Power distribution network partition dynamic voltage control method and system based on wind-solar-storage cooperation

Abstract

The invention relates to the technical field of smart power grids, and discloses a wind-solar-storage coordinated power distribution network partition dynamic voltage control method and system, and the method achieves the precise self-adaptive division of control sub-regions through the dynamic clustering partition based on real-time voltage sensitivity. A double-layer cooperative control framework of'centralized optimization-decentralized execution 'is constructed, and distributed rolling optimization is performed by adopting an alternating direction multiplier method, so that the unification of global coordination and quick local response is realized; a multi-target energy storage optimization configuration model considering wind and light uncertainty is established and is deeply coordinated with an operation control strategy. Therefore, the effects of improving the voltage control precision, the response speed and the robustness of the high-proportion new energy power distribution network are achieved, and meanwhile, the system economy, the power supply reliability and the comprehensive utilization efficiency of wind, light and storage resources are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of smart grid technology, and in particular to a method and system for dynamic voltage control of distribution network zones that integrates wind, solar and energy storage. Background Technology

[0002] Distribution network voltage control refers to adjusting the operating status of various controllable devices in the power grid (such as transformer tap changes, reactive power compensation devices, distributed generation inverters, energy storage systems, etc.) to ensure that the voltage amplitude at all nodes of the distribution network is always maintained within the safe and high-quality range specified by national standards or operating procedures. Its core objective is to cope with disturbances such as load fluctuations and changes in distributed generation output, prevent voltage exceedances, reduce network losses, and ensure power supply quality and the safe and stable operation of the power grid.

[0003] In existing technologies, voltage control methods for distribution networks containing wind, solar, and energy storage can be mainly categorized as follows:

[0004] Centralized optimization control: Data from the entire network is collected at the dispatch center, and a global optimization model is established, encompassing all resources such as wind, solar, and energy storage (typically aiming to minimize network loss and voltage deviation). After solving the model, control commands are issued to each device. Theoretically, this method can achieve global optimization.

[0005] Local distributed (autonomous) control: Each wind, solar, and energy storage device or its aggregation unit is controlled solely based on local measurement information (such as grid connection point voltage). Examples include voltage-based QV droop control for photovoltaic inverters and local signal-based charging and discharging strategies for energy storage. This method offers fast response times and eliminates the need for complex communication.

[0006] Fixed-zone-based coordinated control: Based on the physical topology of the power grid or pre-calculated static electrical relationships (such as electrical distances), the distribution network is divided into several relatively fixed control zones. Coordination and optimization are performed within each zone, or simple power exchange coordination is conducted between zones. This approach attempts to strike a balance between centralized and decentralized control.

[0007] Layered control architecture: Combining the methods described above, the upper layer is typically configured to perform slow global optimization or target distribution, while the lower layer performs rapid local adjustments. However, the coordination strategies between the upper and lower layers in existing methods are often relatively simple or fixed.

[0008] The aforementioned and existing related technologies often suffer from the following drawbacks: strong uncertainty and rapid fluctuations in wind and solar power output, as well as power backfeeding caused by source-load spatiotemporal mismatch, make centralized optimization control, which relies on fixed models and long-cycle communication, difficult to achieve real-time and rapid response, and prone to control lag. Conversely, decentralized control, which relies entirely on local information, lacks a global perspective, potentially leading to conflicts between local optimization and global objectives, and even exacerbating voltage oscillations. This, in turn, affects the speed and stability of voltage control in high-proportion renewable energy scenarios.

[0009] The distribution network is characterized by numerous nodes, complex topology, and dynamic changes in wind, solar, and energy storage connection points. This makes it difficult for zoning methods based on static physical topology or fixed electrical parameters (such as some methods described in the prior art) to accurately reflect the dynamic electrical coupling relationships between nodes in real-time operation. The zoning results may be inaccurate or unreasonable, leading to poor coordination within the zoning and complex coordination at the zoning boundaries. This, in turn, affects the accuracy of voltage control and overall coordination efficiency, making it difficult to achieve optimal local reactive power balance.

