Improved multi-target particle swarm algorithm applied to energy storage optimization of power distribution network

Abstract

The invention belongs to the technical field of power distribution network multi-objective optimization scheduling, and particularly relates to an improved multi-objective particle swarm algorithm applied to power distribution network energy storage optimization. A multi-objective optimization model of node voltage fluctuation, load power fluctuation and energy storage rated capacity is established by combining wind energy, photovoltaic and typical daily load curves, and the safety and economy of the system are comprehensively considered. A dynamic inertia weight updating rule based on Euclidean distance optimal similarity is proposed by taking a target function fitness value of the model as a particle according to a distance between a position vector of a current particle and an optimal particle position vector, so that the algorithm has better global search capability and local search capability in an optimization process. And meanwhile, crossover variation is introduced into the algorithm, so that convergence to local optimum is avoided. The algorithm has a good inhibition effect on node voltage fluctuation, the solving efficiency of a power system scheduling scheme is effectively improved, and a more comprehensive and reasonable scheduling scheme is obtained.

Description

Technical Field

[0001] The present invention belongs to the technical field of multi-objective optimization scheduling of distribution networks, and in particular relates to an improved multi-objective particle swarm algorithm applied to energy storage optimization of distribution networks. Background Art

[0002] The penetration rate of distributed power sources represented by photovoltaic and wind power in active distribution networks has been increasing year by year. However, the access of a large number of distributed power sources to the distribution network will inevitably lead to changes in the structure of the distribution network, resulting in an imbalance in the supply and demand between the source and the load of the distribution network, resulting in voltage and frequency fluctuations, power reverse transmission, increased network losses, and reduced reliability and stability in the distribution network. The introduction of energy storage systems in distribution networks can effectively solve the above problems. Energy storage systems have the characteristics of bidirectional flow, rapid power response, high regulation accuracy, and flexible configuration locations. However, the access location and capacity of the energy storage system and its operation strategy are inseparable from the stable operation of the distribution network. If they are not reasonably optimized and configured, it will be detrimental to the safe and stable operation of the network. At present, the configuration cost of the energy storage system is still high. How to make it produce greater benefits in the distribution network with the smallest capacity is an urgent problem to be solved. The source-load-storage optimization scheduling problem of the distribution network is a multi-objective optimization problem that takes into account the stability, safety and economy of the power grid.

[0003] The distribution network optimization dispatch problem is a nonlinear, multi-constrained, high-dimensional problem. At present, there are generally two ways to deal with multi-objective optimization problems. One is to process the multi-objective function into a single objective function, but the result loses diversity; the other is to use a multi-objective intelligent algorithm to solve it directly. The particle swarm algorithm is widely used in practical problems because of its simple and easy-to-understand principle, efficient cluster parallelism, and the ability to generate multiple non-inferior solutions in each iteration. It also has a memory function, good convergence and global search capabilities.

[0004] According to the existing literature and patent search, in the existing literature, Yan Qunmin, Dong Xinzhou et al. published "Optimal Configuration of Energy Storage in Active Distribution Network Based on Improved Multi-Objective Particle Swarm Algorithm" in Power System Protection and Control (2022, 50(10): 11-19), which introduced a quasi-adversarial learning strategy in the population update process to enhance the coverage and convergence speed of the solution, and adopted an adaptive splitting strategy to separate prematurely aggregated particles according to the number of iterations, thereby enhancing particle diversity and ensuring the ability of the algorithm to jump out of the local optimum. Lu Limin, Chu Guowei et al. published "Optimal Configuration of Energy Storage in Microgrid Based on Improved Multi-Objective Particle Swarm Algorithm" in Power System Protection and Control (2020, 48(15): 116-124), which proposed an adaptive inertia weight update rule and a multi-iteration direction Pareto solution set dynamic update strategy, introducing real-time crossover mutation operations, ensuring the uniformity of the praeto solution set and the ability to jump out of the local optimum. Li Xingshen, Zhang Jing and others published "Multi-objective Optimal Scheduling of Microgrids Based on Improved Particle Swarm Algorithm" in Electric Power Science and Engineering (2021, 37(03): 1-7). They improved the inertia weight and learning factor based on the traditional particle swarm algorithm, and took into account the microgrid operating cost and environmental protection cost, and verified the superiority of the algorithm through simulation.

