Distribution network reconstruction topology-power flow-electric energy quality multi-objective collaborative optimization method

Abstract

The invention relates to the technical field of distribution network optimization, and discloses a distribution network reconstruction topology-power flow-power quality multi-objective collaborative optimization method, which comprises the following steps: acquiring a distribution network node load curve and a photovoltaic output curve, clustering to generate a representative time sequence scene set, and calculating the weight of each scene; constructing a multi-dimensional power quality index system based on the scene weight; constructing a coupling sensitivity matrix of switch action, power flow change and power quality index response, and calculating a switch comprehensive priority; constructing a multi-target collaborative optimization model; precise search is realized based on the comprehensive priority of the switch, an optimal solution is generated, and a reconstruction scheme is obtained; a reconstruction scheme is linked with dynamic governance devices such as SVG / APF / OLTC, harmonic governance, voltage regulation and risk pre-judgment collaboration are achieved, and a full closed loop is formed through effect verification, exception handling and feedback updating. The problem that traditional static modeling is poor in adaptability is effectively solved, the practicability and stability of the optimization effect are remarkably improved, and meanwhile the quality of the optimal solution is guaranteed.

Description

Technical Field

[0001] This invention relates to the field of distribution network optimization technology, and more specifically, to a multi-objective collaborative optimization method for distribution network reconfiguration topology, power flow, and power quality. Background Technology

[0002] In the process of building new power systems, the large-scale grid connection of distributed generation and the widespread application of power electronic equipment have made the operating characteristics of distribution networks increasingly complex. Load and power output exhibit significant temporal fluctuations and uncertainties within 24 hours, posing multiple challenges to the economic operation, reliable power supply, and power quality assurance of distribution networks. Distribution network reconfiguration, as a core means of optimizing power flow distribution and reducing network losses, can effectively improve power supply performance by adjusting switch states to change the network topology. Meanwhile, dynamic management devices such as SVG, APF, and OLTC can quickly respond to power quality issues such as voltage deviation and harmonic pollution. The coordinated regulation of these two systems has become a research hotspot in the field of distribution network optimization.

[0003] Currently, although power quality indicators have been gradually introduced into distribution network optimization research, there are still many shortcomings: some studies only focus on a single objective or a few power quality indicators, lacking a comprehensive consideration of multi-dimensional power quality requirements such as steady state, harmonics, transients, and three-phase imbalance; most methods are based on static single-period modeling, ignoring the time-series characteristics of daily load fluctuations, renewable energy uncertainties, and power quality disturbance events, making it difficult for optimization results to adapt to the actual changing operating conditions; in multi-objective optimization, the method of manually setting weights or geometric averages is often used to transform multiple objectives into a single objective, which cannot adaptively balance the inherent contradictions between economy, reliability, and power quality. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a multi-objective collaborative optimization method for distribution network reconfiguration topology, power flow, and power quality. By constructing time-series scenarios, building a multi-dimensional power quality index system, calculating switch sensitivity matrices and priorities, and constructing a multi-objective collaborative optimization model, this method effectively solves the pain points of existing technologies.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows:

[0006] A multi-objective collaborative optimization method for distribution network reconfiguration topology, power flow, and power quality includes the following steps:

[0007] S1. Collect the 24-hour load curves and photovoltaic output curves of the distribution network nodes, generate a representative time-series scenario set through K-means clustering, and calculate the weight of each scenario;

[0008] S2. Construct a multi-dimensional power quality index system based on the scenario weights. The multi-dimensional power quality index system includes: steady-state voltage deviation index, harmonic distortion index, voltage sag sensitivity index, and three-phase imbalance and flicker comprehensive index.

[0009] S3. Construct a coupling sensitivity matrix of switch action, power flow change and power quality index response. After normalizing the matrix, calculate the overall switch priority.

[0010] S4. Using topological decision variables and dynamic governance synergistic variables as decision variables, and multi-scenario weighted network loss, power outage time, and power quality indicators as objective functions, and incorporating constraints such as radiality, power flow, and equipment capacity, a multi-objective collaborative optimization model is constructed.

[0011] S5. A sensitivity-guided multi-objective particle swarm optimization algorithm is adopted, which relies on the switching comprehensive priority to achieve accurate search, generate the optimal solution for the multi-objective collaborative optimization model, and obtain the reconstruction scheme.

[0012] Furthermore, it also includes step S6:

[0013] S6. Link the reconfiguration scheme with dynamic governance devices such as SVG / APF / OLTC to achieve coordinated harmonic governance, voltage regulation, and risk prediction, and form a closed loop through effect verification, anomaly handling, and feedback updates.

[0014] As a preferred embodiment of the present invention, the load curve in S1 includes an active power load curve. and reactive load curve The sampling interval was 15 minutes, with a total of 96 time points;

[0015] Photovoltaic power output curve The output curve of a photovoltaic power station over 24 hours;

[0016] The specific steps for generating a time-series scene set and calculating the weight of each scene are as follows:

[0017] Determine the number of scenarios using the elbow rule. Construct normalized clustering feature vectors , ,in, The total number of nodes. This represents the maximum active load of node 1. To maximize the output of photovoltaic power;

[0018] The clustering objective is to minimize the deviation between the feature vectors within the scene and the scene center vector. The formula for the clustering objective function is: , For the scene The set of original time points included. For the scene The central feature vector of the data ultimately aggregates 96 original time points into 24 typical scenarios, with each scenario corresponding to a similar running state.

[0019] Finally, the scene weights are calculated. The weight calculation formula is: ,in For the scene The number of original time points included, and satisfying the following conditions. .

[0020] As a preferred embodiment of the present invention, the steady-state voltage deviation index, harmonic distortion index, voltage sag sensitivity index, and three-phase imbalance and flicker comprehensive index in S2 are all calculated with weighted values ​​under multiple scenarios:

[0021] Steady-state voltage deviation index Defined as the average degree to which the voltage of all nodes in the network deviates from the rated value, reflecting voltage stability; its calculation formula is: ,in The system's rated voltage. For the scene Next node voltage amplitude, The total number of nodes;

[0022] Harmonic distortion index Defined as assessing voltage distortion caused by harmonics across the entire frequency band from the 3rd to the 50th order, reflecting the degree of harmonic pollution; its calculation formula is: The total harmonic distortion rate In the formula For nodes In the scene of Second harmonic voltage amplitude This represents the amplitude of the fundamental voltage.

[0023] Voltage sag sensitivity index This is designed for sensitive loads such as precision instruments and data centers, quantifying their sensitivity to voltage sags; the calculation formula is as follows: In the formula, For the set of sensitive load nodes, Its quantity;

[0024] Voltage sag sensitivity By descent depth With duration Graded quantization, voltage sag sensitivity The formula for calculation is:

[0025] ;

[0026] Three-phase imbalance and flicker comprehensive index Its positioning is to integrate three-phase imbalance and voltage flicker, and comprehensively evaluate the three-phase symmetry and voltage stability of the distribution network; its calculation formula is: ;in , ; , , , These represent positive-sequence, negative-sequence, and zero-sequence currents. This is a short-time flicker value.

