Wind power smoothing method based on double-layer MAC and VMD adaptive frequency division

Abstract

Based on double-layer MAC and VMD adaptive frequency of wind power smoothing method, the unbalanced power compensation strategy is formulated. When the unbalanced power is greater than zero, if there is no uncertain load at this time, the positive unbalanced power and the battery A jointly compensate the uncertain load, reduce the action of battery A; if there is no uncertain load at this time, the battery A absorbs the unbalanced power. When the unbalanced power is less than zero, the energy storage system needs to compensate the unbalanced power, and the unbalanced power is compensated first. Formulate the wind power fluctuation suppression strategy, and use hybrid energy storage to realize the wind power fluctuation suppression within five minutes time scale. Extract the compensation power in the fluctuation rate demand, and according to the characteristics of super capacitor and battery, the compensation power is divided into frequency, so that the super capacitor suppresses the high frequency and low amplitude fluctuation, and the battery B suppresses the low frequency and high amplitude fluctuation. Establish the suppression model based on double-layer model algorithm control.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of wind power application, and particularly relates to a wind power smoothing method based on double-layer MAC and VMD adaptive frequency division. BACKGROUND

[0002] As a clean renewable energy, wind power has a good development prospect, but the inherent uncertainty and volatility of wind power bring many problems to the stable operation of power system. (I) The mismatch between the actual output power of wind power and the demand power of power grid leads to unbalanced power, and the positive unbalanced power will lead to wind power curtailment, causing waste of wind resources, while the negative unbalanced power will lead to insufficient grid-connected power, thus the wind farm will suffer corresponding economic losses; (II) While completing the grid-connected task, some wind farms will also directly supply power to the nearby load, but the emergence of uncertain load will affect the actual grid-connected power of wind power, and when the wind farm has excess power, it has enough wind power to be curtailed to cope with the uncertain load; while the wind farm has no excess power, the uncertain load is difficult to meet, affecting the user satisfaction; (III) When the wind power fluctuation exceeds the requirement range of the power grid, it will have a certain impact on the power grid, affecting the stable operation of the power grid.

[0003] As one of the 10 key technologies affecting the future development of power grid, energy storage is a current research hotspot for solving the above problems. The addition of energy storage system in the wind power system can reduce the unbalanced power, maximize the satisfaction of uncertain load, smooth the wind power fluctuation, ensure that the wind power meets the grid fluctuation requirements, improve the safety and stability of the power grid operation, and reduce the economic losses of the wind farm.

[0004] The current application research of wind farm energy storage mainly has the following points: (1) Using energy storage system to compensate the wind power prediction difference and improve the wind power prediction accuracy. Due to the uncertainty of wind power and the inaccuracy of prediction method, the error between the predicted power and the actual power of the wind farm is large, which will affect the real-time scheduling of the power system, thus causing economic losses. (2) Using energy storage to realize wind power consumption, actual output tracking load, and reducing the anti-peaking characteristics of the wind farm. Due to the uncertainty of wind power and load, the wind power curtailment is large, and the load peak-valley difference is large. When the load is large, the actual output power of wind power is small, which cannot meet the load power supply; while the load is small, the actual output power of the wind farm is large, causing wind power curtailment. (3) Using energy storage to smooth the output fluctuation of wind power and maintain the safe and stable operation of the power grid. Due to the fluctuation of wind power, the actual output power of the wind farm has large fluctuation, and direct grid connection will have an impact on the power grid, affecting the stable operation of the power system.

[0005] But a more serious problem of energy storage system is that the energy storage capacity is small, the economic cost is high, and it is difficult to store electric energy on a large scale and long-term effectively, and the absorption and inhibition of high proportion of renewable energy is very limited. Therefore, a research hotspot at present is the capacity planning of energy storage system, that is, using various means and methods to make the energy storage achieve the expected purpose in small capacity. The control strategy of energy storage is formulated according to different optimization objectives of wind farm, and the size of energy storage capacity determines the effect of the implementation of the control strategy. The energy storage capacity optimization method can be divided into single objective optimization method and multi-objective optimization method. The main objective of single objective optimization method is to minimize cost or maximize benefit, and the main objective of multi-objective optimization method is to maximize power system reliability and minimize cost.

[0006] Wind storage combined system and energy storage characteristic modeling

[0007] Wind storage combined system structure

[0008] Figure 1 The wind storage combined system shown comprises a wind farm, a power grid, a user, a storage battery A and a hybrid energy storage. The storage battery A acts on the unbalanced power of the wind farm and the burst load, realizes compensation for the unbalanced power and the burst load, compensates the burst load on the basis that the wind power meets the wind power grid connection demand, and improves user energy use satisfaction. The hybrid energy storage suppresses the compensated wind power, so that the fluctuation is controlled within the grid connection standard range. On the basis of the structure, control strategies for the two groups of energy storage systems are formulated, so that the two groups can better complete the control target, and the capacity of the corresponding energy storage is solved for the control strategies. The effective method of extracting high frequency and low frequency of wind power fluctuation is proposed by considering the wind power fluctuation characteristics of the hybrid energy storage, so that the charge and discharge characteristics of the super capacitor and the battery in the hybrid energy storage are fully utilized.

[0009] Energy storage characteristic modeling

[0010] Energy storage is an effective device for realizing energy time shift, which can store excess energy and release it when needed. In the present application, two groups of energy storage, storage battery A and hybrid energy storage, are configured in the wind farm. The hybrid energy storage is composed of a super capacitor and a storage battery B, and is used for optimizing wind power output and realizing efficient use of energy. When the energy storage is charged, the current time energy storage available charging power is determined by the upper limit of the energy storage SOC and the residual capacity of the energy storage at the previous time. When the energy storage is discharged, the current energy storage available discharging power is determined by the lower limit of the energy storage SOC and the residual capacity of the energy storage at the previous time. The calculation formula is:

[0011]

[0012]

[0013]

[0014] S k0 ≤S k (t)≤S kM (4)

[0015] where k = {A, b, s}, where A represents battery A, b represents battery B, and s represents super capacitor. S k (t) is the state of charge of device k at time t, S kM is the maximum state of charge of device k, S k0 is the minimum state of charge of device k, E k is the rated capacity of device k, P kt (t) is the available charge or discharge power of device k at time t, E kt (t-Δt) is the remaining capacity of device k at the previous time.

[0016] The general constraints of the energy storage capacity configuration process are rated power constraints and rated capacity constraints.

[0017] The rated power constraint is

[0018]

[0019] The rated capacity constraint is

[0020]

[0021]

[0022] where P k is the rated power of device k, η kc / η kd is the charge / discharge efficiency of device k, E k' is the maximum remaining capacity of device k within the scheduling period; E k0 is the initial capacity of device k. SUMMARY

[0023] To solve the problems in the prior art, the purpose of the present application is to provide a wind power smoothing method based on double-layer MAC and VMD adaptive frequency division, and the present application proposes an unbalanced power compensation strategy and a wind power fluctuation suppression strategy. The main contributions are as follows. The unbalanced power compensation strategy will consider the uncertainty of the load, and first compensate the unbalanced power by using the battery A, and then compensate the uncertain load. In order to improve the adaptability of the energy storage control strategy and the capacity, the strategy is embedded into a multi-objective Harris hawk optimization (MOHHO) algorithm for configuring the capacity of the battery A. An improved variational mode decomposition (VMD) adaptive frequency division method is proposed to divide the compensation power, so that the super capacitor acts on the high-frequency scope and the battery B acts on the low-frequency scope. According to the frequency division result, a hybrid algorithm of particle swarm optimization and grey wolf optimization (PSO-GWO) is used to solve the optimal ratio of the hybrid energy storage capacity with the minimum daily average operating cost as the objective function. The wind power fluctuation suppression strategy is mainly completed by the hybrid energy storage system composed of the battery B and the super capacitor. Based on the frequency division result, a hybrid energy storage wind power fluctuation suppression model based on a double-layer model algorithm control (MAC) is proposed.

[0024] To achieve the above purpose, the technical scheme of the present application is:

[0025] The wind power smoothing method based on double-layer MAC and VMD adaptive frequency division is composed of the following steps:

[0026] Step one, battery A control strategy and capacity configuration

[0027] 1.1 Battery A compensation unbalanced power strategy

[0028] The unbalanced power of the wind farm is caused by the mismatch between the actual power of the wind farm and the power demand of the grid, which can be expressed as:

[0029] P(t) = P y (t) - P w (t) (8)

[0030] In the formula, P(t) is the unbalanced power at time t, P y (t) is the actual power of the wind farm at time t, and P w (t) is the power demand of the grid at time t.

