It is to take The largest k value; It is the index of the optimal solution to be selected; the energy storage solution with the highest comprehensive score is selected.

[0294] This step completes the comprehensive evaluation and optimal decision-making of three energy storage schemes. First, the subjective weights of the analytic hierarchy process (AHP) and the objective weights of the improved entropy weight method are combined and weighted at a ratio of β=0.5 to form a combined weight. Then, quantitative indicators (wind curtailment rate, power loss rate, and wind-storage complementarity rate) are converted into standardized scores using membership functions. The comprehensive score of each energy storage scheme is calculated, and finally, the energy storage method with the highest comprehensive score is selected as the optimal energy storage method.

[0295] This invention addresses the mismatch in energy storage schemes caused by weakened feature correlations and static weighting in wind-storage synergy. It constructs a decision-making system that starts with feature selection, uses dynamic weighting as a pivot, and employs differentiated scoring as the final step. Through mutual information, three strongly correlated features—wind curtailment rate, power loss rate, and wind-storage complementarity rate—are extracted from system operation data to form a high-purity feature set. A vector weighting mechanism combining standard entropy weighting and the INFO algorithm is used to generate combined weights that dynamically evolve with wind power fluctuations. A dedicated evaluation model is established based on the performance differences of lithium batteries, flow batteries, and flywheel energy storage. Membership transformation and weighted comprehensive scoring enable quantitative comparison of schemes, ultimately pinpointing the optimal energy storage method based on the highest score.

[0296] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, the phrase "comprising an element defined as..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0297] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their likenesses.

Claims

1. A method for determining the optimal energy storage mode based on the entropy weight method, characterized in that, Includes the following steps: Step S1: Collect historical data of wind turbines and historical data of power grids, and preprocess the historical data of wind turbines and historical data of power grids using the sliding window method to obtain a structured dataset; Based on the characteristics of battery energy storage, a candidate feature set is constructed, and the candidate feature set is filtered by the mutual information feature screening method to obtain a screened feature set, which includes wind curtailment rate, power supply loss rate and wind-storage complementarity rate. Step S2: Based on the autoregressive integral moving average model, construct a wind speed prediction model, input the wind speed data in the structured dataset into the wind speed prediction model to obtain a wind speed prediction sequence, and calculate the wind power output prediction sequence based on the wind speed prediction sequence. Step S3: Discretely model the lithium battery energy storage element using the backward Euler method and the constant admittance discrete adjoint model method to obtain the lithium battery energy storage simulation model; based on the lithium battery energy storage simulation model and the wind power output prediction sequence, calculate the screening feature set of lithium battery energy storage. Step S4: Model the vanadium redox flow battery using the ion concentration field calculation method to obtain the vanadium redox flow battery energy storage simulation model; based on the vanadium redox flow battery energy storage simulation model and the wind power output prediction sequence, calculate the screening feature set of vanadium redox flow battery energy storage. Step S5: Model the flywheel motor in flywheel energy storage using electromechanical state equations to obtain a flywheel energy storage simulation model; based on the flywheel energy storage simulation model and wind power output prediction values, calculate the selection feature set for flywheel energy storage; Step S6: Based on the vector weighting mechanism of the weighted average vector optimization algorithm, the static defects of the entropy weight method are improved by dynamically adjusting the weights, resulting in the improved entropy weight method; Step S7: The improved entropy weight method combined with the analytic hierarchy process (AHP) is used to calculate the screening feature sets for lithium battery energy storage, vanadium redox flow battery energy storage, and flywheel energy storage, respectively, to obtain the comprehensive scores for lithium battery energy storage, vanadium redox flow battery energy storage, and flywheel energy storage. By comparing the comprehensive scores of lithium battery energy storage, vanadium redox flow battery energy storage, and flywheel energy storage, the optimal energy storage method is determined.

