Micro-grid multi-objective optimization control method based on composite energy storage power distribution
By combining pumped storage, batteries and supercapacitor groups in the water-wind storage interconnected microgrid system, using multivariate variational modal decomposition and improved competition and cooperation group optimization algorithm, the problem of intermittent fluctuations in wind power generation is solved, and the problem of intermittent fluctuations in wind power generation is realized. The full frequency band of the power grid is accurate and stable.
Patent Information
- Application Number
- CN202411157769.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-22
- Publication Date
- 2024-12-03
AI Technical Summary
There are serious intermittent fluctuations in the power output of wind power, which has threatened the safe and stable operation of the power grid. The response rate of existing energy storage equipment is slow, making it difficult to effectively suppress high-frequency fluctuations.
Build a water and wind storage interconnected microgrid system, combining pumped storage, battery and supercapacitor group composite energy storage system, and realize the fluctuation of wind power through multivariate variational modal decomposition and improved competition and cooperation group optimization algorithm (ICCSO). Accurate and smooth the entire frequency band.
It achieves accurate suppression of wind power fluctuations in all frequency bands, improves the safety and stability of the power grid and avoids the impact of grid frequency fluctuations on the system.
Smart Images

Figure CN119070354A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the research field of multi-objective optimization control of power systems, and specifically relates to a multi-objective optimization control method for a microgrid based on composite energy storage power distribution. Background Art
[0002] In recent years, countries have gradually optimized their energy structures and continued to advance in the direction of renewable energy. The scale of clean energy generation such as hydropower, wind power, and solar energy has also continued to increase. However, due to the random fluctuations of natural wind, the power output of wind power generation has serious intermittent fluctuations, resulting in poor reliability of its output. Large-scale wind power access will pose a serious threat to the safe and stable operation of the power grid. At present, the method of interconnecting pumped storage units and wind turbines has been widely adopted. Pumped storage units have both power generation and energy storage and can operate in both directions. They can realize high-power and large-capacity power storage in the power grid to play the role of peak regulation and frequency regulation, so as to absorb the impact of intermittent energy such as wind and light on the power grid, and improve the reliability and stability of wind turbines. However, when pumped storage units undertake peak regulation tasks, because the power response needs to go through a series of complex physical processes such as the water diversion system to change the unit output, its response rate is far less than that of chemical battery energy storage and electromagnetic energy storage. When facing higher-frequency fluctuations when smoothing intermittent energy fluctuations in the power grid, its climbing rate is slow and cannot maintain the frequency level, which still affects the safe and stable operation of the power grid system. Supercapacitors have the advantages of long life, many cycles, and fast response time of charge and discharge. However, the dielectric withstand voltage of supercapacitors is very low, and the energy storage level is limited by the withstand voltage. The stored energy is not large and the discharge time is short. To apply it on a large scale in microgrids, it can form a composite energy storage system with battery energy storage to have both power and energy storage characteristics.
[0003] Variational Mode Decomposition (VMD) is a univariate signal decomposition method that aims to decompose a complex signal into a set of intrinsic mode functions (IMFs), each with a clear instantaneous frequency. VMD achieves this goal by minimizing the signal reconstruction error and the modal bandwidth, resulting in a set of pure, mutually orthogonal modal signals. VMD is designed to process single-channel signals, while MVMD (Multivariate Variational Mode Decomposition) can simultaneously process multiple related signals or channels. In a microgrid system, the unsteady power may come from multiple renewable energy generation units and load changes, and there is a high probability that there is a correlation between these signals. MVMD is an extension of VMD designed to handle the decomposition of multivariate or multichannel signals. MVMD optimizes the frequency separation of modes by considering the information of all channels simultaneously, which helps to more accurately identify and separate power fluctuations of different frequencies in complex and dynamic microgrid systems.
[0004] Competition and cooperation swarm optimization (CCSO) is a competition and cooperation swarm optimization method. The update strategy of the CCSO algorithm allows the fitness of one particle to decrease so that the fitness of another particle can be better. CCSO can generate particles from the local optimal area through this special collaborative learning model, and realizes a new position update strategy by allowing the fitness of one particle to deteriorate, that is, sacrificing one particle to make the position of another particle better. However, the traditional CCSO algorithm, whose competition swarm optimization strategy borrows from the particle swarm algorithm (PSO), has a learning rate parameter that is set unchanged. Although this method reduces the parameters, it is not conducive to the convergence of the algorithm. Summary of the invention
[0005] Purpose of the invention: The present invention provides a multi-objective optimization control method for a microgrid based on composite energy storage power allocation, so that a microgrid system containing renewable energy can more accurately and quickly adjust the grid frequency, achieve targeted smoothing of fluctuations in the entire frequency band, and further ensure the safe and stable operation of the system.
