Coordinated optimization method for wind power photovoltaic energy storage ratio

Abstract

The invention discloses a coordinated optimization method for a wind power photovoltaic energy storage ratio, and belongs to the field of power distribution network optimization, and the method comprises the following steps: S1, multi-dimensional data collection and preprocessing; s2, constructing a robustness uncertainty set; s3, constructing a multi-time-scale collaborative distribution model: constructing the multi-time-scale collaborative distribution model, and determining a target power distribution scheme and an energy storage life loss evaluation index of the energy storage equipment; s4, performing dynamic multi-objective optimization solution: constructing a multi-objective function, and performing solution by adopting an improved multi-objective African eagle optimization algorithm; and S5, real-time feedback and dynamic correction: monitoring a system state through real-time acquisition of data, dynamically correcting the multi-time scale collaborative distribution model based on a prediction error, and adjusting an energy storage equipment power distribution strategy to form closed-loop optimization. By adopting the coordinated optimization method for the wind power and photovoltaic energy storage ratio, the renewable energy consumption capability and the system economy are remarkably improved, and the method has great engineering application value.

Description

[0001] Technical Field 3

[0002] The present invention relates to the technical field of distribution network optimization, and in particular, to a coordinated optimization method for the ratio of wind power, photovoltaic power, and energy storage. Background Art

[0003] With the acceleration of the global energy transition, wind power and photovoltaic power, as the core renewable energy sources, have continuously increased their installed capacity ratios. However, due to the significant intermittency and volatility affected by natural conditions such as wind speed and light intensity, the difficulty of power balance regulation in the power grid has increased. The existing energy storage ratio methods have the following core defects:

[0004] 1. Coarse description of uncertainty: Traditional methods mostly rely on a single probability distribution or historical mean for capacity configuration, and do not construct a multi-dimensional uncertainty set (such as the joint distribution of wind speed fluctuations and light prediction errors), resulting in the energy storage system being prone to insufficient or redundant capacity under extreme weather conditions. In typical scenarios, the power fluctuation suppression efficiency is lower than 70%.

[0005] 2. Disconnection of multi-objective optimization: Most studies focus on a single objective, ignoring the coupled optimization of reliability (such as the EENS index) and economy, resulting in an imbalance between the full life cycle cost of the system and power supply stability. For example, a 30MW wind energy storage project did not take into account the battery cycle life, resulting in an increase in operation and maintenance costs by 25% compared to the expected value.

[0006] 3. Lag in dynamic response: Lack of a rolling optimization mechanism based on real-time data, the traditional fixed parameter control strategy has a response delay of more than 15 minutes when the load suddenly changes (such as ±30% fluctuation), causing overcharging and over-discharging of the energy storage device, and significantly shortening the battery life (the number of cycles decreases by 30%).

[0007] 4. Lack of cross-scale coordination: Failure to effectively separate high-frequency power fluctuations (from seconds to minutes) and low-frequency capacity requirements (from hours to days), resulting in blurred division of labor between power-type energy storage (supercapacitor) and energy-type energy storage (battery), and reducing the overall system efficiency by 15%-20%.

[0008] 5. Insufficient data-driven depth: Existing models do not integrate physical mechanisms (such as battery electrochemical decay models) with data intelligence (such as LSTM prediction), and the capacity ratio error under complex climate conditions (such as cloudy and gusty wind scenarios) exceeds 20%, making it difficult to meet the accurate regulation requirements. Summary of the Invention

[0009] The purpose of the present invention is to provide a coordinated optimization method for the ratio of wind power, photovoltaic power, and energy storage to solve the above technical problems.

[0010] To achieve the above purpose, the present invention provides a coordinated optimization method for the ratio of wind power, photovoltaic power, and energy storage, including the following steps:

[0011] S1. Multi-dimensional data collection and preprocessing: Collect historical wind power data, photovoltaic historical power data, load demand data, real-time meteorological data, energy storage equipment parameters and economic parameters to form a multi-dimensional data set, and preprocess the multi-dimensional data set to obtain a standardized data set and load decomposition results;

[0012] S2. Construct robust uncertainty set: Based on the standardized data set and the historical distribution data of wind power prediction errors and photovoltaic prediction errors, an improved box uncertainty set is constructed to describe the power fluctuation range;

[0013] S3. Construct a multi-time scale collaborative allocation model: Based on the box uncertainty set and load decomposition results, construct a multi-time scale collaborative allocation model to determine the target power allocation scheme of the energy storage equipment and the energy storage life loss evaluation index;

[0014] S4. Dynamic multi-objective optimization solution: Taking the minimization of system total cost, minimization of power fluctuation standard deviation, maximization of reliability index and minimization of energy storage life loss cost as multi-objectives, a multi-objective function is constructed, and combined with power balance constraints, energy storage capacity and charging and discharging power constraints and grid interaction power constraints, the improved multi-objective African Vulture optimization algorithm is used to solve and obtain the Pareto frontier solution set;

[0015] S5. Real-time feedback and dynamic correction: Monitor the system status through real-time data collection, dynamically correct the multi-time scale collaborative allocation model based on the prediction error, adjust the power allocation strategy of the energy storage equipment, and form a closed-loop optimization.

[0016] Preferably, step S1 specifically includes the following steps:

[0017] S11, collect historical wind power and photovoltaic power data, real-time meteorological data, load data, energy storage equipment parameters and economic parameters to form a multi-dimensional data set;

[0018] S12, use the 3σ rule to remove abnormal data, and fill in missing values ​​through bidirectional LSTM to construct a spatiotemporal correlation data set;

[0019] S13, decompose the load demand data based on wavelet packets and separate the high-frequency fluctuation component P high (t) and the low-frequency baseload component P low (t);

[0020] S14. Establish the energy storage state update equation:

[0021]

[0022] In the formula, SOC bat (t) and SOC bat(t - 1) represent the state of charge of the battery at times t and t - 1; SOC cap (t) and SOC cap (t - 1) represent the state of charge of the supercapacitor at times t and t - 1 respectively; ΔT represents the sampling interval; P bat (t) and P cap (t) represent the charging and discharging powers of the battery and the supercapacitor at time t respectively; η bat and η cap represent the charging and discharging efficiencies of the battery and the supercapacitor respectively; E bat and E cap represent the rated capacities of the battery and the supercapacitor respectively; and represent the indicator functions for judging the charging and discharging states of the battery and the supercapacitor respectively;

[0023] S15, output the pre - processed standardized data set D after normalization, and the load decomposition results {P high (t), P low (t)};

[0024] Among them,

[0025] D = {P w (t), P p (t), P load (t), SOC bat (t), SOC cap (t)} (2);

[0026] In the formula, P w (t), P p (t) and P load (t) represent the wind power data, photovoltaic power data and load demand data at time t respectively.

