Active power distribution network distributed energy storage optimization configuration method based on improved double-layer multi-target particle swarm optimization

Abstract

The invention provides an active power distribution network distributed energy storage optimization configuration method based on an improved double-layer multi-target particle swarm algorithm, and belongs to the field of novel power system energy storage planning. The method comprises the following steps of a distributed energy storage site selection method, a distributed energy storage capacity configuration method, a particle coding mode, double-layer multi-target particle swarm optimization improvement and active power distribution network distributed energy storage optimization configuration. Optimal configuration of the energy storage system is realized through voltage vulnerability index quantification, maximum interval method capacity calculation of time sequence power curve integral and improved double-layer multi-target particle swarm optimization solution. According to the method, the voltage stability of the system can be remarkably improved, active power loss and energy storage capacity configuration are reduced, and the method has high practical value and economical efficiency. The result analysis shows that the method can significantly reduce the annual comprehensive cost of the system, the vulnerability of the power grid, the active power loss, the energy storage capacity configuration and the node voltage deviation, and ensures that the voltage of the whole grid is in a safe and stable interval.

Description

Technical Field

[0001] The invention discloses an active distribution network distributed energy storage optimization configuration method based on an improved double-layer multi-objective particle swarm algorithm, and belongs to the field of new power system energy storage planning. Background Art

[0002] Distributed energy storage specifically addresses the fluctuating output and intermittent operation characteristics of renewable energy grid integration. Through a bidirectional power regulation mechanism, it enhances the grid's transient stability margin and power supply security, laying the technical foundation for integrating a high proportion of distributed power sources into new power systems. A multi-dimensional constraint-based site selection and sizing planning approach can systematically unleash the technical potential of distributed energy storage in grid frequency and voltage regulation, renewable energy absorption, and other areas, significantly improving the utilization efficiency of clean power assets. Consequently, the global academic community has established a continuous technical research approach focused on the coupled optimization of distributed energy storage spatial layout and capacity planning.

[0003] The article "Optimal Location and Capacity of the Distributed Energy Storage System in a Distribution Network" innovatively constructs an optimization architecture based on the combustion dynamics mechanism, implements regional division by defining metaphorical parameters such as node energy density characteristics, energy diffusion vector field and dynamic propagation rate, and then establishes a capacity configuration model guided by the annual peak value of zoned energy storage revenue; the article "Research on Optimal Configuration of Energy Storage Power Station Sizing and Site Selection Based on Improved Multi-Objective Particle Swarm Optimization" proposes a method based on the improved multi-objective particle swarm algorithm, first designs a variable coefficient active and reactive power droop control strategy, and then constructs a multi-objective optimization model including frequency regulation, voltage regulation comprehensive indicators and system cost. After verification by actual examples, it significantly improves the voltage and frequency regulation effects of the regional power grid; the article "Centralized Energy Storage Site Selection and Sizing Planning Strategy for "Dual High" Power Systems" constructs a comprehensive centralized energy storage site selection and planning indicator system, covering two types of structural and operational first-level indicators and six second-level indicators such as power transmission distribution factor. The entropy weight method and TOPSIS method are used for single evaluation, and the fuzzy Borda method is used for combined evaluation to accurately determine the energy storage site. The article "Energy Storage Site Selection and Sizing Planning Method Considering Dynamic Frequency Support" focuses on dynamic frequency support and innovatively establishes a battery energy storage planning model. By analyzing the system frequency response process and node inertia distribution, it scientifically guides energy storage site selection and plans energy storage capacity with the goal of minimizing investment and system operating costs. The article "Distributed Energy Storage Site Selection and Sizing for Distribution Networks Considering Islanding Time Uncertainty" uses robust optimization to describe the uncertainty of islanding periods, constructing a planning model based on the sum of energy storage investment, distribution network power purchase, and operation and maintenance costs. The column and constraint generation method is used to solve the problem, significantly improving the islanding operation capability and economic efficiency of the distribution network. The article "Optimal Configuration of Distributed Energy Storage Considering Peak Shaving Demand under the Aggregator Model" proposes a two-tiered optimization configuration model for distributed energy storage under the aggregator model. The outer model integrates energy storage arbitrage, peak-shaving subsidy benefits, and investment costs to maximize the annual revenue of the aggregator. The inner model focuses on energy storage-assisted peak-shaving of thermal power units, aiming to reduce the total cost of system peak-shaving. The article "Optimization Method for New Energy Distributed Energy Storage Configuration Considering Wind-Solar Complementarity" proposes a new energy distributed energy storage configuration optimization method considering wind-solar complementarity to address the low self-consumption rate of traditional energy storage configuration. This method analyzes configuration requirements, designs objective functions and constraints, and iteratively optimizes capacity. Experiments have shown that this method can increase the self-consumption rate and improve energy utilization efficiency. The article "Optimal Configuration of Distributed Energy Storage in Distribution Networks Based on Improved MOPSO Algorithm" addresses the problem of distributed photovoltaic access to distribution networks and constructs an energy storage optimization configuration model based on indicators such as voltage stability coefficient.An improved MOPSO algorithm is proposed, adjusting the inertia weight and learning factor, and introducing a gray correlation projection method for optimization. IEEE-33 node simulations have verified that this method can reduce performance indicators, improve efficiency, and improve safety. The article "Capacity Optimization Configuration of Distributed Photovoltaic Distribution Energy Storage Systems Based on an Improved Greedy Algorithm" comprehensively considers factors such as distribution network operating characteristics, photovoltaic output, and load fluctuations to construct a comprehensive optimization model. This paper introduces an improved greedy algorithm, using distribution lines as the starting point to determine the power supply configuration plan. Combined with multiple rounds of optimization adjustments, this method, verified through multi-scenario simulation comparisons, ensures that the system's operating power is more aligned with actual needs, effectively improving overall system performance.