[0005] Among the existing patent documents, the patent document with publication number CN109361237B discloses a method for optimizing the capacity of a microgrid based on an improved hybrid particle swarm algorithm. The patent document proposes an inertia weight adaptive cosine function decreasing method to update. In view of the problem that the particle swarm algorithm is prone to fall into the local optimum, it proposes to rank the subspace based on the rating function to guide the population to produce mutation operations. The patent document with publication number CN110165665A discloses a source-load-storage scheduling method based on an improved multi-objective particle swarm algorithm. In addition to the inertia weight decreasing according to the exponential function with the number of iterations, the patent document adds an inertia weight perturbation amount to make the inertia weight related to the average fitness of the population and the global fitness value. The patent document with publication number CN118589588A discloses a microgrid optimization scheduling system based on an improved particle swarm algorithm. In the patent document, data acquisition, data processing and other modules are introduced into the algorithm to improve the convergence speed of the algorithm and avoid particles from falling into the local optimum.

[0006] In view of this, the inventor hopes to design an improved multi-objective particle swarm algorithm for distribution network energy storage optimization. Summary of the invention

[0007] The purpose of the present invention is to overcome the above problems existing in the traditional technology and provide an improved multi-objective particle swarm algorithm applied to distribution network energy storage optimization, so that the active distribution network can operate stably and the energy storage configuration capacity can be minimized to achieve comprehensive optimization.

[0008] In order to achieve the above technical objectives and the above technical effects, the present invention is implemented through the following technical solutions:

[0009] The present invention provides an improved multi-objective particle swarm algorithm for optimizing energy storage in a distribution network, comprising the following steps:

[0010] S1. Determine the population number N and the maximum number of iterations Iter_Max, randomly and evenly distribute the particle position vectors, and set the current number of iterations to 0;

[0011] S2, based on the flow calculation, construct the objective function fitness value of distribution network node voltage fluctuation, grid connection point load fluctuation and energy storage rated capacity, which is used as the objective function fitness value f of the improved multi-objective particle swarm algorithm. ik (x), grid power constraint, node voltage constraint, energy storage capacity constraint, energy storage state of charge SOC, power and energy balance constraint are used as constraints;

[0012] S3, individual optimal pbest update: the fitness value f of each particle in this iteration ik (x) and the historical optimal fitness value pbest his If the particle has M objective function values ​​that are better than the previous value, it is called f i (x) dominates pbest his , then update the individual optimal value pbest of the current particle. If they do not dominate each other, there is a 50% probability that the current particle will be selected as the individual optimal value;

[0013] S4, global optimal gbest update, determine the external archive set archive, first store the first particle in the archive, then compare each individual in turn, remove the dominated solution, the non-dominated solution will be stored in the archive, and finally perform crowding sorting; when the number of archive solution sets exceeds the maximum limit, use the cyclic crowding sorting method to prune them, and select one of the solutions in the top 20% of the archive solution set as the population optimal solution;

[0014] S5. Propose an inertia weight update rule based on the optimal similarity of Euclidean distance to guide population update and adjust the inertia weight; use the inertia weight, learning factor, individual optimality and population optimality to calculate the speed of the next generation x; perform crossover mutation operations on the particle position vector according to the gap between the particle and the global optimal particle; use the boundary processing strategy to process the next generation of individuals;

[0015] S6. If the maximum number of iterations is met, stop the iteration and output the Pareto solution set; otherwise, Iter=Iter+1 and go to step S2;

[0016] S7. The output Pareto solution set is processed and the ordinal preference method based on information entropy is used to select the optimal solution.