[0027] Preferably, step S2 specifically includes the following steps:

[0028] S21, construct the box - type uncertainty set H:

[0029]

[0030] In the formula, and represent the predicted powers of wind power and photovoltaic power considering uncertainties at time t respectively, and W represents the set of wind farm and photovoltaic array devices; β w (t) and β p (t) represent the variables of wind power and photovoltaic power constrained by the robust coefficient Γ respectively; and respectively represent the maximum absolute deviation of the predicted power of wind power and photovoltaic power; Γ represents the robust coefficient, and 0 ≤ Γ ≤ |W|;

[0031] S22. Introduce the Wasserstein distance to measure the distribution robustness:

[0032]

[0033] In the formula, W(P w , P p ) represents the Wasserstein distance between the wind power P w and the photovoltaic power P p ; Π represents the joint probability distribution; represents the distance from the point w in the wind power P to the point p in the photovoltaic power P ;

[0034] S23. Output the box uncertainty set H and the robustness parameters {Γ, W(P w , P p )}.

[0035] Preferably, step S3 specifically includes the following steps:

[0036] S31. Use the variational mode decomposition algorithm to perform frequency domain decomposition on the high-frequency fluctuation component P high (t), and allocate the components with frequencies greater than a specific threshold to the supercapacitor. The target power P cap-ref (t) allocated to the supercapacitor is calculated by the following formula:

[0037] P cap-ref (t) = λ(t) · P high (t) (5);

[0038] In the formula, λ(t) represents the dynamic allocation coefficient, and 0.8 ≤ λ(t) ≤ 1.0;

[0039] S32. Allocate the low-frequency base load component P low (t) and the remaining power allocated by the supercapacitor to the battery:

[0040] P bat-ref (t) = P low (t) + (1 - λ(t)) · P high (t) (6);

[0041] In the formula, P bat-ref (t) represents the power borne by the battery, and and respectively represent the maximum allowable discharge power of the battery and the maximum allowable charging power of the battery;

[0042] S33. Establish an energy storage life loss evaluation model:

[0043]

[0044] In the formula, L bat and L cap respectively represent the life loss of the battery and the supercapacitor; α and β respectively represent the life loss coefficients of the battery and the supercapacitor; ΔSOC bat (t) represents the change in the state of charge of the battery at time t; DOD max represents the maximum discharge depth of the battery;

[0045] S34. Output the target power distribution scheme {P cap-ref (t), P bat-ref (t)} of the energy storage device, and the energy storage life loss evaluation index {L bat , L cap}.

[0046] Preferably, in step S31, when 0.3 ≤ SOC cap (t) ≤ 0.7, λ(t) = 1.0, indicating that the supercapacitor fully undertakes the high-frequency power fluctuation;

[0047] When SOC cap (t) < 0.3, to avoid over-discharge of the supercapacitor;

[0048] When SOC cap (t) > 0.7, to avoid overcharging of the supercapacitor.

[0049] Preferably, step S4 specifically includes the following steps:

[0050] S41. Construct a multi-objective function:

[0051] minJ = ω1C total + ω2σ fluct + ω3(1 - Ri) + ω4C es-loss (9);

[0052] In the formula, J represents the multi-objective function value; ω1, ω2, ω3, and ω4 all represent weight coefficients, and ω1 + ω2 + ω3 + ω4 = 1; C total represents the total system cost; σ fluct represents the standard deviation of power fluctuation; Ri represents the reliability index; C es-loss represents the energy storage life loss cost;

[0053] Among them,

[0054]

[0055] C es-loss =γ bat L bat +γ cap L cap (13); In the formula, C wind-inv represents the investment cost of wind power equipment; C pv-inv represents the investment cost of photovoltaic equipment; C es-inv represents the investment cost of energy storage equipment; C grid (t) represents the grid electricity price at time t; P grid (t) represents the interaction power with the grid at time t; Δt represents the time interval; P total (t) represents the total power of the system at time t; represents the average value of the total power; P loat (t) represents the load demand power at time t; γ bat and γ cap respectively represent the life loss cost coefficients of the battery and the supercapacitor;

[0056] And add power balance constraints, energy storage capacity and charge-discharge power constraints, and grid interaction power constraints as constraint conditions;

[0057] Among them, the power balance constraint expression is as follows:

[0058] P w (t)+P p (t)+P cap (t)+P bat (t)=P load (t)+P grid (t) (14);

[0059] The energy storage capacity and charge-discharge power constraint expression is as follows:

[0060]

[0061] In the formula, E cap (t) represents the capacity of the supercapacitor at time t; E cap-min and E cap-max respectively represent the minimum and maximum allowable capacities of the supercapacitor; E bat (t) represents the capacity of the battery at time t; E bat-min and E bat-max respectively represent the minimum and maximum allowable capacities of the battery; and respectively represent the maximum allowable discharge and charge powers of the supercapacitor; and respectively represent the maximum allowable discharge and charge power of the battery;

[0062] The grid interaction power constraint expression is as follows:

[0063] P grid-min ≤P grid (t)≤P grid-max (16);

[0064] In the formula, P grid-min and P grid-max respectively represent the minimum and maximum allowable values of the power interacting with the grid;

[0065] S42. Solve using an improved multi-objective African vulture optimization algorithm;

[0066] S421. Initialize the population: Randomly generate a group of initial solutions including wind power installed capacity, photovoltaic power installed capacity, and energy storage capacity as the population;