[0027] As a preferred embodiment of the present invention, the specific steps for constructing the coupling sensitivity matrix in S3 are as follows:

[0028] Constructing coupling sensitivity It is defined as quantifying the marginal impact of a single switch state change on a certain PQ index, realizing coupled modeling of switch action, power flow change, and PQ response; its formula is:

[0029] In the formula For switch The state change, Changes in active power of the line caused by switching action;

[0030] For each switch Calculate separately and The three-phase power flow at that time is obtained from the line In the scene Down, switch Active power amplitude when in disconnected state ,line In the scene Down, switch Active power amplitude when in closed state ,node In the scene Down, switch steady-state fundamental voltage amplitude when in the off state ,node In the scene Down, switch steady-state fundamental voltage amplitude when in closed state Solve for the change in the PQ index before and after the switch action. With line power change ,

[0031] The calculation formula is: , ;

[0032] The final construction dimension is Sensitivity matrix , , This represents the total number of candidate switches.

[0033] As a preferred embodiment of the present invention, the sensitivity matrix in S3 The normalization formula is:

[0034]

[0035] In the formula: For switch For the first Normalized coupling sensitivity of PQ-like indices switch For the first The original coupling sensitivity of the PQ-like index, This represents the minimum original coupling sensitivity of all candidate switches. Maximum original coupling sensitivity of all candidate switches;

[0036] Before calculating the overall priority of the switches, the PQ index weights need to be calculated. First, the sensitivity matrix is ​​normalized. The decision matrix is ​​defined by behavioral switches and columns representing PQ indicators; then, according to... Calculate the proportion of indicators Then calculate the entropy value. The formula for calculating entropy is: , hour Finally, calculate the weights. , ,satisfy ;

[0037] Switch overall priority The calculation formula is: , .

[0038] As a preferred embodiment of the present invention, the decision variables in S4 include topological decision variables and dynamic governance synergy variables, specifically defined as follows:

[0039] Topology decision variables: set as switch state vectors ,in This represents the total number of candidate switches for the distribution network. , Indicates the line When the switch is closed, Indicates the line The switch is open, and the initial value of the vector is taken from the current switch state of the distribution network real-time topology database;

[0040] Dynamic governance coordination variables include SVG reactive power output adjustment and OLTC gear adjustment, specifically:

[0041] SVG reactive power adjustment: set as a vector ,in This represents the total number of distribution network nodes. The total number of representative time-series scenes generated in step S1. Represents a node In the scene The SVG reactive power output adjustment value is set to 0 initially.

[0042] OLTC gear adjustment amount: set as vector ,in Representing a scene The tap change of the on-load tap-changing transformer, the tap adjustment range is: The initial value is set to the current running gear;

[0043] The objective function is to minimize the multi-scene weighted vector. The scene weights of each target component are calculated based on S1. The weighted average is as follows:

[0044] Economic objectives The formula for calculating the weighted network loss across multiple scenarios is as follows:

[0045]

[0046] In the formula, For the collection of distribution network lines, , The lines are respectively In the scene The active power and reactive power are below. For the line In the scene Average voltage under, For the line The resistance, For the scene The duration;

[0047] Reliability targets : Weighted outage time for multiple scenarios, using the system average outage duration. The characterization and calculation formula is as follows:

[0048]

[0049] In the formula, For nodes In the scene The failure rate is as follows For nodes In the scene The following fault repair time, For nodes The number of users;

[0050] Power quality objectives include three components, which directly utilize the multi-dimensional power quality index system constructed in step S2, specifically: , , ,in The steady-state voltage deviation index This is a harmonic distortion index. This is an indicator of voltage sag sensitivity. It is a comprehensive index of three-phase imbalance and flicker.

[0051] In a preferred embodiment of the present invention, in step S4, the constraints of the multi-objective collaborative optimization model include radial constraint, power flow constraint, harmonic constraint, equipment capacity constraint, and switching action constraint, and the specific settings of each constraint are as follows:

[0052] The radial constraint is used to ensure that the distribution network topology is free of loops and islands, thus ensuring power supply safety. The specific verification method is as follows:

[0053] Loop-free verification: Constructing the distribution network topology correlation matrix The matrix dimension is ,in This represents the total number of distribution network nodes. The total number of candidate switches is given. The matrix element values ​​are determined by the following rule: if a node is connected to the corresponding line switch, the value is 1; otherwise, it is 0. When the following conditions are met... When it is determined that there are no loops in the distribution network topology, the formula is as follows: This represents the calculation of the matrix determinant;

[0054] Islanding absence verification: A depth-first search algorithm is used to traverse all nodes starting from the power distribution node. If all nodes can be traversed, it is determined that there are no islands; otherwise, there are islanding regions. For topologies that do not meet the radial constraint, the repair is achieved by adjusting the switch status.

[0055] The power flow constraint is used to ensure node power balance, and the specific formula is as follows:

[0056] In the formula: , They are nodes The set of incoming lines and the set of outgoing lines; , They are nodes In the scene The active and reactive power balance values ​​below; , They are nodes In the scene The active and reactive power of the load; , They are nodes In the scene The active and reactive power outputs of the distributed power sources are taken from the scenario data in step S1. For nodes In the scene The reactive power output adjustment of the SVG is the dynamic governance coordination variable in step S4. , The lines are respectively In the scene The transmitted active and reactive power;

[0057] , The lines are respectively In the scene The active and reactive power losses are calculated using the following formula: , ,in For the line In the scene Average voltage under, , The lines are respectively The resistance and reactance are taken from the power distribution network equipment parameter database;

[0058] The harmonic constraint is used to control harmonic pollution of the 3rd to 50th orders. It is constructed based on harmonic power flow characteristics, and the specific formula is as follows: ;

[0059] In the formula: The harmonic order is 3-50; , They are nodes ,node In the scene Below Second harmonic voltage phasor;

[0060] For the line In the scene Below The subharmonic current phasor includes nonlinear load injection current;

[0061] For the line of The formula for calculating the second harmonic impedance is as follows: ,in The imaginary unit;

[0062] The equipment capacity constraints are used to prevent equipment from operating under overload conditions. These include line current constraints, transformer load rate constraints, and SVG / APF capacity constraints, specifically:

[0063] Line current constraints: ,in For the line In the scene The actual operating current under these conditions For the line The rated current is taken from the power distribution network equipment parameter database;

[0064] Transformer load factor constraint: ,in For transformers in the scene The actual load capacity under the current conditions The rated capacity of the transformer is taken from the distribution network equipment parameter database;

[0065] SVG / APF capacity constraints: ,in For nodes The rated capacity of the SVG is taken from the power distribution network equipment parameter library;

[0066] The switch action constraint is used to avoid equipment damage caused by frequent switch operation. The constraint rule is: count the number of actions of all candidate switches in 24 hours, and the total number of actions per day is ≤3.

