Adaptive mode decomposition power distribution method based on EMISSA algorithm

Abstract

The invention discloses a self-adaptive modal decomposition power distribution method based on an EMISSA algorithm, and the method comprises the following steps: 1, obtaining an optimal value of a modal parameter K and an optimal value of a secondary penalty factor alpha of a VMD algorithm based on an improved sparrow search algorithm; the improved sparrow search algorithm updates a scale factor of a discoverer through fuzzy logic after each iteration, operates a follower subgroup through mixed differential mutation to generate a variant subgroup, and then updates the position of an early warner; after the position of the discoverer is iteratively updated each time, the topological opposite position of the discoverer is obtained based on a topological opposite learning method, and then fitness evaluation is carried out to further update the position of the discoverer; and 2, substituting the optimal value of the modal parameter K and the optimal value of the secondary penalty factor alpha obtained in the step 1 into a VMD algorithm, decomposing the fluctuation power of the photovoltaic unit, and distributing the decomposed fluctuation power to a hybrid energy storage system. According to the invention, the accuracy and effectiveness of power distribution can be improved.

Description

Technical Field

[0001] This invention relates to the field of power allocation methods for direct hybrid energy storage systems, specifically an adaptive mode decomposition power allocation method based on the EMISSA algorithm. Background Technology

[0002] To meet the capacity requirements of DC microgrid systems and the power supply requirements of loads, hybrid energy storage systems are connected to the DC microgrid via bidirectional DC / DC converters. These systems store the electrical energy generated by photovoltaic (PV) generators within the DC microgrid. The hybrid energy storage system comprises vanadium redox flow batteries and lithium-ion batteries. Power distribution within the hybrid energy storage system is crucial for its stable and efficient operation. Typically, a Virtual Mode Decomposition (VMD) algorithm is used to decompose the fluctuating power of the PV generators in the DC microgrid. Based on the decomposition results, the fluctuating power is distributed to the vanadium redox flow batteries and lithium-ion batteries in the hybrid energy storage system. The VMD algorithm effectively suppresses PV power fluctuations, reduces mode aliasing, and achieves reasonable power distribution within the hybrid energy storage system, thus ensuring its stable operation.

[0003] Since the VMD algorithm requires setting the modal parameter K and the secondary penalty factor α, the existing technology generally involves manually setting the optimal value of the modal parameter K and the value of the secondary penalty factor α. However, the manual setting method has the drawback of subjectivity.

[0004] To address the subjective drawbacks of manually setting the modal parameters K and the secondary penalty factor α in the VMD algorithm, existing technologies employ the Sparrow Search Algorithm (SSA) to determine their optimal values. However, in the existing SSA algorithm, the fixed size of different sparrow roles during the optimization process leads to a lack of adaptability, easily causing an imbalance between the algorithm's global and local search capabilities. Furthermore, the position update formula of the SSA algorithm reveals a tendency for individual sparrows to tend towards the origin or optimal position, thus failing to fully explore the space and potentially missing better solutions. Therefore, when using the existing SSA algorithm to determine the optimal values ​​of the modal parameters K and the secondary penalty factor α in the VMD algorithm, the imbalance between global and local search capabilities and the inability to fully explore the space may result in inaccurate determination of the optimal values ​​of the modal parameters K and the secondary penalty factor α, ultimately preventing the VMD algorithm from accurately achieving power allocation in the hybrid energy storage system. Summary of the Invention

[0005] This invention provides an adaptive mode decomposition power allocation method based on the EMISSA algorithm to solve the problem that power allocation cannot be accurately achieved when using the existing SSA algorithm to determine VMD algorithm parameters and then perform power allocation in a hybrid energy storage system.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] An adaptive mode decomposition power allocation method based on the EMISSA algorithm is used for power allocation from photovoltaic units to a hybrid energy storage system in a DC microgrid. The hybrid energy storage system includes vanadium redox flow batteries and lithium-ion batteries, and includes the following steps:

[0008] Step 1: Based on the improved Sparrow Search Algorithm EMISSA, obtain the optimal values ​​of the modal parameter K and the quadratic penalty factor α for the VMD algorithm used for the decomposition of fluctuating power of photovoltaic units.