[0067] S422. Construct a comprehensive fitness function F using the linear weighted method:

[0068]

[0069] In the formula, ω5, ω6, ω7, and ω8 all represent weight coefficients; C total,min and C total,max respectively represent the minimum and maximum values of the total system cost in the current population; σ fluct,min and σ fluct,max respectively represent the minimum and maximum values of the standard deviation of power fluctuations in the current population; Ri min and Ri max respectively represent the minimum and maximum values of the reliability index in the current population; C es-loss,min and C es-loss,max respectively represent the minimum and maximum values of the energy storage life loss cost in the current population;

[0070] S423. Grouping and cooperation: Sort the population according to the fitness value. Assume that the population is divided into k groups, and each group is led by a leader. Then, the relationship between the member i in the jth group and the leader is as follows:

[0071]

[0072] In the formula, D ij represents the distance between the member i and the leader in the jth group in the decision variable space; x id represents the dth decision variable of the member i; x ld represents the dth decision variable of the leader; D represents the dimension of the decision variable;

[0073] S424. Update the position: According to the foraging behavior rules of vultures, update the position of each solution in formula (18), while ensuring that the updated solution satisfies the added constraint conditions;

[0074] S425. Judge the termination condition: When the preset termination condition is met, stop the iteration and output the Pareto front solution set.

[0075] Preferably, step S5 specifically includes the following steps:

[0076] S51. Real-time data acquisition: Obtain the wind power P w (t), photovoltaic power P p (t), load demand P load (t), and the state of charge SOC bat (t) and SOC cap (t) of the energy storage device in real time, and form the real-time state vector X real (t):

[0077] X real (t) = [P w (t), P p (t), P load (t), SOC bat (t), SOC cap (t)] (19);

[0078] S52. Calculate the deviation between the real-time power and the predicted power in step S2:

[0079]

[0080] In the formula, ΔP w (t) and ΔP p (t) respectively represent the deviations of the predicted powers of wind power and photovoltaic power;

[0081] S53. Dynamically adjust the robust coefficient Γ based on the deviation ΔP w (t) of the wind power predicted power and the deviation ΔP p (t) of the photovoltaic predicted power:

[0082]

[0083] In the formula, Γ(t) and Γ(t - 1) respectively represent the robust coefficients at time t and t - 1; κ represents the adaptive adjustment factor, and 0.1 ≤ κ ≤ 0.5;

[0084] At the same time, based on the real-time state of charge SOC bat (t) and SOC cap (t) of the energy storage device, correct the energy storage state update equation:

[0085]

[0086] Wherein, SOC bat-corrected (t) and SOC cap-corrected (t) represent the state of charge of the updated battery and supercapacitor; δ bat and δ cap respectively represent the state correction coefficients of the battery and the supercapacitor;

[0087] S54. Calculate the net load P net (t) to be borne by the energy storage and the power grid according to the real-time load and the renewable energy output, and add the constraint condition:

[0088] P net (t) = P load (t) - P w (t) - P p (t) (24);

[0089]

[0090] Wherein, P bat-min and P bat-max respectively represent the lower and upper limits of the battery charge and discharge power; P cap-min and P cap-max respectively represent the lower and upper limits of the supercapacitor charge and discharge power; SOC bat-min and SOC bat-max respectively represent the lower and upper limits of the safe range of the battery state of charge; SOC cap-min and SOC cap-max respectively represent the lower and upper limits of the safe range of the supercapacitor state of charge;

[0091] S55. Dynamically allocate the energy storage power according to the net load P net (t), the high-frequency fluctuation component P high (t) and the low-frequency base load component P low (t):

[0092] When |P net (t)| ≤ P bat-max + P cap-max at this time, high-frequency fluctuation: P cap-adjust (t) = P high (t), P cap-adjust (t) represents the charge and discharge power of the supercapacitor after dynamic adjustment; low-frequency base load: P bat-adjust (t) = P low (t), P bat-adjust (t) represents the charge and discharge power of the battery after dynamic adjustment;

[0093] Otherwise, trigger grid interaction:

[0094] P grid-adjust P(t) = sgn(P net (t))·max(0, |P net (t)| - (P bat-max + P cap-max )) (26);

[0095] Wherein, P grid-adjust (t) represents the grid interaction power after dynamic adjustment.

[0096] Therefore, the present invention adopts the above - mentioned coordinated optimization method for the ratio of wind power, photovoltaic and energy storage, and has the following beneficial effects:

[0097] 1. Full - dimensional uncertainty modeling, significantly improved robustness: By constructing an improved box - type uncertainty set that integrates wind power / photovoltaic prediction errors and combining the Wasserstein distance to measure distribution robustness, the system can cope with scenarios of ±40% wind speed fluctuations and ±30% sudden changes in light. The power balance qualification rate is increased from 75% of the traditional method to 92%, and the EENS under extreme weather is reduced by 40%, significantly enhancing the anti - interference ability of the power grid;

[0098] 2. Precise multi - time - scale coordination, breakthrough in energy storage efficiency: Based on the VMD - based high - frequency / low - frequency component separation strategy, the division of labor between supercapacitors and batteries is optimized: The high - frequency fluctuation response time is shortened to 2 minutes, the charge - discharge frequency of the battery is reduced by 35%, the cycle life is extended by 45%, the capacity requirement of the supercapacitor is reduced by 20%, and the overall system efficiency is increased by 18% - 22%;

[0099] 3. Dynamic weight adaptive mechanism, deep multi - objective coupling: Introduce the energy storage life loss cost item, and dynamically allocate weights by combining the entropy weight method to search for the Pareto optimal solution in the three - dimensional objective space of cost - reliability - life; Compared with single - objective optimization, the full - life - cycle cost is reduced by 15% - 20%, the battery replacement cycle is extended by 2 years, and at the same time, the reliability index Ri is increased to more than 95%;

[0100] 4. Real - time feedback correction closed - loop, revolutionary response speed: Based on real - time prediction and dynamic mode switching of LSTM - KNN, the control instruction update cycle is shortened to 5 minutes, the power compensation delay during load mutation (±30%) is <1 minute, the energy storage state fluctuation range is controlled within ±10%, avoiding equipment damage caused by over - charging and over - discharging, and the operation and maintenance cost is reduced by 25%;

[0101] 5. Deep integration of data intelligence and physical models: Innovatively integrate the energy storage electrochemistry model with machine learning prediction, construct a "mechanism + data" dual-driven optimization framework, reduce the capacity ratio error from 20% to 8% in complex climate scenarios, and improve the equipment selection accuracy by 30%, providing a universal solution for the high-proportion access of renewable energy.