[0067] All constraints need to be independently verified in each representative time series scenario generated in step S1. Decision variable combinations that do not meet the constraints are determined to be infeasible solutions and need to be returned for adjustment.

[0068] As a preferred embodiment of the present invention, the specific process of S5 is as follows:

[0069] Algorithm initialization:

[0070] Particle generation: Set population size Randomly generated A switch state vector The topology is repaired by disconnecting the lowest priority switch in the loop and closing the highest priority tie switch in the island.

[0071] Parameter initialization: Initial particle velocity Individual optimal solution ; Calculate the objective function value for each particle After non-dominated sorting by NSGA-III, the particle with the largest crowding distance in the first layer is selected as the global optimal solution. ;

[0072] Iterative optimization:

[0073] Set the maximum number of iterations acceleration coefficient , Sensitivity weight speed threshold ;

[0074] Sensitivity-guided speed update: Based on switch priority directional search, the formula is:

[0075]

[0076] Where: linearly decreasing inertia weight , It is a random number. This is the switching priority vector for S3; it is clipped when the speed exceeds the threshold. ;

[0077] Location updates and fixes: via Implement binary position updates and synchronously perform topology repair;

[0078] Cutting-edge update: Computing And verify the constraints, if Dominate Then update Update after reordering Maintaining Pareto frontier diversity;

[0079] Convergence criterion: Rate of change in congestion at the Pareto front for 10 consecutive generations Or the number of iterations reaches Output the Pareto optimal solution set ;

[0080] Optimal solution selection:

[0081] Entropy weight calculation: for the solution set Target matrix Normalization Calculate the entropy value ( Entropy weight ; To solve for the quantity;

[0082] Proximity calculation: Determining the ideal solution Negative ideal solution Calculate distance , Proximity ;

[0083] Optimal solution output: Selecting proximity The largest solution is used as the final reconstruction scheme. Synchronous output , .

[0084] As a preferred embodiment of the present invention, the harmonic mitigation coordination specifically refers to: when the scenario Next node Total Harmonic Distortion This triggers APF compensation, the formula is:

[0085]

[0086] Where: harmonic order 3-50 times, hour Higher harmonics ; For nodes of Subharmonic injection current; For nodes Fundamental voltage amplitude;

[0087] The voltage regulation coordination specifically refers to:

[0088] OLTC range optimization: Aiming for minimum voltage deviation, the formula is as follows: , This is the system's rated voltage;

[0089] SVG reactive power adjustment: calculated based on voltage deviation, the formula is as follows. , For adjustment coefficients, The node voltage after reconstruction;

[0090] The risk prediction collaboration specifically involves: predicting the probability of PQ exceeding the standard using an LSTM model. , The risk level is low; maintain the parameters; 10% ≤ For medium risk, reserve SVG margin. ; High risk, closed A high-sensitivity switch, and .

[0091] As a preferred embodiment of the present invention, the exception handling mechanism in S6 is specifically as follows:

[0092] Troubleshooting Equipment Failures:

[0093] SVG / APF fault: Compensation for capacity expansion of adjacent node devices, the formula is as follows: In the formula Adjacent nodes These are the rated capacities of the adjacent and faulty nodes, respectively;

[0094] OLTC fault: SVG emergency voltage regulation, formula is as follows , , For rated voltage, The node voltage after reconstruction;

[0095] Topology anomaly handling:

[0096] Switch fails to operate: Select priority Replace with a spare switch. The original switch priority;

[0097] Sudden Islanding: Closure Sensitivity The interconnection switch synchronously cuts off non-critical loads to restore power supply;

[0098] The feedback update closed loop is specifically as follows: Update trigger: Regular update at 2:00 AM daily; Emergency update when PQ index exceeds the standard for 3 consecutive points or photovoltaic fluctuation > 20%;

[0099] Update content:

[0100] Scenario Update: Re-execute S1 clustering with new data; adjust the number of scenarios if the clustering error is > 0.1. ;

[0101] Sensitivity Update: Recalculate the S3 coupling sensitivity matrix and remove standard deviation. outliers;

[0102] Algorithm parameter update: When the number of scenes increases, S5's and Simultaneous increase of 20%;

[0103] Closed-loop iteration: The updated data and parameters are re-entered into S4-S6 to generate a new solution to replace the original solution.

[0104] The beneficial technical effects of this invention are:

[0105] K-means clustering is used to divide the 24-hour load and photovoltaic output curves into scenarios. The elbow rule is used to determine the reasonable number of scenarios, generate a representative time series scenario set and calculate the scenario weights. This allows the optimization scheme to be specifically adapted to high-probability operating scenarios, effectively solving the problem of poor adaptability of traditional static modeling and significantly improving the practicality and stability of the optimization effect.

[0106] A multi-dimensional power quality index system covering steady-state voltage deviation, harmonic distortion rate, voltage sag sensitivity, three-phase imbalance and flicker is constructed. Each index is comprehensively evaluated through multi-scenario weighted calculation, which fully covers the power demand of different loads and makes up for the shortcomings of the one-sided coverage of existing technical indicators.

[0107] The constructed coupling sensitivity matrix, through normalization and entropy weight method to calculate the comprehensive priority of the switches, guides the sensitivity-guided multi-objective particle swarm optimization algorithm to perform directional search, solving the problem of blind search in traditional algorithms, improving the convergence speed by more than 30%, and ensuring the quality of the optimal solution.

[0108] By adopting Pareto optimization instead of fixed weighting, a multi-objective optimization model is constructed to calculate multi-scenario weighted network loss, outage time, and multi-dimensional power quality indicators. The objective weights are adaptively determined to avoid the defects of subjective human setting and achieve the synergistic optimization of distribution network economy, reliability, and power quality, thus overcoming the subjective problem of weights in existing multi-objective optimization.

[0109] This approach achieves complementary synergy between slow-dynamic topology optimization and fast-dynamic device control, fully leveraging the advantages of both to effectively address the limitations of single control methods and improve distribution network operational stability and power quality compliance rates. A robust anomaly handling mechanism is designed, along with backup compensation or alternative strategies to ensure power supply continuity. A closed-loop update mechanism is established, dynamically adjusting the scenario set, sensitivity matrix, and algorithm parameters based on real-time operational data. This ensures the optimization scheme continuously adapts to changes in distribution network operating conditions, maintaining optimal control performance over the long term. Attached Figure Description

[0110] Figure 1 This is a schematic diagram of the overall process of the present invention.

