Power distribution network dynamic partitioning method and system based on multi-objective ant colony optimization

Abstract

The invention discloses a power distribution network dynamic partitioning method and system based on multi-objective ant colony optimization, and relates to the technical field of power system optimization, and the method comprises the steps: carrying out the power grid topology parameter set synchronization and three-dimensional field intensity mapping through a dynamic state estimation algorithm and a finite element electromagnetic field simulation algorithm based on a dynamic data flow, and generating a virtual power grid mirror image; a multi-target ant colony optimization algorithm is embedded in the virtual power grid mirror image, new energy sudden drop and harmonic resonance disturbance are injected into the virtual power grid through Monte Carlo simulation, and a Pareto optimal solution set is generated; and based on the Pareto optimal solution set and in combination with real-time power grid multi-dimensional state data, dynamically adjusting pheromone parameters of a multi-target ant colony optimization algorithm through a depth deterministic strategy gradient algorithm, and generating optimization parameters. Through virtual mirror image generation and parameter dynamic adjustment, deep coupling of data-strategy-execution is realized, and the technical bottlenecks of dynamic data synchronization misalignment and optimization parameter stiffness in a traditional method are solved.

Description

Technical Field

[0001] This invention relates to the field of power system optimization technology, and in particular to a dynamic zoning method and system for distribution networks based on multi-objective ant colony optimization. Background Technology

[0002] With the increasing integration of high-proportion renewable energy sources and the growing penetration of power electronic equipment, the operating environment of distribution networks exhibits characteristics of strong randomness, multi-timescale coupling, and complex electromagnetic transient processes. Traditional static zoning methods are insufficient to meet the voltage-reactive power coordinated control requirements in dynamic scenarios. Currently, dynamic zoning technology for distribution networks mainly relies on a centralized data acquisition architecture, constructs grid connection relationships through topology identification algorithms, and generates zoning strategies using heuristic optimization algorithms.

[0003] At the level of dynamic data processing and optimization decision-making, existing methods have significant limitations. Traditional data synchronization mechanisms rely on a central server to centrally process heterogeneous data sources, which is limited by communication latency and computing resource bottlenecks, making it difficult to respond to millisecond-level changes in the power grid status in a timely manner. When the output of new energy sources drops sharply or harmonic resonance causes voltage fluctuations, the static parameter model cannot quickly adjust the optimization direction, resulting in a disconnect between control commands and disturbance suppression requirements, and even exacerbating the risk of system oscillation. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a dynamic zoning method for power distribution networks based on multi-objective ant colony optimization, which can solve the problem of control command lag caused by insufficient real-time synchronization accuracy of multi-source heterogeneous data.

[0006] To address the aforementioned technical problems, this invention provides the following technical solution: a dynamic zoning method for distribution networks based on multi-objective ant colony optimization, comprising: real-time acquisition of multi-dimensional state data of the power grid; synchronizing the multi-dimensional state data of the power grid to an edge cluster through a distributed data synchronization mechanism based on timestamp alignment, and generating a dynamic data stream by combining data cleaning; based on the dynamic data stream, synchronizing the power grid topology parameter set and mapping the three-dimensional field strength through a dynamic state estimation algorithm and a finite element electromagnetic field simulation algorithm to generate a virtual power grid image; embedding a multi-objective ant colony optimization algorithm into the virtual power grid image, and injecting new energy sudden drop and harmonic resonance disturbances into the virtual power grid through Monte Carlo simulation to generate a Pareto optimal solution set; based on the Pareto optimal solution set and combined with real-time multi-dimensional state data of the power grid, dynamically adjusting the pheromone parameters of the multi-objective ant colony optimization algorithm through a deep deterministic strategy gradient algorithm to generate optimized parameters; based on the optimized parameters, performing command parsing and dynamic strategy mapping through QV droop control and voltage regulation to generate reactive power compensation commands and topology switching commands.

[0007] As a preferred embodiment of the dynamic zoning method for power distribution networks based on multi-objective ant colony optimization described in this invention, the multi-dimensional state data of the power grid includes equipment operating parameters, electrical parameters, environmental disturbance data, and topology parameters.

[0008] As a preferred embodiment of the dynamic zoning method for distribution networks based on multi-objective ant colony optimization described in this invention, the step of synchronizing multi-dimensional state data of the power grid to the edge cluster through a distributed data synchronization mechanism based on timestamp alignment, and generating a dynamic data stream by combining data cleaning, includes the following specific steps:

[0009] Based on multi-dimensional power grid status data, a multi-source data stream is generated by using a clock synchronization mechanism and a distributed transmission method to perform time-series alignment and security encryption of cross-node data.

[0010] Based on multi-source data streams, cross-node timing deviations are eliminated through timestamp alignment, and data is cleaned using the isolated forest algorithm and standardization processing to generate dynamic data streams.

[0011] As a preferred embodiment of the dynamic zoning method for distribution networks based on multi-objective ant colony optimization described in this invention, the step of generating a virtual power grid mirror by synchronizing the power grid topology parameter set and mapping three-dimensional field strength based on dynamic data flow through a dynamic state estimation algorithm and a finite element electromagnetic field simulation algorithm is as follows.

[0012] Based on dynamic data streams, a real-time power grid topology parameter set is generated through a dynamic state estimation algorithm.

[0013] Based on the power grid topology parameter set, the three-dimensional spatial field strength distribution is inverted through the finite element electromagnetic field simulation algorithm to generate field strength gradient data;

[0014] Based on field strength gradient data combined with dynamic data streams, spatiotemporal data fusion and standardized encapsulation are performed through power grid multi-source data fusion specifications and wide-area time synchronization methods to generate a virtual power grid image.