[0006] Technical solution: The multi-objective optimization control method of a microgrid based on composite energy storage power allocation described in the present invention comprises the following steps:
[0007] (1) constructing a water-wind-storage interconnected microgrid system, and obtaining a power signal required to be smoothed by a composite energy storage system according to a power allocation strategy of the microgrid system; the composite energy storage system includes a pumped storage power station, a battery group, and a supercapacitor group;
[0008] (2) Decomposing the power signal that needs to be smoothed by the composite energy storage system through MVMD to obtain a multivariate modulation component signal;
[0009] (3) Reconstruct the multi-modulation component signals to obtain power signals in three relative frequency bands: high, medium and low;
[0010] (4) Establish a multi-objective optimization model for the microgrid system based on composite energy storage power allocation, including optimization variables, objective functions and constraints;
[0011] (5) The CCSO algorithm is improved to obtain the ICCSO algorithm; the ICCSO algorithm introduces the Levy flight random walk strategy and the learning rate adjustment strategy that dynamically adapts to the current iteration number to address the CCSO algorithm's proneness to local optimality. The ICCSO algorithm quickly expands the search space in the early stage to fully explore, and ensures fine adjustment near the optimal solution in the later stage; and introduces Chebyshev mapping initialization to improve the global convergence efficiency of the algorithm;
[0012] (6) The ICCSO algorithm is used to solve the established multi-objective optimization model to obtain the optimal signal reconstruction summation mode number and penalty factor parameter group that meet the expected objective function, so as to achieve accurate smoothing of wind power fluctuations in the microgrid system in the entire frequency band.
[0013] Furthermore, the water-wind-storage interconnected microgrid system in step (1) includes a hydropower station, a wind farm, a load system, a composite energy storage system and a control system.
[0014] Furthermore, the power signal required to be smoothed by the composite energy storage system in step (1) is:
[0015] P c (t) = P w (t)-P grid (t)
[0016] Among them, P c (t) is the power signal that the interconnected microgrid needs to be smoothed by the composite energy storage system, P w (t) is the wind power output power, P grid (t) is the expected power of wind power connected to the grid; and:
[0017] P c (t) = P p (t)+P b (t)+P sc (t)
[0018] Among them, P p (t) is the required pumped storage unit smoothing power, P b (t) is the required battery pack power, P sc (t) is the power required by the supercapacitor bank to smooth the power.
[0019] Furthermore, the implementation process of step (2) is as follows:
[0020] The power signal to be decomposed P c The number of channels of (t) is C, the number of multivariate modulation components to be decomposed is K, and the decomposition results in K multivariate modulation components u k (t), k = 1, 2, ..., K, 3 ≤ K ≤ 20, and the variational problem is constructed as follows:
[0021]
[0022] Where c is the number of original signals input to the algorithm; ω k is the center frequency of each mode after MVMD decomposition; P c (t) is the set of input multivariate time series characteristic signals; u k,c (t) is the set of decomposed time series characteristic signals, The multivariate modulation signal u is obtained by Hilbert transform k,c (t) is the analytical signal.
[0023] Furthermore, the implementation process of step (3) is as follows:
[0024] The K multivariate modulation components u k (t), k = 1, 2, ..., K, 3 ≤ K ≤ 20 is reconstructed into three high, medium and low frequency reconstructed signals, according to the center frequency ω of each mode j The modes are classified according to the following principles:
[0025]
[0026] In the formula, ω max ,ω min They are respectively the last model component u except for the decomposition K The highest and lowest center frequencies of the modes other than (t), ω c The average value of the two is used as the high-frequency critical value of signal reconstruction, divided by u K In all modes except (t), the center frequency ω k ≤ω c It is classified as an intermediate frequency signal, with a center frequency of ω k >ω c It is classified as a high-frequency signal, and the last model component u decomposed by MVMD is K (t) is a low-frequency signal;
[0027] Add the modal signals belonging to the same frequency band to obtain the reconstructed signals of the corresponding three frequency bands:
[0028]
[0029] Wherein, H(t) is the reconstructed high-frequency component signal allocated to the supercapacitor group for smoothing, M(t) is the reconstructed intermediate-frequency component signal allocated to the battery group for smoothing, L(t) is the reconstructed low-frequency component signal allocated to the pumped storage for smoothing, H and M are the number of modes summed when reconstructing the high and intermediate frequency signals of the decomposed multivariate modulated signal, respectively, and 3≤K≤20 is the total number of modes of the multivariate modulated signal decomposed by MVMD.