[0111] Figure 2 This is a schematic diagram of the dynamic governance collaboration and closed-loop update mechanism of the present invention.

[0112] Figure 3 This is a schematic diagram of the multi-objective particle swarm optimization algorithm of the present invention. Detailed Implementation

[0113] In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, the specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0114] Combination Figure 1-3 The present invention provides the following embodiments:

[0115] A multi-objective collaborative optimization method for distribution network reconfiguration topology, power flow, and power quality includes the following steps:

[0116] S1. Collect the 24-hour load curves and photovoltaic output curves of the distribution network nodes, generate a representative time-series scenario set through K-means clustering, and calculate the weight of each scenario;

[0117] S2. Construct a multi-dimensional power quality index system based on the scenario weights. The multi-dimensional power quality index system includes: steady-state voltage deviation index, harmonic distortion index, voltage sag sensitivity index, and three-phase imbalance and flicker comprehensive index.

[0118] S3. Construct a coupling sensitivity matrix of switch action, power flow change and power quality index response. After normalizing the matrix, calculate the overall switch priority.

[0119] S4. Using topological decision variables and dynamic governance synergistic variables as decision variables, and multi-scenario weighted network loss, power outage time, and power quality indicators as objective functions, and incorporating constraints such as radiality, power flow, and equipment capacity, a multi-objective collaborative optimization model is constructed.

[0120] S5. A sensitivity-guided multi-objective particle swarm optimization algorithm is adopted, which relies on the switch comprehensive priority to achieve accurate search, generate the optimal solution for the multi-objective collaborative optimization model, and obtain the reconstruction scheme.

[0121] S6. Link the reconfiguration scheme with dynamic governance devices such as SVG / APF / OLTC to achieve coordinated harmonic governance, voltage regulation, and risk prediction, and form a closed loop through effect verification, anomaly handling, and feedback updates.

[0122] Furthermore, the load curve in S1 includes the active power load curve. and reactive load curve The sampling interval was 15 minutes, with a total of 96 time points;

[0123] Photovoltaic power output curve The output curve of a photovoltaic power station over 24 hours;

[0124] The specific steps for generating a time-series scene set and calculating the weight of each scene are as follows:

[0125] Determine the number of scenarios using the elbow rule. Construct normalized clustering feature vectors , ,in, The total number of nodes. This represents the maximum active load of node 1. To maximize the output of photovoltaic power;

[0126] The clustering objective is to minimize the deviation between the feature vectors within the scene and the scene center vector. The formula for the clustering objective function is: , For the scene The set of original time points included. For the scene The central feature vector of the data ultimately aggregates 96 original time points into 24 typical scenarios, with each scenario corresponding to a similar running state.

[0127] Finally, the scene weights are calculated. The weight calculation formula is: ,in For the scene The number of original time points included, and satisfying the following conditions. .

[0128] A 15-minute sampling interval can fully capture the daily peak and valley load and photovoltaic output fluctuations; K-means clustering minimizes the bias within the scene, retaining dynamic characteristics while reducing dimensionality, and S=24 balances scene representativeness and computational efficiency; scene weights directly reflect the probability of the operating condition, enabling subsequent optimization schemes to be adapted to high-proportion operating scenarios first, solving the problem of poor adaptability of traditional static modeling.

[0129] Furthermore, in S2, the steady-state voltage deviation index, harmonic distortion index, voltage sag sensitivity index, and three-phase imbalance and flicker comprehensive index are all calculated with weighted values ​​under multiple scenarios:

[0130] Steady-state voltage deviation index Defined as the average degree to which the voltage of all nodes in the network deviates from the rated value, reflecting voltage stability; its calculation formula is: ,in The system's rated voltage. For the scene Next node voltage amplitude, The total number of nodes;

[0131] Harmonic distortion index Defined as assessing voltage distortion caused by harmonics across the entire frequency band from the 3rd to the 50th order, reflecting the degree of harmonic pollution; its calculation formula is: The total harmonic distortion rate In the formula For nodes In the scene of Second harmonic voltage amplitude This represents the amplitude of the fundamental voltage.

[0132] Voltage sag sensitivity index This is designed for sensitive loads such as precision instruments and data centers, quantifying their sensitivity to voltage sags; the calculation formula is as follows: In the formula, For the set of sensitive load nodes, Its quantity;

[0133] Voltage sag sensitivity By descent depth With duration Graded quantization, voltage sag sensitivity The formula for calculation is:

[0134] ;

[0135] Three-phase imbalance and flicker comprehensive index Its positioning is to integrate three-phase imbalance and voltage flicker, and comprehensively evaluate the three-phase symmetry and voltage stability of the distribution network; its calculation formula is: ;in , ; , , , These represent positive-sequence, negative-sequence, and zero-sequence currents. This is a short-time flicker value.

[0136] Steady-state voltage deviation index The global average degree of voltage deviation from the rated value directly reflects voltage stability; harmonic distortion index For nonlinear loads and inverter-type distributed generators (DGs) exhibiting 3rd to 50th harmonic pollution, the degree of distortion is accurately assessed using THD (Total Harmonic Distortion) indicators; voltage sag sensitivity index. Classified by transient sag depth and duration, suitable for the transient tolerance requirements of sensitive loads such as precision instruments; comprehensive index of three-phase imbalance and flicker. It integrates three-phase imbalance and flicker, while balancing three-phase symmetry and voltage stability; multi-scenario weighted fusion. This avoids the limitations of single-moment indicators, enables comprehensive evaluation of power quality throughout the entire time period, and makes up for the shortcomings of traditional indicators in terms of coverage.

[0137] Furthermore, the specific steps for constructing the coupling sensitivity matrix in S3 are as follows:

[0138] Constructing coupling sensitivity It is defined as quantifying the marginal impact of a single switch state change on a certain PQ index, realizing coupled modeling of switch action, power flow change, and PQ response; its formula is:

[0139] In the formula For switch The state change, Changes in active power of the line caused by switching action;

[0140] For each switch Calculate separately and The three-phase power flow at that time is obtained from the line In the scene Down, switch Active power amplitude when in disconnected state ,line In the scene Down, switch Active power amplitude when in closed state ,node In the scene Down, switch steady-state fundamental voltage amplitude when in the off state ,node In the scene Down, switch steady-state fundamental voltage amplitude when in closed state Solve for the change in the PQ index before and after the switch action. With line power change ,

[0141] The calculation formula is: , ;

[0142] The final construction dimension is Sensitivity matrix , , This represents the total number of candidate switches.