[0009] The improved sparrow search algorithm EMISSA updates the proportion factor of discoverers using fuzzy logic after each iteration of the sparrow search algorithm. Based on the updated proportion factor, the proportion of discoverers in the sparrow population is adjusted after each iteration. After updating the positions of discoverers and followers in each iteration, a mixed differential mutation operation is performed on the follower subgroup to generate a mutated subgroup, and then the position of the early warning bird is updated. Furthermore, after updating the position of discoverers in each iteration, the topological opposition position of discoverers is obtained based on the topological opposition learning method. Then, the fitness of the topological opposition position of discoverers and the current updated position is evaluated, and the position of discoverers is further updated based on the fitness evaluation results.

[0010] Step 2: Substitute the optimal values ​​of the modal parameter K and the quadratic penalty factor α obtained in Step 1 into the VMD algorithm. Use the VMD algorithm with the optimal values ​​to decompose the fluctuating power of the photovoltaic unit, and allocate the decomposed power to the vanadium redox flow battery and lithium-ion battery in the hybrid energy storage system.

[0011] Furthermore, in step 1, the fitness function used by the improved sparrow search algorithm EMISSA includes sample entropy, aggregation algebra, and Pearson correlation coefficient.

[0012] Furthermore, in step 1, the inputs to the fuzzy logic are the current iteration stage and the population diversity. Based on the current iteration stage and the population diversity, the proportion factor of the discoverers after each iteration is obtained through fuzzy logic.

[0013] Furthermore, in step 1, the mixed differential mutation operation employs two differential mutation strategies: SSA-DE / rand / 1 and SSA-DE / best / 1.

[0014] Furthermore, in step 1, if the fitness of the discoverer's topologically opposed position is better than the fitness of the current iteration update position, then the discoverer's current iteration update position is swapped with the topologically opposed position; otherwise, no swap is performed.

[0015] In this invention, by using a capture mechanism that combines archiving during the discoverer search phase to expand the search range, sparrow individuals can be fully searched in the search space, thus solving the problem of imbalance between global and local search capabilities in existing sparrow search algorithms when determining VMD algorithm parameters.

[0016] In this invention, an adaptive neighborhood search method is used to fully explore the location information around high-quality individuals, thus solving the problem that existing sparrow search algorithms may lose better solutions due to insufficient space exploration when determining VMD algorithm parameters.

[0017] Therefore, the improved Sparrow Search Algorithm EMISSA in this invention can accurately obtain the optimal values ​​of the modal parameter K and the quadratic penalty factor α of the VMD algorithm. When the VMD algorithm with the optimal values ​​is used to allocate the power load of the hybrid energy storage system, the accuracy and effectiveness of power allocation can be improved. It can also effectively smooth out the fluctuations of photovoltaic grid connection, which is conducive to extending the service life of the hybrid energy storage system and improving the energy utilization rate of the hybrid energy storage system. Attached Figure Description

[0018] Figure 1 This is a grid connection schematic diagram of the hybrid energy storage system in an embodiment of the present invention. Detailed Implementation

[0019] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0020] This embodiment discloses an adaptive mode decomposition power allocation method based on the EMISSA algorithm, used for power allocation from photovoltaic units to hybrid energy storage systems in DC microgrids. Figure 1 As shown, the hybrid energy storage system includes a vanadium redox flow battery (VRB) and a lithium-ion battery (LB). The VRB and LB are connected in parallel to the photovoltaic units in the DC microgrid via a DC / DC converter and a DC bus.

[0021] The adaptive mode decomposition power allocation method based on the EMISSA algorithm in this embodiment is as follows:

[0022] Step 1: Based on the improved Sparrow Search Algorithm EMISSA, obtain the optimal values ​​of the modal parameter K and the quadratic penalty factor α for the VMD algorithm used for the decomposition of fluctuating power of photovoltaic units.

[0023] In this embodiment, the improved Sparrow Search Algorithm EMISSA is derived from the existing Sparrow Search Algorithm SSA. The process by which the improved Sparrow Search Algorithm EMISSA determines the optimal values ​​of the modal parameter K and the quadratic penalty factor α is as follows:

[0024] Step 1.1) Determine the sparrow population size N, the problem dimension D, and the maximum number of iterations T. Individual sparrows in the population are divided into discoverers, followers, and early warning systems, where:

[0025] Producer: The top 20% of sparrows with the best fitness, responsible for searching for high-quality areas.