[0010] Distributed energy storage possesses unique capabilities for rapid bidirectional power regulation and spatiotemporal energy shifting, while wind and solar power generation has certain reactive power regulation potential. This leads existing control strategies to often simply treat energy storage as an active power source or backup power, failing to achieve refined spatiotemporal coordination with the reactive / active power regulation capabilities of wind and solar power. Control objectives are often limited to voltage stability at a single moment, failing to comprehensively consider the economic efficiency (such as peak shaving and valley filling benefits, and reduced grid losses) and reliability improvement throughout the entire timeframe. This consequently affects the overall utilization efficiency of energy storage and wind / solar resources, limiting their full potential in improving grid flexibility, promoting renewable energy consumption, and reducing overall system operating costs, and also impacting the economic feasibility of promoting control strategies. Summary of the Invention

[0011] The technical problem to be solved by the present invention is that the existing technology has the following drawbacks: poor adaptability of control mode leading to insufficient ability to cope with disturbances, rigid zoning strategy leading to low control accuracy and efficiency, and lack of deep synergy of wind, solar and energy storage resources leading to limited overall benefits. To this end, we propose a dynamic voltage control method and system for distribution network zoning with wind, solar and energy storage synergy.

[0012] To achieve the above objectives, this application adopts the following technical solution: a dynamic voltage control method for distribution network zones integrating wind, solar, and energy storage, comprising the following steps:

[0013] Collect and preprocess operational data from key nodes of the power distribution network and wind, solar and energy storage equipment;

[0014] The current operating status of the system is assessed based on the preprocessed data, and ultra-short-term forecasts of wind and solar power output and load power are made.

[0015] With the goal of achieving optimal economic efficiency and highest operational reliability throughout the system's life cycle, an energy storage optimization configuration model is established and solved.

[0016] The voltage sensitivity matrix is ​​obtained based on real-time power flow calculation, and the K-means clustering algorithm is used for dynamic clustering and partitioning to form several voltage control sub-regions;

[0017] A two-layer collaborative control architecture is constructed. The upper layer performs centralized optimization with the goal of minimizing the voltage deviation and network loss of the entire network, generates a global coordination target and distributes it. The lower layer, each voltage control sub-region, performs local rapid adjustment based on the global coordination target.

[0018] Based on the alternating direction multiplier method, the two-layer cooperative control problem is reconstructed into a distributed optimization problem. Each sub-region solves its local sub-problem in parallel and iterates until convergence by exchanging boundary information, generating the optimal scheduling instruction sequence for the controllable equipment.

[0019] The optimal scheduling instruction sequence is sent to the controllable device for execution, and closed-loop control is formed based on the execution feedback.

[0020] Furthermore, the step of obtaining the voltage sensitivity matrix based on real-time power flow calculation and performing dynamic clustering partitioning using the K-means clustering algorithm includes:

[0021] Inverting the Jacobian matrix J generated from real-time power flow calculations yields the voltage sensitivity matrix J. -1 This includes voltage active power sensitivity indicators. and voltage reactive power sensitivity index ;: Construct a feature vector for each node i ,in: and Let be the voltage active and reactive power sensitivity vectors of node i with respect to all other nodes, respectively. : The electrical distance vector of node i;

[0022] The K-means clustering algorithm is used to dynamically cluster the feature vectors of all nodes, dividing the distribution network nodes into K voltage control sub-regions.

[0023] The objective function of the K-means algorithm is: , The set of node samples in cluster i. : The feature vector of the j-th node The center of cluster i.

[0024] Furthermore, the dynamic clustering using the K-means clustering algorithm includes:

[0025] Select K node samples as the initial cluster centers;

[0026] Calculate the Euclidean distance between the remaining samples and the center of each cluster, and assign each sample to the cluster with the closest distance.

[0027] For each cluster, the average of the feature vectors of all samples within it is used as the new center. All samples were re-divided based on the new center;

[0028] Repeat the process of updating the center and repartitioning until the cluster partition remains unchanged or the maximum number of iterations is reached, to obtain the final partition result.

[0029] Furthermore, prior to the dynamic clustering partitioning, a reactive power voltage partitioning criterion is also included:

[0030] Define the expected voltage change at node i caused by the photovoltaic system at node j adjusting to its maximum reactive power capacity. ,in , : The maximum reactive power regulation capacity of photovoltaic system at node j at time τ within the time period t; : Expected value; Node j: Apparent photovoltaic capacity; : Photovoltaic active power output at node j at time τ Expectation operator;

[0031] Set voltage change threshold ,in: Total number of nodes; : Partition size coefficient, if If so, then node i will be assigned to the reactive voltage control sub-region centered on node j.