[0102] Through the above technical solutions, the present invention breaks through the bottlenecks of traditional methods in aspects such as uncertainty processing, multi-objective balance, and dynamic response, forming a complete technical chain of "precision modeling - intelligent allocation - real-time regulation", which is applicable to scenarios such as distributed microgrids and centralized wind-solar-storage power stations.

[0103] The technical solutions of the present invention will be further described in detail below with reference to the drawings and embodiments. Description of the Drawings

[0104] Figure 1 It is a flowchart of a coordinated optimization method for the ratio of wind power, photovoltaic power, and energy storage of the present invention. Detailed Embodiments

[0105] In order to make the purpose, technical solutions, and advantages of the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the embodiments of the present invention and are not used to limit the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of this application. The examples of the embodiments are shown in the drawings, where the same or similar reference numerals represent the same or similar elements or elements with the same or similar functions throughout.

[0106] It should be noted that the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units does not necessarily have to be limited to those clearly listed steps or units, but may include other steps or units that are not clearly listed or are inherent to these processes, methods, products, or devices.

[0107] The embodiments of the present invention will be described in detail below with reference to the drawings.

[0108] As Figure 1 shown, a coordinated optimization method for the ratio of wind power, photovoltaic power, and energy storage includes the following steps:

[0109] S1. Multi-dimensional data collection and preprocessing: Collect historical wind power data, photovoltaic historical power data, load demand data, real-time meteorological data, energy storage equipment parameters and economic parameters to form a multi-dimensional data set, and preprocess the multi-dimensional data set to obtain a standardized data set and load decomposition results;

[0110] Step S1 specifically includes the following steps:

[0111] S11. Collect historical wind power and photovoltaic power data, real-time meteorological data, load data, energy storage equipment parameters and economic parameters to form a multi-dimensional data set; in this embodiment, the historical wind power data has a time span of no less than 3 years and a time resolution of 5 minutes, covering power output data in different seasons, day and night, and extreme weather (such as typhoons, rainstorms, sandstorms and other special meteorological conditions). Photovoltaic historical power data is consistent with the time span and resolution of wind power data, and records in detail the power values ​​under different light intensities and temperature changes. Load demand data, with a time resolution of 1 minute, distinguishes between industrial, commercial and residential electricity load characteristics, and also includes load change data under different time modes such as holidays and working days. Real-time meteorological data, including wind speed, light intensity, temperature, humidity and air pressure, with an accuracy of wind speed ±0.5m / s, light intensity ±20W / m 2 , temperature ±1℃, humidity ±5%, air pressure ±5hPa. Energy storage equipment parameters, including the rated capacity of the battery, the rated capacity of the supercapacitor, the charge and discharge efficiency of the battery, the charge and discharge efficiency of the supercapacitor, the internal resistance of the battery, the equivalent series resistance of the supercapacitor, and the life attenuation curve of the battery and supercapacitor. Economic parameters, including the investment cost of wind power equipment, photovoltaic equipment, and energy storage equipment, in units of RMB / W; operation and maintenance cost coefficient, in units of RMB / (W·year); peak and valley electricity prices of the power grid.

[0112] S12, use the 3σ rule to remove abnormal data, and fill in missing values ​​through bidirectional LSTM to construct a spatiotemporal correlation data set;

[0113] S13, decompose the load demand data based on wavelet packets and separate the high-frequency fluctuation component P high (t) and (period < 30min) low frequency base load component P low (t) (period ≥ 30 min);

[0114] S14. Establish the energy storage state update equation:

[0115]

[0116] In the formula, SOC bat (t) and SOC bat (t-1) represents the battery state of charge at time t and t-1 respectively; SOC cap(t) and SOC cap (t - 1) represent the state of charge of the supercapacitor at times t and t - 1 respectively; ΔT represents the sampling interval; P bat (t) and P cap (t) represent the charging and discharging powers of the battery and the supercapacitor at time t respectively; η bat and η cap represent the charging and discharging efficiencies of the battery and the supercapacitor respectively; E bat and E cap represent the rated capacities of the battery and the supercapacitor respectively; and represent the indicator functions for judging the charging and discharging states of the battery and the supercapacitor respectively;

[0117] S15. After normalization processing, output the pre - processed standardized data set D, and the load decomposition result {P high (t), P low (t)};

[0118] Among them,

[0119] D = {P w (t), P p (t), P load (t), SOC bat (t), SOC cap (t)} (2);

[0120] In the formula, P w (t), P p (t) and P load (t) represent the wind power data, photovoltaic power data and load demand data at time t respectively.

[0121] S2. Construct a robustness uncertainty set: Based on the standardized data set, the historical distribution data of wind power prediction errors and the historical distribution data of photovoltaic prediction errors, construct an improved box - type uncertainty set to describe the power fluctuation range; among them, the historical distribution data of wind power prediction errors is obtained by comparing and analyzing the wind power prediction data and actual data over the past 3 years or more to obtain the standard deviation of the prediction error and the probability density function of the error distribution. The historical distribution data of photovoltaic prediction errors is also obtained through long - term data comparison to obtain the standard deviation of the prediction error and the probability density function of the error distribution.