[0030] Furthermore, the implementation process of step (4) is as follows:
[0031] The optimization variables are selected as the penalty factor ε for constructing the Lagrangian function in the MVMD decomposition and the modal numbers H and M respectively summed when reconstructing the high and medium frequency signals through the MVMD decomposition;
[0032] The objective function f1 is the weighted value of the mean square error of the reconstructed signal obtained in the signal decomposition and reconstruction process compared with the original signal and the average kurtosis of the decomposed mode, which ensures the fidelity of the signal during the decomposition process and reduces the cross-interference between modes:
[0033] f1=α×MSE+(1-α)×Q
[0034] Among them, α is the weight of the mean square error, Q is the average kurtosis of the decomposed mode, and MSE is the mean square error;
[0035] The objective function f2 is the time multiplied by the absolute error integral of the system grid frequency:
[0036]
[0037] Among them, e(t) represents the error between the actual output and the expected output of the system, and t represents time;
[0038] The constraints are:
[0039]
[0040] Among them, H t (t), M t (t), L t (t) are the values of the reconstructed high, medium and low frequency power signals at any time t, respectively. sc,max , P sc,min Indicates the upper and lower limits of the supercapacitor group output, P soc,max , P soc,min Indicates the upper and lower limits of the battery pack output, and its positive and negative signs represent charging and discharging, P p,max , P p,min It represents the upper and lower limits of the pumped storage unit output. Its positive and negative signs indicate pumping and releasing of water. K is the total number of multivariate modulation signal modes obtained by MVMD decomposition.
[0041] Furthermore, the implementation process of step (5) is as follows:
[0042] Use Chebyshev mapping to initialize the population and determine the initial layout of the population:
[0043] x n+1 = cos(μarccos x n ),x n ∈[-1,1]
[0044] Among them, μ is the order of Chebyshev polynomial, which can generate unrelated chaotic sequences with ergodic properties;
[0045] The introduction of the Levy flight strategy encourages individuals to spread widely in the search space, avoids local optimal traps, and significantly improves the global optimization efficiency; the position update formula of the Levy flight is:
[0046]
[0047] Among them, X i (t) represents the i-th solution of the t-th generation; represents point-to-point multiplication; α represents the weight of controlling the step size, α=0.01(X i (t)-X b ), X b is the current optimal solution; Levy(λ) represents the path that obeys the Levy distribution and satisfies: Levy(λ)~μ=t -λ , 1≤λ≤3;
[0048] The exponential decay method is used to adaptively adjust the learning rate. The dynamic adjustment mechanism ensures that the learning rate can be appropriately reduced as the iteration progresses, which is conducive to model convergence. The details are as follows:
[0049] η t =η0·Υ t
[0050] Among them, t is the number of real-time iterations, η t Represents the learning rate value when the algorithm iterates to the tth round, η0 represents the initial learning rate, and γ is the attenuation coefficient, which reflects the rate at which the learning rate gradually decreases with the iteration process.
[0051] Furthermore, the mean square error is:
[0052]
[0053] Among them, P c (t) is the original power signal to be decomposed, is the signal after decomposition and reconstruction, and N is the length of the signal, that is, the number of sample points in the time series.
[0054] Furthermore, the average kurtosis of the decomposed mode is:
[0055]
[0056] Among them, 3≤K≤20 is the number of decomposition modes, N is the signal length, u j is the modal signal u j The average value of (t):
[0057]
[0058] Furthermore, the process of implementing population initialization using Chebyshev mapping is as follows:
[0059] Set the population size to N, randomly generate a d-dimensional vector in the d-dimensional space as the initial individual, Y = (y1, y2, ..., y d ),v i ∈[-1,1];
[0060] Perform N-1 iterations on each dimension of Y to generate the remaining N-1 individuals;
[0061] Map the generated N individuals to the solution search space in turn:
[0062] x id = l d +(1+y id )×(u d -l d ) / 2
[0063] Among them, u d , l d are the upper and lower bounds of the dth dimension of the search space, y id is the dth dimension of the ith individual, x id is the d-dimensional coordinate value of the i-th individual in the search space.