[0004] Existing research has largely focused on quantitative analysis of the economic benefits of distributed energy storage, while research on the mechanisms by which distributed energy storage contributes to the voltage support capabilities of distribution networks is relatively limited. In particular, there is a lack of a theoretical framework for distributed energy storage site selection and sizing based on voltage support effectiveness. To address this research gap, this paper innovatively incorporates the dual constraints of economic indicators and grid vulnerability into a distributed energy storage planning model, achieving optimal equilibrium decisions based on these two characteristics through Pareto frontier analysis. Summary of the Invention

[0005] The current high penetration of distributed power sources in distribution networks has had a serious impact on the safe and stable operation of distribution networks. Active distribution network technology systems can effectively solve these problems. Distributed energy storage, with its inherent storage and release characteristics of electrical energy, significantly improves the active distribution network's ability to absorb distributed power sources, laying a solid foundation for building a reliably operating power system. Against this background, the present invention proposes a two-layer optimization model for active distribution networks that integrates distributed energy storage site selection and sizing decisions, providing an innovative solution for energy storage planning in new power systems.

[0006] The object of the present invention is achieved like this:

[0007] The method for optimizing the configuration of distributed energy storage in an active distribution network based on an improved double-layer multi-objective particle swarm optimization algorithm is characterized by comprising the following steps:

[0008] a) Distributed energy storage pre-site selection method;

[0009] b) Distributed energy storage capacity configuration method;

[0010] c) Particle encoding method;

[0011] d) Improved two-layer multi-objective particle swarm optimization algorithm;

[0012] e) Optimal configuration of distributed energy storage in active distribution networks.

[0013] 2. As a further improvement of the present invention, the distributed energy storage pre-site selection method in step a is performed, and the specific steps are:

[0014] Step a1: Perform power flow calculation based on the Newton-Raphson method, inputting network topology parameters and DG output data;

[0015] In step a2, combined with the power flow calculation results, the voltage vulnerability index of each node is quantified using the following formula:

[0016]

[0017] Among them, U t,i is the voltage of node i at time t; U i,o is the rated voltage at node i; 0.07 is the maximum voltage offset in a 10 kV power system;

[0018] Step a3: sort the vulnerability index of each node in descending order. A larger index value indicates a higher voltage collapse risk, and energy storage devices should be configured first.

[0019] In step a4, based on the improved algorithm, multi-objective optimization is performed on the basis of the energy storage installation priority sequence to solve the optimal configuration solution that meets the safety and economic constraints.

[0020] As a further improvement of the present invention, the distributed energy storage capacity configuration method in step b comprises the following specific steps:

[0021] Step b1, based on the DES charge and discharge power data, the improved algorithm is used to calculate the net power change of the energy storage system during the charge and discharge cycle to obtain the active power output curve of the energy storage;

[0022] In step b2, based on the charge and discharge status of the energy storage system during its operation cycle, the time series power curve is divided into continuous charge and discharge intervals based on the following formula, the active power variation characteristics of each independent period are identified, and the capacity demand of each period is calculated.

[0023]

[0024] Where: It represents the charging power and discharging power of the i-th DES at time t. It is positive when the energy storage is discharging and negative when the energy storage is charging. Δt is the time interval.

[0025] In step b3, the rated capacity parameters of a single-node DES are determined through extreme value analysis. A multi-node DES capacity collaborative optimization model is then constructed based on the vector superposition principle. Its mathematical representation is the linear superposition operation of the DES capacity parameters of each node, and the global optimal solution set of the system's total capacity configuration plan is ultimately output.

[0026] Wi ”=max(W i '(t)),t=1,2,…,T

[0027]

[0028] Where: W” i is the capacity required for energy storage i; W is the total capacity of DES.