[0015] As a preferred embodiment of the dynamic zoning method for distribution networks based on multi-objective ant colony optimization described in this invention, the steps of embedding the multi-objective ant colony optimization algorithm into a virtual power grid mirror and injecting new energy surges and harmonic resonance disturbances into the virtual power grid through Monte Carlo simulation to generate a Pareto optimal solution set are as follows.

[0016] A multi-objective ant colony optimization algorithm is embedded in a virtual power grid image to simulate multiple optimization objectives under power grid operation conditions and generate a set of candidate strategies.

[0017] Based on the candidate strategy set, a set of events for sudden drop in new energy and harmonic resonance disturbance is generated by combining random parameter perturbation with dynamic data stream through Monte Carlo simulation and perturbation scenario generation method. Then, the effective strategy subset is analyzed and screened by Gaussian kernel density estimation method.

[0018] The Pareto front screening method is used to perform a multi-objective comprehensive evaluation of the effective policy subset and generate the Pareto optimal solution set.

[0019] As a preferred embodiment of the dynamic zoning method for distribution networks based on multi-objective ant colony optimization described in this invention, the steps for generating optimization parameters by dynamically adjusting the pheromone parameters of the multi-objective ant colony optimization algorithm based on Pareto optimal solution sets and combined with real-time multi-dimensional power grid state data are as follows:

[0020] Based on the Pareto optimal solution set and real-time power grid state data, dynamic feature vectors of the power grid are extracted through Kalman filtering and principal component analysis.

[0021] Based on the dynamic feature vector of the power grid, the correlation between the power grid state and multi-objective optimization is analyzed by a deep deterministic strategy gradient algorithm to generate dynamic adjustment signals;

[0022] Based on the dynamically adjusted signal, the pheromone parameters are corrected through a multi-target ant colony optimization algorithm to generate optimized parameters.

[0023] As a preferred embodiment of the dynamic zoning method for distribution networks based on multi-objective ant colony optimization described in this invention, the steps of generating reactive power compensation commands and topology switching commands by parsing and dynamically mapping commands through QV droop control and voltage regulation based on optimization parameters are as follows:

[0024] Based on optimized parameters, the voltage regulation reference value and the reference reactive power compensation requirement are obtained through QV droop control analysis. The initial control command is formed by weighting the command and mapping the dynamic strategy through voltage-reactive power sensitivity analysis combined with the real-time power grid topology parameter set.

[0025] By combining the real-time power grid status, the initial control command is dynamically corrected through a dynamic state estimation algorithm to generate the final reactive power compensation command and topology switching command. The reactive power compensation command drives the reactive power compensation device to adjust the reactive power output of the region, and the topology switching command controls the circuit breaker to perform opening and closing operations, thereby realizing dynamic zoning of the distribution network.

[0026] Another objective of this invention is to provide a dynamic zoning system for power distribution networks based on multi-objective ant colony optimization.

[0027] As a preferred embodiment of the dynamic zoning system for distribution networks based on multi-objective ant colony optimization described in this invention, it includes: a data preprocessing module, a virtual image generation module, a disturbance optimization module, a parameter optimization module, and an instruction generation module;

[0028] The data processing module collects multi-dimensional status data of the power grid in real time, synchronizes the multi-dimensional status data of the power grid to the edge cluster through a distributed data synchronization mechanism based on timestamp alignment, and generates dynamic data streams by combining data cleaning.

[0029] The virtual image generation module, based on dynamic data flow, uses dynamic state estimation algorithm and finite element electromagnetic field simulation algorithm to synchronize the power grid topology parameter set and map three-dimensional field strength to generate a virtual power grid image.

[0030] The disturbance optimization module embeds a multi-objective ant colony optimization algorithm into the virtual power grid mirror and injects new energy sudden drop and harmonic resonance disturbances into the virtual power grid through Monte Carlo simulation to generate a Pareto optimal solution set.

[0031] The parameter optimization module, based on the Pareto optimal solution set and combined with real-time multi-dimensional power grid state data, dynamically adjusts the pheromone parameters of the multi-objective ant colony optimization algorithm through a deep deterministic strategy gradient algorithm to generate optimized parameters.

[0032] The instruction generation module, based on optimized parameters, performs instruction parsing and dynamic strategy mapping through QV droop control and voltage regulation to generate reactive power compensation instructions and topology switching instructions.

[0033] The present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of a dynamic zoning method for a power distribution network based on multi-objective ant colony optimization.

[0034] The present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of a dynamic zoning method for a power distribution network based on multi-objective ant colony optimization.

[0035] The beneficial effects of this invention are as follows: By generating virtual images and dynamically adjusting parameters, deep coupling of data, strategy, and execution is achieved, which solves the technical bottlenecks of dynamic data synchronization inaccuracy and rigid optimization parameters in traditional methods, and provides a high-precision, adaptive dynamic zoning control scheme for high-proportion new energy distribution networks. Attached Figure Description

[0036] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 This is a flowchart illustrating a dynamic zoning method for power distribution networks based on multi-objective ant colony optimization, provided as an embodiment of the present invention.

[0038] Figure 2 This is a schematic diagram of a dynamic zoning system for a power distribution network based on multi-objective ant colony optimization, provided as an embodiment of the present invention.

[0039] Figure 3 This is a data preprocessing flowchart of a dynamic zoning method for power distribution networks based on multi-target ant colony optimization, provided as an embodiment of the present invention.