[0031] P f (t) represents the uncertain load at time t, P c (t) represents the positive imbalance power at time t, P At1 (t) represents the battery A power consumption for compensating the imbalance power at time t, P At2 (t) represents the battery A power consumption for compensating the uncertain load at time t.

[0032] For the imbalance power of the wind farm, and considering the uncertain load, a strategy of compensating the imbalance power by the energy storage is developed. The strategy can be summarized as:

[0033] (1) P(t) > 0, P f (t) > 0, the imbalance power is positive, and there is an uncertain load

[0034] When P(t) > P f (t), the imbalance power can directly compensate the uncertain load, P c (t) = P f (t). The battery A charges to absorb the positive imbalance power, P At1 (t) = min{ (P(t) - P f (t)), E At (t - Δt)}.

[0035] When 0 < P(t) < P f (t), the imbalance power can compensate part of the uncertain load, P c (t) = P f (t). The battery A discharges to compensate the remaining load, P At2 (t) = -min{ (P f (t) - P(t)), E At (t - Δt)}.

[0036] (2) P(t) > 0, P f (t) = 0, the imbalance power is positive, and there is no uncertain load, the battery A only needs to charge to absorb the positive imbalance power, P At1 (t) = min{P(t), E At (t - Δt)}.

[0037] (3) P(t) < 0, P f (t) > 0, the imbalance power is negative, and there is an uncertain load, at this time the battery A first compensates the negative imbalance power, P At1 (t) = -min{ |P(t)|, E At (t - Δt)}

[0038] When E At(t-Δt)-P At1 (t)>0, discharge compensation for uncertain load, P At2 (t)=-min{P f (t),(E At (t-Δt)-P At1 (t))

[0039] When E At (t-Δt)-P At1 (t)=0, stop charging and discharging when there is no surplus power in battery A, P At2 (t)=0

[0040] (4)P(t)<0, P f (t)=0, unbalanced power is negative, and there is no uncertain load, battery A discharges to compensate for unbalanced power, P At1 (t)=-min{|P(t)|,E At (t-Δt)}

[0041] The charging and discharging process of battery A follows formulas (1)-(4).

[0042] 1.2 Battery A capacity configuration

[0043] Based on the compensation strategy for unbalanced power of battery A formulated in step one, considering the compensation for unbalanced power and uncertain load, a battery A capacity configuration model is constructed. Because the positive unbalanced power is large, a very large capacity of energy storage is needed to ensure complete absorption. Considering the economic impact, mainly absorbing part of the positive unbalanced power, for the purpose of compensating for negative unbalanced power and uncertain load.

[0044] 1.2.1 Objective function

[0045] Considering the economic impact, the minimum daily operating cost of battery A is taken as the objective function.

[0046]

[0047] In the formula, r is the annual interest rate; y A is the life of battery A; P A , E A are the rated power (kW) and rated capacity (kWh) of battery A, respectively; C bp , C be are the unit power cost (yuan / kW) and unit capacity cost (yuan / kWh) of battery, respectively; C rb is the maintenance cost of battery (yuan / kWh).

[0048] Considering the compensation effect of unbalanced power and uncertain load, the maximum wind energy utilization rate is taken as the objective function.

[0049]

[0050] In the formula, is the unbalanced power absorbed by the battery A; is the unbalanced power greater than the on-line requirement.

[0051] 1.2.2 Constraint conditions

[0052] The battery A must satisfy the charge-discharge power constraint at any time t:

[0053] P At (t)=P At1 (t)+P At2 (t) (11)

[0054] The battery A rated power constraint, rated capacity constraint, SOC state constraint follow formula (5)-(7).

[0055] 1.2.3 Solution algorithm

[0056] Harris Hawks Optimizer (HHO) is a gradient-free optimization algorithm that simulates the hunting behavior of Harris Hawks, mainly composed of three parts: search phase, search and development conversion and development phase. Multi-objective Harris Hawks Optimizer (MOHHO) is an optimization algorithm suitable for multi-objective problems based on HHO, which adds an archive to save and retrieve Pareto optimal results in HHO algorithm, and has been proved to have good global search ability. The mixed energy storage double-layer planning model of the mixed energy storage operation strategy is embedded in the MOHHO algorithm to improve the matching of the control strategy and the configuration method.

[0057] The battery A compensation unbalanced power strategy of section 3.1 is embedded into MOHHO, after each iteration, the algorithm can update the rated power and rated capacity of battery A according to the power, capacity and SOC constraint in the operation strategy, until the constraint condition is met, and the configured energy storage can absorb part of the positive unbalanced power to compensate for the negative unbalanced power and uncertain load.

[0058] Step two, adaptive frequency division of hybrid energy storage capacity configuration based on VMD

[0059] Section 1.1 uses the battery A to achieve compensation for unbalanced power, and obtains the grid-connected power that meets the grid demand. According to the grid-connected power fluctuation requirement, the compensation power required to be stored and discharged to meet the fluctuation requirement is extracted. According to the characteristics of supercapacitors and batteries, the VMD adaptive frequency division method is first used to divide the compensation power, so that the hybrid energy storage operates in its respective frequency range. According to the frequency division result, a hybrid energy storage capacity optimal matching model is constructed.

[0060] 2.1 VMD adaptive frequency division model

[0061] Variational mode decomposition is a signal decomposition estimation method. In the process of obtaining the decomposition components, the frequency center and bandwidth of each component are determined by iteratively searching for the optimal solution of the variational model, so that the frequency domain of the signal can be adaptively divided and each component can be effectively separated. The decomposition formula is:

[0062]

[0063] In the formula, {u k}={u1,u2,...,u k},{ω k}={ω1,ω2,...,ω k}, The subscripts in the formula represent the square sum L2 norm, and F(t) is the original compensation power to be decomposed.

[0064] The compensation power is divided into high-frequency components and low-frequency components, and the characteristics of supercapacitors and batteries are fully utilized, so that the supercapacitors can suppress high-frequency low-amplitude fluctuations, and the battery B can suppress low-frequency high-amplitude fluctuations. The reason for choosing VMD is that it can minimize the sum of the estimated bandwidth of each mode. The alternating direction multiplier method is used to update the modes and their center frequencies, and gradually demodulate each mode to the corresponding base frequency band. Finally, each mode, i.e. the corresponding center frequency, is extracted, which has very good frequency division effect.

[0065] In order to achieve the purpose of suppressing high-frequency low-amplitude fluctuations by supercapacitors and suppressing low-frequency high-amplitude fluctuations by battery B, a VMD component adaptive reconstruction method is proposed, that is, different reconstruction ranges are selected for each component at each time during VMD component reconstruction. K represents the number of components required to reconstruct low-frequency components. During the reconstruction process, if the amplitude of the high-frequency component is greater than that of the low-frequency component, the value of K is increased. The high-frequency and low-frequency reconstruction formulas are:

[0066]

[0067]

[0068]

[0069] wherein u k (t) is the kth component of the compensation power decomposed by VMD at time t; u L (t), u L (t) are the high and low frequency components of the initial reconstructed compensation power at time t, respectively.

[0070] However, the above VMD component reconstruction method is not accurate because the selection of the number N of VMD decomposition is not accurate, which can lead to over-decomposition of the original signal. In addition, the sum of the high and low frequency components after VMD component reconstruction is smaller than the amplitude of the original signal. The high and low frequency components reconstructed by the present application are used as the basis for configuring the mixed energy storage capacity. Inaccurate decomposition can lead to inaccurate configuration of the energy storage capacity, and ultimately lead to the inability of energy storage to smooth wind power fluctuations. Therefore, the reconstruction process is adjusted as follows:

[0071] (1) If over-decomposition occurs, the high frequency component value at this moment is adjusted to be the same as the amplitude of the original signal;

[0072] (2) At the same time, the original signal is subtracted from the adjusted high frequency component to obtain the adjusted low frequency component, which is used to solve the problem that the amplitude of the VMD component after reconstruction is different from that of the original signal.