2. The method for determining the optimal energy storage mode based on the entropy weight method according to claim 1, characterized in that, Historical data on wind turbines and power grids were collected. These historical data were preprocessed using a sliding window method to obtain a structured dataset, including: The power grid dispatch center retrieves historical data from wind turbines and the power grid through a dedicated data interface. The historical data from wind turbines includes wind speed, real-time power, and start / stop status; the historical data from the power grid includes load demand, voltage frequency, and grid-connected power. The data is encrypted and transmitted to the local server, where it is stored after being sorted by timestamp. The data preprocessing process is as follows: Sliding window filtering of neighboring points: In the temporal alignment stage of data preprocessing, the sliding window method is used to provide a neighboring point filtering framework for Lagrange interpolation; Missing value imputation: For missing points in historical data, Lagrange interpolation is used for imputation. An interpolation function is constructed using four known data points near the missing time point. Based on the spatiotemporal relationship and observation values ​​of the four neighboring points, the imputation value is calculated through a weighted combination. The generated imputation value will completely restore the historical data value at that moment, resulting in the imputed historical data. Standardization process: ； in, is the standardized eigenvalue, a dimensionless variable whose function is to eliminate dimensional differences; x is the original eigenvalue. It is the feature mean, representing the historical mean of a certain feature; It is the standard deviation of the feature, reflecting the range of fluctuation of the feature; Output structured dataset .

3. The method for determining the optimal energy storage mode based on the entropy weight method according to claim 2, characterized in that, The candidate feature set is filtered using a mutual information feature filtering method to obtain a filtered feature set, including: Based on the characteristics of battery energy storage, namely, comprehensively considering the core performance parameters of the energy storage system during charging and discharging, such as power response speed, energy density, and cycle life, and combining its dynamic adaptation characteristics when operating in conjunction with wind farms, such as the ability to smooth wind power fluctuations and the effect of regulating load peaks and valleys, a candidate feature set C={ is constructed to comprehensively reflect the system-level response and collaborative performance of energy storage devices under actual operating conditions. , , }; Structured datasets column vectors in Through feature transformation function () Generate candidate features: ； in, These are candidate features; h is the total number of parameters; It is a structured dataset The column vectors in the model; m is the total number of column vectors; the candidate features include the original parameters: wind speed fluctuation. Change rate, etc.; composite indicators: wind curtailment rate, power supply loss rate, wind-storage complementarity rate; The feature selection process is as follows: enter The candidate feature variables in the data are analyzed using mutual information measurement technology to quantify the statistical correlation between battery energy storage parameters and power supply quality indicators, thereby enabling the extraction of high-contribution features. Discrete variables: ； Wherein, the independent variable X is Candidate feature variables; the target variable Y is the power supply stability index; It represents the correlation between discrete variable X and target variable Y, and its function is to quantify the statistical correlation between battery energy storage parameters and power supply quality indicators. It represents the discrete values ​​of the characteristic variable X, which are actually the set of wind turbine states; It represents the discrete values ​​of the target variable Y, which are actually the power grid frequency state; The joint probability is the joint probability of the grid frequency being at various frequencies when the wind turbine is in various states; It is the marginal probability, representing the probability of the wind turbine being in various states; It is the marginal probability, representing the probability of the power grid operating at various frequencies; Continuous variables: ； in, It represents the correlation between the continuous variable X and the target variable Y; p(x) is the probability density function; Feature selection, based on mutual information values, involves classifying feature relevance levels as follows: MI is obtained as described above. and When MI is less than 0.4, the feature correlation level is weak and it is directly removed; when MI is greater than 0.4 and less than 0.7, the feature correlation level is moderate and it can be retained for auxiliary analysis; when MI is greater than 0.7, the feature correlation level is strong and it can be used as a core decision feature. With the stability of power supply from wind-storage power generation as the optimization objective, the candidate feature set C={ , , Mutual information values ​​with key indicators: Calculate the MI value of each parameter in relation to the power supply quality index; By setting a threshold τ=0.6, features with I(X,Y)≥τ are filtered to obtain wind curtailment rate, power loss rate, and wind-storage complementarity rate. Output the filtered feature set F: ； in, It refers to the rate of curtailment of new energy sources, specifically the wind curtailment rate. = Curtailed wind power / Theoretical power generation × 100%; It is the power supply loss rate. =Power shortage / Total load demand × 100%; It refers to the wind-storage complementarity ratio, or simply wind-storage complementarity ratio. =Energy storage power supply / Total power supply × 100%.