[0040] Figure 4 This is a flowchart illustrating the virtual power grid mirror generation process of a power distribution network dynamic zoning method based on multi-objective ant colony optimization, as provided in one embodiment of the present invention. Detailed Implementation

[0041] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0042] Example 1, referring to Figures 1-4 This is the first embodiment of the present invention, which provides a dynamic zoning method for distribution networks based on multi-objective ant colony optimization, including:

[0043] S1: Real-time acquisition of multi-dimensional status data of the power grid, synchronization of multi-dimensional status data of the power grid to the edge cluster through a distributed data synchronization mechanism based on timestamp alignment, and generation of dynamic data stream by combining data cleaning.

[0044] The multi-dimensional status data of the power grid includes equipment operating parameters, electrical parameters, environmental disturbance data, and topology parameters;

[0045] It should be noted that the multi-dimensional status data acquisition of the power grid involves acquiring various types of parameters of the physical entities and operating environment of the power grid in real time through sensors and monitoring terminals deployed on power equipment, lines, and the environment. These parameters include equipment operating parameters, electrical parameters, environmental disturbance data, and topology parameters. Equipment operating parameters are collected in real time using temperature sensors, vibration sensors, and oil chromatography monitoring units installed on equipment such as transformers, circuit breakers, and reactive power compensation devices. These parameters include data characterizing the health status of the equipment, such as equipment body temperature, mechanical vibration amplitude, insulating oil gas content, and switch opening and closing status. Electrical parameters are obtained using smart meters, synchronous phasor measurement devices, fault recorders, etc. The monitoring terminal collects dynamic data reflecting the electrical characteristics of the power grid, such as voltage amplitude and phase at power grid nodes, effective value of line current, active / reactive power, power factor, and harmonic distortion rate. Environmental disturbance data is collected through devices such as meteorological sensors, light intensity meters, and noise monitors, including data on natural conditions such as ambient temperature, humidity, wind speed, and light intensity, as well as data on external disturbance events such as power output fluctuations, harmonic resonance, and load abrupt changes at renewable energy power plants. Topology parameters are based on real-time status signals from circuit breakers and disconnectors, combined with topology analysis methods for power grid dispatch automation networks, to dynamically update topology information such as the connection relationships of power grid nodes, regional division structure, and equipment commissioning / decommissioning status.

[0046] Based on multi-dimensional power grid status data, a multi-source data stream is generated by using a clock synchronization mechanism and a distributed transmission method to perform time-series alignment and security encryption of cross-node data.

[0047] It should be noted that the clock synchronization mechanism adopts a dual-mode clock synchronization method of BeiDou satellite navigation and global positioning to establish a master-slave clock architecture, assigning a unified timestamp to all node data and eliminating time sequence deviations from data sources in different geographical locations. The distributed transmission method relies on the fiber optic longitudinal differential protection channel and the fifth-generation mobile communication technology wireless private network in the power dedicated communication network to transmit equipment operating parameters, electrical parameters, environmental disturbance data, and topology parameters across nodes. During transmission, the data is encrypted end-to-end through the transport layer security protocol. After time sequence alignment and security encryption, the equipment operating parameters, electrical parameters, environmental disturbance data, and topology parameters are converted into a format according to the standardized data classes and attributes defined by the power industry public information standardization structure, generating a dynamic data stream containing multi-dimensional status information.

[0048] Based on multi-source data streams, cross-node timing deviations are eliminated through timestamp alignment, and data is cleaned using the isolated forest algorithm and standardization processing to generate dynamic data streams.

[0049] It should be noted that, based on multi-source data streams, timestamp alignment adopts a dual-mode clock synchronization method of BeiDou satellite navigation and global positioning under a master-slave clock architecture. This assigns a unified timestamp to the equipment operating parameters, electrical parameters, environmental disturbance data, and topology parameters in the multi-source data stream, eliminating timing deviations caused by geographical differences in the acquisition devices at different nodes. During data cleaning, anomaly data points that significantly deviate from the normal range in the multi-source data stream are identified and removed based on the isolated forest algorithm. Simultaneously, the temperature and vibration amplitude in the equipment operating parameters, the voltage and current in the electrical parameters, and the temperature and humidity in the environmental disturbance data are normalized using the Z-score standardization method, unifying the units and numerical ranges. After timing alignment, anomaly detection, and standardization, the multi-source data stream is dynamically smoothed according to the minute-level time window division and moving average method, generating a spatiotemporally continuous and noise-free dynamic data stream.

[0050] In one feasible embodiment, data cleaning can be achieved through anomaly detection based on DBSCAN clustering. Specifically, based on the dual-mode clock synchronization method of BeiDou satellite navigation and Global Positioning, a unified timestamp is assigned to multi-source data streams (equipment operating parameters, electrical parameters, environmental disturbance data) to eliminate cross-node timing deviations. Input data includes parameters such as temperature, vibration amplitude, voltage, current, and humidity within a minute-level window. The DBSCAN (density-based spatial clustering) algorithm is used to identify anomalies. Power grid parameters (such as voltage amplitude and temperature) are used as feature vectors, and a neighborhood radius and a minimum point threshold are set. Core point density is defined based on historical power grid normal operation data (such as voltage fluctuation range and temperature threshold). Data points are classified as core points, boundary points, or noise points; noise points (such as isolated temperature peaks or voltage spikes) are considered anomalous data and are directly removed. For example, in new energy output fluctuation events, the algorithm can identify current anomalies caused by harmonic resonance.