[0073]

[0074] P H (t) = u H1 (t) (17)

[0075] wherein u H1 (t) is the adjusted high frequency component; P L (t) is the low frequency component of the compensation power at time t; P H (t) is the high frequency component of the compensation power at time t.

[0076] 2.2 Mixed energy storage capacity configuration model

[0077] According to the VMD adaptive frequency division result in section 2.1, the mixed energy storage is used to smooth in the respective frequency range, and the minimum daily operation cost of the mixed energy storage is used as the objective function.

[0078] 2.2.1 Objective function

[0079] The energy storage operation cost includes investment cost and maintenance cost:

[0080] min C = C b + C sc + C r (18)

[0081] wherein C is the energy storage operation cost; C b , Cs The investment costs for battery B and supercapacitor are respectively; C r The maintenance cost of hybrid energy storage is expressed as follows:

[0082]

[0083]

[0084]

[0085] In the formula, y b y s The lifespans of battery B and supercapacitor are respectively; C sp C se These represent the unit power cost (yuan / kW) and unit capacity cost (yuan / kWh) of supercapacitors, respectively; C rs Maintenance cost of supercapacitors (RMB / kWh);

[0086] 2.2.2 Constraints

[0087] Supercapacitors operate within their respective frequency ranges, subject to the following constraints:

[0088]

[0089]

[0090] P L (t)+P H (t)=P x (t) (24)

[0091] In the formula, P x (t) represents the compensation power that satisfies the volatility at time t.

[0092] The rated power constraint, rated capacity constraint, and SOC state constraint of hybrid energy storage follow formulas (5)-(7).

[0093] 2.2.3 Solution Method

[0094] The Gray-Wolf algorithm is an optimization search method inspired by the predatory behavior of gray wolves. This algorithm boasts strong convergence, few parameters, and ease of implementation. However, its position update equation suffers from strong exploitation capabilities but weak exploration capabilities. Optimal allocation of dual-repository capacity requires simultaneously satisfying different constraints, making it a nonlinear, multi-constraint, and multivariate problem-solving process. Particle Swarm Optimization (PSO) updates local and global optimizations by updating position and velocity until the optimal value is reached. The PSO-GWO algorithm is a new algorithm that improves upon the GWO algorithm through PSO, giving it better global search capabilities.

[0095] The basic idea of the algorithm is: starting from a randomly generated initial population, a new individual is generated by summing the vector difference of any two individuals in the population with a third individual, and then comparing the new individual with the corresponding individual in the current population. If the fitness of the new individual is better than that of the current individual, the new individual replaces the old individual in the next generation, otherwise the old individual is still saved. Through continuous evolution, good individuals are retained and poor individuals are eliminated, guiding the search to approach the optimal solution. In this optimization process, the VMD adaptive frequency division result and the constraint conditions of equations (21)-(26) are embedded in the algorithm, and the PSO-GWO is used to solve the optimal proportioning of mixed energy storage capacity according to equation (20).

[0096] Step three, the suppression model based on the double-layer model algorithm control

[0097] According to the capacity of supercapacitor and battery B solved in section 1, a suppression model based on double-layer model algorithm control is constructed. The upper model mainly uses supercapacitor to suppress high-frequency low-amplitude fluctuations, and the lower model uses battery B to suppress low-frequency high-amplitude fluctuations.

[0098] MAC is a form of MPC, which can realize open-loop optimal control within a limited step length. In terms of energy storage suppression of wind power grid fluctuation rate, its time domain rolling optimization process is: 1) establish an optimization model according to the SOC state constraint of energy storage and the grid fluctuation constraint; 2) according to the constraint conditions, solve the control instruction sequence of energy storage in the future period of time, with the SOC state variable value and the actual output value of wind power at the current time; 3) apply the first value of the control instruction sequence to the mixed energy storage control system; 4) roll to the next time, update the SOC state variable value and the grid power value, and repeat the above process. MAC algorithm uses the idea of rolling prediction and advance control, which can effectively solve the output of energy storage in the future period of time according to the current actual output of wind power and the constraint conditions.

[0099] 3.1 Upper high-frequency fluctuation suppression model

[0100] For the use of supercapacitor to suppress high-frequency fluctuations, the relationship between power and energy storage SOC change is:

[0101] P Hf (k)=P w (k)-P L (k) (25)

[0102] P Hy (k+1)=P Hf (k)-P st (k) (26)

[0103]

[0104] where P Hf (k) is the actual power without low frequency components at time k; P Hy (k+1) is the power after suppressing high frequency components at time k+1; T represents the control period of the super capacitor.

[0105]

[0106] where S st0 is the initial SOC state of the super capacitor; M is the prediction time length, i.e., the length of the control command sequence.

[0107] P smax ≤ P st (k+i)≤ P smax (29)

[0108] S s0 ≤ S s (k+i)≤ S sM (30)

[0109]

[0110] where P rate is the rated capacity of the wind power; δ is the limit value of the wind power grid fluctuation rate.

[0111] Suppose the state variables x1(k) = P Hy (k), x2(k) = S s (k); the control variable u(k) = P st (k); and r(k) = P Hf (k) are the disturbance variables. Then the state space equation of the wind power system containing the super capacitor is:

[0112]

[0113] Let:

[0114] The objective function formula (28) and the constraint condition formula (29)-(31) are converted into the standard type of quadratic programming:

[0115]

[0116] where z = [u(k), x T (k+1), u(k+1),..., u(k+M-1), x T (k+M)] T , H is a matrix composed of the quadratic term weight Q of the energy storage control sequence and the output level.

[0117] 3.2 Lower layer high frequency fluctuation suppression model

[0118] For the use of super capacitor to suppress high frequency fluctuation, power and energy storage SOC change relationship:

[0119] P Lf (k) = P Hy (k) + P L (k) (34)

[0120] P Ly (k+1) = P Lf (k) - P bt (k) (35)

[0121]

[0122] In the formula, P Lf (k) is the upper layer after the suppression of low frequency component of power at time k; P Ly (k+1) is the power after the suppression of low frequency component at time k+1.

[0123] Under the premise of suppressing low frequency fluctuation to meet the grid fluctuation requirements, the minimum output of battery B and charge-discharge balance, the lower MAC rolling optimization objective function is established:

[0124]

[0125] In the formula, S bt0 is the initial SOC state of battery B.

[0126] -P bmax ≤ P bt (k+i) ≤ P bmax (38)

[0127] S b0 ≤ S b (k+i) ≤ S bM (39)

[0128]

[0129] Assume that state variables y1(k) = P Ly (k), y2(k) = S b (k); control variable v(k) = P bt (k); f(k) = P Hf (k) as the disturbance. The state space equation of wind power system containing battery B is:

[0130]

[0131] Record:

[0132] The objective function formula (37) and the constraint formulas (38) to (40) are transformed into the standard form of a quadratic programming problem:

[0133]

[0134] Where w = [v(k), y T (k+1),v(k+1)....,v(k+M-1),y T (k+M)] T H is a matrix composed of the quadratic weights R and Q of the energy storage control sequence and the output level. R = 2.

[0135] According to formulas (4) to (7), we can obtain:

[0136] B b =[P bmax ,P bmax ,δ,δ,-S b0 ,S bM ]

[0137] B bq =[P bmax ,b bmax ,-S b0 ,S bM ]

[0138]

[0139] The above model is optimized and solved using MATLAB quadratic programming in each control cycle to obtain a sequence of control commands. The first value is taken as the current total control command for energy storage. This process is repeated iteratively to obtain the daily control command sequence for the supercapacitor and battery B [u(k), u(k+1), ..., u(k+M-1)]. T [v(k),v(k+1)....,v(k+M-1)] T .

[0140] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0141] This invention addresses a series of problems caused by the volatility of wind power by proposing the configuration of two sets of energy storage to solve the problems of unbalanced power, uncertain load, and wind power fluctuation in wind farms.

[0142] To address the issues of unbalanced power and uncertain loads, a collaborative strategy for battery A to compensate for these issues was developed. Considering cost and wind energy utilization, the capacity of battery A was configured using the MOHHO (Modular Energy Management System). Experiments show that the strategy of this invention can completely compensate for negative unbalanced power and uncertain loads.

[0143] An improved VMD adaptive frequency division method is proposed to divide the compensation power, so that the super capacitor acts on the high-frequency low-amplitude part, and the battery acts on the low-frequency high-amplitude part. The PSO-GWO algorithm is used to solve the optimal proportion of mixed energy storage capacity. The simulation results show that this method can well decompose the compensation power and retain the characteristics of the original signal.