[0064] Beneficial effects: Compared with the prior art, the present invention has the following beneficial effects:
[0065] 1. The present invention constructs a water-wind-storage interconnected microgrid system model based on a composite energy storage system of a pumped-storage power station, a battery group, and a supercapacitor group. Taking into account the defect that the pumped-storage unit has a slow response speed when facing high-frequency fluctuations of wind power in the microgrid, it is proposed to decompose and reconstruct the unstable power signal of wind power in the microgrid and then use the composite energy storage system for targeted smoothing, reconstruct it into high, medium and low frequencies and distribute them to the supercapacitor group, battery group, and pumped-storage power station for response, thereby further realizing the detailed full-frequency smoothing of wind power grid-connected fluctuations;
[0066] 2. The present invention decomposes and reconstructs the unstable power signal of wind power by MVMD, takes the decomposition parameter penalty factor and the summed modal number of the reconstructed signal as optimization variables, takes the time multiplied by the integral of absolute error (ITAE) of the grid frequency of the microgrid system as the first goal, considers the fidelity in signal decomposition and reduces the cross-overlapping influence between modes, takes the mean square error and average kurtosis in signal decomposition as the second goal, and establishes a multi-objective optimization model for the microgrid;
[0067] 3. In order to solve the established multi-objective optimization model of the microgrid, the present invention proposes an improved solution algorithm ICCSO, which solves the defect that the CCSO algorithm is prone to fall into local optimality and improves the global convergence efficiency of the algorithm. By solving the established multi-objective optimization model, the optimal signal reconstruction summation modal number and penalty factor parameter group that meet the expected objective function can be obtained, thereby achieving accurate smoothing of wind power fluctuations in the microgrid system in the entire frequency band. BRIEF DESCRIPTION OF THE DRAWINGS
[0068] Figure 1 is a flow chart of the present invention;
[0069] Figure 2 It is a schematic diagram of the topological structure of the microgrid system;
[0070] Figure 3 Schematic diagram of the process flow decomposition and reconstruction using MVMD;
[0071] Figure 4 Flowchart of the proposed ICCSO algorithm for solving the multi-objective optimization model. DETAILED DESCRIPTION
[0072] The present invention will be further described in detail below in conjunction with the accompanying drawings.
[0073] like Figure 1 As shown, the present invention proposes a multi-objective optimization control method for a microgrid based on composite energy storage power allocation, which mainly includes the following steps:
[0074] Step 1: Construct a water-wind storage interconnected microgrid system model based on a pumped storage power station, battery group, and supercapacitor group composite energy storage system. The microgrid system topology is as follows: Figure 2 As shown in the figure, according to the power allocation strategy of the microgrid system, the power signal that needs to be smoothed by the energy storage system is obtained:
[0075] P c (t) = P w (t)-P grid (t) (1)
[0076] Among them, P c (t) is the power signal that the interconnected microgrid needs to be smoothed by the composite energy storage system, P w (t) is the wind power output power, P grid (t) is the expected power of wind power grid-connected. And:
[0077] P c (t) = P p (t)+P b (t)+P sc (t)
[0078] Among them, P p(t) is the required pumped storage unit smoothing power, P b (t) is the required battery pack power, P sc (t) is the power required by the supercapacitor bank to smooth the power.
[0079] Step 2: Decompose the power signal obtained in step 1 that needs to be smoothed by the energy storage system through the MVMD decomposition method so that it can be reconstructed and allocated to the composite energy storage system for targeted smoothing and optimize the stable operation of the microgrid. The MVMD decomposition method is as follows: Figure 3 shown.
[0080] The power signal to be decomposed is P c (t), the number of channels is C, the number of multivariate modulation modal components to be decomposed is K, and the decomposition results in K multivariate modulation components u k (t), k = 1, 2, ..., K, 3 ≤ K ≤ 20, the specific steps of constructing the variational problem are as follows:
[0081] First, the multivariate modulated vibration signal u k (t) Obtain the analytical signal by performing Hilbert transform
[0082]
[0083] in, Respectively represent the analytical signals of the first to Cth channels; u k,1 (t)~u k,C (t) is u k (t) is the data of the first to Cth channels; δ(t) is the impulse function; j is the complex number symbol.
[0084] calculate The center frequency ω k , through the frequency shift operation The spectrum is transferred to the baseband, specifically:
[0085]
[0086] Then use the gradient L 2 norm, the variational problem is formulated as:
[0087]
[0088] Where K is the number of separation modes set in the algorithm and 3≤K≤20; ω k is the center frequency of each mode after MVMD decomposition; P c (t) is the set of multivariate time series characteristic signals to be decomposed; u k,c (t) is the kth component of the decomposition k (t) is the data of the cth channel, The multivariate modulation signal u is obtained by Hilbert transform k,c The analytical signal of (t), represents the partial derivative with respect to time.