[0051] The data after anomaly detection is standardized using Min-Max. Equipment operating parameters (temperature, vibration amplitude), electrical parameters (voltage, current), and environmental parameters (humidity) are mapped to the [0,1] interval. Based on the historical minimum and maximum values ​​of each parameter (e.g., voltage standard range 0.9-1.1 pu), a scaling value is calculated for each data point to unify the dimensions. For example, temperature data is linearly transformed according to the normal range of the transformer (e.g., 20-100℃).

[0052] Dividing the data into minute-level time windows, the standardized data is smoothed using a moving average method to eliminate residual noise and generate a spatiotemporally continuous and noise-free dynamic data stream for subsequent steps.

[0053] In another feasible embodiment, data cleaning can also be achieved through anomaly detection based on LOF (Local Outlier Factor). Specifically, data timestamps are aligned using a dual-mode clock synchronization method combining BeiDou satellite navigation and Global Positioning, and multi-source data streams are input. The LOF algorithm is then used to identify outliers. A feature space is constructed using electrical parameters (such as power factor and harmonic distortion rate) and environmental parameters (such as wind speed and light intensity). The local density deviation of each data point is calculated: based on the k-nearest neighbor distance, a density benchmark for normal data points is defined (such as harmonic content when voltage is stable). Points with LOF values ​​greater than a threshold (such as harmonic exceedance thresholds in historical events) are considered outliers (such as power anomalies caused by load surges) and are removed. This algorithm is particularly suitable for identifying outliers in locally dense areas, such as local current anomalies caused by power output fluctuations in renewable energy power plants.

[0054] Data after anomaly detection is standardized using Robust Scaler. Scaling is performed based on the median and interquartile range (IQR): the median and IQR are calculated for each parameter (e.g., temperature, voltage) (e.g., ignoring extreme values), transforming the data to a uniform range (e.g., [-1, 1]). For example, ambient humidity data is benchmarked against the historical median to reduce the residual impact of anomalies. A smooth dynamic data stream is generated using a moving average method.

[0055] S2. Based on dynamic data flow, a virtual power grid image is generated by synchronizing the power grid topology parameter set and mapping the three-dimensional field strength through dynamic state estimation algorithm and finite element electromagnetic field simulation algorithm.

[0056] Based on dynamic data streams, a real-time power grid topology parameter set is generated through a dynamic state estimation algorithm.

[0057] It should be noted that, based on dynamic data streams, the dynamic state estimation algorithm adopts a weighted least squares state estimation method. It takes the node voltage amplitude, line active power, line reactive power, and transformer tap position in the dynamic data stream as measurement inputs, and establishes measurement equations by combining the grid topology connection relationship and branch parameters. By iteratively solving the objective function of minimizing the residual of the measurement equations, the estimated values ​​of node voltage phase angle and amplitude are updated. Based on the updated node voltage estimates and the admittance matrix parameters in the measurement equations, the power flow distribution of the lines and the equipment commissioning and decommissioning status are inferred to generate a real-time grid topology parameter set that includes node connection status, branch admittance parameters, and transformer turns ratio parameters.

[0058] In a feasible embodiment, dynamic state estimation can be achieved through extended Kalman filtering. Specifically, based on a dynamic data stream, input measurement data such as node voltage amplitude, line active power, and reactive power are combined with real-time grid topology parameters (e.g., node admittance matrix, branch parameters). A grid state-space model is constructed: the state variables are node voltage phase angle and amplitude, the measurement variables are power and voltage, and the state transition equation considers topology changes (e.g., circuit breaker opening and closing) and environmental disturbances (e.g., temperature effects). Extended Kalman filtering is used for state estimation. Based on the state estimate and state transition equation from the previous time step (linearizing the nonlinear parts, such as transformer turns ratio), the current state (e.g., voltage amplitude) is predicted. Using new measurement data (e.g., PMU synchronization phasors), the Kalman gain is calculated, and the state estimate is updated by minimizing the residual (the deviation between the measured and predicted values). For example, in the event of a sudden drop in renewable energy, the algorithm dynamically adjusts the prediction model to adapt to power fluctuations.

[0059] Based on the updated state estimates, the power flow distribution and equipment commissioning / discharging status are inferred (e.g., current distribution is calculated using the admittance matrix), generating a real-time power grid topology parameter set, including node connection status, branch admittance, and transformer turns ratio parameters. The estimation error is monitored using the covariance matrix, and when the error exceeds a threshold (e.g., voltage deviation limit), re-initialization or data recovery is triggered to ensure the accuracy of the parameter set.

[0060] In another feasible embodiment, dynamic state estimation can also be achieved through unscented Kalman filtering. Specifically, a nonlinear state-space model is constructed by inputting a dynamic data stream and topology parameters. The state variables include voltage phase angle and amplitude, and the measurement variables are power data. The model integrates the effects of electromagnetic disturbances (such as voltage distortion caused by harmonics). Unscented Kalman filtering is used for state estimation. Sigma points (representing the state distribution) are selected through unscented transformation, and these points are propagated nonlinearly (e.g., simulating voltage fluctuations under harmonic resonance) to predict the state mean. Based on actual measurement data, the Kalman gain is calculated to update the state estimate. For example, in harmonic events, the algorithm directly handles non-Gaussian noise, avoiding linearization errors.

[0061] By using updated state estimates, power flow and equipment states are inferred to generate a set of topology parameters. The focus is on handling nonlinear effects (such as the impact of insulating oil gas content on transformer parameters), and adaptive noise covariance adjustment is used to address sudden grid changes (such as load step changes) to ensure the stability of the parameter set under disturbances.