[0144] In order to reduce the fluctuation of wind power network, a smoothing model based on double-layer model algorithm control is proposed. Compared with the original fluctuation rate, the grid-connected power fluctuation rate after smoothing is reduced by 16%. BRIEF DESCRIPTION OF DRAWINGS

[0145] Figure 1 Structure diagram of wind storage combined system

[0146] Figure 2 Model diagram of load power supply and energy storage A compensation unbalanced power

[0147] Figure 3 Battery A compensation unbalanced power strategy

[0148] Figure 4 Strategy flow

[0149] Figure 5 A set of Pareto solutions meeting the double objectives

[0150] Figure 6 Φ1, Φ2 corresponding to the Pareto solution set meeting the double objectives

[0151] Figure 7 Unbalanced power caused by actual power of wind farm and grid-connected demand power

[0152] Figure 8 Effect of battery A compensation unbalanced power

[0153] Figure 9 Effect of battery A compensation uncertain load

[0154] Figure 10 Effect of battery A compensation unbalanced power and uncertain load output

[0155] Figure 11 Battery SOC state change

[0156] Figure 12 Signal diagram before adjusting adaptive VMD frequency division result

[0157] Figure 13 Signal diagram after adjusting adaptive VMD frequency division result

[0158] Figure 14 Actual output of mixed energy storage to smooth wind power fluctuation

[0159] Figure 15 Hybrid SOC state change

[0160] Figure 16 Hybrid energy storage smooths wind power fluctuations. DETAILED DESCRIPTION

[0161] The technical solutions of the present application will be described in further detail below in combination with the drawings and specific embodiments:

[0162] As Figures 1-16 shown, based on the wind power smoothing method of double-layer MAC and VMD adaptive frequency division, the addition of energy storage in the wind farm can realize the compensation of wind power imbalance and the smoothing of wind power fluctuations. The imbalance power compensation strategy is formulated, and the battery is used to realize the imbalance power compensation within 5 minutes time scale and the uncertainty load compensation within 1 hour time scale. When the imbalance power is greater than zero, if there is no uncertainty load at this time, the positive imbalance power and the battery A jointly compensate the uncertainty load, reducing the action of battery A; if there is no uncertainty load at this time, the battery A absorbs the imbalance power. When the imbalance power is less than zero, the energy storage system needs to compensate the imbalance power. Because the importance of wind power grid connection is greater than that of uncertainty load, the imbalance power is preferentially compensated, and the uncertainty load is compensated when the battery A has surplus power. The wind power fluctuation smoothing strategy is formulated, and the hybrid energy storage is used to realize the wind power fluctuation smoothing within 5 minutes time scale. The compensation power within the fluctuation rate demand is extracted, and the compensation power is frequency-divided according to the characteristics of the super capacitor and the battery, so that the super capacitor smoothes the high-frequency low-amplitude fluctuation, and the battery B smoothes the low-frequency high-amplitude fluctuation. The smoothing model based on double-layer model algorithm control is established. The overall idea of the present application is shown in Figure 1 .

[0163] 1. Battery A control strategy and capacity configuration

[0164] 1.1 Battery A compensation imbalance power strategy

[0165] The wind farm imbalance power is caused by the mismatch between the wind farm actual power and the grid demand power, which can be expressed as:

[0166] P(t) = P y (t) - P w (t) (8)

[0167] In the formula, P(t) is the imbalance power at time t, P y (t) is the wind farm actual power at time t, and P w (t) is the grid demand power at time t.

[0168] P f (t) represents the uncertainty load at time t, and P c(t) represents the power consumption of the battery A to compensate for the unbalanced power at time t, P At1 (t) represents the power consumption of the battery A to compensate for the unbalanced power at time t, P At2 (t) represents the power consumption of the battery A to compensate for the unbalanced power at time t, P

[0169] For the unbalanced power of the wind farm, and considering the uncertain load, a strategy for energy storage to compensate for unbalanced power is developed. This strategy can be summarized as:

[0170] (1) P(t) > 0, P f (t) > 0, the unbalanced power is positive, and there is an uncertain load

[0171] When P(t) > P f (t), the unbalanced power can directly compensate for the uncertain load, P c (t) = P f (t). The battery A charges to absorb the positive unbalanced power, P At1 (t) = min{ (P(t) - P f (t)), E At (t - Δt)}.

[0172] When 0 < P(t) < P f (t), the unbalanced power can compensate for part of the uncertain load, P c (t) = P f (t). The battery A discharges to compensate for the remaining load, P At2 (t) = -min{ (P f (t) - P(t)), E At (t - Δt)}.

[0173] (2) P(t) > 0, P f (t) = 0, the unbalanced power is positive, and there is no uncertain load, the battery A only needs to charge to absorb the positive unbalanced power, P At1 (t) = min{P(t), E At (t - Δt)}.

[0174] (3) P(t) < 0, P f (t) > 0, the unbalanced power is negative, and there is an uncertain load, at this time the battery A first compensates for the negative unbalanced power, P At1 (t) = -min{ |P(t)|, E At (t - Δt)}

[0175] When E At (t - Δt) - P At1 (t) > 0, the battery A has remaining capacity to discharge to compensate for the uncertain load, PAt2 (t) = -min{P f (t),(E At (t-Δt)-P At1 (t))

[0176] When E At (t-Δt)-P At1 (t) = 0, the charging and discharging of battery A is stopped, P At2 (t) = 0

[0177] (4) P(t) < 0, P f (t) = 0, the unbalanced power is negative, and there is no uncertain load, battery A discharges to compensate for the unbalanced power, P At1 (t) = -min{|P(t)|, E At (t-Δt)}

[0178] The charging and discharging process of battery A follows formulas (1)-(4). The compensation strategy for unbalanced power of battery A is shown in Figure 3 1.2. Capacity configuration of battery A

[0179] Based on the compensation strategy for unbalanced power of battery A formulated in Section 1.1, considering the compensation of unbalanced power and uncertain load, a capacity configuration model of battery A is constructed. Since the positive unbalanced power is large, a very large capacity of energy storage is needed to ensure complete absorption. Considering the economic impact, the main purpose is to absorb part of the positive unbalanced power to compensate for the negative unbalanced power and uncertain load.

[0180] 1.2.1 Objective function

[0181] Considering the economic impact, the minimum daily operating cost of battery A is taken as the objective function.

[0182]

[0183] In the formula, r is the annual interest rate; y A is the service life of battery A; P A , E A are the rated power (kW) and rated capacity (kWh) of battery A, respectively; C bp , C be are the unit power cost (yuan / kW) and unit capacity cost (yuan / kWh) of battery, respectively; C rb is the maintenance cost of battery (yuan / kWh).

[0184] Considering the compensation effect of unbalanced power and uncertain load, the maximum wind energy utilization rate is taken as the objective function.

[0185] In the formula,

[0186] In the formula, is the unbalanced power absorbed by the battery A; is the unbalanced power greater than the on-grid requirement.

[0187] 1.2.2 Constraint conditions

[0188] The battery A must satisfy the charge-discharge power constraint at any time t:

[0189] P At (t)=P At1 (t)+P At2 (t) (11)

[0190] The battery A rated power constraint, rated capacity constraint, SOC state constraint follow formula (5)-(7).

[0191] 1.2.3 Solution algorithm

[0192] Harris Hawks Optimizer (HHO) is a gradient-free optimization algorithm that simulates the hunting behavior of Harris Hawks, mainly composed of three parts: search phase, conversion of search and development, and development phase. Multi-objective Harris Hawks Optimizer (MOHHO) is an optimization algorithm suitable for multi-objective problems based on HHO, which adds an archive to save and retrieve Pareto optimal results in the HHO algorithm, and has been proven to have good global search ability. The mixed energy storage double-layer planning model of the mixed energy storage operation strategy in the smoothing layer is embedded in the MOHHO algorithm to improve the matching of the control strategy and the configuration method.

[0193] The battery A compensation unbalanced power strategy of section 1.1 is embedded into MOHHO, after each iteration, the algorithm can update the rated power and rated capacity of the battery A according to the power, capacity and SOC constraint in the operation strategy, until the constraint condition is met, and the configured energy storage can absorb part of the positive unbalanced power to compensate for the negative unbalanced power and uncertain load.