[0143] Sensitivity matrix in S3 The normalization formula is:

[0144]

[0145] In the formula: For switch For the first Normalized coupling sensitivity of PQ-like indices switch For the first The original coupling sensitivity of the PQ-like index, This represents the minimum original coupling sensitivity of all candidate switches. Maximum original coupling sensitivity of all candidate switches;

[0146] Before calculating the overall priority of the switches, the PQ index weights need to be calculated. First, the sensitivity matrix is ​​normalized. The decision matrix is ​​defined by behavioral switches and columns representing PQ indicators; then, according to... Calculate the proportion of indicators Then calculate the entropy value. The formula for calculating entropy is: , hour Finally, calculate the weights. , ,satisfy ;

[0147] Switch overall priority The calculation formula is: , .

[0148] By quantifying chain associations, this approach overcomes the limitation of traditional reconstruction, which only indirectly improves PQ (Potential Quality). It ultimately achieves precise quantification of the optimization value of switches, providing a basis for targeted searches in subsequent algorithms and resolving the problem of blind searches. Normalization eliminates the dimensional differences between different PQ indicators, making sensitivity comparable horizontally. The entropy weight method adaptively calculates indicator weights based on the sensitivity matrix, avoiding subjective manual setting. The comprehensive priority of switches integrates the improvement value of multiple PQ indicators, and after sorting, it can quickly locate the key switches that best improve PQ, guiding the algorithm to prioritize high-priority switches and improving search efficiency and solution quality.

[0149] Furthermore, the decision variables described in S4 include topological decision variables and dynamic governance synergy variables, specifically defined as follows:

[0150] Topology decision variables: set as switch state vectors ,in This represents the total number of candidate switches for the distribution network. , Indicates the line When the switch is closed, Indicates the line The switch is open, and the initial value of the vector is taken from the current switch state of the distribution network real-time topology database;

[0151] Dynamic governance coordination variables include SVG reactive power output adjustment and OLTC gear adjustment, specifically:

[0152] SVG reactive power adjustment: set as a vector ,in This represents the total number of distribution network nodes. The total number of representative time-series scenes generated in step S1. Represents a node In the scene The SVG reactive power output adjustment value is limited to a range of 100%. kVar, initialized to 0;

[0153] OLTC gear adjustment amount: set as vector ,in Representing a scene The tap change of the on-load tap-changing transformer, the tap adjustment range is: Each pressure adjustment range is ±0.5%, and the initial value is set to the current operating level.

[0154] The objective function is to minimize the multi-scene weighted vector. The scene weights of each target component are calculated based on S1. The weighted average is as follows:

[0155] Economic objectives The formula for calculating the weighted network loss across multiple scenarios is as follows:

[0156]

[0157] In the formula, For the collection of distribution network lines, , The lines are respectively In the scene The active power and reactive power are below. For the line In the scene Average voltage under, For the line The resistance, For the scene The duration;

[0158] Reliability targets : Weighted outage time for multiple scenarios, using the system average outage duration. The characterization and calculation formula is as follows:

[0159]

[0160] In the formula, For nodes In the scene The failure rate is as follows For nodes In the scene The following fault repair time, For nodes The number of users;

[0161] Power quality objectives include three components, which directly utilize the multi-dimensional power quality index system constructed in step S2, specifically: , , ,in The steady-state voltage deviation index This is a harmonic distortion index. This is an indicator of voltage sag sensitivity. It is a comprehensive index of three-phase imbalance and flicker.

[0162] The constraints of the multi-objective collaborative optimization model in step S4 include radial constraint, power flow constraint, harmonic constraint, equipment capacity constraint, and switching action constraint. The specific settings of each constraint are as follows:

[0163] The radial constraint is used to ensure that the distribution network topology is free of loops and islands, thus ensuring power supply safety. The specific verification method is as follows:

[0164] Loop-free verification: Constructing the distribution network topology correlation matrix The matrix dimension is ,in This represents the total number of distribution network nodes. The total number of candidate switches is given. The matrix element values ​​are determined by the following rule: if a node is connected to the corresponding line switch, the value is 1; otherwise, it is 0. When the following conditions are met... When it is determined that there are no loops in the distribution network topology, the formula is as follows: This represents the calculation of the matrix determinant;

[0165] Islanding absence verification: A depth-first search algorithm is used to traverse all nodes starting from the power distribution node. If all nodes can be traversed, it is determined that there are no islands; otherwise, there are islanding regions. For topologies that do not meet the radial constraint, the repair is achieved by adjusting the switch status.

[0166] The power flow constraint is used to ensure node power balance, and the specific formula is as follows:

[0167] In the formula: , They are nodes The set of incoming lines and the set of outgoing lines; , They are nodes In the scene The active and reactive power balance values ​​below; , They are nodes In the scene The active and reactive power of the load; , They are nodes In the scene The active and reactive power outputs of the distributed power sources are taken from the scenario data in step S1. For nodes In the scene The reactive power output adjustment of the SVG is the dynamic governance coordination variable in step S4. , The lines are respectively In the scene The transmitted active and reactive power;

[0168] , The lines are respectively In the scene The active and reactive power losses are calculated using the following formula: , ,in For the line In the scene Average voltage under, , The lines are respectively The resistance and reactance are taken from the power distribution network equipment parameter database;

[0169] The harmonic constraint is used to control harmonic pollution of the 3rd to 50th orders. It is constructed based on harmonic power flow characteristics, and the specific formula is as follows: ;

[0170] In the formula: The harmonic order is 3-50; , They are nodes ,node In the scene Below Second harmonic voltage phasor;

[0171] For the line In the scene Below The subharmonic current phasor includes nonlinear load injection current;

[0172] For the line of The formula for calculating the second harmonic impedance is as follows: ,in The imaginary unit;

[0173] The equipment capacity constraints are used to prevent equipment from operating under overload conditions. These include line current constraints, transformer load rate constraints, and SVG / APF capacity constraints, specifically:

[0174] Line current constraints: ,in For the line In the scene The actual operating current under these conditions For the line The rated current is taken from the power distribution network equipment parameter database;

[0175] Transformer load factor constraint: ,in For transformers in the scene The actual load capacity under the current conditions The rated capacity of the transformer is taken from the distribution network equipment parameter database;

[0176] SVG / APF capacity constraints: ,in For nodes The rated capacity of the SVG is taken from the power distribution network equipment parameter library;

[0177] The switch action constraint is used to avoid equipment damage caused by frequent switch operation. The constraint rule is: count the number of actions of all candidate switches in 24 hours, and the total number of actions per day is ≤3.

[0178] All constraints need to be independently verified in each representative time series scenario generated in step S1. Decision variable combinations that do not meet the constraints are determined to be infeasible solutions and need to be returned for adjustment.

[0179] Dual decision variables enable synergy between slow topology adjustment and fast device control, overcoming the limitations of single control methods; multiple objective functions cover economy, reliability, and power quality, and Pareto optimization is used instead of fixed weighting to adaptively balance the inherent contradictions between objectives; objective functions are weighted based on scenario weights to ensure that the optimization scheme adapts to dynamic operating conditions throughout the entire time period and avoids local optima.