[0062] Based on the power grid topology parameter set, the three-dimensional spatial field strength distribution is inverted through the finite element electromagnetic field simulation algorithm to generate field strength gradient data;

[0063] It should be noted that, based on the power grid topology parameter set, the finite element electromagnetic field simulation algorithm constructs a three-dimensional spatial electromagnetic field control equation based on Maxwell's equations. The algorithm uses the node coordinates, branch spatial layout, transformer turns ratio parameters, and conductor material properties from the power grid topology parameter set as input conditions. Conductor and insulating medium regions are divided in a three-dimensional spatial discretized grid. The electric and magnetic field intensity distributions in the discretized grid are iteratively solved using the Galerkin weighted residual method, calculating the electric field intensity amplitude and magnetic field intensity direction at each grid node. Based on the electric and magnetic field intensity distribution results, the rate of change of field intensity between adjacent grid nodes is calculated using a spatial gradient operator, generating field intensity gradient data characterizing the trend of electromagnetic field intensity changes in three-dimensional space.

[0064] Based on field strength gradient data combined with dynamic data streams, spatiotemporal data fusion and standardized encapsulation are performed through power grid multi-source data fusion specifications and wide-area time synchronization methods to generate a virtual power grid image.

[0065] It should be noted that, based on field strength gradient data combined with dynamic data streams, the power grid multi-source data fusion specification, according to the standardized data classes and attributes defined by the power industry's public information standardization structure, semantically aligns and unifies the three-dimensional spatial electromagnetic field intensity variation trend in the field strength gradient data with the equipment operating parameters, electrical parameters, environmental disturbance data, and topology parameters in the dynamic data stream. The wide-area time synchronization method adopts a dual-mode clock synchronization method using BeiDou satellite navigation and Global Positioning to assign a unified timestamp to the field strength gradient data and the equipment operating parameters, electrical parameters, environmental disturbance data, and topology parameters in the dynamic data stream, eliminating cross-source data timing deviations. During the spatiotemporal data fusion process, a sliding time window mechanism is used to perform spatial interpolation and time series matching on the timestamp-aligned field strength gradient data and the dynamic data stream, associating the field strength distribution and power grid operating status under the same time segment. Standardized encapsulation, based on the data class coding rules defined by the power industry's public information standardization structure, encapsulates the fused field strength gradient data, equipment operating parameters, electrical parameters, environmental disturbance data, and topology parameters into a virtual power grid image with a unified data interface.

[0066] S3. Embed a multi-objective ant colony optimization algorithm in the virtual power grid mirror, and inject new energy sudden drop and harmonic resonance disturbance into the virtual power grid through Monte Carlo simulation to generate Pareto optimal solution set.

[0067] A multi-objective ant colony optimization algorithm is embedded in a virtual power grid image to simulate multiple optimization objectives under power grid operation conditions and generate a set of candidate strategies.

[0068] It should be noted that, based on the virtual power grid mirror, the multi-objective ant colony optimization algorithm uses the three-dimensional spatial field strength distribution in the field strength gradient data layer, the node voltage and line power flow in the electrical parameter layer, and the harmonic spectrum characteristics in the environmental disturbance data layer as input parameters. Voltage stability, harmonic distortion rate, and network loss minimization are defined as optimization objectives. The pheromone concentration matrix and heuristic factor matrix of the multi-objective ant colony optimization algorithm are initialized. Based on the voltage deviation range and harmonic limits in the national power industry standards, real-time topology parameters, and ambient temperature data, combined with Monte Carlo simulation and finite element electromagnetic field simulation results, voltage limits, harmonic distortion rate thresholds, and line current carrying capacity constraints are obtained. The harmonic distortion rate threshold is dynamically corrected and set through harmonic spectrum analysis and resonance risk identification in the virtual power grid mirror. Voltage limits, harmonic distortion rate thresholds, and line current carrying capacity constraints are set as path search constraints. During the iteration process, individual ants select the switching capacity of reactive power compensation equipment and the combination of topology switching strategies as feasible paths based on pheromone concentration and heuristic factor probability. The pheromone concentration is updated by screening the non-dominated solution set through Pareto front sorting. Based on the equipment monitoring parameters (such as voltage, current, and harmonics), topology status, and environmental disturbance records in the historical operation of the power grid, and the convergence analysis, the path search of the pre-set multi-objective ant colony optimization algorithm with the number of optimization units and the number of iterations is generated to generate a candidate strategy set that satisfies multiple optimization objectives and meets the constraints. This set includes reactive power compensation instructions, topology switching instructions, and corresponding quantitative indicators such as voltage stability improvement rate, harmonic suppression rate, and network loss reduction rate.

[0069] Based on the candidate strategy set, a set of events for sudden drop in new energy and harmonic resonance disturbance is generated by combining random parameter perturbation with dynamic data stream through Monte Carlo simulation and perturbation scenario generation method. Then, the effective strategy subset is analyzed and screened by Gaussian kernel density estimation method.

[0070] It should be noted that, based on the candidate strategy set, the Monte Carlo simulation adopts the wind turbine output probability characteristics defined by the International Electrotechnical Commission (IEC) standard. Random parameter perturbations are applied to the reactive power compensation command and topology switching command in the candidate strategy set to generate a set of sudden drop events where the output of new energy sources is greater than or equal to the critical threshold jointly set by industry standards and grid tolerance. The perturbation scenario generation method is based on the harmonic test conditions defined by the Institute of Electrical and Electronics Engineers (IEEE) standard, combined with the harmonic content time-series data in the dynamic data stream, injecting a set of resonant perturbation events where the total harmonic distortion rate exceeds the grid's safe operation limit. The Latin hypercube sampling method is used to combine the new energy sudden drop event set and the harmonic resonant perturbation event set in a high-dimensional space to generate a composite event set containing multiple perturbation couplings. The Gaussian kernel density estimation method calculates the strategy failure probability density function based on the number of voltage exceedances, harmonic exceedances, and network loss increment datasets recorded during the perturbation test for each candidate strategy in the composite event set. Candidate strategies whose cumulative failure probability density function value exceeds the grid's safe operation limit are screened, generating an effective strategy subset containing reactive power compensation commands, topology switching commands, and corresponding quantitative indicators.