[0026] Scrounger: The remaining sparrows, excluding the discoverer and the early warning sparrow, learn from the discoverer.

[0027] Scouts: 10%-20% of the sparrows are randomly selected and are responsible for escaping local optima.

[0028] Step 1.2) Initialize the positions of individuals in the population and evaluate the fitness of each individual in the population using the fitness function.

[0029] In this embodiment, the parameter combination [K, α] is used as the sparrow population position for iterative optimization, and a signal evaluation index that fits the VMD algorithm is introduced to construct the fitness function of the improved sparrow search algorithm EMISSA. The fitness function in this embodiment includes three evaluation indices: sample entropy, aggregation algebra, and Pearson correlation coefficient, where:

[0030] Sample entropy is a quantitative measure of whether data exhibits regularity. It better reflects the frequency characteristics of each IMF (Information Management Function). The smaller the sample entropy value, the lower the signal uncertainty and the more pronounced the frequency characteristics of the IMF. The estimation of sample entropy can be expressed as follows:

[0031] sampEn(data,q,r)=ln B q (r)-ln B q+1 (r) (1)

[0032] In the formula: sampleEn() represents the sample entropy function; data represents a random signal time series; q represents the embedding dimension; r represents the similarity tolerance; B represents the probability that two vectors match q or q+1 real numbers under the similarity tolerance. q (r) represents the probability that a vector matches q real numbers under similarity tolerance, B q+1 (r) represents the probability that a vector matches q+1 real numbers under similarity tolerance.

[0033] The convergence algebra is the length of the optimal center frequency signal after VMD decomposition, and it is an evaluation metric for the IMF frequency characteristics. A lower convergence algebra indicates faster signal convergence and a more pronounced IMF frequency characteristic. The convergence algebra can be calculated as follows:

[0034] omega = length(ω)k (2)

[0035] In the formula: omega represents the aggregation algebra of the optimal center frequency; length(ω) k ) indicates the extraction of the optimal center frequency ω k The signal length.

[0036] The Pearson correlation coefficient is a method for measuring the degree of linear correlation between two signals and can be used as an evaluation index of the deviation between the reconstructed signal and the original signal. A higher Pearson correlation coefficient indicates a smaller deviation. The Pearson correlation coefficient is defined as follows:

[0037]

[0038] In the formula, x and y are assumed to be two time series of length N, x = (x1, x2, ..., xy). n ), y = (y1, y2, ..., y n ); P * Let x be the Pearson correlation coefficient between the time series x and y; and These represent the means of x and y, respectively; x i Let x be an element in the sequence, and y be an element in the sequence. i P is an element in the sequence y; * ∈[-1,1],P * The larger the absolute value, the stronger the correlation between x and y.

[0039] Taking into account the three evaluation indicators, the fitness function f of the EMSSA constructed in this embodiment is... SSA As shown in the following formula:

[0040]

[0041] In the formula: lg(omega) is the relative aggregation function.

[0042] Step 1.3) Initialize the current iteration number t = 1 and perform the current iteration. After the current iteration, based on the fitness function f... SSA Calculate the fitness of the individual positions in the population at the current iteration, sort the fitness, and obtain the best and worst individual positions in the population.

[0043] Step 1.4): After the current iteration t, update the discoverer proportion factor using fuzzy logic, and adjust the proportion of discoverers in the sparrow population after each iteration based on the updated proportion factor. Specific details are as follows:

[0044] In the existing SSA algorithm, the sparrow finder occupies the optimal resource in the entire population, and its population size is determined by the scaling factor PD, which remains unchanged during the algorithm's iterative search process. In this embodiment, to enable the finder to dynamically adjust during the algorithm's iterative search process to adapt to the optimization mechanism, a fuzzy inference system is introduced to nonlinearly update the sparrow finder's scaling factor. The two input variables of the fuzzy inference system are the current iteration stage and population diversity. Based on the current iteration stage and population diversity, the updated scaling factor of the finder is obtained through fuzzy logic iteration.