[0194] 2. Hybrid energy storage capacity configuration based on VMD adaptive frequency division

[0195] ​Section 1 uses the battery A to achieve compensation for unbalanced power, to get the grid demand of the power, according to the power fluctuation rate, extract the compensation power needed to store energy to meet the fluctuation rate requirement. According to the characteristics of super capacitor and battery, first of all, the VMD adaptive frequency division method is adopted to divide the compensation power, so that the hybrid energy storage runs in its own frequency range. According to the frequency division result, the optimal capacity matching model of hybrid energy storage is constructed. The transition from unbalanced power compensation strategy to wind power fluctuation suppression strategy is shown in Figure 4

[0196] 2.1 VMD adaptive frequency division model

[0197] Variational mode decomposition is a signal decomposition estimation method. In the process of obtaining the decomposition components, the frequency center and bandwidth of each component are determined by iteratively searching for the optimal solution of the variational model, so as to adaptively realize the frequency domain partitioning of the signal and the effective separation of each component. Its decomposition formula is:

[0198]

[0199] In the formula, {u k}={u1,u2,...,u k},{ω k}={ω1,ω2,...,ω k}, The subscripts in the formula represent the square sum L2 norm respectively, and F(t) is the original compensation power to be decomposed.

[0200] The compensation power is divided into high frequency component and low frequency component, and the characteristics of super capacitor and battery are fully utilized, so that the super capacitor can suppress the high frequency and low amplitude fluctuation, and the battery B can suppress the low frequency and high amplitude fluctuation. The reason for choosing VMD is that it can minimize the sum of the estimated bandwidth of each mode. The alternating direction multiplier method is adopted to update the center frequency of each mode and gradually demodulate each mode to the corresponding base frequency band. Finally, each mode is extracted, which has very good frequency division effect.

[0201] In order to achieve the purpose of suppressing the high frequency and low amplitude fluctuation of super capacitor and the low frequency and high amplitude fluctuation of battery B, a VMD component adaptive reconstruction method is proposed, that is, different reconstruction range is selected for each component at each time during VMD component reconstruction. K represents the number of components required for low frequency component reconstruction. During the reconstruction process, if the amplitude of high frequency component is greater than that of low frequency component, the value of K is increased. The high and low frequency reconstruction formula is:

[0202]

[0203] ​

[0204]

[0205] In the formula, u k (t) represents the k-th component of the compensation power in the VMD decomposition at time t; u L (t), u L (t) represents the high and low frequency components of the compensation power initially reconstructed at time t.

[0206] However, the above-mentioned VMD component reconstruction method is inaccurate. Inaccurate selection of the number of VMD decompositions, N, leads to over-decomposition of the original signal. Secondly, after VMD component reconstruction, the sum of the high and low frequency components has a smaller amplitude than the original signal. Since the reconstructed high and low frequency components are used as the basis for hybrid energy storage capacity configuration, inaccurate decomposition leads to inaccurate energy storage capacity configuration, ultimately resulting in energy storage's inability to smooth wind power fluctuations. Therefore, the reconstruction process is adjusted as follows:

[0207] (1) If over-decomposition occurs, adjust the high-frequency component value at this moment to be the same as the original signal amplitude.

[0208] (2) At the same time, the adjusted high-frequency component is subtracted from the original signal to obtain the adjusted low-frequency component, which is used to solve the problem that the amplitude of the VMD component after reconstruction is different from that of the original signal.

[0209]

[0210] P H (t)=u H1 (t) (17)

[0211] In the formula, u H1 (t) represents the adjusted high-frequency component; P L (t) represents the low-frequency component of the compensation power at time t; P H (t) represents the high-frequency component of the compensation power at time t.

[0212] 2.2 Hybrid Energy Storage Capacity Configuration Model

[0213] Based on the VMD adaptive frequency division results in Section 2.1, the hybrid energy storage is smoothed out within its respective frequency range, with the objective function being to minimize the average daily operating cost of the hybrid energy storage.

[0214] 2.2.1 Objective Function

[0215] Energy storage operating costs include investment costs and maintenance costs:

[0216] minC=C b +C sc +C r (18)

[0217] C is the operation cost of energy storage; C b C s are the investment cost of battery B and super capacitor, respectively; C r is the maintenance cost of hybrid energy storage, which is expressed as

[0218]

[0219]

[0220]

[0221] y b y s are the lifetime of battery B and super capacitor, respectively; C sp C se are the unit power cost (yuan / kW) and unit capacity cost (yuan / kWh) of super capacitor, respectively; C rs is the maintenance cost of super capacitor (yuan / kWh);

[0222] 2.2.2 Constraints

[0223] The super capacitor operates in its respective frequency range, with the constraint that:

[0224]

[0225]

[0226] P L (t) + P H (t) = P x (t) (24)

[0227] where P x (t) is the compensation power at time t that meets the fluctuation rate.

[0228] The rated power constraint, rated capacity constraint, and SOC state constraint of the hybrid energy storage follow equations (5)-(7).

[0229] 2.2.3 Solution method

[0230] Gray-Wolf algorithm is an optimization search method inspired by the hunting behavior of gray wolves. The algorithm has strong convergence, few parameters, and is easy to implement. However, its position update equation has the disadvantage of strong development ability and weak exploration ability. The optimal allocation of dual reservoir capacity needs to meet different constraint conditions at the same time, which is a nonlinear, multi-constraint, and multi-variable problem solving process. Particle swarm optimization (PSO) updates the local and global optimization by updating the position and velocity until the optimal value is reached. The PSO-GWO algorithm is a new algorithm that improves the GWO algorithm through the PSO algorithm to have better global search ability.

[0231] The basic idea of the algorithm is: starting from a randomly generated initial population, a new individual is generated by summing the vector difference of any two individuals in the population with the third individual, and then comparing the new individual with the corresponding individual in the current population. If the fitness of the new individual is better than that of the current individual, the new individual replaces the old individual in the next generation, otherwise the old individual is still saved. Through continuous evolution, good individuals are retained and poor individuals are eliminated, guiding the search to approach the optimal solution. In this optimization process, the VMD adaptive frequency division result and the constraint conditions of equations (21)-(26) are embedded in the algorithm, and the PSO-GWO is used to solve the optimal allocation of mixed energy storage capacity according to equation (20).

[0232] 3. Suppression model based on double-layer model algorithm control

[0233] According to the capacity of supercapacitor and battery B solved in section 1, a suppression model based on double-layer model algorithm control is constructed. The upper model mainly uses supercapacitor to suppress high-frequency low-amplitude fluctuations, and the lower model uses battery B to suppress low-frequency high-amplitude fluctuations.

[0234] MAC is a form of MPC that can achieve open-loop optimal control within a limited step size. In terms of energy storage for suppressing wind power grid fluctuation rate, the time domain rolling optimization process is: 1) establish an optimization model according to the SOC state constraint of the energy storage and the grid fluctuation constraint; 2) according to the constraint conditions, solve the energy storage control instruction sequence for a period of time in the future based on the current SOC state variable value and the actual wind power output value; 3) apply the first value of the control instruction sequence to the mixed energy storage control system; 4) roll to the next time, update the SOC state variable value and the grid power value, and repeat the above process. The MAC algorithm uses the idea of rolling prediction and advance control, which can effectively solve the output of energy storage in the future according to the current wind power actual output and constraint conditions.

[0235] 3.1 Upper layer high-frequency fluctuation suppression model

[0236] For the use of supercapacitors to suppress high-frequency fluctuations, the relationship between power and energy storage SOC changes:

[0237] P Hf (k)=P w (k)-P L (k) (25)

[0238] P Hy (k+1)=P Hf (k)-P st (k) (26)

[0239]

[0240] In the formula, P Hf (k) represents the actual power at time k after removing low-frequency components; P Hy (k+1) represents the power after suppressing high-frequency components at time k+1; T represents the control period of the supercapacitor.

[0241]

[0242] In the formula, S st0 M represents the initial state of the supercapacitor at SOC; M is the prediction time length, i.e., the length of the control command sequence.

[0243] -P smax ≤P st (k+i)≤P smax (29)

[0244] S s0 ≤S s (k+i)≤S sM (30)

[0245]

[0246] In the formula, P rate δ represents the rated capacity of wind power; δ is the limit value for the grid connection volatility of wind power.