[0071] The Pareto front screening method is used to perform a multi-objective comprehensive evaluation of the effective policy subset and generate the Pareto optimal solution set.

[0072] It should be noted that, based on the effective strategy subset, the Pareto front screening uses the voltage stability improvement rate corresponding to the reactive power compensation command, the harmonic suppression rate corresponding to the topology switching command, and the network loss reduction rate as multi-objective evaluation functions. A non-dominated ranking method is used to compare the pairwise objective space dominance relationships of all candidate strategies in the effective strategy subset. The Pareto front level is initialized to the first level, and combinations of reactive power compensation commands and topology switching commands not dominated by other candidate strategies are marked as Pareto non-dominated solutions and included in the Pareto optimal solution set. The remaining candidate strategies are further divided into subsequent Pareto front levels using a fast non-dominated ranking method, eliminating candidate strategies dominated by higher-level solutions. Finally, based on a preset Pareto front level threshold, the combination of reactive power compensation commands and topology switching commands with the best comprehensive voltage stability improvement rate, harmonic suppression rate, and network loss reduction rate is selected to generate a Pareto optimal solution set that satisfies the multi-objective collaborative optimization.

[0073] S4. Based on the Pareto optimal solution set and combined with real-time multi-dimensional power grid state data, the pheromone parameters of the multi-objective ant colony optimization algorithm are dynamically adjusted through the deep deterministic strategy gradient algorithm to generate optimization parameters.

[0074] Based on the Pareto optimal solution set and real-time power grid state data, dynamic feature vectors of the power grid are extracted through Kalman filtering and principal component analysis.

[0075] It should be noted that, based on the Pareto optimal solution set and real-time collected power grid state data, Kalman filtering uses the voltage stability improvement rate corresponding to reactive power compensation commands, the harmonic suppression rate corresponding to topology switching commands, and the network loss reduction rate as observation variables, and the node voltage amplitude, line active power, and reactive power in the power grid state data as state variables. It estimates the optimal value of the power grid dynamic state through a prediction-correction iterative process. Principal component analysis decomposes the state estimate output by Kalman filtering with the quantitative indicators of each strategy in the Pareto optimal solution set using a covariance matrix, calculating the eigenvalues ​​and eigenvectors of the principal components of node voltage fluctuations, harmonic spectrum, and network loss. Based on the variance contribution rate in descending order, the principal component eigenvectors whose cumulative variance contribution rate exceeds the fidelity requirement based on the power system dynamic characteristics and are cross-validated using a historical power grid disturbance event dataset are selected to form the power grid dynamic feature vector.

[0076] Based on the dynamic feature vector of the power grid, the correlation between the power grid state and multi-objective optimization is analyzed by a deep deterministic strategy gradient algorithm to generate dynamic adjustment signals;

[0077] It should be noted that, based on the dynamic feature vector of the power grid, a deep deterministic policy gradient algorithm constructs an actor-commentator framework. The principal component feature vectors of node voltage fluctuations, harmonic spectrum, and network loss in the dynamic feature vector of the power grid are used as state inputs. The reactive power compensation capacity adjustment magnitude and topology switching command are defined as a continuous action space. The commentator network calculates the Q-value function of the state-action pair through time difference error, evaluating a multi-objective weighted reward for voltage stability improvement, harmonic suppression, and network loss reduction. The actor network, based on the deterministic policy gradient update rule, backpropagates the Q-value gradient to the action policy parameters through chain-like differentiation, generating reactive power compensation device switching capacity adjustment signals and topology switching priority adjustment signals. An experience replay buffer stores historical state transition samples, and random batch sampling eliminates data correlation, improving policy convergence stability. Finally, a dynamic adjustment signal is output, containing the dynamic capacity correction coefficient of the reactive power compensation command and the priority ranking list of the topology switching command.

[0078] Based on the dynamically adjusted signal, the pheromone parameters are corrected through a multi-target ant colony optimization algorithm to generate optimized parameters.

[0079] It should be noted that, based on the dynamically adjusted signal, the multi-objective ant colony optimization algorithm uses the dynamic capacity correction coefficient of reactive power compensation instructions and the priority ranking list of topology switching instructions in the dynamically adjusted signal as input to update the pheromone parameters. Pheromones are virtual chemical markers simulating the path selection behavior of optimization units, and their parameters include pheromone evaporation coefficient, pheromone heuristic factor, and expected heuristic factor, which are used to quantify the path quality of candidate strategies. The pheromone evaporation coefficient adaptively adjusts the local pheromone update rules according to the dynamic capacity correction coefficient of reactive power compensation instructions, enhancing the pheromone concentration of paths corresponding to high-priority topology switching instructions. The pheromone heuristic factor and expected heuristic factor dynamically adjust the weight allocation of voltage stability, harmonic suppression rate, and network loss reduction rate in multi-objective optimization based on the priority ranking list of topology switching instructions, thereby guiding optimization units to favor strategy combinations with better overall performance during the search process. After the pheromone parameters are dynamically updated, the multi-objective ant colony optimization algorithm generates optimization parameters containing the updated pheromone evaporation coefficient, pheromone heuristic factor, and expected heuristic factor through iterative path search.