[0122] Step S2 specifically includes the following steps:

[0123] S21. Construct a box - type uncertainty set H:

[0124]

[0125] In the formula, and respectively represent the predicted power of wind power and photovoltaic power considering uncertainty at time t, and W represents the set of wind farms and photovoltaic array devices; β w (t) and β p (t) respectively represent the variables of wind power and photovoltaic power constrained by the robustness coefficient Γ; and respectively represent the maximum absolute deviation of the predicted power of wind power and photovoltaic power; Γ represents the robustness coefficient, and 0 ≤ Γ ≤ |W|, and when Γ = 0, it means that the wind power and photovoltaic power are within the expected range, and the uncertainty in the smoothing process can be ignored;

[0126] S22. Introduce the Wasserstein distance to measure distribution robustness:

[0127]

[0128] In the formula, W(P w , P p ) represents the Wasserstein distance between wind power P w and photovoltaic power P p ; Π represents the joint probability distribution; represents the distance from the point w in wind power P to the point p in photovoltaic power P ;

[0129] S23. Output the box - type uncertainty set H and the robustness parameters {Γ, W(P w , P p )}.

[0130] S3. Construct a multi - time - scale collaborative allocation model: Based on the box - type uncertainty set and the load decomposition result, construct a multi - time - scale collaborative allocation model to determine the target power allocation scheme of energy storage devices and the evaluation index of energy storage life loss;

[0131] Step S3 specifically includes the following steps:

[0132] S31. According to the characteristics that the supercapacitor has a fast response speed and is suitable for dealing with high - frequency power fluctuations, allocate the components with a frequency greater than a specific threshold (such as 1 Hz) to the supercapacitor. To give full play to the advantage of the supercapacitor in quickly absorbing and releasing energy, suppress high - frequency power fluctuations, and ensure the stability of the system. Therefore, the variational mode decomposition algorithm is used to perform frequency - domain decomposition on the high - frequency fluctuation component P high (t), and allocate the components with a frequency greater than the specific threshold to the supercapacitor, and the target power P cap-ref (t) allocated to the supercapacitor is calculated by the following formula:

[0133] P cap-ref P(t) = λ(t)·P high (t) (5);

[0134] Wherein, λ(t) represents the dynamic distribution coefficient, and 0.8 ≤ λ(t) ≤ 1.0;

[0135] Preferably, in step S31, when 0.3 ≤ SOC cap (t) ≤ 0.7, λ(t) = 1.0, indicating that the supercapacitor fully bears the high-frequency power fluctuation;

[0136] When SOC cap (t) < 0.3, to avoid over-discharging of the supercapacitor;

[0137] When SOC cap (t) > 0.7, to avoid overcharging of the supercapacitor.

[0138] S32. The battery has the characteristics of high energy density, being suitable for storing and providing continuous energy. Therefore, it bears the low-frequency base load component and the remaining power distributed by the supercapacitor. At the same time, to protect the battery, it is necessary to constrain the power distributed to it to ensure that the battery operates within a safe charge and discharge range and extend the battery life. Therefore, the low-frequency base load component P low (t) and the remaining power distributed by the supercapacitor are distributed to the battery:

[0139] P bat-ref (t) = P low (t) + (1 - λ(t))·P high (t) (6);

[0140] Wherein, P bat-ref (t) represents the power borne by the battery, and and respectively represent the maximum allowable discharge power of the battery and the maximum allowable charge power of the battery;

[0141] S33. Establish an evaluation model for energy storage life loss:

[0142]

[0143] Wherein, L bat and L cap respectively represent the life losses of the battery and the supercapacitor; α and β respectively represent the life loss coefficients of the battery and the supercapacitor, and 0.5 ≤ α ≤ 1.0, 0.001 ≤ β ≤ 0.01; ΔSOC bat (t) represents the change in the state of charge of the battery at time t; DOD maxIndicates the maximum depth of discharge of the battery;

[0144] It can be seen that the energy storage life loss is closely related to the charge and discharge process. By establishing a life loss assessment model, the impact of different charge and discharge strategies on the life of energy storage equipment can be quantified, providing an important basis for system optimization. For batteries, their life loss is mainly related to the change range and frequency of the state of charge (SOC); for supercapacitors, the life loss is related to the magnitude of the charge and discharge power.

[0145] S34. Output the target power distribution scheme {P cap-ref (t), P bat-ref (t)} of the energy storage equipment, and the energy storage life loss assessment indexes {L bat , L cap}.

[0146] S4. Dynamic multi-objective optimization solution: Taking the minimization of the total system cost, the minimization of the standard deviation of power fluctuations, the maximization of the reliability index, and the minimization of the energy storage life loss cost as multiple objectives, a multi-objective function is constructed, and combined with the power balance constraint, the energy storage capacity and charge and discharge power constraints, and the grid interaction power constraint, an improved multi-objective African vulture optimization algorithm is used for solution to obtain the Pareto front solution set;

[0147] Step S4 specifically includes the following steps:

[0148] S41. Construct a multi-objective function:

[0149] minJ = ω1C total + ω2σ fluct + ω3(1 - Ri)+ ω4C es-loss (9);

[0150] In the formula, J represents the multi-objective function value; ω1, ω2, ω3, and ω4 all represent weight coefficients, and ω1 + ω2 + ω3 + ω4 = 1; C total represents the total system cost; σ fluct represents the standard deviation of power fluctuations; Ri represents the reliability index; C es-loss represents the energy storage life loss cost;

[0151] Among them,

[0152]

[0153] C es-loss = γ bat L bat + γ cap L cap (13);

[0154] In the formula, C wind-inv represents the investment cost of wind power equipment; Cpv-inv Denotes the investment cost of the photovoltaic equipment; C es-inv Denotes the investment cost of the energy storage equipment; C grid (t) denotes the grid electricity price at time t; P grid (t) denotes the interaction power with the grid at time t; Δt denotes the time interval; P total (t) denotes the total power of the system at time t; Denotes the average value of the total power; P loat (t) denotes the load demand power at time t; γ bat and γ cap Respectively denote the life loss cost coefficients of the battery and the supercapacitor;

[0155] And add the power balance constraint, the energy storage capacity and charge-discharge power constraint, and the grid interaction power constraint as constraint conditions;

[0156] Among them, the power balance constraint expression is as follows:

[0157] P w (t) + P p (t) + P cap (t) + P bat (t) = P load (t) + P grid (t) (14);

[0158] The energy storage capacity and charge-discharge power constraint expression is as follows:

[0159]

[0160] In the formula, E cap (t) denotes the capacity of the supercapacitor at time t; E cap-min and E cap-max Respectively denote the minimum and maximum allowable capacities of the supercapacitor; E bat (t) denotes the capacity of the battery at time t; E bat-min and E bat-max Respectively denote the minimum and maximum allowable capacities of the battery; and Respectively denote the maximum allowable discharge and charge powers of the supercapacitor; and Respectively denote the maximum allowable discharge and charge powers of the battery;

[0161] The grid interaction power constraint expression is as follows:

[0162] P grid-min ≤ P grid (t) ≤ P grid-max (16);

[0163] In the formula, Pgrid-min and P grid-max respectively represent the minimum and maximum allowable values of the power interacting with the power grid;

[0164] S42. Solve using an improved multi-objective African vulture optimization algorithm;

[0165] S421. Initialize the population: Randomly generate a set of initial solutions including wind power installed capacity, photovoltaic installed capacity, and energy storage capacity as the population;

[0166] S422. Construct a comprehensive fitness function F using the linear weighted method:

[0167]

[0169] In the formula, ω5, ω6, ω7, and ω8 all represent weight coefficients; C total,min and C total,max respectively represent the minimum and maximum values of the total system cost in the current population; σ fluct,min and σ fluct,max respectively represent the minimum and maximum values of the standard deviation of power fluctuations in the current population; Ri min and Ri max respectively represent the minimum and maximum values of the reliability index in the current population; C es-loss,min and C es-loss,max respectively represent the minimum and maximum values of the energy storage life loss cost in the current population;

[0170] S423. Grouping and cooperation: Sort the population according to the fitness value. Assume the population is divided into k groups, and each group is led by a leader. Then, the relationship between the member i in the j-th group and the leader is as follows:

[0171]

[0172] In the formula, D ij represents the distance between the member i in the j-th group and the leader in the decision variable space; x id represents the d-th decision variable of the member i; x ld represents the d-th decision variable of the leader; D represents the dimension of the decision variable;

[0173] S424. Update the position: According to the foraging behavior rules of the vultures, update the position of each solution in formula (18), and at the same time ensure that the updated solutions satisfy the added constraint conditions;

[0174] S425. Judge the termination condition: When the preset termination condition is met, stop the iteration and output the Pareto front solution set.

[0175] S5. Real-time Feedback and Dynamic Calibration: Monitor the system status by collecting real-time data, dynamically calibrate the multi-time-scale collaborative allocation model based on the prediction error, adjust the power allocation strategy of the energy storage device, and form a closed-loop optimization.

[0176] Step S5 specifically includes the following steps:

[0177] S51. Real-time Data Collection: Obtain the wind power P w (t), photovoltaic power P p (t), load demand P load (t), and the state of charge SOC bat (t) and SOC cap (t) of the energy storage device at the current moment t to form a real-time state vector X real (t):

[0178] X real (t) = [P w (t), P p (t), P load (t), SOC bat (t), SOC cap (t)] (19);

[0179] S52. Calculate the Deviation between the Real-time Power and the Predicted Power in Step S2:

[0180]

[0181] In the formula, ΔP w (t) and ΔP p (t) respectively represent the deviations of the predicted powers of wind power and photovoltaic power;

[0182] S53. Dynamically Adjust the Robust Coefficient Γ Based on the Deviation ΔP w (t) of the Predicted Wind Power and the Deviation ΔP p (t) of the Predicted Photovoltaic Power:

[0183]

[0184] In the formula, Γ(t) and Γ(t - 1) respectively represent the robust coefficients at time t and t - 1; k represents the adaptive adjustment factor, and 0.1 ≤ κ ≤ 0.5;

[0185] Meanwhile, based on the real-time state of charge SOC bat (t) and SOC cap (t) of the energy storage device, correct the energy storage state update equation:

[0186]

[0187] In the formula, SOCbat-corrected (t) and SOC cap-corrected (t) represents the state of charge of the updated battery and supercapacitor; δ bat and δ cap respectively represent the state correction coefficients of the battery and supercapacitor;

[0188] S54. Calculate the net load P net (t) to be borne by the energy storage and the power grid according to the real-time load and renewable energy output, and add the constraint condition:

[0189] P net (t) = P load (t) - P w (t) - P p (t) (24);

[0190]

[0191] In the formula, P bat-min and P bat-max respectively represent the lower and upper limits of the battery charge and discharge power; P cap-min and P cap-max respectively represent the lower and upper limits of the supercapacitor charge and discharge power; SOC bat-min and SOC bat-max respectively represent the lower and upper limits of the safe range of the battery state of charge; SOC cap-min and SOC cap-max respectively represent the lower and upper limits of the safe range of the supercapacitor state of charge;

[0192] S55. Dynamically allocate the energy storage power according to the net load P net (t), the high-frequency fluctuation component P high (t) and the low-frequency base load component P low (t):

[0193] When |P net (t)| ≤ P bat-max + P cap-max At this time, for high-frequency fluctuations: P cap-adjust (t) = P high (t), P cap-adjust (t) represents the charge and discharge power of the supercapacitor after dynamic adjustment; for low-frequency base load: P bat-adjust (t) = P low (t), P bat-adjust (t) represents the charge and discharge power of the battery after dynamic adjustment;

[0194] Otherwise, trigger grid interaction:

[0195] P grid-adjust (t) = sgn(P net(t))·max(0,|P net (t)|-(P bat-max +P cap-max )) (26);

[0196] In the formula, P grid-adjust (t) represents the grid interaction power after dynamic adjustment.