[0089] In order to simplify the calculation process, the constrained variation problem is transformed into an unconstrained variation problem and constructed using the Lagrange multiplier method. The form of its Lagrange function is as follows:
[0090]
[0091] Where: ε is the penalty factor, λ c for In the conversion process, the introduction of penalty factors and Lagrange multipliers will affect the optimization process of the objective function, resulting in deviations in the final variational mode decomposition results. Therefore, it is necessary to optimize and adjust the penalty factors or Lagrange multipliers in the decomposition operation.
[0092] To solve the unconstrained variational problem after this transformation, the alternating direction multiplier method (ADMM) is used to iteratively calculate u in equation (8): k,c (t) and ω k ,u k,c (t0 and ω k Calculate alternately. The frequency domain update formula is:
[0093]
[0094] The center frequency update formula is:
[0095]
[0096] Lagrange multiplier function λ c The update formula is:
[0097]
[0098] In the formula, τ is the update coefficient, which is updated through algorithm iteration, that is, the decomposed K multivariate modulation components u are obtained. k (t),k=1,2....K, 3≤K≤20.
[0099] Step 3: Reconstruct the signal of the decomposed multivariate modulation components to obtain power signals of three relative frequency bands: high, medium and low.
[0100] The K multivariate modulation components u k (t≤, k=1,2,...,K, 3≤K≤20 is reconstructed into three high, medium and low frequency reconstructed signals, according to the center frequency ω of each mode j The modes are classified according to the following principles:
[0101]
[0102] In the formula, ω max ,ω min They are respectively the last model component u except for the decomposition K The highest and lowest center frequencies of the modes other than (t), ω c The average value of the two is used as the high-frequency critical value of signal reconstruction, divided by u K In all modes except (t), the center frequency ω k ≤ω c It is classified as an intermediate frequency signal, with a center frequency of ω k >ω c It is classified as a high-frequency signal, and the last model component u decomposed by MVMD is K (t) is a low frequency signal.
[0103] Add the modal signals belonging to the same frequency band to obtain the corresponding reconstructed signals of the three frequency bands of high, medium and low. The mathematical formula is as follows:
[0104]
[0105] Wherein, H(t) is the reconstructed high-frequency component signal allocated to the supercapacitor group for smoothing, M(t) is the reconstructed intermediate-frequency component signal allocated to the battery group for smoothing, L(t) is the reconstructed low-frequency component signal allocated to the pumped storage for smoothing, H and M are the number of modes summed when reconstructing the high and intermediate frequency signals of the decomposed multivariate modulated signal, respectively, and 3≤K≤20 is the total number of modes of the multivariate modulated signal decomposed by MVMD.
[0106] Step 4: Establish a multi-objective optimization model based on composite energy storage power allocation for the microgrid system, including optimization variables, objective functions and constraints.
[0107] The optimization variables are the penalty factor ε for constructing the Lagrangian function in the MVMD decomposition and the modal numbers H and M respectively summed when reconstructing the high and medium frequency signals through MVMD decomposition.
[0108] The objective function f1 is selected as the weighted value of the mean square error of the reconstructed signal obtained in the signal decomposition and reconstruction process compared with the original signal and the average kurtosis of the decomposed mode, aiming to ensure the fidelity of the signal during the decomposition process and reduce the cross-interference between modes. The objective function f1 is calculated as follows:
[0109] f1=α×MSE+(1-α)×Q (11)
[0110] Among them, α is the weight of the mean square error, which is set to 0.5.
[0111] The objective function f2 is the time multiplied by the integral of absolute error (ITAE) of the system grid frequency, and the expression is as follows:
[0112]
[0113] Among them, e(t) represents the error between the actual output and the expected output of the system, and t represents time.
[0114] The mean square error is:
[0115]
[0116] Among them, P c (t) is the original power signal to be decomposed, is the decomposed and reconstructed signal, namely, H(t), M(t), and L(t), and N is the length of the signal, namely, the number of sample points in the time series.
[0117] The average kurtosis of the decomposed modes is:
[0118]
[0119] Among them, 3≤K≤20 is the number of decomposition modes, N is the signal length, is the modal signal u j The average value of (t) is calculated as:
[0120]
[0121] The constraints are as follows:
[0122]
[0123] Among them, H t (t), M t (t), L t (t) are the values of the reconstructed high, medium and low frequency power signals at any time t, respectively. sc,max , P sc,min Indicates the upper and lower limits of the supercapacitor group output, P soc,max , P soc,min Indicates the upper and lower limits of the battery pack output, and its positive and negative signs represent charging and discharging, P p,max , P p,min It represents the upper and lower limits of the pumped storage unit output. Its positive and negative signs indicate pumping and releasing of water. K is the total number of multivariate modulation signal modes obtained by MVMD decomposition.