[0180] Radial constraints, through topological correlation matrix determinant and depth-first search, ensure no loops and no islands, which is the foundation of power supply safety; switch action constraints limit the number of actions to no more than 3 per day to protect switchgear; all constraints are independently verified in each scenario to ensure that the optimization scheme is adapted to dynamic operating conditions and has no safety hazards.

[0181] Furthermore, the specific process of S5 is as follows:

[0182] Algorithm initialization:

[0183] Particle generation: Set population size Randomly generated A switch state vector The topology is repaired by disconnecting the lowest priority switch in the loop and closing the highest priority tie switch in the island.

[0184] Parameter initialization: Initial particle velocity Individual optimal solution ; Calculate the objective function value for each particle After non-dominated sorting by NSGA-III, the particle with the largest crowding distance in the first layer is selected as the global optimal solution. ;

[0185] Iterative optimization:

[0186] Set the maximum number of iterations acceleration coefficient , Sensitivity weight speed threshold ;

[0187] Sensitivity-guided speed update: Based on switch priority directional search, the formula is:

[0188]

[0189] Where: linearly decreasing inertia weight , It is a random number. This is the switching priority vector for S3; it is clipped when the speed exceeds the threshold. ;

[0190] Location updates and fixes: via Implement binary position updates and synchronously perform topology repair;

[0191] Cutting-edge update: Computing And verify the constraints, if Dominate Then update Update after reordering Maintaining Pareto frontier diversity;

[0192] Convergence criterion: Rate of change in congestion at the Pareto front for 10 consecutive generations Or the number of iterations reaches Output the Pareto optimal solution set ;

[0193] Optimal solution selection:

[0194] Entropy weight calculation: for the solution set Target matrix Normalization Calculate the entropy value ( Entropy weight ; To solve for the quantity;

[0195] Proximity calculation: Determining the ideal solution Negative ideal solution Calculate distance , Proximity ;

[0196] Optimal solution output: Selecting proximity The largest solution is used as the final reconstruction scheme. Synchronous output , .

[0197] During the initialization phase, loop breaking and island repair ensure the topological feasibility of particles. This embodiment can also guide particles to search towards high-value switches, improving convergence speed. NSGA-III non-dominated sorting and crowding calculation maintain the diversity of the Pareto front, avoiding premature convergence. The entropy-weighted TOPSIS method selects the comprehensive optimal solution from the solution set, balancing the needs of multiple objectives.

[0198] The harmonic mitigation coordination in S6 specifically refers to: when the scenario... Next node Total Harmonic Distortion This triggers APF compensation, the formula is:

[0199]

[0200] Where: harmonic order 3-50 times, hour Higher harmonics ; For nodes of Subharmonic injection current; For nodes Fundamental voltage amplitude;

[0201] The voltage regulation coordination specifically refers to:

[0202] OLTC range optimization: Aiming for minimum voltage deviation, the formula is as follows: , This is the system's rated voltage;

[0203] SVG reactive power adjustment: calculated based on voltage deviation, the formula is as follows. , For adjustment coefficients, The node voltage after reconstruction;

[0204] The risk prediction collaboration specifically involves: predicting the probability of PQ exceeding the standard using an LSTM model. , The risk level is low; maintain the parameters; 10% ≤ For medium risk, reserve SVG margin. ; High risk, closed A high-sensitivity switch, and .

[0205] Furthermore, the exception handling mechanism in S6 is as follows:

[0206] Troubleshooting Equipment Failures:

[0207] SVG / APF fault: Compensation for capacity expansion of adjacent node devices, the formula is as follows: In the formula Adjacent nodes These are the rated capacities of the adjacent and faulty nodes, respectively;

[0208] OLTC fault: SVG emergency voltage regulation, formula is as follows , , For rated voltage, The node voltage after reconstruction;

[0209] Topology anomaly handling:

[0210] Switch fails to operate: Select priority Replace with a spare switch. The original switch priority;

[0211] Sudden Islanding: Closure Sensitivity The interconnection switch synchronously cuts off non-critical loads to restore power supply;

[0212] The feedback update closed loop is specifically as follows: Update trigger: Regular update at 2:00 AM daily; Emergency update when PQ index exceeds the standard for 3 consecutive points or photovoltaic fluctuation > 20%;

[0213] Update content:

[0214] Scenario Update: Re-execute S1 clustering with new data; adjust the number of scenarios if the clustering error is > 0.1. ;

[0215] Sensitivity Update: Recalculate the S3 coupling sensitivity matrix and remove standard deviation. outliers;

[0216] Algorithm parameter update: When the number of scenes increases, S5's and Simultaneous increase of 20%;

[0217] Closed-loop iteration: The updated data and parameters are re-entered into S4-S6 to generate a new solution to replace the original solution.

[0218] The LSTM model predicts the probability of PQ exceeding the limit. For low-risk cases, parameters are maintained; for medium-risk cases, SVG margin is reserved; and for high-risk cases, the topology and devices are adjusted in advance. The hierarchical strategy balances response speed and operating cost.

[0219] In the event of a device failure, adjacent nodes' SVG / APF capacity expansion compensation and SVG emergency voltage regulation ensure power supply continuity; in the event of topology anomalies, backup switches replace refusing switches, and high-sensitivity tie switches close to eliminate islanding and prevent power outages; feedback updates dynamically adjust the scenario, sensitivity matrix, and algorithm parameters through daily routine updates and event-triggered updates, so that the optimization scheme continuously adapts to changes in the distribution network's operating status and maintains optimal performance in the long term.

[0220] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A multi-objective collaborative optimization method for distribution network reconfiguration topology, power flow, and power quality, characterized in that, Includes the following steps: S1. Collect the 24-hour load curves and photovoltaic output curves of the distribution network nodes, generate a representative time-series scenario set through K-means clustering, and calculate the weight of each scenario; S2. Construct a multi-dimensional power quality index system based on the scenario weights. The multi-dimensional power quality index system includes: steady-state voltage deviation index, harmonic distortion index, voltage sag sensitivity index, and three-phase imbalance and flicker comprehensive index. S3. Construct a coupling sensitivity matrix of switch action, power flow change and power quality index response. After normalizing the matrix, calculate the overall switch priority. S4. Using topological decision variables and dynamic governance synergistic variables as decision variables, and multi-scenario weighted network loss, power outage time, and power quality indicators as objective functions, and incorporating constraints such as radiality, power flow, and equipment capacity, a multi-objective collaborative optimization model is constructed. S5. A sensitivity-guided multi-objective particle swarm optimization algorithm is adopted, which relies on the switching comprehensive priority to achieve accurate search, generate the optimal solution for the multi-objective collaborative optimization model, and obtain the reconstruction scheme.