[0029] As a further improvement of the present invention, the particle encoding method in step c is specifically as follows:

[0030] This paper constructs a hierarchical optimization framework and solves the model by improving the two-layer MOPSO algorithm. The algorithm uses a hierarchical decoupling encoding mechanism, in which the upper-layer particle swarm encoding corresponds to the main network purchase electricity cost of the DES in the distribution network node, and the lower-layer particle swarm dynamically represents the access location and configuration capacity of the DES equipment. The specific particle encoding method is designed as follows:

[0031] Step c1, upper layer particles:

[0032]

[0033] Where: represents the main grid electricity purchase cost of the i-th energy storage;

[0034] Step c2, lower layer particles:

[0035]

[0036] The lower particle consists of two parts. The first part represents the energy storage access location, i.e. represents the access location of the i-th energy storage; the second part represents the energy storage configuration capacity, that is, represents the capacity of the i-th energy storage;

[0037] Step c3, in addition, the number of DES accesses in the present invention is 2, therefore, The following constraints should be met:

[0038]

[0039] in, Takes 0 or 1, Indicates that the i-th DES is connected at node j, Indicates that node j has no DES access.

[0040] As a further improvement of the present invention, the improved double-layer multi-objective particle swarm algorithm is used in step d, and the specific steps are as follows:

[0041] Step d1, Parameter Initialization Protocol: Execute the power grid basic data loading protocol to obtain the initial operation parameter set and complete the collaborative configuration of algorithm parameters;

[0042] Step d2, Upper-layer Particle Swarm Initialization Protocol: Construct the particle population distribution based on the upper-layer model parameter space, define the initial position vector and velocity vector through constraint conditions, initialize the individual optimal solution vector and the global optimal solution vector, and set the iteration counter tu = 0;

[0043] Step d3, Upper-layer Particle Swarm Evolution Mechanism: Execute the dynamic update of the particle multi-dimensional state vector, implement the solution space scaling and boundary attraction mutation strategy for out-of-limit particles, and the iteration counter tu = tu + 1;

[0044] Step d4, Cross-layer Parameter Interaction Protocol: Transmit the upper-layer optimization solution vector to the lower-layer model through the parameter mapping interface, execute the lower-layer particle swarm initialization protocol, define the initial position vector and velocity vector, initialize the individual optimal solution vector and the global optimal solution vector, and set the iteration counter td = 0;

[0045] Step d5, Lower-layer Particle Swarm Evolution Mechanism: Implement the iterative optimization of the particle state vector, and the boundary processing follows the ε-neighborhood constraint rule, and the iteration counter td = td + 1;

[0046] Step d6, Fitness Evaluation Protocol: Based on the particle parameter update, distribute the DES power vector space, and obtain the real-time fitness evaluation value through the dynamic power flow calculation engine;

[0047] Step d7, Parameter Update Strategy: If td < max_td, trigger the loop iteration mechanism, introduce the quasi-oppositional learning strategy, the adaptive particle splitting strategy, and the adaptive inertia weight, establish the non-dominated solution screening mechanism, and optimize the solution space through the dynamic correction of the individual optimal solution vector and the collaborative update of the global optimal solution vector, and return to Step 5;

[0048] Step d8, Lower-layer Termination Judgment Protocol: If td ≥ max_td, execute the lower-layer non-dominated solution set output protocol;

[0049] Step d9, Reverse Parameter Interaction Protocol: Transmit the lower-layer optimization solution vector back to the upper-layer model through the feedback channel and execute the upper-layer fitness evaluation protocol;

[0050] Step d10, Parameter Update Strategy: If tu < max_tu, trigger the loop iteration mechanism, introduce the quasi-oppositional learning strategy, the adaptive particle splitting strategy, and the adaptive inertia weight, establish the non-dominated solution screening mechanism, and optimize the solution space through the dynamic correction of the individual optimal solution vector and the collaborative update of the global optimal solution vector, and return to Step 4;

[0051] Step d11, global termination decision protocol: if tu≥max_tu, execute the global non-inferior solution set output protocol.

[0052] Among them, tu and td represent the hierarchical iteration counter; tu max 、td max Defines the maximum iteration threshold.

[0053] As a further improvement of the present invention, in step e, the improved double-layer multi-objective particle swarm algorithm is used to optimize the configuration of distributed energy storage in the active distribution network, and the specific steps are as follows:

[0054] This paper proposes to use a two-layer multi-objective particle swarm model to carry out distributed site selection and sizing planning of active distribution networks. In this model, the planning is divided into two layers, the upper and lower layers, and different objective functions are obtained respectively:

[0055] Step e1, upper-level model: Consider the annual comprehensive cost of the energy storage system. Constraints include power balance constraints and distributed generation capacity constraints.

[0056] Step e2, lower-level model: Dynamically optimizes the DES's sequential charging and discharging strategy, targeting grid vulnerability at each node, minimizing total active power losses in the distribution network, and minimizing energy storage capacity. Constraints include distribution network power flow constraints and energy storage component constraints. The optimization framework utilizes an improved two-layer MOPSO algorithm, deeply coupled with the AC power flow calculation module, to achieve coordinated optimization of the upper-layer annual comprehensive cost minimization and the lower-layer energy storage site selection, sizing, and operation strategy.