[0124] Step 5: Figure 4As shown in the figure, the ICCSO algorithm for solving the multi-objective optimization control model of the microgrid is designed. The ICCSO algorithm is designed to solve the problem that the CCSO algorithm is prone to fall into the local optimal defect. The update strategy introduces the Levy flight random walk strategy and the learning rate adjustment strategy that dynamically adapts to the current iteration number. In the early stage, the search space is quickly expanded to fully explore, and in the later stage, fine adjustments are ensured near the optimal solution. The Chebyshev map initialization is introduced to improve the global convergence efficiency of the algorithm.
[0125] In order to obtain an initial population with wide coverage, balanced distribution and strong global search capability, the Chebyshev mapping method is used to determine the initial layout of the population, and the formula is as follows:
[0126] x n+1 = cos(μarccos x n ),x n ∈[-1,1](17)
[0127] Where μ is the order of the Chebyshev polynomial, which can generate unrelated chaotic sequences with ergodic properties.
[0128] The steps for population initialization using Chebyshev mapping are as follows:
[0129] Set the population size to N, randomly generate a d-dimensional vector in the d-dimensional space as the initial individual, Y = (y1, y2, ..., y d ),v i ∈[-1,1].
[0130] Use formula (4) to iterate each dimension of Y N-1 times to generate the remaining N-1 individuals.
[0131] Map the generated N individuals to the solution search space in turn:
[0132] x id = l d +(1+y id )×(u d -l d ) / 2(18)
[0133] Among them, u d , l d are the upper and lower bounds of the dth dimension of the search space, y id is the dth dimension of the ith individual, x id is the d-dimensional coordinate value of the i-th individual in the search space.
[0134] Levy flight is a non-Gaussian random process, whose stable increment follows the Levy stable distribution and exhibits the characteristics of a random walk: it contains both dense small-amplitude movements and occasional large-amplitude jumps, and the two are intertwined. The introduction of the Levy flight strategy can encourage individuals to spread widely in the search space, avoid local optimal traps, and significantly improve the global optimization efficiency. The position update formula of Levy flight is:
[0135]
[0136] Among them, X i (t) represents the i-th solution of the t-th generation; represents point-to-point multiplication; α represents the weight of controlling the step size, α=0.01(X i (t)-X b ), X b is the current optimal solution; Levy(λ) represents the path that obeys the Levy distribution and satisfies: Levy(λ)~μ=t -λ , 1≤λ≤3.
[0137] The exponential decay method is used to adaptively adjust the learning rate. The dynamic adjustment mechanism ensures that the learning rate can be appropriately reduced as the iteration progresses, which is conducive to model convergence. The specific formula is as follows:
[0138] η t =η0·Υ t (20)
[0139] Among them, t is the number of real-time iterations, η t Represents the learning rate value when the algorithm iterates to the tth round, η0 represents the initial learning rate, and γ is the attenuation coefficient, which reflects the rate at which the learning rate gradually decreases with the iteration process.
[0140] Step 6: Use the designed ICCSO algorithm to solve the established microgrid multi-objective optimization model to obtain the optimal signal reconstruction summation mode number and penalty factor parameter group that meet the expected objective function, so as to achieve accurate smoothing of wind power fluctuations in the microgrid system across the entire frequency band. The algorithm optimization implementation steps are as follows.
[0141] 1) Set the parameters required for system optimization: including population size, maximum number of iterations, expansion factor, initial learning rate and decay factor, upper and lower limits of optimization variables, etc. Use Chebyshev mapping to initialize and generate the initial population, including competitive groups and cooperative groups.
[0142] 2) Calculate population fitness and perform non-dominated sorting.
[0143] 3) Main loop, evolutionary algebra G=1.
[0144] 4) The competitive group executes the competitive group optimization strategy to generate offspring. Its speed update method draws on the particle swarm algorithm (PSO) and introduces the Levy flight strategy and adaptive learning rate adjustment strategy to generate offspring populations.
[0145] 5) The cooperative group executes the collaborative group optimization strategy to generate offspring. The offspring population is generated based on the arithmetic crossover operator.
[0146] 6) Merge the original population with the newly generated offspring, perform fast non-dominated hierarchical sorting, and calculate the crowding degree.
[0147] 7) Update the population size to the initial N with the help of the elite preservation strategy to complete the iteration.
[0148] 8) Determine whether the evolutionary generation G is less than the maximum evolutionary generation. If so, execute G=G+1 and return to step 4). Otherwise, the operation ends.