[0017] Furthermore, the specific expression of the objective function of step S2 is as follows:

[0018]

[0019] Among them, N bus is the number of system nodes; T is the number of observation time; V ij is the voltage value of node i at time j; is the average value of node i at the time of inspection; P s (i) is the grid input power at time i; is the average value of grid input power at the time of investigation;

[0020] The maximum charge / discharge energy E' of the energy storage system during the investigation period store and the accumulated charge and discharge energy E' store The larger value of is taken as its rated capacity, and the output curve of the energy storage device is divided into n segments according to charging and discharging. There is no change in the charging and discharging state in each segment. The charging / discharging energy of the energy storage system in the i-th segment is C' storei , then the maximum charge / discharge energy of the energy storage device is E' store The calculation is as follows

[0021]

[0022] Among them, t is is the initial time of the i-th segment; t ie is the end time of the i-th segment; P store (t) is the charge / discharge power of the energy storage system at time t; Δt is the time interval between the two times;

[0023] Cumulative charge and discharge energy E" store The calculation method is as follows

[0024]

[0025] Among them, C st ” orej is the accumulated charge and discharge energy of the energy storage system at time j;

[0026] The rated capacity of energy storage is:

[0027] E store =max{E' store ,E” store} (5)

[0028] The total capacity of the energy storage system is:

[0029]

[0030] Among them, N store is the number of energy storage units in the system, E storek is the rated capacity of the kth energy storage.

[0031] Furthermore, the crossover operation in step 3 is:

[0032]

[0033] The mutation operation is:

[0034] x id =x min +(x max -x min )·r (11)

[0035] Where r is a random number between [0,1].

[0036] Furthermore, in step S4, the cyclic crowding degree sorting method is firstly used to sort the particles in the archive according to the crowding distance, and then the solution with the smallest crowding distance is deleted in each cycle until the number of solution sets meets the condition.

[0037] Furthermore, the calculation formula of the crowding distance of the mesoparticle is:

[0038]

[0039] Among them, x j 、x k is the distance x i The two closest particles, f m (x j ), f m (x k ) is particle x j The mth objective function value of mmax is the maximum value of the mth objective function of all particles.

[0040] Furthermore, the dynamic inertia weight update rule based on the Euclidean distance optimal similarity in step S5 is:

[0041]

[0042] Where D is the dimension of the solution space; W i (k) is the inertia weight of the i-th particle in the k-th generation; W max , W min is the maximum and minimum value of W; x max 、x minare the maximum and minimum values ​​of the particle position variable.

[0043] Furthermore, the crossover mutation operation steps in step S5 are:

[0044] 1) Calculate the difference X between each particle and the optimal particle according to formula (2): i , crossover probability p c and mutation probability p m ;

[0045] 2) Determine the difference threshold X, if X i >X, then perform crossover mutation operation, otherwise go to step 3;

[0046] 3) For each dimension of each particle i, select a random number r in [0,1] id1 , if r id1 <p c , perform mutation operation on this dimension component, and then select a random number r id2 , if r id2 <p m , perform a crossover operation, and the crossover object is the global optimal solution.

[0047] Furthermore, the boundary processing strategy in step S5 is:

[0048]

[0049] Among them, x dmax , x dmin is the maximum and minimum value of the d-dimensional independent variable, x d (t+1) represents the d-dimensional component of the particle, v d (t+1) represents the particle's d-dimensional flying speed, and t represents the number of iterations.

[0050] The beneficial effects of the present invention are:

[0051] 1. The present invention applies the energy storage system to the active distribution network, adopts a dynamic optimization scheduling scheme, takes a typical day of 24 hours as a scheduling cycle, establishes an energy storage optimization configuration model of node voltage fluctuation, load fluctuation and energy storage rated capacity within a scheduling cycle, comprehensively considers the safety and economy of the system, and provides certain technical support for the access of energy storage to the distribution network.

[0052] 2. The present invention proposes an inertia weight adaptive adjustment strategy based on the optimal similarity of Euclidean distance. When the difference is large, the inertia weight is also large, so that the particles have better global search ability. When the difference is small, the inertia weight is also small, so that the particles have local search ability. The convergence speed of the algorithm is improved.