[0057] The two-layer optimization model realizes information exchange between layers through a collaborative mechanism. The upper-layer model inputs the optimized decision variables (rated power parameters of the DG system) into the lower-layer model. The lower-layer model constructs operating constraints based on these parameters and initializes state variables. After the lower-layer model performs the optimization calculation of the DES access location, it transmits fitness indicators such as the distribution network voltage vulnerability index, active network loss value and energy storage capacity back to the upper-layer model. The upper-layer model evaluates the total objective function (annual comprehensive cost of the distribution network) based on these fitness indicators.

[0058] The beneficial effects of the present invention are:

[0059] First, in response to the slow convergence speed and premature local extreme value phenomenon of the traditional two-layer MOPSO algorithm when dealing with multi-objective high-dimensional optimization problems such as the optimal configuration of distributed energy storage systems, the present invention adopts a quasi-adversarial learning strategy, an adaptive particle splitting strategy and an adaptive inertia weight to improve the algorithm, and tests the algorithm performance using a multi-objective test function. Experimental data show that the optimized algorithm has significantly improved the uniformity of the Pareto front distribution, effectively suppressing the local convergence phenomenon.

[0060] Second, the improved dual-layer multi-objective particle swarm optimization algorithm designed in this invention significantly improves convergence efficiency and optimization accuracy, helping to obtain a more optimal system configuration solution. The optimization solution proposed in this invention significantly reduces the system's annual comprehensive cost, grid vulnerability, active network losses, energy storage capacity configuration, and node voltage deviation, ensuring that the voltage of the entire network remains within a safe and stable range. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 This is a flow chart of the method for optimizing the configuration of distributed energy storage in active distribution networks based on an improved double-layer multi-objective particle swarm algorithm according to the present invention;

[0062] Figure 2 This is the flowchart of the improved double-layer multi-objective particle swarm algorithm program of the present invention;

[0063] Figure 3 This is a diagram of the framework structure of the double-layer optimization model;

[0064] Figure 4 The results of the operation and state of charge change trend of distributed energy storage using the double-layer MOPSO solution algorithm for energy storage site selection and sizing for 11 nodes connected to energy storage;

[0065] Figure 5 The results of the operation and state of charge change trend of distributed energy storage using the double-layer MOPSO solution algorithm for energy storage site selection and sizing for 26 nodes connected to energy storage;

[0066] Figure 6 The operating strategy and state of charge change trend of distributed energy storage for energy storage site selection and capacity determination using the improved double-layer MOPSO solution algorithm of the present invention for 5 nodes connected to energy storage;

[0067] Figure 7 The operating strategy and state of charge change trend results of distributed energy storage using the improved double-layer MOPSO solution algorithm of the present invention for energy storage site selection and capacity determination for 24 nodes connected to energy storage;

[0068] Figure 8 The following is a comparison of simulation results of 24-hour voltage deviation of 17 nodes under different scenarios.

[0069] Figure 9 Figure 2 shows the daily load simulation results before and after energy storage configuration. DETAILED DESCRIPTION

[0070] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings. Specific implementation method 1

[0072] The flow chart of the distributed energy storage optimization configuration method for active distribution network based on the improved double-layer multi-objective particle swarm algorithm in this specific implementation is as follows: Figure 1As shown, the following steps are included:

[0073] a) Distributed energy storage pre-site selection method;

[0074] b) Distributed energy storage capacity configuration method;

[0075] c) Particle encoding method;

[0076] d) Improved two-layer multi-objective particle swarm optimization algorithm;

[0077] e) Optimal configuration of distributed energy storage in active distribution networks. Specific implementation method 2

[0079] The distributed energy storage optimization configuration method for active distribution network based on the improved double-layer multi-objective particle swarm optimization algorithm in this specific embodiment is further defined on the basis of the first specific embodiment:

[0080] As a further improvement of the present invention, the distributed energy storage pre-site selection method is performed in step a, and the specific steps are:

[0081] Step a1: Perform power flow calculation based on the Newton-Raphson method, inputting network topology parameters and DG output data;

[0082] In step a2, combined with the power flow calculation results, the voltage vulnerability index of each node is quantified using the following formula:

[0083]

[0084] Among them, U t,i is the voltage of node i at time t; U i,o is the rated voltage at node i; 0.07 is the maximum voltage offset in a 10 kV power system;

[0085] Step a3: sort the vulnerability index of each node in descending order. A larger index value indicates a higher voltage collapse risk, and energy storage devices should be configured first.

[0086] In step a4, based on the improved algorithm, multi-objective optimization is performed on the basis of the energy storage installation priority sequence to solve the optimal configuration solution that meets the safety and economic constraints.