4. The method for determining the optimal energy storage mode based on the entropy weight method according to claim 3, characterized in that, The calculated wind power output prediction sequence includes: Calculate wind power output, model wind turbine power characteristics, and input wind speed prediction sequence; Segmented output model : ； in, It is the maximum power output that the wind turbine can continuously and stably generate at rated wind speed; v is the wind power forecast value, derived from... ; It is the wind turbine start-up threshold; It is the rated wind speed; u is the aerodynamic characteristic coefficient of the fan; This is the mechanical limit of the wind turbine; Output wind power prediction sequence : ； Wind power output prediction sequence It serves as the unified input for S3-S5 energy storage simulations.

5. The method for determining the optimal energy storage mode based on the entropy weight method according to claim 4, characterized in that, Based on the lithium battery energy storage simulation model and wind power output prediction sequence, a selection feature set for lithium battery energy storage is calculated, including: Based on S2 output And lithium battery energy storage simulation model, calculate and screen feature set { , , } Lithium battery energy storage wind curtailment rate : ； in, It is the wind curtailment rate of lithium battery energy storage, which is the ratio of the difference between the theoretical power generation of the wind turbine and the actual grid-connected power. Its function is to reflect the ability of the lithium battery energy storage device to absorb fluctuating wind power. The lower the value, the better the regulation effect of lithium battery energy storage. It represents the actual grid-connected wind power output and the simulated output of lithium battery energy storage. It is the theoretical maximum wind power, calculated from the wind speed-power curve; Lithium battery energy storage power supply loss rate : ； in, It is the lithium battery energy storage power supply loss rate, which is the percentage difference between the load demand power and the actual power supplied. It is the power demand of the power grid, which is estimated from historical data; It is the grid-connected output power of the lithium battery energy storage device, and the simulated output of the lithium battery energy storage. Lithium-ion battery energy storage wind-storage complementarity : ； in, It refers to the complementarity ratio of lithium battery energy storage to wind and energy storage, which is the proportion of power supplied by wind power and lithium battery energy storage in a coordinated manner, maximizing... To improve the utilization rate of renewable energy; It is the theoretical output of the wind turbine, derived from the wind power output prediction sequence; It is the total load, which comes from the real-time metering system of the power grid dispatch center.

6. The method for determining the optimal energy storage mode based on the entropy weight method according to claim 5, characterized in that, Based on the vanadium redox flow battery energy storage simulation model and wind power output prediction sequence, a selection feature set for vanadium redox flow battery energy storage is calculated, including: Based on S2 output And a simulation model of vanadium redox flow battery energy storage, calculating characteristic variables { , , }; Vanadium redox flow battery energy storage wind curtailment rate : ； in, It is the wind curtailment rate of vanadium redox flow battery energy storage, which is the ratio of the difference between the theoretical power generation of the wind turbine and the actual grid-connected power. Its function is to reflect the ability of the vanadium redox flow battery energy storage device to absorb fluctuating wind power. The lower the value, the better the energy storage regulation effect. It is the theoretical maximum wind power, calculated from the wind speed-power curve; It represents the actual grid-connected wind power output and the simulated output of vanadium redox flow battery energy storage. Vanadium redox flow battery energy storage power supply loss rate : ； in, It is the power supply loss rate, which is the percentage difference between the load demand power and the actual power supplied. It is the power demand of the power grid, which is estimated from historical data; It is the grid-connected output power of the vanadium redox flow battery energy storage device, and the simulated output of the vanadium redox flow battery energy storage. Vanadium redox flow battery energy storage wind-storage complementarity : ； in, It refers to the wind-storage complementarity ratio of vanadium redox flow battery energy storage, which is the proportion of power supplied by wind power and vanadium redox flow battery energy storage in a coordinated manner, maximizing... To improve the utilization rate of renewable energy; It is the theoretical output of the wind turbine, derived from the wind power output prediction sequence; It is the total load, which comes from the real-time metering system of the power grid dispatch center.