[0247] Assume the state variable x1(k) = P Hy (k), x2(k)=S s (k); Control quantity u(k) = P st (k); r(k) = P Hf (k) is used as a disturbance. The state-space equation for the wind power system including the supercapacitor is then:

[0248]

[0249] remember:

[0250] Transform the objective function formula (28) and constraint formulas (29) to (31) into the standard form of a quadratic programming problem:

[0251]

[0252] where z = [u(k), x T (k+1),u(k+1)....,u(k+M-1),x T (k+M)] T H is a matrix composed of the quadratic term weight Q of the energy storage control sequence and the output level.

[0253] 3.2 Lower layer high-frequency fluctuation suppression model

[0254] For the use of super capacitor to suppress high-frequency fluctuation, the power and energy storage SOC change relationship:

[0255] P Lf (k) = P Hy (k) + P L (k) (34)

[0256] P Ly (k+1) = P Lf (k) - P bt (k) (35)

[0257]

[0258] where P Lf (k) is the upper layer power after suppression of low-frequency components; P Ly (k+1) is the power after suppressing low-frequency components at k+1 time.

[0259] Under the premise of suppressing low-frequency fluctuations to meet the grid fluctuation requirements, the lower layer MAC rolling optimization objective function is established to minimize the battery B output and charge-discharge balance:

[0260]

[0261] where S bt0 is the initial SOC state of battery B.

[0262] -P bmax ≤ P bt (k+i) ≤ P bmax (38)

[0263] S b0 ≤ S b (k+i) ≤ S bM (39)

[0264]

[0265] Assume that the state variable y1(k) = PLy (k), y2(k) = S b (k); control variable v(k) = P bt (k); f(k) = P Hf (k) as a disturbance variable. Then the state space equation of the wind power system containing the battery B is:

[0266]

[0267] Note:

[0268] The objective function formula (37) and the constraint condition formula (38)~(40) are converted into the standard type of quadratic programming:

[0269]

[0270] where w = [v(k), y T (k+1), v(k+1)...., v(k+M-1), y T (k+M)] T , H is a matrix composed of the quadratic terms of the energy storage control sequence and the output level. R = 2,

[0271] According to formula (4)~(7), we have:

[0272] B b = [P bmax , P bmax , δ, δ, -S b0 , S bM ]

[0273] B bq = [P bmax , b bmax , -S b0 , S bM ]

[0274]

[0275] The above model is optimized and solved in each control period by using the MATLAB quadratic programming method to obtain a sequence of control commands. The first value is taken as the current total control command of the energy storage, and the iteration is repeated to obtain the control command sequence of the super capacitor and the battery B [u(k), u(k+1)...., u(k+M-1)] T , [v(k), v(k+1)...., v(k+M-1)] T

[0276] 4. Simulation results

[0277] 4.1 Battery A Capacity Configuration Results

[0278] To ensure the applicability of the method, measured wind power data from a wind farm in Xinjiang in 2020 were selected. The actual power generation for each day was analyzed, and the day with the largest standard deviation of unbalanced power and uncertain load was selected as the research object. The sampling period was one day, with an unbalanced power sampling period of 5 minutes and an uncertain load sampling period of 1 hour. The relevant parameters and costs of battery A are shown in Table 1. The unbalanced power and uncertain load compensation strategy formulated in Section 2 was embedded into the MOHHO algorithm. Based on the constraints (5)-(7) and (11), the battery capacity with equations (9) and (10) as the objective function was solved.

[0279] Table 1. Relevant parameters of Battery System A

[0280]

[0281] Find the objective value corresponding to a Pareto solution set that satisfies both objectives, as follows: Figure 5 As shown.

[0282] Depend on Figure 5 It can be seen that the average daily operating cost of battery A (C) A ) and battery capacity (E) A It is directly proportional to the wind energy utilization rate (Y). However, the wind energy utilization rate (Y) is directly proportional to the battery capacity (E). A Not proportional, in E A When the power is 13.45 MW h, Y = 90.91%, C A =8142 yuan, when E A >13.45MW h, with E A As increases, Y no longer changes. Therefore, the answer is E. A =13.45MWh, which can ensure the maximization of wind energy utilization with minimal cost. To verify the rationality of the capacity configuration, two indicators were established to analyze the effect of unbalanced power and uncertain load compensation under different rated capacities of battery A:

[0283]

[0284]

[0285] In the formula, Φ1 represents the effect of battery A in compensating for negative unbalanced power, and Φ2 represents the effect of battery A in compensating for uncertain loads. Both should be as small as possible. When Φ1 = 0 and Φ2 = 0, it means that battery A has completely compensated for negative unbalanced power and uncertain loads. P - (t) represents the negative unbalanced power.

[0286] The Φ1, Φ2 corresponding to the Pareto solution set meeting the double target are as shown in Figure 6

[0287] It can be seen from Figure 6 that when E A = 13.45 MW h, Φ1 = 0, Φ2 = 0, which further proves that the selection of E A = 13.45 MW h can guarantee the maximization of wind energy utilization rate at the minimum cost, and the compensation effect of unbalanced power and uncertain load is the best at this time.

[0288] 4.2 Compensation effect of battery A on unbalanced power and uncertain load strategy

[0289] According to the rated capacity of battery A configured in section 4.1, the compensation effect of battery A on unbalanced power and uncertain load strategy is analyzed. The unbalanced power caused by the actual power of the wind farm and the grid-connected demand power is as shown in Figure 7

[0290] The compensation effect of battery A on unbalanced power is as shown in Figure 8

[0291] It can be seen from Figure 8 that due to the consideration of economy, the configured battery A does not compensate for all unbalanced power, but only absorbs part of the positive unbalanced power, which is used to compensate for all negative unbalanced power and uncertain load. The compensation effect of battery A on uncertain load is as shown in Figure 9

[0292] It can be seen from Figure 9 that the uncertain load is mainly compensated by the current positive unbalanced power, and the remaining uncertain load is compensated by battery A. The positive unbalanced power and battery A can cooperate to meet the compensation of all uncertain loads.

[0293] The output of battery A compensating for unbalanced power and uncertain load and the change of SOC state are as shown in Figure 10 and Figure 11

[0294] It can be seen from Figure 10 that battery A is mainly used to compensate for unbalanced power, and a small part is used to compensate for uncertain load. Because the unbalanced power of the wind farm is large, in order to consider economy, according to the strategy, the positive unbalanced power is preferentially used to compensate for the uncertain load. The SOC value of the battery is maintained between 20% and 80%, avoiding overcharging and overdischarging.

[0295] 4.3 Adaptive VMD frequency division result and hybrid energy storage capacity configuration result

[0296] ​​​​​To ensure the applicability of the method, the measured wind power data of a wind farm in Xinjiang in 2020 is selected, the fluctuation of the actual power of each day is analyzed, and the day with the maximum standard deviation of fluctuation rate is selected as the research object, the sampling time is one day, and the sampling period is 5 minutes. The adaptive VMD frequency division method is proposed to divide the compensation power, so as to achieve the purpose of super capacitor suppressing high-frequency low-amplitude part and battery suppressing low-frequency high-amplitude part. The adaptive VMD frequency division results before and after adjustment are shown in Figure 12 and Figure 13 .

[0297] As can be seen from Figure 12 , before the adaptive VMD frequency division result is adjusted, the sum of high and low frequencies does not correspond to the original signal, and the part of the original signal being 0 is greater than 0 after reconstruction, which will affect the suppression effect of wind power fluctuation. The sum of high and low frequencies after adjustment corresponds to the original signal, which shows that the frequency division result is reasonable. From the frequency division result, the super capacitor suppresses the high-frequency low-amplitude part, and the battery suppresses the low-frequency high-amplitude part.

[0298] The related parameters and costs of hybrid energy storage are shown in Table 2. The adaptive VMD frequency division result is embedded in the PSO-GWO algorithm, and according to the constraints (5)-(7), (22)-(24), the optimal capacity matching of hybrid energy storage is solved with formula (18) as the objective function.

[0299] Table 2 Related parameters of hybrid energy storage system

[0300]

[0301] The configuration result is shown in Table 3.