[0045] The current iteration is defined as the percentage between the current iteration count and the maximum iteration count, as shown in the following formula:

[0046] I iteration =t / T (5)

[0047] Among them: I iteration This represents the iteration phase; t is the current iteration number; T is the maximum iteration number.

[0048] Population diversity is measured by the Euclidean distance between each individual sparrow and the current best individual, which indicates the dispersion of sparrows in the search space. If sparrows are relatively dispersed, the population diversity is high; conversely, if sparrows are relatively concentrated, the population diversity is low. The population diversity is shown in the following formula:

[0049]

[0050] Where: N is the population size; D is the dimension of the variable; x i,j (t) represents the j-th dimension value of the i-th sparrow at iteration t; x best,j (t) represents the j-th dimension value of the optimal sparrow individual in t iterations; I diversity For population diversity.

[0051] In this embodiment, after the current iteration, I is calculated using formulas (5) and (6). iteration with I diversity The data is then input into fuzzy logic to obtain an updated scaling factor PD, which is used to adjust the proportion of discoverers in the sparrow population.

[0052] Step 1.5) After the current iteration t, update the positions of the discoverer and follower using formulas (7) and (8), as shown in the following equations:

[0053]

[0054] In formula (7): Updated location for the discoverer; α is the position of the discoverer before the update; i is the i-th sparrow; α is a random number in the interval (0,1]; T is the maximum number of iterations; R is a random number in the interval [0,1]; S is a random number in the interval [0.5,1], representing the safety threshold; Q is a random number that follows a normal distribution; L is a 1×d matrix with all elements equal to 1.

[0055] In formula (8): Update the position of the followers; The position of the follower before the update; i is the i-th follower; The current worst position; n is the total number of sparrows; The updated location for the discoverer; A + is a pseudo-inverse matrix; Q is a random number following a normal distribution; L is a 1×d matrix with all elements equal to 1.

[0056] After each iteration updates the positions of the discoverer and followers, a mixed differential mutation operation is performed on the follower subgroup to generate a mutated subgroup, and then the position of the early warning discoverer is updated.

[0057] As can be seen from the follower position update formula, in each iteration of the SSA algorithm, the follower moves towards the optimal position of the discoverer or the origin. This update method can accelerate the convergence of the algorithm, but it may also affect the diversity of the sparrow population. Differential mutation is an efficient optimization strategy used in the Differential Evolution (DE) algorithm that uses a differential variable mechanism to change the position of candidate solutions. It is often combined with other algorithms to improve the diversity of solutions.

[0058] In the improved sparrow search algorithm EMISSA of this embodiment, two differential mutation strategies are proposed to perform mixed mutation operations on the follower subgroup to generate a new subgroup, allowing individual sparrows to fully search the search space and thus escape local optima. The two differential mutation strategies are SSA-DE / rand / 1 and SSA-DE / best / 1, and the specific implementations of SSA-DE / rand / 1 and SSA-DE / best / 1 are shown in the following equations:

[0059] u i (t)=x i (t)+r1(x q (t)-x i (t))+r2·(x m (t)-x n (t)) (9)

[0060] u i+k (t)=xi+k (t)+r1(x best (t)-x i+k (t))+r2(x m (t)-x n (t)) (10)

[0061] Where: u i (t) and u i+k (t) represent the i-th and i+k-th mutant individuals generated in the t-th iteration, respectively; r1 and r2 are the weight coefficients in [0, 1]; x i (t) represents the position of the i-th sparrow follower in the t-th iteration; x i+k (t) represents the position of the (i+k)th sparrow follower in the t-th iteration; x q (t) represents the discoverer randomly selected from the non-optimal discoverer individuals in the t-th iteration; x best (t) represents the optimal discoverer individual; x m (t) and x n (t) represents two random positions different from the current follower's position, and m ≠ n.

[0062] In this embodiment, after generating the mutant subgroup of followers by performing a mixed differential mutation operation using formulas (9) and (10), the position of the early warning unit is updated using formula (11), which is shown below:

[0063]

[0064] in: The updated location of the person who issued the warning; The current optimal position is represented by β, which is the step size control parameter. For...; f i f is the current fitness value of the sparrow. g The global optimal fitness value is denoted as K; K is a random number located in the interval [-1, 1]. This represents the worst possible position; f worst ε is the worst fitness value globally; ε is a very small number used to avoid a denominator of 0.