7. The method for determining the optimal energy storage mode based on the entropy weight method according to claim 6, characterized in that, Based on the aforementioned flywheel energy storage simulation model and predicted wind power output, a selection feature set for flywheel energy storage is calculated, including: Based on S2 output And flywheel energy storage simulation model, calculate characteristic variables { , , }; Flywheel energy storage wind curtailment rate : ； in, It is the flywheel energy storage wind curtailment rate, which is the ratio of the difference between the theoretical power generation of the wind turbine and the actual grid-connected power. Its function is to reflect the ability of the flywheel energy storage device to absorb fluctuating wind power. The lower the value, the better the energy storage regulation effect. It is the theoretical maximum wind power, calculated from the wind speed-power curve; It represents the actual grid-connected wind power output and the simulated output of flywheel energy storage. Flywheel energy storage power supply loss rate : ； in, It is the flywheel energy storage power supply loss rate, which is the percentage difference between the load demand power and the actual power supplied. It is the power demand of the power grid, which is estimated from historical data; It is the grid-connected output power of the flywheel energy storage device, and the simulated output of the flywheel energy storage. Flywheel energy storage wind-storage complementarity : ； in, It refers to the flywheel energy storage-wind-storage complementarity ratio, which is the proportion of power supplied by wind power and flywheel energy storage in a coordinated manner, maximizing... To improve the utilization rate of renewable energy; It is the theoretical output of the wind turbine, derived from the wind power output prediction sequence; It is the total load, which comes from the real-time metering system of the power grid dispatch center.

8. The method for determining the optimal energy storage mode based on the entropy weight method according to claim 7, characterized in that, Based on the weighted average vector optimization algorithm, a vector weighting mechanism is used to improve the static defects of the entropy weight method by dynamically adjusting the weights, resulting in an improved entropy weight method, including: Standard entropy weighting : ； Where n is the total number of data items; b is the temporary index subscript; The vector weighting mechanism of the INFO algorithm: Vector weighted mean definition: ； in, It is a random weighting factor, and its function is to dynamically balance the two mean structures; and These are mean vector 1 and mean vector 2; Improved implementation process of the entropy weight method: Time-sensitive weights generated using an improved INFO algorithm : ； ； in, m is the input vector; Quantity; The input weights are α, which is obtained from the above improvements; t corresponds to the wind power fluctuation period; the INFO operator dynamically adjusts the weights using a vector weighted average. It is the standard entropy weight method weight; It is the feature difference penalty weight; Divide the time period according to the wind power fluctuation cycle and match the weights. : ； in, It is a fluctuation cycle. =4h; each time period Internally adopts independent ; Combine static entropy weights and dynamic weights to calculate the combined weights. : ； ； Where T is the total number of time periods; It is a static weighting percentage. ∈[0.4, 0.6], take =0.5; It is a static weight; It is the dynamic weight for the current time period.

9. The method for determining the optimal energy storage mode based on the entropy weight method according to claim 8, characterized in that, The corresponding comprehensive scores for lithium battery energy storage, vanadium redox flow battery energy storage, and flywheel energy storage are obtained, including: Subjective weight Solution: ； ； ； Where A is the judgment matrix; It is an eigenvector, the vector corresponding to the largest eigenvalue; It is the largest eigenvalue; n is the matrix order; It determines the elements of a matrix; It is the geometric mean; Combined weights synthesis: ； in, It is a combined weight; verified through sensitivity analysis, when When the ratio is 0.5, the combined weights exhibit the best adaptability to wind power fluctuation cycles. Overall score of the plan : Quantitative indicators are converted into standardized scores using cost-based and benefit-based membership functions. ∈[0,1], together with the qualitative index transformation value, participates in the comprehensive score. calculate: ； in, It is the overall score of scheme k; is the membership degree of index j to scheme k; n is the number of features in the selected feature set; j is the feature index; It is a comprehensive score for lithium battery energy storage; It is a comprehensive score for vanadium redox flow battery energy storage; It is a comprehensive score for flywheel battery energy storage.

10. The method for determining the optimal energy storage mode based on the entropy weight method according to claim 9, characterized in that, By comparing the comprehensive scores of lithium battery energy storage, vanadium redox flow battery energy storage, and flywheel energy storage, the optimal energy storage method is determined, including: Optimal decision rule: Select the energy storage solution with the highest comprehensive score; Scheme Ranking: ； Where k is the energy storage scheme type index; It is the comprehensive score of the k-th scheme; It is to take The largest k value; It is the index of the optimal solution that is ultimately selected.

Images