[0197] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that they can still modify or equivalently replace the technical solutions of the present invention, and these modifications or equivalent replacements cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A coordinated optimization method for the ratio of wind power, photovoltaic power and energy storage, characterized in that: The following steps are involved: S1. Multi-dimensional data collection and preprocessing: Collect historical wind power data, photovoltaic historical power data, load demand data, real-time meteorological data, energy storage equipment parameters and economic parameters to form a multi-dimensional data set, and preprocess the multi-dimensional data set to obtain a standardized data set and load decomposition results; S2. Construct robust uncertainty set: Based on the standardized data set and the historical distribution data of wind power prediction errors and photovoltaic prediction errors, an improved box uncertainty set is constructed to describe the power fluctuation range; S3. Construct a multi-time scale collaborative allocation model: Based on the box uncertainty set and load decomposition results, construct a multi-time scale collaborative allocation model to determine the target power allocation scheme of the energy storage equipment and the energy storage life loss evaluation index; S4. Dynamic multi-objective optimization solution: Taking the minimization of system total cost, minimization of power fluctuation standard deviation, maximization of reliability index and minimization of energy storage life loss cost as multi-objectives, a multi-objective function is constructed, and combined with power balance constraints, energy storage capacity and charging and discharging power constraints and grid interaction power constraints, the improved multi-objective African Vulture optimization algorithm is used to solve and obtain the Pareto frontier solution set; S5. Real-time feedback and dynamic correction: Monitor the system status through real-time data collection, dynamically correct the multi-time scale collaborative allocation model based on the prediction error, adjust the power allocation strategy of the energy storage equipment, and form a closed-loop optimization.

2. The coordinated optimization method for the ratio of wind power, photovoltaic power and energy storage according to claim 1, wherein: Step S1 specifically includes the following steps: S11, collect historical wind power and photovoltaic power data, real-time meteorological data, load data, energy storage equipment parameters and economic parameters to form a multi-dimensional data set; S12, use the 3σ rule to remove abnormal data, and fill in missing values ​​through bidirectional LSTM to construct a spatiotemporal correlation data set; S13. Based on wavelet packet decomposition of load demand data, separate the high-frequency fluctuation component P high (t) and the low-frequency base load component P low (t); S14. Establish the energy storage state update equation: Where, SOC bat (t) and SOC bat (t - 1) represent the state of charge of the battery at times t and t - 1 respectively; SOC cap (t) and SOC cap (t - 1) represent the state of charge of the supercapacitor at times t and t - 1 respectively; ΔT represents the sampling interval; P bat (t) and P cap (t) represent the charging and discharging powers of the battery and the supercapacitor at time t respectively; η bat and η cap represent the charging and discharging efficiencies of the battery and the supercapacitor respectively; E bat and E cap represent the rated capacities of the battery and the supercapacitor respectively; and represent the indicator functions used to judge the charging and discharging states of the battery and the supercapacitor respectively; S15. Output the preprocessed standardized dataset D and the load decomposition results {P high (t), P low (t)} after normalization; in, D = {P w (t), P p (t), P load (t), SOC bat (t), SOC cap (t)} (2); Where, P w (t), P p (t) and P load (t) represent the wind power data, photovoltaic power data, and load demand data at time t, respectively.

3. The coordinated optimization method for the wind power-photovoltaic energy storage ratio according to claim 2, wherein: Step S2 specifically includes the following steps: S21. Construct a box uncertainty set H: In the formula, and respectively represent the predicted power of wind power and photovoltaic considering uncertainties at time t, and W represents the set of wind farms and photovoltaic array devices; β w (t) and β p (t) respectively represent the variables of wind power and photovoltaic constrained by the robust coefficient Γ; and respectively represent the maximum absolute deviations of the predicted power of wind power and photovoltaic; Γ represents the robust coefficient, and 0 ≤ Γ ≤ |W|; S22. Introduce the robustness of Wasserstein distance metric: Where, W(P w , P p ) represents the Wasserstein distance between wind power P w and photovoltaic power P p ; Π represents the joint probability distribution; represents the distance from the point w in wind power P to the point p in photovoltaic power P ; S23. Output the box uncertainty set H and the robustness parameters {Γ, W(P w , P p )}.

4. The coordinated optimization method for the ratio of wind power, photovoltaic power and energy storage according to claim 3, characterized in that: Step S3 specifically includes the following steps: S31. Use the variational mode decomposition algorithm to perform frequency-domain decomposition on the high-frequency fluctuation component P high (t), and allocate the components with frequencies greater than a specific threshold to the supercapacitor. The target power P cap-ref (t) formula is as follows: P cap-ref (t) = λ(t)·P high (t) (5); Wherein, λ(t) represents the dynamic allocation coefficient, and 0.8≤λ(t)≤1.0; S32. Allocate the low-frequency base load component \(P_{ low}(t)\) and the remaining power distributed by the supercapacitor to the battery: low (t) and the remaining power distributed by the supercapacitor to the battery: P bat-ref P(t) = P low P(t) + (1 - λ(t))·P high P(t) (6); Wherein, P bat-ref (t) represents the power borne by the battery, and and respectively represent the maximum allowable discharge power of the battery and the maximum allowable charge power of the battery; S33. Establishing an energy storage life loss assessment model: Wherein, L bat and L cap respectively represent the life loss of the battery and the supercapacitor; α and β respectively represent the life loss coefficients of the battery and the supercapacitor; ΔSOC bat (t) represents the change in the state of charge of the battery at time t; DOD max represents the maximum depth of discharge of the battery; S34. Output the target power distribution scheme of the energy storage device {P cap-ref (t), P bat-ref (t)}, and the evaluation index of energy storage life loss {L bat , L cap}.

5. The coordinated optimization method for the wind power photovoltaic energy storage ratio according to claim 4, characterized in that: In step S31, when 0.3 ≤ SOC cap (t) ≤ 0.7, λ(t) = 1.0, indicating that the supercapacitor fully bears the high-frequency power fluctuation; When the SOC cap (t) < 0.3, to avoid over-discharging of the super capacitor; When the SOC cap (t) > 0.7, to avoid overcharging the supercapacitor.