[0053] Of course, any product implementing the present invention does not necessarily need to achieve all of the above advantages at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0055] Figure 1 is the algorithm flow chart;

[0056] Figure 2 It is the IEEE-33 node distribution network diagram;

[0057] Figure 3 is the inertia weight diagram;

[0058] Figure 4 Iterative convergence graph for the objective function f1;

[0059] Figure 5 Iterative convergence diagram for the objective function f2;

[0060] Figure 6 is the Pareto solution set of the standard particle swarm algorithm;

[0061] Figure 7 To improve the Pareto solution set of particle swarm algorithm;

[0062] Figure 8 This is the voltage fluctuation diagram without energy storage configuration;

[0063] Fig. 9 This is the voltage fluctuation diagram of the standard particle swarm algorithm;

[0064] Fig.10 Voltage fluctuation diagram for improving particle swarm algorithm. DETAILED DESCRIPTION

[0065] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0066] This embodiment proposes an improved multi-objective particle swarm algorithm for optimizing energy storage in distribution networks, so that the active distribution network can operate stably and the energy storage configuration capacity can be minimized to achieve comprehensive optimization.

[0067] (1) A mathematical model is established with the node voltage fluctuation, power fluctuation and energy storage rated capacity function of the distribution network as the optimization objectives.

[0068] The objective function of the energy storage optimization configuration model is:

[0069]

[0070] Among them, N bus is the number of system nodes, T is the number of observation time, V ij is the voltage value of node i at time j, is the average value of node i at the time of observation. s (i) is the grid input power at time i, is the average value of the grid input power at the time of investigation. The maximum charge / discharge energy E' of the energy storage system during the investigation period store and accumulated charge and discharge energy.

[0071] E' store The larger value of is taken as its rated capacity, and the output curve of the energy storage device is divided into n segments according to charging and discharging. There is no change in the charging and discharging state in each segment. The charging / discharging energy of the energy storage system in the i-th segment is C' storei , then the maximum charge / discharge energy of the energy storage device is E' store The calculation is as follows

[0072]

[0073] Where: t is is the initial time of the i-th segment; t ie is the end time of the i-th segment; P store (t) is the charge / discharge power of the energy storage system at time t; Δt is the time interval between the two times.

[0074] Cumulative charge and discharge energy E" store The calculation method is as follows

[0075]

[0076] Among them, C st ” orej is the accumulated charging and discharging energy of the energy storage system at time j.

[0077] The rated capacity of energy storage is:

[0078] E store =max{E' store ,E” store} (5)

[0079] The total capacity of the energy storage system is:

[0080]

[0081] Among them, N bus is the number of system nodes, T is the number of observation time, V ij is the voltage value of node i at time j, is the average value of node i at the time of observation. s (i) is the grid input power at time i, is the average value of the grid input power at the time of investigation. store is the number of energy storage units in the system, E storek is the rated capacity of the kth energy storage.

[0082] In the particle swarm algorithm optimization process, it is necessary to constrain the power of the grid, the energy storage assembly capacity, the energy storage SOC, the energy balance, the power constraint, the node voltage constraint and its constraints include:

[0083]

[0084] SOC min ≤SOC store ≤SOC max (twenty two)

[0085]

[0086] P store_min ≤P store ≤P store_max (twenty four)

[0087] V min ≤V ij ≤V max (25)

[0088] Among them, P s is the grid input power, P loadi is the load power of node i at a certain moment, P DGj is the output of the jth distributed generation at a certain moment, P storek It is the kth energy storage output at a certain moment. SOC max , SOC min , P store_max and P stoer_min are the upper and lower limits of energy storage SOC and energy storage output, V max 、V min are the upper and lower limits of the node voltage.

[0089] (2) Using the above optimization objective function as the fitness value of the improved multi-objective particle swarm algorithm, the energy storage model is solved to obtain a set of Pareto solutions.