[0087] As a further improvement of the present invention, the distributed energy storage capacity configuration method in step b comprises the following specific steps:

[0088] Step b1, based on the DES charge and discharge power data, the improved algorithm is used to calculate the net power change of the energy storage system during the charge and discharge cycle to obtain the active power output curve of the energy storage;

[0089] In step b2, based on the charge and discharge status of the energy storage system during its operation cycle, the time series power curve is divided into continuous charge and discharge intervals based on the following formula, the active power variation characteristics of each independent period are identified, and the capacity demand of each period is calculated.

[0090]

[0091] Where: It represents the charging power and discharging power of the i-th DES at time t. It is positive when the energy storage is discharging and negative when the energy storage is charging. Δt is the time interval.

[0092] In step b3, the rated capacity parameters of a single-node DES are determined through extreme value analysis. A multi-node DES capacity collaborative optimization model is then constructed based on the vector superposition principle. Its mathematical representation is the linear superposition operation of the DES capacity parameters of each node, and the global optimal solution set of the system's total capacity configuration plan is ultimately output.

[0093] W i ”=max(W i '(t)),t=1,2,…,T

[0094]

[0095] Where: W i ” is the capacity of energy storage i that needs to be configured; W is the total capacity of DES.

[0096] As a further improvement of the present invention, the particle encoding method in step c is specifically as follows:

[0097] This paper constructs a hierarchical optimization framework and solves the model by improving the two-layer MOPSO algorithm. The algorithm uses a hierarchical decoupling encoding mechanism, in which the upper-layer particle swarm encoding corresponds to the main network purchase electricity cost of the DES in the distribution network node, and the lower-layer particle swarm dynamically represents the access location and configuration capacity of the DES equipment. The specific particle encoding method is designed as follows:

[0098] Step c1, upper layer particles:

[0099]

[0100] Where: represents the main grid electricity purchase cost of the i-th energy storage;

[0101] Step c2, lower layer particles:

[0102]

[0103] The lower particle consists of two parts. The first part represents the energy storage access location, i.e. represents the access location of the i-th energy storage; the second part represents the energy storage configuration capacity, that is, represents the capacity of the i-th energy storage;

[0104] Step c3, in addition, the number of DES accesses in the present invention is 2, therefore, The following constraints should be met:

[0105]

[0106] in, Takes 0 or 1, Indicates that the i-th DES is connected at node j, Indicates that node j has no DES access.

[0107] As a further improvement of the present invention, the improved double-layer multi-objective particle swarm algorithm is used in step d, and the specific steps are as follows:

[0108] Step d1, parameter initialization protocol: execute the grid basic data loading protocol to obtain the initial operating parameter set and complete the collaborative configuration of algorithm parameters;

[0109] Step d2, upper particle swarm initialization protocol: construct particle population distribution based on the upper model parameter space, define the initial position vector and velocity vector through constraint conditions, initialize the individual optimal solution vector and the global optimal solution vector, and set the iteration counter tu = 0;

[0110] Step d3, upper particle swarm evolution mechanism: perform dynamic update of particle multidimensional state vector, implement solution space scaling and boundary attraction mutation strategy for out-of-limit particles, iteration counter tu=tu+1;

[0111] Step d4, cross-layer parameter interaction protocol: pass the upper layer optimization solution vector to the lower layer model through the parameter mapping interface, execute the lower layer particle swarm initialization protocol, define the initial position vector and velocity vector, initialize the individual optimal solution vector and the global optimal solution vector, and set the iteration counter td = 0;

[0112] Step d5, lower-layer particle swarm evolution mechanism: implement iterative optimization of particle state vector, boundary processing follows the ε neighborhood constraint rule, iteration counter td = td + 1;

[0113] Step d6, fitness evaluation protocol: update the DES power vector spatial distribution based on particle parameters, and obtain real-time fitness evaluation values through the dynamic power flow calculation engine;

[0114] Step d7, Parameter update strategy: If td < max_td, trigger the loop iteration mechanism, introduce the quasi-opposition learning strategy, adaptive particle splitting strategy, and adaptive inertia weight, establish a non-dominated solution screening mechanism, and optimize the solution space through the dynamic correction of the individual optimal solution vector and the collaborative update of the global optimal solution vector, then return to Step 5;

[0115] Step d8, Lower-layer termination determination protocol: If td ≥ max_td, execute the lower-layer non-dominated solution set output protocol;

[0116] Step d9, Reverse parameter interaction protocol: Transmit the lower-layer optimized solution vector back to the upper-layer model through the feedback channel, and execute the upper-layer fitness evaluation protocol;

[0117] Step d10, Parameter update strategy: If tu < max_tu, trigger the loop iteration mechanism, introduce the quasi-opposition learning strategy, adaptive particle splitting strategy, and adaptive inertia weight, establish a non-dominated solution screening mechanism, and optimize the solution space through the dynamic correction of the individual optimal solution vector and the collaborative update of the global optimal solution vector, then return to Step 4;

[0118] Step d11, Global termination determination protocol: If tu ≥ max_tu, execute the global non-dominated solution set output protocol.