6. The coordinated optimization method for the wind power photovoltaic energy storage ratio according to claim 5, characterized in that: Step S4 The specific steps include: S41. Construct multi-objective function: minJ = ω1C total + ω2σ fluct + ω3(1 - Ri)+ ω4C es-loss (9); In the formula, J represents the multi-objective function value; ω1, ω2, ω3, and ω4 all represent weight coefficients, and ω1 + ω2 + ω3 + ω4 = 1; C total represents the total system cost; σ fluct represents the standard deviation of power fluctuation; Ri represents the reliability index; C es-loss represents the cost of energy storage life loss; in, C es-loss = γ bat L bat + γ cap L cap (13); Where, C wind-inv represents the investment cost of wind power equipment; C pv-inv represents the investment cost of photovoltaic equipment; C es-inv represents the investment cost of energy storage equipment; C grid (t) represents the grid electricity price at time t; P grid (t) represents the interaction power with the grid at time t; Δt represents the time interval; P total (t) represents the total power of the system at time t; represents the average value of the total power; P loat (t) represents the load demand power at time t; γ bat and γ cap respectively represent the life loss cost coefficients of the battery and the supercapacitor; And add power balance constraints, energy storage capacity and charging and discharging power constraints, and grid interaction power constraints as constraints; The power balance constraint expression is as follows: P w (t) + P p (t) + P cap (t) + P bat (t) = P load (t) + P grid (t) (14); The energy storage capacity and charge / discharge power constraint expressions are as follows: where, E cap (t) represents the capacity of the supercapacitor at time t; E cap-min and E cap-max represent the minimum and maximum allowable capacities of the supercapacitor, respectively; E bat (t) represents the capacity of the battery at time t; E bat-min and E bat-max represent the minimum and maximum allowable capacities of the battery, respectively; and represent the maximum allowable discharge and charge powers of the supercapacitor, respectively; and represent the maximum allowable discharge and charge powers of the battery, respectively; The grid interaction power constraint expression is as follows: P grid-min ≤P grid (t)≤P grid-max (16); where P grid-min and P grid-max respectively represent the minimum and maximum allowable values of the power interacting with the power grid; S42, using an improved multi-objective African vulture optimization algorithm to solve; S421, initializing the population: randomly generating a set of initial solutions including wind power installed capacity, photovoltaic installed capacity, and energy storage capacity as the population; S422, using linear weighted method to construct comprehensive fitness function F: In the formula, ω5, ω6, ω7, and ω8 all represent weight coefficients; C total,min and C total,max respectively represent the minimum and maximum values of the total system cost in the current population; σ fluct,min and σ fluct,max respectively represent the minimum and maximum values of the standard deviation of power fluctuation in the current population; Ri min and Ri max respectively represent the minimum and maximum values of the reliability index in the current population; C es-loss,min and C es-loss,max respectively represent the minimum and maximum values of the energy storage life loss cost in the current population; S423, Grouping and collaboration: Sort the population according to the fitness value. Assuming that the population is divided into k groups, each group is led by a leader, then the relationship between member i and the leader in the jth group is as follows: where D ij represents the distance between member i in the j-th group and the leader in the decision variable space; x id represents the d-th decision variable of member i; x ld represents the d-th decision variable of the leader; D represents the dimension of the decision variable; S424. Update position: According to the foraging behavior rules of vultures, update the position of each solution in formula (18), while ensuring that the updated solution satisfies the added constraint conditions; S425. Judge the termination condition: When the preset termination condition is met, stop the iteration and output the Pareto front solution set.

7. The coordinated optimization method for the wind power and photovoltaic energy storage ratio according to claim 6, characterized in that: Step S5 specifically includes the following steps: S51. Real-time data acquisition: Obtain the wind power P w (t), photovoltaic power P p (t), load demand P load (t), and the state of charge SOC of the energy storage device bat (t) and SOC cap (t), to form the real-time state vector X real (t): X real (t) = [P w (t), P p (t), P load (t), SOC bat (t), SOC cap (t)] (19); S52. Calculate the deviation between the real-time power and the predicted power in step S2: where ΔP w (t) and ΔP p (t) represent the deviations of the predicted power of wind power and photovoltaic power respectively; S53. Deviation ΔP w of the predicted power of wind power and deviation ΔP p (t) of the predicted power of photovoltaic power to dynamically adjust the robust coefficient Γ: In the formula, Γ(t) and Γ(t - 1) respectively represent the robustness coefficients at time t and t - 1; k represents the adaptive adjustment factor, and 0.1 ≤ k ≤ 0.5; Meanwhile, based on the state of charge SOC of the real-time energy storage device bat (t) and SOC cap (t), correct the energy storage state update equation: where SOC bat-corrected (t) and SOC cap-corrected (t) represent the state of charge of the updated battery and supercapacitor; δ bat and δ cap respectively represent the state correction coefficients of the battery and supercapacitor; S54. Calculate the net load P net (t) to be borne by the energy storage and the power grid according to the real-time load and the renewable energy output, and add the constraint condition: P net P(t) = P load P(t) - P w P(t) - P p P(t)(24); Wherein, P bat-min and P bat-max respectively represent the lower limit and the upper limit of the charging and discharging power of the battery; P cap-min and P cap-max respectively represent the lower limit and the upper limit of the charging and discharging power of the supercapacitor; SOC bat-min and SOC bat-max respectively represent the lower limit and the upper limit of the safe range of the state of charge of the battery; SOC cap-min and SOC cap-max respectively represent the lower limit and the upper limit of the safe range of the state of charge of the supercapacitor; S55. According to the net load P net (t), the high-frequency fluctuation component P high (t) and the low-frequency base load component P low (t) dynamically allocate the energy storage power: When |P net (t)| ≤ P bat-max + P cap-max At this time, high-frequency fluctuation: P cap-adjust (t) = P high (t), P cap-adjust (t) represents the charge-discharge power of the supercapacitor after dynamic adjustment; low-frequency base load: P bat-adjust (t) = P low (t), P bat-adjust (t) represents the charge-discharge power of the battery after dynamic adjustment; Otherwise, trigger grid interaction: P grid-adjust (t) = sgn(P net (t))·max(0, |P new (t)| - (P bat-max + P cap-max ))(26); where P grid-adjust (t) represents the grid interaction power after dynamic adjustment.

Images