[0080] S5. Based on optimized parameters, QV droop control and voltage regulation are used to perform command parsing and dynamic strategy mapping to generate reactive power compensation commands and topology switching commands.

[0081] Based on optimized parameters, the voltage regulation reference value and the reference reactive power compensation requirement are obtained through QV droop control analysis. The initial control command is formed by weighting the command and mapping the dynamic strategy through voltage-reactive power sensitivity analysis combined with the real-time power grid topology parameter set.

[0082] It should be noted that QV droop control is based on the voltage-reactive power droop curve defined by the Institute of Electrical and Electronics Engineers (IEEE) standard. The pheromone evaporation coefficient and pheromone heuristic factor in the optimization parameters are used as weights to adjust the slope of the droop curve, analyzing the voltage regulation baseline value and baseline reactive power compensation demand under the current grid operating state. The voltage-reactive power sensitivity analysis method calculates the voltage sensitivity coefficient matrix of each reactive power compensation node to the target bus based on the node admittance matrix and branch impedance parameters in the real-time grid topology parameter set, generating a node voltage-reactive power compensation sensitivity ranking list. During the command weighting process, the node priority weights are allocated to the baseline reactive power compensation demand through the voltage-reactive power sensitivity ranking list, and strategy mapping is performed in conjunction with the topology switching command priority ranking list in the dynamic adjustment signal. The dynamic strategy mapping generates an initial control command based on the voltage regulation baseline value and the weighted reactive power compensation demand, including the target bus voltage adjustment amount, reactive power compensation node switching capacity, and topology switching command execution timing.

[0083] By combining the real-time power grid status, the initial control command is dynamically corrected through a dynamic state estimation algorithm to generate the final reactive power compensation command and topology switching command. The reactive power compensation command drives the reactive power compensation device to adjust the reactive power output of the region, and the topology switching command controls the circuit breaker to perform opening and closing operations, thereby realizing dynamic zoning of the distribution network.

[0084] It should be noted that, based on the initial control command, the dynamic state estimation algorithm adopts the weighted least squares state estimation method, using the real-time collected node voltage amplitude, line active power, and reactive power as measurement inputs. Through a state prediction-correction iterative process, it updates the target bus voltage adjustment and reactive power compensation node switching capacity in the initial control command. During the dynamic correction process, the voltage-reactive power sensitivity analysis method, based on the node admittance matrix parameters in the real-time power grid topology parameter set, recalculates the sensitivity coefficient matrix of the reactive power compensation nodes to the target bus, generating corrected node priority weights. Combining the voltage adjustment deviation and sensitivity weight deviation from the dynamic state estimation output, it linearly weights and adjusts the reactive power compensation capacity allocation ratio and topology switching command execution timing in the initial control command, generating the final reactive power compensation command and topology switching command. The final reactive power compensation command drives reactive power compensation devices such as static var generators and capacitor banks to adjust the injected or absorbed reactive power, while the topology switching command controls the circuit breaker opening and closing operations to change the connection relationship of power grid nodes, realizing dynamic zoning of the distribution network based on electrical coupling strength and dynamic operating status.

[0085] Example 2 is the second embodiment of the present invention, which differs from the previous embodiment in that:

[0086] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0087] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0088] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0089] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0090] Example 3, an embodiment of the present invention, provides a dynamic zoning system for distribution networks based on multi-objective ant colony optimization, including a data preprocessing module, a virtual image generation module, a disturbance optimization module, a parameter optimization module, and an instruction generation module;

[0091] The data processing module collects multi-dimensional status data of the power grid in real time, synchronizes the multi-dimensional status data of the power grid to the edge cluster through a distributed data synchronization mechanism based on timestamp alignment, and generates dynamic data streams by combining data cleaning.

[0092] The virtual image generation module, based on dynamic data flow, uses dynamic state estimation algorithm and finite element electromagnetic field simulation algorithm to synchronize the power grid topology parameter set and map three-dimensional field strength to generate a virtual power grid image.

[0093] The disturbance optimization module embeds a multi-objective ant colony optimization algorithm into the virtual power grid mirror and injects new energy sudden drop and harmonic resonance disturbances into the virtual power grid through Monte Carlo simulation to generate a Pareto optimal solution set.

[0094] The parameter optimization module, based on the Pareto optimal solution set and combined with real-time multi-dimensional power grid state data, dynamically adjusts the pheromone parameters of the multi-objective ant colony optimization algorithm through a deep deterministic strategy gradient algorithm to generate optimized parameters.

[0095] The instruction generation module, based on optimized parameters, performs instruction parsing and dynamic strategy mapping through QV droop control and voltage regulation to generate reactive power compensation instructions and topology switching instructions.

[0096] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A dynamic zoning method for distribution networks based on multi-objective ant colony optimization, characterized in that: include, Real-time acquisition of multi-dimensional status data of the power grid; synchronization of multi-dimensional status data of the power grid to the edge cluster through a distributed data synchronization mechanism based on timestamp alignment; and generation of dynamic data stream by combining data cleaning. Based on dynamic data flow, a virtual power grid mirror is generated by synchronizing the power grid topology parameter set and mapping three-dimensional field strength through dynamic state estimation algorithm and finite element electromagnetic field simulation algorithm. A multi-objective ant colony optimization algorithm is embedded in a virtual power grid mirror, and a Pareto optimal solution set is generated by injecting new energy sudden drop and harmonic resonance disturbance into the virtual power grid through Monte Carlo simulation. Based on the Pareto optimal solution set and combined with real-time multi-dimensional power grid state data, the pheromone parameters of the multi-objective ant colony optimization algorithm are dynamically adjusted through a deep deterministic strategy gradient algorithm to generate optimization parameters. Based on optimized parameters, QV droop control and voltage regulation are used to perform command parsing and dynamic strategy mapping to generate reactive power compensation commands and topology switching commands.