[0119] Where, tu and td represent the hierarchical iteration counters; tu max , td max define the maximum iteration threshold.

[0120] As a further improvement of the present invention, in Step e, an improved double-layer multi-objective particle swarm algorithm is used for the optimal configuration of distributed energy storage in an active distribution network, and the specific steps are as follows:

[0121] The present invention intends to adopt a double-layer multi-objective particle swarm model for the distributed siting and sizing planning of an active distribution network. In this model, the planning is divided into two layers, the upper layer and the lower layer, and different objective functions are obtained respectively:

[0122] Step e1, Upper-layer model: Consider the annual comprehensive cost of the energy storage system. The constraint conditions are power balance constraints and capacity constraints of distributed power sources.

[0123] Step e2, Lower-layer model: Take the grid vulnerability of each node, the minimum total active power loss of the distribution network, and the minimum energy storage capacity configuration as the optimization objectives, and dynamically optimize the time-series charging and discharging strategies of DES. The constraint conditions cover distribution network power flow constraints, energy storage element constraints, etc. The optimization framework adopts an improved double-layer MOPSO algorithm, and through deep coupling with the AC power flow calculation module, it realizes the collaborative optimization of the minimum annual comprehensive cost of the upper layer and the energy storage siting, sizing, and operation strategies of the lower layer.

[0124] The two-layer optimization model realizes information exchange between layers through a collaborative mechanism. The upper-layer model inputs the optimized decision variables (rated power parameters of the DG system) into the lower-layer model. The lower-layer model constructs operating constraints based on these parameters and initializes state variables. After the lower-layer model performs the optimization calculation of the DES access location, it transmits fitness indicators such as the distribution network voltage vulnerability index, active network loss value and energy storage capacity back to the upper-layer model. The upper-layer model evaluates the total objective function (annual comprehensive cost of the distribution network) based on these fitness indicators.

[0125] Figure 2 In order to improve the program flow chart of the double-layer multi-objective particle swarm optimization algorithm, the solving steps of the improved double-layer multi-objective particle swarm optimization algorithm are summarized.

[0126] Figure 3 This is the framework structure diagram of the double-layer optimization model of the present invention, which better describes how the upper-layer planning model determines the parameter model and uses the upper-layer model results as input parameters of the lower-layer model, and the results of the lower-layer model react to the double-layer model.

[0127] Figure 4 Simulation results of the operation strategy and state of charge change trend of distributed energy storage using a double-layer multi-objective particle swarm optimization algorithm to select the site and size for 11 nodes connected to energy storage. Figure 5 The simulation results of the operation strategy and charge state change trend of distributed energy storage using a double-layer multi-objective particle swarm algorithm to select the site and size for 26 nodes connected to the energy storage.

[0128] Figure 6 Simulation results of the operation strategy and state of charge change trend of distributed energy storage for 5 nodes using an improved double-layer multi-objective particle swarm optimization algorithm for site selection and capacity determination. Figure 7 Simulation results of the operation strategy and state of charge change trend of distributed energy storage for 26 nodes using an improved double-layer multi-objective particle swarm optimization algorithm for site selection and capacity determination.

[0129] Depend on Figure 4 、 Figure 5 Figure 6 and Figure 7 The comparative analysis shows that the operation sequences of the two distributed energy storage units connected to the energy storage system using the improved double-layer multi-objective particle swarm optimization algorithm for site selection and capacity determination show similar rules.

[0130] Figure 8 The following is a comparison of simulation results of 24-hour voltage deviation of 17 nodes under different scenarios.

[0131] Scenario 1: No energy storage is connected for optimization; Scenario 2: Energy storage is connected, and the two-layer MOPSO solution algorithm is used for energy storage site selection and sizing; Scenario 3: Energy storage is connected, and the improved two-layer MOPSO solution algorithm is used for energy storage site selection and sizing.

[0132] In Scenario 1, without the DES, voltage excursions at 17 distribution network nodes were significant, particularly during peak load periods. The maximum voltage excursion reached 18.12% at 20:00, posing a serious threat to system stability. A comparative analysis of Scenario 2 and Scenario 3 shows that the optimized DES configuration effectively improves system voltage levels and reduces voltage excursions. The system optimized with the two-layer MOPSO algorithm reduced the voltage excursion rate to 10.40% at 20:00, while the improved algorithm further reduced it to 7.84%, achieving voltage regulation performance far superior to the traditional algorithm. These results demonstrate the effectiveness of the improved two-layer MOPSO algorithm in optimizing DES configuration and its significant superiority over the traditional two-layer MOPSO algorithm. DES effectively suppresses voltage excursions by providing dynamic power support to load nodes and altering the power flow distribution characteristics of the distribution network. Specifically, the energy storage system, through a bidirectional power regulation mechanism, releases energy to maintain voltage stability during peak load periods and absorbs excess power during low load periods to prevent voltage violations, ultimately keeping voltage excursions within the system's safe operating range.