[0302] Table 3 Hybrid energy storage optimization results

[0303]

[0304] 4.4 Suppression effect of double-layer MAC

[0305] According to the capacity of super capacitor and battery B, a suppression model based on double-layer model algorithm control is constructed. The upper layer model mainly uses super capacitor to suppress high-frequency low-amplitude fluctuation, and the lower layer model uses battery B to suppress low-frequency high-amplitude fluctuation. The actual output and SOC state change of hybrid energy storage suppressing wind power fluctuation are shown in Figure 14 and Figure 15 .

[0306] As can be seen from Figure 14It can be seen that the hybrid energy storage is not completely allowed according to the VMD frequency division result, because the VMD frequency division result only provides a basis for configuring a reasonable hybrid energy storage capacity. The MAC method performs rolling optimization according to the configured capacity, realizes real-time feedback, and realizes that the super capacitor suppresses the high-frequency low-amplitude compensation power, and the battery B suppresses the low-frequency high-amplitude compensation power. The super capacitor SOC is maintained between 5% and 95%, and the battery B SOC is maintained between 20% and 80%. The effect of the hybrid energy storage on the wind power fluctuation is shown in Figure 16

[0307] Figure 16 It can be seen that the maximum fluctuation rate of the suppressed grid-connected power curve is 10%, and compared with the fluctuation rate of the original curve, the fluctuation rate of the suppressed grid-connected power curve is reduced by 16%.

[0308] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, any change or replacement without creative labor should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be limited by the protection scope defined in the claims.​​

Claims

1. A wind power smoothing method based on dual-layer MAC and VMD adaptive frequency division, characterized by comprising the following steps: Step 1: Battery A Control Strategy and Capacity Configuration 1.1 Battery A Unbalanced Power Compensation Strategy The unbalanced power of a wind farm is caused by the mismatch between the actual generated power of the wind farm and the power demand of the grid, which can be expressed as: P(t)=P y (t)-P w (t) (8) In the formula, P(t) is the unbalanced power at time t, P y (t) represents the actual generated power of the wind farm at time t, P w (t) represents the power demand of the power grid at time t; With P f (t) represents the uncertain load at time t, P c (t) represents the power consumption of the unbalanced load at time t due to power compensation uncertainty, P At1 (t) represents the power consumption of battery A at time t to compensate for unbalanced power, P At2 (t) represents the power consumption of battery A at time t to compensate for uncertain loads; To address the unbalanced power output of wind farms and considering uncertain loads, an energy storage strategy for compensating for unbalanced power output is formulated. This strategy can be summarized as follows: (1) P(t)>0, P f (t)>0, the unbalanced power is positive, and there is an uncertain load. When P(t) > P f At time (t), the unbalanced power directly compensates for the uncertain load, P c (t)=P f (t); Battery A absorbs positive unbalanced power during charging, P At1 (t)=min{(P(t)-P f (t)),E At (t-Δt)}; When 0 < P(t) < P f At time (t), the unbalanced power compensates for the uncertain load, P c (t)=P f (t); Battery A discharges to compensate for the remaining load, P At2 (t)=-min{(P f (t)-P(t)),E At (t-Δt)}; (2) P(t>0, P f When (t) = 0, the unbalanced power is positive, and there is no uncertain load. Battery A only needs to be charged to absorb the positive unbalanced power. At1 (t)=min{P(t),E At (t-Δt)}; (3) P(t) < 0, P f When (t) > 0, the unbalanced power is negative, and there is an uncertain load. In this case, battery A first compensates for the negative unbalanced power, P At1 (t)=-min{|P(t)|,E At (t-Δt)} When E At (t-Δt)-P At1 (t)>0, when battery A has remaining charge, it discharges to compensate for uncertain loads, P At2 (t)=-min{P f (t),(E At (t-Δt)-P At1 (t))} When E At (t-Δt)-P At1 When (t) = 0, battery A stops charging and discharging when there is no remaining charge. At2 (t)=0 (4) P(t) < 0, P f (t) = 0, the unbalanced power is negative, and there is no uncertain load. Battery A discharges to compensate for the unbalanced power, P At1 (t)=-min{|P(t)|,E At (t-Δt)} The charging and discharging process of battery A follows formulas (1)-(4); 1.2 Battery A Capacity Configuration Based on the unbalanced power compensation strategy for battery A established in Step 1, and considering both unbalanced power compensation and uncertain load, a capacity configuration model for battery A is constructed. Since the positive unbalanced power is large, a very large capacity of energy storage is required to ensure complete absorption. Considering the economic impact, the main purpose is to absorb part of the positive unbalanced power to compensate for the negative unbalanced power and uncertain load. 1.2.1 Objective Function Considering the economic impact, the objective function is to minimize the average daily operating cost of battery A. In the formula, r is the annual interest rate; y A For the lifespan of battery A; P A E A These are the rated power (kW) and rated capacity (kWh) of battery A, respectively; C bp C be These represent the unit power cost (yuan / kW) and unit capacity cost (yuan / kWh) of the battery, respectively; C rb Battery maintenance cost (RMB / kWh); Considering the compensation effect of unbalanced power and uncertain load, the objective function is to maximize wind energy utilization. In the formula, The unbalanced power absorbed by battery A; This is an unbalanced power output greater than the power required for internet access. 1.2.2 Algorithm Selection Step 2: Hybrid Energy Storage Capacity Configuration Based on VMD Adaptive Frequency Division Section 1.1 utilizes battery A to compensate for unbalanced power, obtaining the grid-connected power that meets the grid demand. Based on the grid power fluctuation rate requirements, the compensation power required for energy storage charging and discharging to meet the fluctuation rate requirements is extracted. According to the characteristics of supercapacitors and batteries, the VMD adaptive frequency division method is first used to divide the compensation power into frequencies, so that the hybrid energy storage can operate within their respective frequency ranges. Based on the frequency division results, an optimal ratio model for hybrid energy storage capacity is constructed. 2.1 VMD Adaptive Frequency Division Model Variational mode decomposition (VMD) is a signal decomposition and estimation method. This method determines the frequency center and bandwidth of each component by iteratively searching for the optimal solution of the variational model during the acquisition of decomposed components, thereby adaptively achieving frequency domain partitioning of the signal and effective separation of each component. Its decomposition formula is as follows: In the formula, {u k }={u1,u2,...,u k },{ω k }={ω1,ω2,...,ω k }, The subscripts and superscripts in the text represent the square and L2 norm, respectively, and F(t) is the original compensation power to be decomposed; The compensation power is divided into high-frequency and low-frequency components, making full use of the characteristics of supercapacitors and batteries. The supercapacitors smooth out high-frequency low-amplitude fluctuations, and the battery B smooths out low-frequency high-amplitude fluctuations. VMD is chosen because it can minimize the sum of the estimated bandwidths of each mode. The alternating direction multiplier method is adopted to continuously update each mode and its center frequency, gradually demodulating each mode to the corresponding fundamental frequency band. Finally, each mode and its corresponding center frequency are extracted together, which has a very good frequency division effect. To achieve the goal of smoothing high-frequency, low-amplitude fluctuations in the supercapacitor and smoothing low-frequency, high-amplitude fluctuations in battery B, an adaptive VMD component reconstruction method is proposed. This method involves selecting a different reconstruction range for each component at each time step during VMD component reconstruction, where K represents the number of components required to reconstruct the low-frequency components. During reconstruction, if the amplitude of the high-frequency component is greater than that of the low-frequency component, the value of K is increased. The high- and low-frequency reconstruction formula is as follows: In the formula, u k (t) represents the k-th component of the compensation power in the VMD decomposition at time t; u H (t), u L (t) represents the high and low frequency components of the compensation power initially reconstructed at time t; However, the above-mentioned VMD component reconstruction method is inaccurate because an inaccurate selection of the number N of VMD decompositions can lead to over-decomposition of the original signal. Secondly, after VMD component reconstruction, the sum of the high and low frequency components has an amplitude smaller than the original signal. Since the reconstructed high and low frequency components are used as the basis for hybrid energy storage capacity configuration, inaccurate decomposition will lead to inaccurate energy storage capacity configuration, ultimately resulting in energy storage being unable to smooth wind power fluctuations. Therefore, the reconstruction process is adjusted as follows: (1) If over-decomposition occurs, adjust the high-frequency component value at this moment to be the same as the original signal amplitude. (2) At the same time, the adjusted high-frequency component is subtracted from the original signal to obtain the adjusted low-frequency component, which is used to solve the problem that the amplitude of the reconstructed VMD component is different from that of the original signal. P H (t)=u H1 (t) (17) In the formula, u H1 (t) represents the adjusted high-frequency component; P L (t) represents the low-frequency component of the compensation power at time t; P H (t) represents the high-frequency component of the compensation power at time t; 2.2 Hybrid Energy Storage Capacity Configuration Model Based on the VMD adaptive frequency division results in Section 2.1, the hybrid energy storage is smoothed out within its respective frequency range, with the objective function being to minimize the average daily operating cost of the hybrid energy storage. 2.2.1 Objective Function Energy storage operating costs include investment costs and maintenance costs: minC=C b +C sc +C r (18) In the formula, C represents the operating cost of energy storage; C b C s The investment costs for battery B and supercapacitor are respectively; C r The maintenance cost of hybrid energy storage is expressed as follows: In the formula, y b y s The lifespans of battery B and supercapacitor are respectively; C sp C se These represent the unit power cost (yuan / kW) and unit capacity cost (yuan / kWh) of supercapacitors, respectively; C rs Maintenance cost of supercapacitors (RMB / kWh); 2.2.2 Constraints Supercapacitors operate within their respective frequency ranges, subject to the following constraints: P L (t)+P H (t)=P x (t) (24) In the formula, P x (t) represents the compensation power that satisfies the volatility at time t; The rated power constraint, rated capacity constraint, and SOC state constraint of hybrid energy storage follow formulas (5)-(7). 2.2.3 Solution Method The Gray-Wolf algorithm is an optimization search method inspired by the predatory behavior of gray wolves. It boasts strong convergence, few parameters, and ease of implementation. However, its position update equation suffers from strong development capability but weak exploration capability. Optimal allocation of dual-repository capacity requires simultaneously satisfying different constraints, making it a nonlinear, multi-constraint, and multi-variable problem-solving process. Particle Swarm Optimization (PSO) updates local and global optimizations by updating position and velocity until the optimal value is reached. The PSO-GWO algorithm is a new algorithm that improves upon the GWO algorithm through PSO, giving it better global search capabilities. The basic idea of ​​the algorithm is: starting from a randomly generated initial population, a new individual is generated by summing the vector difference between any two individuals in the population and the third individual. Then, the new individual is compared with the corresponding individual in the current population. If the fitness of the new individual is better than that of the current individual, the new individual is used to replace the old individual in the next generation. Otherwise, the old individual is retained. Through continuous evolution, the excellent individuals are retained and the inferior individuals are eliminated, guiding the search to approach the optimal solution. In this optimization process, the VMD adaptive frequency division result and the constraints of equations (21)-(26) are embedded into the algorithm. According to equation (20), PSO-GWO is used to solve the optimal ratio of mixed energy storage capacity. Step 3: Smoothing Model Based on Two-Level Model Algorithm Control Based on the capacities of the supercapacitor and battery B obtained in Section 1, a smoothing model based on a two-layer model algorithm is constructed. The upper-layer model mainly uses the supercapacitor to smooth high-frequency, low-amplitude fluctuations, while the lower-layer model uses battery B to smooth low-frequency, high-amplitude fluctuations. MAC is a type of MPC that can achieve open-loop optimal control within a finite step size. In terms of energy storage mitigating wind power grid connection volatility, its time-domain rolling optimization process is as follows: 1) Establish an optimization model based on energy storage SOC state constraints and grid connection volatility constraints; 2) Using the current SOC state variable value and the actual wind power output value, solve for the energy storage control command sequence for a future period based on the constraints; 3) Apply the first value of the control command sequence to the hybrid energy storage control system; 4) Roll to the next time step, update the SOC state variable value and the grid connection power value, and repeat the above process. The MAC algorithm adopts the idea of ​​rolling prediction and advance control, which can effectively solve the output of energy storage in the future based on the current actual wind power output and constraints. 3.1 Upper-level high-frequency fluctuation smoothing model Regarding the use of supercapacitors to smooth high-frequency fluctuations, the relationship between power and SOC changes is as follows: P Hf (k)=P w (k)-P L (k) (25) P Hy (k+1)=P Hf (k)-P st (k) (26) In the formula, P Hf (k) represents the actual power at time k after removing low-frequency components; P Hy (k+1) represents the power after suppressing high-frequency components at time k+1; T represents the control period of the supercapacitor. In the formula, S st0 The initial state of charge (SOC) of the supercapacitor is denoted by M; M represents the prediction time length, i.e., the length of the control command sequence. -P s ≤P st (k+i)≤P s (29) S s0 ≤S s (k+i)≤S sM (30) In the formula, P rate This refers to the rated capacity of wind power. δ is the limit value for the grid connection volatility of wind power; Assume the state variable x1(k) = P Hy (k), x2(k)=S s (k); Control quantity u(k) = P st (k); r(k) = P Hf (k) is used as a disturbance; then the state-space equation of the wind power system including the supercapacitor is: remember: Transform the objective function formula (28) and constraint formulas (29) to (31) into the standard form of a quadratic programming problem: In the formula, z = [u(k), x T (k+1),u(k+1)....,u(k+M-1),x T (k+M)] T H is a matrix composed of the quadratic weights Q of the energy storage control sequence and the output level; 3.2 Lower-level high-frequency fluctuation suppression model Regarding the use of supercapacitors to smooth high-frequency fluctuations, the relationship between power and SOC changes is as follows: P Lf (k)=P Hy (k)+P L (k) (34) P Ly (k+1)=P Lf (k)-P bt (k) (35) In the formula, P Lf (k) represents the low-frequency component superimposed on the power after upper-level smoothing at time k; P Ly (k+1) represents the power after suppressing low-frequency components at time k+1; To mitigate low-frequency fluctuations and meet grid connection fluctuation requirements, a lower-level MAC rolling optimization objective function was established based on the minimum output of battery B and charge-discharge balance. In the formula, S bt0 The initial state of charge (SOC) of battery B; -P b ≤P bt (k+i)≤P b (38) S b0 ≤S b (k+i)≤S bM (39) Assume the state variable y1(k) = P Ly (k), y2(k)=S b (k); Control quantity v(k) = P bt (k); f(k) = P Hf (k) is taken as a disturbance; then the state-space equation of the wind power system including battery B is: remember: The objective function formula (37) and the constraint formulas (38) to (40) are transformed into the standard form of a quadratic programming problem: Where w = [v(k), y T (k+1),v(k+1)....,v(k+M-1),y T [(k+M)]T and H are matrices composed of the quadratic weights R and Q of the energy storage control sequence and the output level; R = 2. According to formulas (4) to (7), we can obtain: B b =[P b ,P b ,d,d,-S b0 ,S bM ] B bq =[P b ,b b ,-S b0 ,S bM ] The above model is optimized and solved using MATLAB quadratic programming in each control cycle to obtain a sequence of control commands. The first value is taken as the current total control command for energy storage. This process is repeated iteratively to obtain the daily control command sequence for the supercapacitor and battery B [u(k), u(k+1), ..., u(k+M-1)]. T [v(k),v(k+1)....,v(k+M-1)] T .