[0090] The improved multi-objective particle swarm algorithm mentioned above has the following specific improvement measures:

[0091] Adaptive inertia weight: According to the distance between the current particle and the optimal particle, a dynamic inertia weight update rule based on the Euclidean distance optimal similarity is proposed, and its formula is:

[0092]

[0093] Where D is the dimension of the solution space, W i (k) is the inertia weight of the i-th particle in the k-th generation; W max , W min is the maximum and minimum value of W, x max 、x min are the maximum and minimum values ​​of the particle position variable.

[0094] (3) The TOPSIS method based on information entropy is used to select the optimal solution for energy storage optimization configuration from the obtained Pareto solution set.

[0095] First, the objective function is standardized:

[0096]

[0097] Among them, n is the size of the Pareto solution set, f ik 、f' ik are the actual objective function value and the standardized objective function value of the kth particle of the ith particle, respectively; are the negative ideal value and positive ideal value of the kth objective function respectively (the minimum value of the objective function is sought, so the minimum value is the positive ideal value).

[0098] The weight of the kth objective function is:

[0099]

[0100] in

[0101]

[0102] Calculate the distance scale S and relative proximity C i :

[0103]

[0104] in, are the positive and negative ideal distances of the particle, respectively; is the maximum and minimum value of the kth objective function after standardization, C i is the relative proximity, C i The bigger the better.

[0105] Step 1: Input photovoltaic, wind power and typical daily load curves. When selecting the energy storage site for the distribution network, it is necessary to encode the energy storage location, capacity, and 1-24 hour output. The encoding format is as follows:

[0106]

[0107] Among them, N s is the number of energy storage systems, x Ns Indicates N s The energy storage positions are encoded (the positions are integers), x sNs Indicates N s The energy storage capacity is encoded, x Ns,T Indicates the Nth s The power of the energy storage at time T.

[0108] Step 2: Set the population size N and the maximum number of iterations Iter_max.

[0109] Step 3: Energy storage requires grid power constraints, energy storage assembly capacity constraints, energy storage SOC, energy balance, power constraints, node voltage constraints and other constraints including:

[0110]

[0111] SOC min ≤SOC store ≤SOC max (38)

[0112]

[0113] P store_min ≤P store ≤P store_max (40)

[0114] V min ≤V ij ≤V max (41)

[0115] Among them, P s is the grid input power, P loadi is the load power of node i at a certain moment, P DGj is the output of the jth distributed generation at a certain moment, P storek It is the kth energy storage output at a certain moment. The upper and lower limits of energy storage capacity, SOC max , SOC min , P store_max and P stoer_min They are the upper and lower limits of energy storage SOC and energy storage output, V max 、V minare the upper and lower limits of the node voltage.

[0116] Step 4: Adjust relevant parameters on the IEEE-33 node system, such as Figure 2 As shown, the total active load of the system is 3715kW, the reactive load is 2300kVar, the reference voltage is 12.66kV, the node voltage allowable range is 0.95~1.05pu, and the photovoltaic power with a rated power of 200kW is connected to nodes 7 and 8, and the wind power with a rated power of 200kW is connected to nodes 25 and 32. The power flow calculation is performed for the photovoltaic, wind power, load, energy storage location, output conditions and constraints of each node at each moment, and the mathematical model of node voltage fluctuation, load fluctuation and energy storage capacity is established. The configuration model is:

[0117]

[0118] Among them, N bus is the number of system nodes, T is the number of observation time, V ij is the voltage value of node i at time j, is the average value of node i at the time of observation. s (i) is the grid input power at time i, is the average value of the grid input power at the time of investigation. storek is the maximum charge / discharge energy E' of the kth energy storage system during the observation period store and accumulated charge and discharge energy E” store The output curve of the energy storage device is divided into n segments according to charging and discharging. There is no change in the charging and discharging state in each segment. The charging / discharging energy of the energy storage system in the i-th segment is C' storei , then the maximum charge / discharge energy of the energy storage device is E" store The calculation is as follows

[0119]

[0120] Where: t is is the initial time of the i-th segment; t ie is the end time of the i-th segment; store (t) is the charge / discharge power of the energy storage system at time t; Δt is the time interval between the two times.

[0121] Cumulative charge and discharge energy E" store The calculation method is as follows

[0122]

[0123] Among them, C storej is the accumulated charging and discharging energy of the energy storage system at time j.