[0149] The above embodiments are only for illustrating the technical concept and features of the present invention, and their purpose is to enable people familiar with the technology to understand the content of the present invention and implement it accordingly, and they cannot be used to limit the protection scope of the present invention. Any equivalent transformation or modification made according to the spirit of the present invention should be included in the protection scope of the present invention.
Claims
1. A multi-objective optimization control method for a microgrid based on composite energy storage power distribution, characterized in that: The following steps are involved: (1) constructing a water-wind-storage interconnected microgrid system, and obtaining a power signal required to be smoothed by a composite energy storage system according to a power allocation strategy of the microgrid system; the composite energy storage system includes a pumped storage power station, a battery group, and a supercapacitor group; (2) Decomposing the power signal that needs to be smoothed by the composite energy storage system through MVMD to obtain a multivariate modulation component signal; (3) Reconstruct the multi-modulation component signals to obtain power signals in three relative frequency bands: high, medium and low; (4) Establish a multi-objective optimization model for the microgrid system based on composite energy storage power allocation, including optimization variables, objective functions and constraints; (5) The CCSO algorithm is improved to obtain the ICCSO algorithm; the ICCSO algorithm introduces the Levy flight random walk strategy and the learning rate adjustment strategy that dynamically adapts to the current iteration number to address the CCSO algorithm's proneness to local optimality. The ICCSO algorithm quickly expands the search space in the early stage to fully explore, and ensures fine adjustment near the optimal solution in the later stage; and introduces Chebyshev mapping initialization to improve the global convergence efficiency of the algorithm; (6) The ICCSO algorithm is used to solve the established multi-objective optimization model to obtain the optimal signal reconstruction summation mode number and penalty factor parameter group that meet the expected objective function, so as to achieve accurate smoothing of wind power fluctuations in the microgrid system in the entire frequency band.
2. A multi-objective optimization control method for a microgrid based on composite energy storage power allocation according to claim 1, characterized in that: The water-wind-storage interconnected microgrid system in step (1) includes a hydropower station, a wind farm, a load system, a composite energy storage system and a control system.
3. A multi-objective optimization control method for a microgrid based on composite energy storage power allocation according to claim 1, characterized in that: The power signal required to be smoothed by the composite energy storage system in step (1) is: P c (t)=P w (t)-P grid (t) Among them, P c (t) is the power signal that the interconnected microgrid needs to be smoothed by the composite energy storage system, P w (t) is the wind power output power, P grid (t) is the expected power of wind power connected to the grid; and: P c (t)=P p (t)+P b (t)+P sc (t) Among them, P p (t) is the required pumped storage unit smoothing power, P b (t) is the required battery pack smoothing power, P sc (t) is the power required by the supercapacitor bank to smooth the power.
4. A multi-objective optimization control method for a microgrid based on composite energy storage power allocation according to claim 1, characterized in that: The implementation process of step (2) is as follows: The power signal to be decomposed P c The number of channels of (t) is C, the number of multivariate modulation components to be decomposed is K, and the decomposition results in K multivariate modulation components u k (t), k = 1, 2, ..., K, 3 ≤ K ≤ 20, and the variational problem is constructed as follows: Where c is the number of original signals input to the algorithm; ω k is the center frequency of each mode after MVMD decomposition; P c (t) is the set of input multivariate time series characteristic signals; u k,c (t) is the set of decomposed time series characteristic signals, The multivariate modulation signal u is obtained by Hilbert transform k,c (t) is the analytical signal.
5. The multi-objective optimization control method for a microgrid based on composite energy storage power allocation according to claim 1 is characterized in that: The implementation process of step (3) is as follows: The K multivariate modulation components u k (t), k = 1, 2, ..., K, 3 ≤ K ≤ 20 is reconstructed into three high, medium and low frequency reconstructed signals, according to the center frequency ω of each mode j The modes are classified according to the following principles: In the formula, ω max ,ω min They are respectively the last model component u except for the decomposition K The highest and lowest center frequencies of the modes other than (t), ω c The average value of the two is used as the high-frequency critical value of signal reconstruction, divided by u K In all modes except (t), the center frequency ω k ≤ω c It is classified as an intermediate frequency signal, with a center frequency of ω k >ω c It is classified as a high-frequency signal, and the last model component u decomposed by MVMD is K (t) is a low-frequency signal; Add the modal signals belonging to the same frequency band to obtain the reconstructed signals of the corresponding three frequency bands: Wherein, H(t) is the reconstructed high-frequency component signal allocated to the supercapacitor group for smoothing, M(t) is the reconstructed intermediate-frequency component signal allocated to the battery group for smoothing, L(t) is the reconstructed low-frequency component signal allocated to the pumped storage for smoothing, H and M are the number of modes summed when reconstructing the high and intermediate frequency signals of the decomposed multivariate modulated signal, respectively, and 3≤K≤20 is the total number of modes of the multivariate modulated signal decomposed by MVMD.