2. The method according to claim 1, characterized in that: in step one, the constraint condition for configuring the capacity of battery A is: Battery A must satisfy the charge / discharge power constraint at any time t: P At (t)=P At1 (t)+P At2 (t) (11) The rated power constraint, rated capacity constraint, and SOC state constraint of battery A follow formulas (5)-(7).

3. The method according to claim 1, characterized in that: In step one, the algorithm selection is as follows: the Harris Eagle Optimization Algorithm is a gradient-free optimization algorithm that simulates the predation behavior of the Harris Eagle, mainly composed of three parts: the search phase, the search-development transition, and the development phase; while the Multi-Objective Harris Eagle Optimization Algorithm (MOHHO) is an optimization algorithm based on HHO that is suitable for multi-objective problems. It adds an archive to the HHO algorithm to store and retrieve Pareto optimal results, and has been proven to have good global search capabilities; the hybrid energy storage operation strategy formulated by the smoothing layer of the hybrid energy storage bi-layer planning model is embedded into the MOHHO algorithm to improve the matching between the control strategy and the configuration method; By embedding the unbalanced power compensation strategy of battery A into MOHHO, the algorithm can update the rated power and rated capacity of battery A according to the power, capacity and SOC constraints in the running strategy after each iteration until the constraints are met, so that the configured energy storage can absorb part of the positive unbalanced power to compensate for the negative unbalanced power and uncertain load.

Images