[0065] Step 1.6) After updating the discoverer's position in the current iteration t, the topological opposition position of the discoverer is obtained based on the topological opposition learning method. Then, the fitness of the discoverer's topological opposition position and the current iteration update position is evaluated, and the discoverer's position is further updated based on the fitness evaluation results.

[0066] In this embodiment, the Topological Opposites Learning (TOBL) method is employed. This method is an intelligent optimization and enhancement strategy proposed by Dai et al. based on the opposition learning mechanism, which can significantly enhance the algorithm's spatial search capability. Compared to traditional opposition learning methods where each individual has only one counterpart with the exact opposite position to its original position, the TOBL method allows each original individual to obtain 2... D There are 1 candidate individuals, defined as follows:

[0067] Let x i =(x i,1 ,x i,2 ,...,x i,j ,x i,D If is a point in D-dimensional space, then its topological opposite point T is... i =(T i,1 ,T i,2 ,...,T i,j ,T i,D The formula is shown below:

[0068]

[0069] Where: x i,1 ,x i,2 ,...,x i,j ,x i,D Let i be the position of the i-th individual in the 1, 2, ..., J, D dimensions;

[0070] T i,1 ,T i,2 ,...,T i,j ,T i,D Let i be the position of the i-th topological opposite point in the 1, 2, ..., J, D dimensions;

[0071] x i,j Let i be the position of the i-th individual in the j-th dimension;

[0072] x best,j The j-th dimension value of the optimal individual;

[0073] o i,j Let o be the j-th dimension value of the i-th opposing point. i,j As shown below:

[0074]

[0075] in: and These represent the upper and lower bounds of the search space, respectively.

[0076] To enhance the discoverer's ability to explore the space, the improved sparrow search algorithm EMISSA in this embodiment incorporates the above mechanism into the SSA algorithm. After updating the positions of all sparrow individuals, the original opposing point of each discoverer individual is first generated using formula (13). Then, the Manhattan distance between the optimal sparrow individual, the original opposing point, and the current sparrow discoverer individual is compared. If the distance between the original opposing point and the optimal individual is the smallest, it indicates that this opposing point is the topological opposing point position of the current sparrow discoverer individual.

[0077] The purpose of calculating the topological opposition points is to find better sparrow discoverers to explore the search space more fully, while other sparrows can follow the discoverers to explore better foraging areas, thereby improving the optimization efficiency.

[0078] After obtaining the topological counterparts of the sparrow finder, the fitness function f is then used. SSA The fitness of the discoverer's topologically opposed position and the current iteration update position are evaluated, and the discoverer's position is further updated based on the fitness evaluation results. If the fitness of the discoverer's topologically opposed position is better than the fitness of the current iteration update position, the discoverer's current iteration update position is swapped with the topologically opposed position; otherwise, no swap is performed. The swap formula is as follows:

[0079]

[0080] Where: f(x) i (t)) and f(T) i (t) represents the fitness values ​​of the current sparrow discoverer's current iteration update position and the corresponding topological point position, respectively; x i (t) represents the current iteration update position of the sparrow discoverer; T i (t) represents the current topological position of the individual sparrow discoverer; This is the final solution.

[0081] Therefore, in this embodiment, the topological opposition point position of the sparrow discoverer is generated by formula (12) and formula (13), and the discoverer position is further updated by formula (14).

[0082] Step 1.7) Set the current iteration number t to t+1, that is, let t = t+1, and perform the (t+1)th iteration according to steps 1.3)-1.6).

[0083] Step 1.8) Determine whether the number of iterations of the algorithm has reached the maximum number of iterations T. If the number of iterations has reached the maximum number of iterations, output the best individual position and its fitness value; otherwise, continue the iteration loop from step 1.3).

[0084] Therefore, based on the improved Sparrow Search Algorithm EMISSA described in steps 1.1)-1.8), this embodiment obtains the optimal values ​​of the modal parameter K and the quadratic penalty factor α for the VMD algorithm used for the decomposition of fluctuating power of photovoltaic units.