2. The multi-objective collaborative optimization method for distribution network reconfiguration topology, power flow, and power quality according to claim 1, characterized in that, The load curve in S1 includes the active power load curve. and reactive load curve The sampling interval was 15 minutes, with a total of 96 time points; Photovoltaic power output curve The output curve of a photovoltaic power station over 24 hours; The specific steps for generating a time-series scene set and calculating the weight of each scene are as follows: Determine the number of scenarios using the elbow rule. Construct normalized clustering feature vectors , ,in, The total number of nodes. This represents the maximum active load of node 1. To maximize the output of photovoltaic power; The clustering objective is to minimize the deviation between the feature vectors within the scene and the scene center vector. The formula for the clustering objective function is: , For the scene The set of original time points included. For the scene The central feature vector of the data ultimately aggregates 96 original time points into 24 typical scenarios, with each scenario corresponding to a similar running state. Finally, the scene weights are calculated. The weight calculation formula is: ,in For the scene The number of original time points included, and satisfying the following conditions. .

3. The multi-objective collaborative optimization method for distribution network reconfiguration topology, power flow, and power quality according to claim 1, characterized in that, In S2, the steady-state voltage deviation index, harmonic distortion index, voltage sag sensitivity index, and three-phase imbalance and flicker comprehensive index are all calculated with weighted values ​​under multiple scenarios: Steady-state voltage deviation index Defined as the average degree to which the voltage of all nodes in the network deviates from the rated value, reflecting voltage stability; its calculation formula is: ,in The system's rated voltage. For the scene Next node voltage amplitude, The total number of nodes; Harmonic distortion index Defined as assessing voltage distortion caused by harmonics across the entire frequency band from the 3rd to the 50th order, reflecting the degree of harmonic pollution; The formula is: The total harmonic distortion rate In the formula For nodes In the scene of Second harmonic voltage amplitude This represents the amplitude of the fundamental voltage. Voltage sag sensitivity index It is designed for sensitive loads such as precision instruments and data centers, quantifying their sensitivity to voltage dips; The formula is: In the formula, For the set of sensitive load nodes, Its quantity; Voltage sag sensitivity By descent depth With duration Graded quantization, voltage sag sensitivity The formula for calculation is: ； Three-phase imbalance and flicker comprehensive index Its positioning is to integrate three-phase imbalance and voltage flicker, and comprehensively evaluate the three-phase symmetry and voltage stability of the distribution network; its calculation formula is: ;in , ; , , , These represent positive-sequence, negative-sequence, and zero-sequence currents. This is a short-time flicker value.

4. The multi-objective collaborative optimization method for distribution network reconfiguration topology, power flow, and power quality according to claim 1, characterized in that, The specific steps for constructing the coupling sensitivity matrix in S3 are as follows: Constructing coupling sensitivity It is defined as quantifying the marginal impact of a single switch state change on a certain PQ index, and realizing the coupled modeling of switch action, power flow change and PQ response; The formula is: In the formula For switch The state change, Changes in active power of the line caused by switching action; For each switch Calculate separately and The three-phase power flow at that time is obtained from the line In the scene Down, switch Active power amplitude when in disconnected state ,line In the scene Down, switch Active power amplitude when in closed state ,node In the scene Down, switch steady-state fundamental voltage amplitude when in the off state ,node In the scene Down, switch steady-state fundamental voltage amplitude when in closed state Solve for the change in the PQ index before and after the switch action. With line power change , The calculation formula is: , ; The final construction dimension is Sensitivity matrix , , This represents the total number of candidate switches.

5. The multi-objective collaborative optimization method for distribution network reconfiguration topology, power flow, and power quality according to claim 4, characterized in that, Sensitivity matrix in S3 The normalization formula is: In the formula: For switch For the first Normalized coupling sensitivity of PQ-like indices switch For the first The original coupling sensitivity of the PQ-like index, This represents the minimum original coupling sensitivity of all candidate switches. Maximum original coupling sensitivity of all candidate switches; Before calculating the overall priority of the switches, the PQ index weights need to be calculated. First, the sensitivity matrix is ​​normalized. The decision matrix is ​​defined by behavioral switches and columns representing PQ indicators; then, according to... Calculate the proportion of indicators Then calculate the entropy value. The formula for calculating entropy is: , hour Finally, calculate the weights. , ,satisfy ; Switch overall priority The calculation formula is: , .

6. The multi-objective collaborative optimization method for distribution network reconfiguration topology, power flow, and power quality according to claim 1, characterized in that, The decision variables described in S4 include topological decision variables and dynamic governance synergy variables, specifically defined as follows: Topology decision variables: set as switch state vectors ,in This represents the total number of candidate switches for the distribution network. , Indicates the line When the switch is closed, Indicates the line The switch is open, and the initial value of the vector is taken from the current switch state of the distribution network real-time topology database; Dynamic governance coordination variables include SVG reactive power output adjustment and OLTC gear adjustment, specifically: SVG reactive power adjustment: set as a vector ,in This represents the total number of distribution network nodes. The total number of representative time-series scenes generated in step S1. Represents a node In the scene The SVG reactive power output adjustment value is set to 0 initially. OLTC gear adjustment amount: set as vector ,in Representing a scene The tap change of the on-load tap-changing transformer, the tap adjustment range is: The initial value is set to the current running gear; The objective function is to minimize the multi-scene weighted vector. The scene weights of each target component are calculated based on S1. The weighted average is as follows: Economic objectives The formula for calculating the weighted network loss across multiple scenarios is as follows: In the formula, For the collection of distribution network lines, , The lines are respectively In the scene The active power and reactive power are below. For the line In the scene Average voltage under, For the line The resistance, For the scene The duration; Reliability targets : Weighted outage time for multiple scenarios, using the system average outage duration. The characterization and calculation formula is as follows: In the formula, For nodes In the scene The failure rate is as follows For nodes In the scene The following fault repair time, For nodes The number of users; Power quality objectives include three components, which directly utilize the multi-dimensional power quality index system constructed in step S2, specifically: , , ,in The steady-state voltage deviation index This is a harmonic distortion index. This is an indicator of voltage sag sensitivity. It is a comprehensive index of three-phase imbalance and flicker.