2. The dynamic zoning method for distribution networks based on multi-objective ant colony optimization as described in claim 1, characterized in that: The multi-dimensional status data of the power grid includes equipment operating parameters, electrical parameters, environmental disturbance data, and topology parameters.

3. The dynamic zoning method for distribution networks based on multi-objective ant colony optimization as described in claim 2, characterized in that: The generation of dynamic data streams includes generating multi-source data streams by using a clock synchronization mechanism and a distributed transmission method to perform time-series alignment and security encryption of cross-node data based on multi-dimensional power grid status data. Based on multi-source data streams, cross-node timing deviations are eliminated through timestamp alignment, and data is cleaned using the isolated forest algorithm and standardization processing to generate dynamic data streams.

4. The dynamic zoning method for distribution networks based on multi-objective ant colony optimization as described in claim 3, characterized in that: The generation of the virtual power grid image includes generating a real-time power grid topology parameter set based on dynamic data streams and a dynamic state estimation algorithm; Based on the power grid topology parameter set, the three-dimensional spatial field strength distribution is inverted through the finite element electromagnetic field simulation algorithm to generate field strength gradient data; Based on field strength gradient data combined with dynamic data streams, spatiotemporal data fusion and standardized encapsulation are performed through power grid multi-source data fusion specifications and wide-area time synchronization methods to generate a virtual power grid image.

5. The dynamic zoning method for distribution networks based on multi-objective ant colony optimization as described in claim 4, characterized in that: The generation of the Pareto optimal solution set includes embedding a multi-objective ant colony optimization algorithm in a virtual power grid image to simulate multiple optimization objectives under power grid operation conditions and generate a candidate strategy set. Based on the candidate strategy set, a set of events for sudden drop in new energy and harmonic resonance disturbance is generated by combining random parameter perturbation with dynamic data stream through Monte Carlo simulation and perturbation scenario generation method. Then, the effective strategy subset is analyzed and screened by Gaussian kernel density estimation method. The Pareto front screening method is used to perform a multi-objective comprehensive evaluation of the effective policy subset and generate the Pareto optimal solution set.

6. The dynamic zoning method for distribution networks based on multi-objective ant colony optimization as described in claim 5, characterized in that: The generated optimization parameters include extracting dynamic feature vectors of the power grid through Kalman filtering and principal component analysis based on the Pareto optimal solution set and real-time collected power grid state data. Based on the dynamic feature vector of the power grid, the correlation between the power grid state and multi-objective optimization is analyzed by a deep deterministic strategy gradient algorithm to generate dynamic adjustment signals; Based on the dynamically adjusted signal, the pheromone parameters are corrected through a multi-target ant colony optimization algorithm to generate optimized parameters.

7. The dynamic zoning method for distribution networks based on multi-objective ant colony optimization as described in claim 6, characterized in that: The generation of reactive power compensation instructions and topology switching instructions includes: based on optimized parameters, obtaining voltage regulation reference values ​​and reference reactive power compensation requirements through QV droop control analysis, and forming initial control instructions by weighting instructions and mapping dynamic strategies using voltage-reactive power sensitivity analysis combined with real-time power grid topology parameter sets. By combining the real-time power grid status, the initial control command is dynamically corrected through a dynamic state estimation algorithm to generate the final reactive power compensation command and topology switching command. The reactive power compensation command drives the reactive power compensation device to adjust the reactive power output of the region, and the topology switching command controls the circuit breaker to perform opening and closing operations, thereby realizing dynamic zoning of the distribution network.

8. A system for dynamic zoning of a distribution network based on multi-objective ant colony optimization, employing the method for dynamic zoning of a distribution network based on multi-objective ant colony optimization as described in any one of claims 1 to 7, characterized in that, include: Data preprocessing module, virtual image generation module, disturbance optimization module, parameter optimization module and instruction generation module; The data processing module collects multi-dimensional status data of the power grid in real time, synchronizes the multi-dimensional status data of the power grid to the edge cluster through a distributed data synchronization mechanism based on timestamp alignment, and generates dynamic data streams by combining data cleaning. The virtual image generation module, based on dynamic data flow, uses a dynamic state estimation algorithm and a finite element electromagnetic field simulation algorithm to synchronize the power grid topology parameter set and map the three-dimensional field strength to generate a virtual power grid image. The disturbance optimization module embeds a multi-objective ant colony optimization algorithm into the virtual power grid mirror and injects new energy sudden drop and harmonic resonance disturbances into the virtual power grid through Monte Carlo simulation to generate a Pareto optimal solution set. The parameter optimization module, based on the Pareto optimal solution set and combined with real-time multi-dimensional power grid state data, dynamically adjusts the pheromone parameters of the multi-objective ant colony optimization algorithm through a deep deterministic strategy gradient algorithm to generate optimized parameters. The instruction generation module, based on optimized parameters, performs instruction parsing and dynamic strategy mapping through QV droop control and voltage regulation to generate reactive power compensation instructions and topology switching instructions.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the dynamic zoning method for power distribution networks based on multi-objective ant colony optimization as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the dynamic zoning method for power distribution networks based on multi-objective ant colony optimization as described in any one of claims 1 to 7.

Images