[0133] Figure 9 Figure 2 shows the simulation results of the daily load curve before and after energy storage configuration. Comparative analysis results show that without energy storage, the system load curve exhibits significant peak-valley characteristics. At 4:00 a.m., the load power drops to a daily minimum of 0.6517 pu, and at 9:00 p.m., the load power climbs to a daily maximum of 1.0000 pu, with a peak-to-valley difference of 0.3483 pu. With the energy storage system configured, the system load level during low-load periods increases to 0.7123 pu, and the peak load during peak load periods decreases to 0.9455 pu, narrowing the daily peak-to-valley difference to 0.2332 pu. This result verifies that the energy storage system achieves peak-to-valley filling of the load curve through a dynamic charging and discharging strategy, significantly improving the system's power supply and demand balance.

[0134] The distributed energy storage optimization configuration method for active distribution networks based on an improved double-layer multi-objective particle swarm algorithm significantly reduces the system's annual comprehensive cost, grid vulnerability, active network losses, energy storage capacity configuration, and node voltage deviation, ensuring that the voltage of the entire network is within a safe and stable range. Analysis of the charge and discharge characteristics during the operating cycle shows that the energy storage system's SOC meets operating constraints throughout the entire process, helping to extend the equipment's service life.

Claims

1. An optimized configuration method for distributed energy storage in active distribution network based on improved double-layer multi-objective particle swarm optimization algorithm is characterized by: The following steps are involved: a) Distributed energy storage pre-site selection method; b) Distributed energy storage capacity configuration method; c) Particle encoding method; d) Improved two-layer multi-objective particle swarm optimization algorithm; e) Optimal configuration of distributed energy storage in active distribution networks.

2. The method for optimizing the configuration of distributed energy storage in active distribution networks based on an improved double-layer multi-objective particle swarm optimization algorithm according to claim 1, characterized in that: The distributed energy storage pre-site selection method is performed in step a, and the specific steps are: Step a1: Perform power flow calculation based on the Newton-Raphson method, inputting network topology parameters and DG output data; In step a2, combined with the power flow calculation results, the voltage vulnerability index of each node is quantified using the following formula: Among them, U t,i is the voltage of node i at time t; U i,o is the rated voltage at node i; 0.07 is the maximum voltage offset in a 10 kV power system; Step a3: sort the vulnerability index of each node in descending order. A larger index value indicates a higher voltage collapse risk, and energy storage devices should be configured first. In step a4, based on the improved algorithm, multi-objective optimization is performed on the basis of the energy storage installation priority sequence to solve the optimal configuration solution that meets the safety and economic constraints.

3. The method for optimizing the configuration of distributed energy storage in active distribution networks based on an improved double-layer multi-objective particle swarm optimization algorithm according to claim 2, characterized in that: The distributed energy storage capacity configuration method in step b comprises the following specific steps: Step b1, based on the DES charge and discharge power data, the improved algorithm is used to calculate the net power change of the energy storage system during the charge and discharge cycle to obtain the active power output curve of the energy storage; Step b2: Based on the charge and discharge status of the energy storage system during its operation cycle, the time-series power curve is divided into continuous charge and discharge intervals according to the following formula, the active power variation characteristics of each independent period are identified, and the capacity demand of each period is calculated; Where: P i DES (t) represents the charging power and discharging power of the i-th DES at time t, which is positive when the energy storage is discharging and negative when the energy storage is charging; Δt is the time interval; Finally, in step b3, the rated capacity parameters of a single-node DES are determined through extreme value analysis. A multi-node DES capacity collaborative optimization model is then constructed based on the vector superposition principle. This model is mathematically represented as a linear superposition operation of the DES capacity parameters of each node, ultimately outputting the global optimal solution set for the system's total capacity configuration. W i ”=max(W i '(t)),t=1,2,…,T Where: W i ” is the capacity of energy storage i that needs to be configured; W is the total capacity of DES.

4. The method for optimizing the configuration of distributed energy storage in active distribution networks based on an improved double-layer multi-objective particle swarm optimization algorithm according to claim 3, characterized in that: The particle encoding method in step c is as follows: This paper constructs a hierarchical optimization framework and solves the model by improving the two-layer MOPSO algorithm. The algorithm adopts a hierarchical decoupling coding mechanism, in which the upper-layer particle swarm encoding corresponds to the main network purchase electricity cost of the DES in the distribution network node, and the lower-layer particle swarm dynamically represents the access location and configuration capacity of the DES equipment. The specific particle coding method is designed as follows: Step c1, upper layer particles: Where: represents the main grid electricity purchase cost of the i-th energy storage; Step c2, lower layer particles: The lower particle consists of two parts. The first part represents the energy storage access location, i.e. represents the access location of the i-th energy storage; the second part represents the energy storage configuration capacity, that is, represents the capacity of the i-th energy storage; Step c3, in addition, the number of DES accesses in the present invention is 2, therefore, The following constraints should be met: in, Takes 0 or 1, Indicates that the i-th DES is connected at node j, Indicates that node j has no DES access.