[0124] The rated capacity of energy storage is:

[0125] E store =max{E' store ,E” store} (47)

[0126] Step 5: Once screened, the iteration value of each particle is compared with the previous iteration value. If the three objective function values ​​of this particle are all smaller than the previous values, f(x i (k)) dominates f(x i (k-1)), then update the individual optimal value of this particle. If they do not dominate each other, there is a 50% probability of selecting the current particle as the individual optimal value.

[0127] Step 6: Secondary screening, determine the external archive set archive, first store the first particle in the archive, then compare each individual in turn, remove the dominated solution, and store the non-dominated solution in the archive; then sort the particles by crowding distance, the crowding distance of the particle is

[0128]

[0129] Among them, x j 、x k is the distance x i The two closest particles, f m (x j ), f m (x k ) is particle x j The mth objective function value of mmax is the maximum value of the mth objective function of all particles.

[0130] Step 7: When the number of archive solution sets exceeds the maximum limit, they need to be deleted. The cyclic congestion sorting method is used to sort all the particles in the solution set according to the congestion, and delete the solution with the highest congestion each time until the remaining solution sets meet the number requirements.

[0131] Step 8: Select one of the solutions in the top 20% of the archive solution set as the population optimal solution;

[0132] Step 9: Propose an inertia weight update rule based on the optimal similarity of Euclidean distance to guide population update. It updates the inertia weight according to the distance similarity between the current particle and the optimal particle, such as Figure 2 As shown, the formula is as follows:

[0133]

[0134] Where D is the dimension of the solution space, W i(k) is the inertia weight of the i-th particle in the k-th generation; W max , W min is the maximum and minimum value of W, x max 、x min are the maximum and minimum values ​​of the particle position variable.

[0135] Step 10: Use the inertia weight, learning factor, individual optimality and population optimality to calculate the speed of the next generation x. The formula is as follows:

[0136]

[0137] Step 11: According to formula (40), determine whether the particle should undergo crossover mutation operation, and perform crossover operation on the particle:

[0138]

[0139] The mutation operation is:

[0140] x id =x min +(x max -x min )·r (54)

[0141] Where r is a random number between [0,1].

[0142] Step 12: Perform boundary processing on particles. The boundary processing strategy is:

[0143]

[0144] Among them, x dmax , x dmin is the maximum and minimum value of the d-dimensional independent variable, x d (t+1) represents the d-dimensional component of the particle, v d (t+1) represents the particle's d-dimensional flying speed, and t represents the number of iterations.

[0145] Step 13: If the maximum number of iterations is met, stop the iteration and output the Pareto solution set; otherwise, Iter=Iter+1 and go to step 3;

[0146] Step 14: Output the Pareto solution set and use the ordinal preference method based on information entropy to select the optimal solution. The specific steps are as follows

[0147] First, the objective function is standardized:

[0148]

[0149] Among them, n is the size of the Pareto solution set, f ik、f' ik are the actual objective function value and the standardized objective function value of the kth particle of the ith particle, respectively; are the negative ideal value and positive ideal value of the kth objective function respectively (the minimum value of the objective function is sought, so the minimum value is the positive ideal value).

[0150] The weight of the kth objective function is:

[0151]

[0152] in

[0153]

[0154]

[0155] Calculate the distance scale S and relative proximity C i :

[0156]

[0157] Among them, S i + , are the positive and negative ideal distances of the particle, respectively; is the maximum and minimum value of the kth objective function after standardization, Ci is the relative proximity, and the larger Ci is, the better.