7. A multi-objective collaborative optimization method for distribution network reconfiguration topology, power flow, and power quality according to claim 6, characterized in that, In step S4, the constraints of the multi-objective collaborative optimization model include radial constraint, power flow constraint, harmonic constraint, equipment capacity constraint, and switching action constraint. The specific settings of each constraint are as follows: The radial constraint is used to ensure that the distribution network topology is free of loops and islands, thus ensuring power supply safety. The specific verification method is as follows: Loop-free verification: Constructing the distribution network topology correlation matrix The matrix dimension is ,in This represents the total number of distribution network nodes. The total number of candidate switches is given. The matrix element values ​​are determined by the following rule: if a node is connected to the corresponding line switch, the value is 1; otherwise, it is 0. When the following conditions are met... When it is determined that there are no loops in the distribution network topology, the formula is as follows: This represents the calculation of the matrix determinant; Islanding absence verification: A depth-first search algorithm is used to traverse all nodes starting from the power distribution node. If all nodes can be traversed, it is determined that there are no islands; otherwise, there are islanding regions. For topologies that do not meet the radial constraint, the repair is achieved by adjusting the switch status. The power flow constraint is used to ensure node power balance, and the specific formula is as follows: In the formula: , They are nodes The set of incoming lines and the set of outgoing lines; , They are nodes In the scene The active and reactive power balance values ​​below; , They are nodes In the scene The active and reactive power of the load; , They are nodes In the scene The active and reactive power outputs of the distributed power sources are taken from the scenario data in step S1. For nodes In the scene The reactive power output adjustment of the SVG is the dynamic governance coordination variable in step S4. , The lines are respectively In the scene The transmitted active and reactive power; , The lines are respectively In the scene The active and reactive power losses are calculated using the following formula: , ,in For the line In the scene Average voltage under, , The lines are respectively The resistance and reactance are taken from the power distribution network equipment parameter database; The harmonic constraint is used to control harmonic pollution of the 3rd to 50th orders. It is constructed based on harmonic power flow characteristics, and the specific formula is as follows: ; In the formula: The harmonic order is 3-50; , They are nodes ,node In the scene Below Second harmonic voltage phasor; For the line In the scene Below The subharmonic current phasor includes nonlinear load injection current; For the line of The formula for calculating the second harmonic impedance is as follows: ,in The imaginary unit; The equipment capacity constraints are used to prevent equipment from operating under overload conditions. These include line current constraints, transformer load rate constraints, and SVG / APF capacity constraints, specifically: Line current constraints: ,in For the line In the scene The actual operating current under these conditions For the line The rated current is taken from the power distribution network equipment parameter database; Transformer load factor constraint: ,in For transformers in the scene The actual load capacity under the current conditions The rated capacity of the transformer is taken from the distribution network equipment parameter database; SVG / APF capacity constraints: ,in For nodes The rated capacity of the SVG is taken from the power distribution network equipment parameter library; The switch action constraint is used to avoid equipment damage caused by frequent switch operation. The constraint rule is: count the number of actions of all candidate switches in 24 hours, and the total number of actions per day is ≤3. All constraints need to be independently verified in each representative time series scenario generated in step S1. Decision variable combinations that do not meet the constraints are determined to be infeasible solutions and need to be returned for adjustment.

8. The multi-objective collaborative optimization method for distribution network reconfiguration topology, power flow, and power quality according to claim 1, characterized in that, The specific process for S5 is as follows: Algorithm initialization: Particle generation: Set population size Randomly generated A switch state vector The topology is repaired by disconnecting the lowest priority switch in the loop and closing the highest priority tie switch in the island. Parameter initialization: Initial particle velocity Individual optimal solution ; Calculate the objective function value for each particle After non-dominated sorting by NSGA-III, the particle with the largest crowding distance in the first layer is selected as the global optimal solution. ; Iterative optimization: Set the maximum number of iterations acceleration coefficient , Sensitivity weight speed threshold ; Sensitivity-guided speed update: Based on switch priority directional search, the formula is: Where: linearly decreasing inertia weight , It is a random number. This is the switching priority vector for S3; it is clipped when the speed exceeds the threshold. ; Location updates and fixes: via Implement binary position updates and synchronously perform topology repair; Cutting-edge update: Computing And verify the constraints, if Dominate Then update Update after reordering Maintaining Pareto frontier diversity; Convergence criterion: Rate of change in congestion at the Pareto front for 10 consecutive generations Or the number of iterations reaches Output the Pareto optimal solution set ; Optimal solution selection: Entropy weight calculation: for the solution set Target matrix Normalization Calculate the entropy value ( Entropy weight ; To solve for the quantity; Proximity calculation: Determining the ideal solution Negative ideal solution Calculate distance , Proximity ; Optimal solution output: Selecting proximity The largest solution is used as the final reconstruction scheme. Synchronous output , .

9. A multi-objective collaborative optimization method for distribution network reconfiguration topology, power flow, and power quality according to claim 1, characterized in that, It also includes step S6: S6. Link the reconfiguration scheme with dynamic governance devices such as SVG / APF / OLTC to achieve coordinated harmonic governance, voltage regulation and risk prediction, and form a closed loop through effect verification, anomaly handling and feedback updates. The harmonic mitigation collaboration specifically refers to: when the scenario... Next node Total Harmonic Distortion This triggers APF compensation, the formula is: Where: harmonic order 3-50 times, hour Higher harmonics ; For nodes of Subharmonic injection current; For nodes Fundamental voltage amplitude; The voltage regulation coordination specifically refers to: OLTC range optimization: Aiming for minimum voltage deviation, the formula is as follows: , This is the system's rated voltage; SVG reactive power adjustment: calculated based on voltage deviation, the formula is as follows. , For adjustment coefficients, The node voltage after reconstruction; The risk prediction collaboration specifically involves: predicting the probability of PQ exceeding the standard using an LSTM model. , The risk level is low; maintain the parameters; 10% ≤ For medium risk, reserve SVG margin. ; High risk, closed A high-sensitivity switch, and .

10. A multi-objective collaborative optimization method for distribution network reconfiguration topology, power flow, and power quality according to claim 9, characterized in that, The exception handling mechanism in S6 is as follows: Troubleshooting Equipment Failures: SVG / APF fault: Compensation for capacity expansion of adjacent node devices, the formula is as follows: In the formula Adjacent nodes These are the rated capacities of the adjacent and faulty nodes, respectively; OLTC fault: SVG emergency voltage regulation, formula is as follows , , For rated voltage, The node voltage after reconstruction; Topology anomaly handling: Switch fails to operate: Select priority Replace with a spare switch. The original switch priority; Sudden Islanding: Closure Sensitivity The interconnection switch synchronously cuts off non-critical loads to restore power supply; The feedback update closed loop is specifically as follows: Update trigger: Regular update at 2:00 AM daily; Emergency update when PQ index exceeds the standard for 3 consecutive points or photovoltaic fluctuation > 20%; Update content: Scenario Update: Re-execute S1 clustering with new data; adjust the number of scenarios if the clustering error is > 0.

1. ; Sensitivity Update: Recalculate the S3 coupling sensitivity matrix and remove standard deviation. outliers; Algorithm parameter update: When the number of scenes increases, S5's and Simultaneous increase of 20%; Closed-loop iteration: The updated data and parameters are re-entered into S4-S6 to generate a new solution to replace the original solution.

Images