5. The method for optimizing the configuration of distributed energy storage in active distribution networks based on an improved double-layer multi-objective particle swarm optimization algorithm according to claim 4, characterized in that: In step d, an improved double-layer multi-objective particle swarm algorithm is used, and the specific steps are as follows: Step d1, parameter initialization protocol: execute the grid basic data loading protocol to obtain the initial operating parameter set and complete the collaborative configuration of algorithm parameters; Step d2, Upper-layer Particle Swarm Initialization Protocol: Construct the particle population distribution based on the upper-layer model parameter space, define the initial position vector and velocity vector through constraint conditions, initialize the individual optimal solution vector and the global optimal solution vector, and set the iteration counter tu = 0; Step d3, Upper-layer Particle Swarm Evolution Mechanism: Execute the dynamic update of the particle multi-dimensional state vector, implement the solution space scaling and boundary attraction mutation strategy for out-of-limit particles, and the iteration counter tu = tu + 1; Step d4, Cross-layer Parameter Interaction Protocol: Transfer the upper-layer optimized solution vector to the lower-layer model through the parameter mapping interface, execute the lower-layer particle swarm initialization protocol, define the initial position vector and velocity vector, initialize the individual optimal solution vector and the global optimal solution vector, and set the iteration counter td = 0; Step d5, Lower-layer Particle Swarm Evolution Mechanism: Implement the iterative optimization of the particle state vector, and the boundary handling follows the ε-neighborhood constraint rule, and the iteration counter td = td + 1; Step d6, Fitness Evaluation Protocol: Based on the particle parameter update, distribute the DES power vector space, and obtain the real-time fitness evaluation value through the dynamic power flow calculation engine; Step d7, Parameter Update Strategy: If td < max_td, trigger the loop iteration mechanism, introduce the quasi-oppositional learning strategy, adaptive particle splitting strategy, and adaptive inertia weight, establish the non-dominated solution screening mechanism, and optimize the solution space through the dynamic correction of the individual optimal solution vector and the collaborative update of the global optimal solution vector, and return to Step 5; Step d8, Lower-layer Termination Judgment Protocol: If td ≥ max_td, execute the lower-layer non-dominated solution set output protocol; Step d9, Reverse Parameter Interaction Protocol: Transfer the lower-layer optimized solution vector back to the upper-layer model through the feedback channel, and execute the upper-layer fitness evaluation protocol; Step d10, Parameter Update Strategy: If tu < max_tu, trigger the loop iteration mechanism, introduce the quasi-oppositional learning strategy, adaptive particle splitting strategy, and adaptive inertia weight, establish the non-dominated solution screening mechanism, and optimize the solution space through the dynamic correction of the individual optimal solution vector and the collaborative update of the global optimal solution vector, and return to Step 4; Step d11, Global Termination Judgment Protocol: If tu ≥ max_tu, execute the global non-dominated solution set output protocol; Among them, tu and td represent the hierarchical iteration counter; tu max 、td max Defines the maximum iteration threshold.

6. A method for optimizing the configuration of distributed energy storage in an active distribution network based on an improved double-layer multi-objective particle swarm algorithm according to claim 5, characterized in that in step e, the improved double-layer multi-objective particle swarm algorithm is used to optimize the configuration of distributed energy storage in an active distribution network, and the specific steps are as follows: The present invention intends to adopt a double-layer multi-objective particle swarm model for the active distribution network distributed site selection and sizing planning. In this model, the planning is divided into two layers, the upper layer and the lower layer, and different objective functions are obtained respectively: Step e1, Upper-layer Model: Consider the annual comprehensive cost of the energy storage system, and the constraint conditions are power balance constraints and capacity constraints of distributed power sources; Step e2, lower-level model: Dynamically optimize the DES's sequential charging and discharging strategy, taking the grid vulnerability of each node, minimizing the total active power loss of the distribution network, and minimizing the energy storage capacity configuration as optimization objectives. Constraints include distribution network power flow constraints and energy storage component constraints. The optimization framework uses an improved two-layer MOPSO algorithm, which is deeply coupled with the AC power flow calculation module to achieve coordinated optimization of the upper-layer annual comprehensive cost minimization and the lower-layer energy storage site selection, sizing, and operation strategy. The two-layer optimization model realizes information exchange between layers through a collaborative mechanism. The upper-layer model inputs the optimized decision variables (rated power parameters of the DG system) into the lower-layer model. The lower-layer model constructs operating constraints based on these parameters and initializes state variables. After the lower-layer model performs the optimization calculation of the DES access location, it transmits fitness indicators such as the distribution network voltage vulnerability index, active network loss value and energy storage capacity back to the upper-layer model. The upper-layer model evaluates the total objective function (annual comprehensive cost of the distribution network) based on these fitness indicators.

Images