[0085] Step 2: Substitute the optimal values ​​of the modal parameter K and the quadratic penalty factor α obtained in Step 1 into the VMD algorithm. Use the VMD algorithm with the optimal values ​​to decompose the fluctuating power of the photovoltaic unit, and allocate the decomposed power to the vanadium redox flow battery and lithium-ion battery in the hybrid energy storage system.

[0086] This embodiment addresses the drawback of the traditional VMD algorithm, where the subjectivity of manually setting the modal parameter K and the secondary penalty factor α may affect signal decomposition. As a result, it can effectively suppress photovoltaic power fluctuations, reduce mode aliasing, achieve reasonable power allocation in the hybrid energy storage system, and realize the stable operation of the hybrid energy storage system.

[0087] The preferred embodiments of the present invention have been described in detail above with reference to the accompanying drawings. These embodiments are merely descriptions of preferred embodiments and are not intended to limit the scope or concept of the invention. The specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. Such combinations, as long as they do not violate the spirit of the present invention, should also be considered as part of this disclosure. To avoid unnecessary repetition, the present invention will not further describe the various possible combinations.

[0088] This invention is not limited to the specific details of the above embodiments. Within the scope of the technical concept of this invention and without departing from the design idea of ​​this invention, all modifications and improvements made by those skilled in the art to the technical solutions of this invention should fall within the protection scope of this invention. The technical content for which protection is sought in this invention has been fully described in the claims.

Claims

1. An adaptive mode decomposition power allocation method based on the EMISSA algorithm, used for power allocation from photovoltaic units to a hybrid energy storage system in a DC microgrid, wherein the hybrid energy storage system includes vanadium redox flow batteries and lithium-ion batteries, characterized in that... Includes the following steps: Step 1: Based on the improved Sparrow Search Algorithm EMISSA, obtain the modal parameters of the VMD algorithm used for the decomposition of fluctuating power in photovoltaic units. K The optimal value, the second penalty factor α The optimal value; The improved sparrow search algorithm EMISSA updates the proportion factor of the discoverer through fuzzy logic after each iteration of the sparrow search algorithm, and adjusts the proportion of the discoverer in the sparrow population after each iteration based on the updated proportion factor. After each iteration updates the positions of the discoverer and followers, a mixed differential mutation operation is performed on the follower subgroup to generate a mutated subgroup, and then the position of the early warning discoverer is updated. Furthermore, after each iteration of updating the discoverer's position, the topological opposition position of the discoverer is obtained based on the topological opposition learning method. Then, the fitness of the discoverer's topological opposition position and the current iteration update position is evaluated, and the discoverer's position is further updated based on the fitness evaluation results. Step 2: Convert the modal parameters obtained in Step 1 into... K The optimal value, the second penalty factor α The optimal value is substituted into the VMD algorithm to decompose the fluctuating power of the photovoltaic unit, and the decomposed power is allocated to the vanadium redox flow battery and lithium-ion battery in the hybrid energy storage system.

2. The adaptive mode decomposition power allocation method based on the EMISSA algorithm according to claim 1, characterized in that, In step 1, the fitness function used in the improved sparrow search algorithm EMISSA includes sample entropy, aggregation algebra, and Pearson correlation coefficient.

3. The adaptive mode decomposition power allocation method based on the EMISSA algorithm according to claim 1, characterized in that, In step 1, the inputs to the fuzzy logic are the current iteration stage and the population diversity. Based on the current iteration stage and the population diversity, the proportion factor of the discoverers after each iteration is obtained through fuzzy logic.

4. The adaptive mode decomposition power allocation method based on the EMISSA algorithm according to claim 1, characterized in that, In step 1, the mixed differential mutation operation employs two differential mutation strategies: SSA-DE / rand / 1 and SSA-DE / best / 1.

5. The adaptive mode decomposition power allocation method based on the EMISSA algorithm according to claim 1, characterized in that, In step 1, if the fitness of the discoverer's topologically opposed position is better than the fitness of the current iteration update position, then the discoverer's current iteration update position is swapped with the topologically opposed position; otherwise, no swap is performed.

Images