[0158] In order to illustrate the advantages of the improved multi-objective particle swarm algorithm in this embodiment in energy storage site selection and capacity determination in active distribution networks, two scenarios are set for comparative analysis:

[0159] Scenario 1: Classical multi-objective particle swarm algorithm for energy storage site selection and capacity determination

[0160] Scenario 2: Improved multi-objective particle swarm algorithm for energy storage site selection and capacity determination

[0161] Table 1: Simulation parameter settings

[0162] Parameter Type Value Iterations 200 Population size 100 Initial value of inertia weight 0.9 Final value of inertia weight 0.4 Pareto solution set size 100 Threshold X 1.2 <![CDATA[Crossing probability p c > 0.1 <![CDATA[Mutation probability p m > 0.05

[0163] The results of different optimization scenarios are shown in Table 2:

[0164] Table 2

[0165]

[0166] It can be seen from Table 2 that the optimization results obtained by the improved multi-objective particle swarm algorithm are better than those of the basic particle swarm algorithm. The two algorithms are run 10 times each, and the convergence of the objective functions f1 and f2 obtained by the basic multi-objective particle swarm algorithm and the improved multi-objective particle swarm algorithm external solutions is statistically obtained. Figure 4 and Figure 5 As shown. The Pareto solution set corresponding to the external solution is statistically Figure 6 and Figure 7 As shown in Figure 2, it can be seen that the improved multi-objective particle swarm algorithm is superior to the basic particle swarm algorithm in terms of convergence speed and convergence accuracy, and the improved multi-objective particle swarm algorithm is superior to the basic particle swarm algorithm in terms of search range and uniformity of solution set. The voltage obtained by the basic particle swarm algorithm and the improved multi-objective particle swarm algorithm through power flow calculation are shown in Figure 2. Figure 6 and Figure 7 As shown, it can be seen that the improved multi-objective particle swarm optimization algorithm has a good inhibitory effect on node voltage fluctuations.

[0167] Combined with wind energy, photovoltaics and typical daily load curves, a multi-objective optimization model of node voltage fluctuations, load power fluctuations and energy storage rated capacity is established to comprehensively consider the safety and economy of the system. Taking the fitness value of the objective function of the model as the particle, and then based on the distance between the position vector of the current particle and the position vector of the optimal particle, a dynamic inertia weight update law based on the optimal similarity of the Euclidean distance is proposed, so that the algorithm has better global search ability and local search ability during the optimization process. At the same time, cross-mutation is introduced into the algorithm to prevent it from converging to the local optimum. This embodiment uses an improved multi-objective particle swarm algorithm to have a good inhibitory effect on node voltage fluctuations, effectively improve the efficiency of solving the power system scheduling plan, and obtain a more comprehensive and reasonable scheduling plan.

[0168] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. An improved multi-objective particle swarm algorithm for energy storage optimization in distribution networks, characterized in that: The following steps are involved: S1. Determine the population number N and the maximum number of iterations Iter_Max, randomly and evenly distribute the particle position vectors, and set the current number of iterations to 0; S2, based on the flow calculation, construct the objective function fitness value of distribution network node voltage fluctuation, grid connection point load fluctuation and energy storage rated capacity, which is used as the objective function fitness value f of the improved multi-objective particle swarm algorithm. ik (x), grid power constraint, node voltage constraint, energy storage capacity constraint, energy storage state of charge SOC, power and energy balance constraint are used as constraints; S3, individual optimal pbest update: the fitness value f of each particle in this iteration ik (x) and the historical optimal fitness value pbest his If the particle has M objective function values ​​that are better than the previous value, it is called f i (x) dominates pbest his , then update the individual optimal value pbest of the current particle. If they do not dominate each other, there is a 50% probability that the current particle will be selected as the individual optimal value; S4, global optimal gbest update, determine the external archive set archive, first store the first particle in the archive, then compare each individual in turn, remove the dominated solution, the non-dominated solution will be stored in the archive, and finally perform crowding sorting; when the number of archive solution sets exceeds the maximum limit, use the cyclic crowding sorting method to prune them, and select one of the solutions in the top 20% of the archive solution set as the population optimal solution; S5. Propose an inertia weight update rule based on the optimal similarity of Euclidean distance to guide population update and adjust the inertia weight; use the inertia weight, learning factor, individual optimality and population optimality to calculate the speed of the next generation x; perform crossover mutation operations on the particle position vector according to the gap between the particle and the global optimal particle; use the boundary processing strategy to process the next generation of individuals; S6. If the maximum number of iterations is met, stop the iteration and output the Pareto solution set; otherwise, Iter=Iter+1 and go to step S2; S7. The output Pareto solution set is processed and the ordinal preference method based on information entropy is used to select the optimal solution.