6. A multi-objective optimization control method for a microgrid based on composite energy storage power allocation according to claim 1, characterized in that: The implementation process of step (4) is as follows: The optimization variables are selected as the penalty factor ε for constructing the Lagrangian function in the MVMD decomposition and the modal numbers H and M respectively summed when reconstructing the high and medium frequency signals through the MVMD decomposition; The objective function f1 is the weighted value of the mean square error of the reconstructed signal obtained in the signal decomposition and reconstruction process compared with the original signal and the average kurtosis of the decomposed mode, which ensures the fidelity of the signal during the decomposition process and reduces the cross-interference between modes: f1=α×MSE+(1-α)×Q Among them, α is the weight of the mean square error, Q is the average kurtosis of the decomposed mode, and MSE is the mean square error; The objective function f2 is the time multiplied by the absolute error integral of the system grid frequency: Among them, e(t) represents the error between the actual output and the expected output of the system, and t represents time; The constraints are: Among them, H t (t), M t (t), L t (t) are the values of the reconstructed high, medium and low frequency power signals at any time t, respectively. sc,max , P sc,min Indicates the upper and lower limits of the supercapacitor group output, P soc,max , P soc,min Indicates the upper and lower limits of the battery pack output, and its positive and negative signs represent charging and discharging, P p,max , P p,min It represents the upper and lower limits of the pumped storage unit output. Its positive and negative signs indicate pumping and releasing of water. K is the total number of multivariate modulation signal modes obtained by MVMD decomposition.
7. A multi-objective optimization control method for a microgrid based on composite energy storage power allocation according to claim 1, characterized in that: The implementation process of step (5) is as follows: Use Chebyshev mapping to initialize the population and determine the initial layout of the population: x n+1 =cos(μarccos x n ),x n ∈[-1,1] Among them, μ is the order of Chebyshev polynomial, which can generate unrelated chaotic sequences with ergodic properties; The introduction of the Levy flight strategy encourages individuals to spread widely in the search space, avoids local optimal traps, and significantly improves the global optimization efficiency; the position update formula of the Levy flight is: Among them, X i (t) represents the i-th solution of the t-th generation; represents point-to-point multiplication; α represents the weight of controlling the step size, α=0.01(X i (t)-X b ), X b is the current optimal solution; Levy(λ) represents the path that obeys the Levy distribution and satisfies: Levy(λ)~μ=t -λ , 1≤λ≤3; The exponential decay method is used to adaptively adjust the learning rate. The dynamic adjustment mechanism ensures that the learning rate can be appropriately reduced as the iteration progresses, which is conducive to model convergence. The details are as follows: or t =η0·γ t Among them, t is the number of real-time iterations, η t Represents the learning rate value when the algorithm iterates to the tth round, η0 represents the initial learning rate, and γ is the attenuation coefficient, which reflects the rate at which the learning rate gradually decreases with the iteration process.
8. A multi-objective optimization control method for a microgrid based on composite energy storage power allocation according to claim 6, characterized in that: The mean square error is: Among them, P c (t) is the original power signal to be decomposed, is the signal after decomposition and reconstruction, and N is the length of the signal, that is, the number of sample points in the time series.
9. A multi-objective optimization control method for a microgrid based on composite energy storage power allocation according to claim 6, characterized in that: The average kurtosis of the decomposed mode is: Among them, 3≤K≤20 is the number of decomposition modes, N is the signal length, is the modal signal u j The average value of (t):
10. A multi-objective optimization control method for a microgrid based on composite energy storage power allocation according to claim 7, characterized in that: The process of implementing population initialization using Chebyshev mapping is as follows: Set the population size to N, randomly generate a d-dimensional vector in the d-dimensional space as the initial individual, Y = (y1, y2, ..., y d ),v i ∈[-1,1]; Perform N-1 iterations on each dimension of Y to generate the remaining N-1 individuals; Map the generated N individuals to the solution search space in turn: x id =l d +(1+y id )×(u d -l d ) / 2 Among them, u d , l d are the upper and lower bounds of the dth dimension of the search space, y id is the dth dimension of the ith individual, x id is the d-dimensional coordinate value of the i-th individual in the search space.
Citation Information
Cited By
Distributed energy power control method and system
CN120749805A
A distributed energy power control method and system
CN120749805B
New energy grid-connected optimization method and system considering generated power fluctuation
CN121602539A