2. The improved multi-objective particle swarm algorithm for distribution network energy storage optimization according to claim 1 is characterized in that: The specific expression of the objective function of step S2 is as follows: Among them, N bus is the number of system nodes; T is the number of observation time; V ij is the voltage value of node i at time j; V i is the average value of node i at the time of inspection; P s (i) is the grid input power at time i; is the average value of grid input power at the time of investigation; The maximum charge / discharge energy E of the energy storage system during the investigation period st ' ore and the accumulated charge and discharge energy E s ' tore The larger value of is taken as its rated capacity, and the output curve of the energy storage device is divided into n segments according to charging and discharging. There is no change in the charging and discharging state in each segment. The charging / discharging energy of the energy storage system in the i-th segment is C s ' torei , then the maximum charge / discharge energy E of the energy storage device st ' ore The calculation is as follows Among them, t is is the initial time of the i-th segment; t ie is the end time of the i-th segment; P store (t) is the charge / discharge power of the energy storage system at time t; Δt is the time interval between the two times; Cumulative charge and discharge energy E s ' to ' re The calculation method is as follows Among them, C st ” orej is the accumulated charging and discharging energy of the energy storage system at time j; The rated capacity of energy storage is: AND store =max{E s ' tore ,AND s ” tore } (5) The total capacity of the energy storage system is: Among them, N store is the number of energy storage units in the system, E storek is the rated capacity of the kth energy storage.

3. The improved multi-objective particle swarm algorithm for distribution network energy storage optimization according to claim 1 is characterized in that: The crossover operation in step 3 is: The mutation operation is: x id =x min +(x max -x min )·r (11) Where r is a random number between [0,1].

4. The improved multi-objective particle swarm algorithm for distribution network energy storage optimization according to claim 1 is characterized in that: In step S4, the cyclic crowding degree sorting method first sorts the particles in the archive according to the crowding distance, and then deletes the solution with the smallest crowding distance in each cycle until the number of solution sets meets the conditions.

5. The improved multi-objective particle swarm algorithm for distribution network energy storage optimization according to claim 4 is characterized in that: The calculation formula of the crowding distance of particles in the spherical nanostructured particle is: Among them, x j 、x k is the distance x i The two closest particles, f m (x j ), f m (x k ) is particle x j The mth objective function value; f mmax is the maximum value of the mth objective function of all particles.

6. The improved multi-objective particle swarm algorithm for distribution network energy storage optimization according to claim 1, characterized in that: The dynamic inertia weight update rule based on the optimal similarity of Euclidean distance in step S5 is: Where D is the dimension of the solution space; W i (k) is the inertia weight of the i-th particle in the k-th generation; W max , W min is the maximum and minimum value of W; x max 、x min are the maximum and minimum values ​​of the particle position variable.

7. The improved multi-objective particle swarm algorithm for distribution network energy storage optimization according to claim 1 is characterized in that: The crossover mutation operation steps in step S5 are: 1) Calculate the difference X between each particle and the optimal particle according to formula (2): i , crossover probability p c and mutation probability p m ; 2) Determine the difference threshold X, if X i >X, then perform crossover mutation operation, otherwise go to step 3; 3) For each dimension of each particle i, select a random number r in [0,1] id1 , if r id1 <p c , perform mutation operation on this dimension component, and then select a random number r id2 , if r id2 <p m , perform a crossover operation, and the crossover object is the global optimal solution.

8. The improved multi-objective particle swarm algorithm for distribution network energy storage optimization according to claim 1 is characterized in that: The boundary processing strategy in step S5 is: Among them, x dmax , x dmin is the maximum and minimum value of the d-dimensional independent variable, x d (t+1) represents the d-dimensional component of the particle, v d (t+1) represents the particle's d-dimensional flying speed, and t represents the number of iterations.

Images