[0032] Furthermore, the construction of the two-layer collaborative control architecture includes:

[0033] The upper centralized control layer establishes a global optimization model with the goal of minimizing the sum of squares of voltage deviations across the entire network and minimizing the total active power loss of the system. The decision variables include the exchange power between each sub-region and the backbone network and the overall output of key controllable resources. The power exchange target or key node voltage reference value is solved and issued in the first optimization cycle.

[0034] The lower-level distributed control layer establishes a local controller in each voltage control sub-region to receive instructions from the upper layer and collect information on voltage, wind and solar power output, and load in the sub-region. With the goal of quickly eliminating voltage deviations within the sub-region and minimizing regulation losses, it coordinates and controls the reactive power output of photovoltaic inverters and wind power converters, as well as the charging and discharging power of distributed energy storage.

[0035] Furthermore, the distributed optimization problem reconstructed based on the alternating direction multiplier method includes:

[0036] The global objective function is decomposed into sub-problems in each sub-region, and the decision variables for each sub-problem are... It only includes scheduling instructions for controllable devices within this sub-region;

[0037] In the k-th iteration, subregion s solves the local optimization subproblem: ,in The local objective function of subregion s. Its decision variables, Globally consistent variables Dual variable, Penalty parameter, Local constraint matrix;

[0038] After solving each sub-region, the boundary coupling variable information is exchanged with the adjacent sub-regions, z and u are updated, and the process is iterated until convergence, generating the optimal scheduling instruction sequence for each controllable device.

[0039] Furthermore, the establishment and solution of the energy storage optimal configuration model includes:

[0040] Establish a multi-objective optimization model, with the objective function including economic objectives. and reliability objectives ,in: Initial investment cost for energy storage; Annual replacement cost of energy storage; Annual operation and maintenance costs of energy storage; Costs related to unit start-up and shutdown during time period t; Carbon emission costs during period t; : Cost of electricity for energy storage charging and discharging during time period t; Carbon trading revenue during period t Average system outage frequency Average system outage duration System average power supply reliability These are the weighting coefficients for each reliability index.

[0041] Furthermore, the improved particle swarm optimization algorithm is used to solve the energy storage optimization configuration model:

[0042] Position vector of each particle This represents a set of energy storage configuration solutions;

[0043] The fitness function is and The weighted sum or Pareto ranking evaluation value;

[0044] The particle update formula is: ,

[0045] ,

[0046] in: Number of iterations; and Let be the position and velocity vectors of particle i, respectively. and These are the individual's historical best position and the population's global historical best position, respectively. , A random number within the range [0,1], acceleration factor , It is used to balance individual and group learning abilities.

[0047] Furthermore, the step of issuing scheduling instructions to controllable devices for execution and forming closed-loop control includes:

[0048] The optimal scheduling instruction sequence is converted into control signals and sent to photovoltaic inverters, wind power converters and energy storage systems through standard communication protocols;

[0049] Real-time monitoring of command reception status, deviation between actual device output and set value, and device operating status;

[0050] The actual execution results are fed back to the upper control layer, forming a closed loop;

[0051] When the deviation exceeds the threshold or the equipment malfunctions, the control strategy is readjusted.

[0052] Furthermore, the collection and preprocessing of operational data includes:

[0053] Perform time synchronization and alignment on the raw runtime data;

[0054] A strategy combining physical thresholding and statistical outlier detection is employed to identify and remove outliers and noise.

[0055] For short-term missing data, interpolation based on time series prediction is used to fill in the gaps.

[0056] A wind-solar-storage coordinated distribution network zone dynamic voltage control system includes:

[0057] Data acquisition module: Collects and preprocesses operational data from key nodes in the power distribution network and wind, solar, and energy storage equipment;

[0058] Prediction module: Based on the preprocessed data, assess the current operating status of the system and make ultra-short-term predictions of wind and solar power output and load power;

[0059] Partitioning Module: With the goal of achieving optimal economic efficiency and highest operational reliability throughout the system's lifecycle, an energy storage optimization configuration model is established and solved; the voltage sensitivity matrix is ​​obtained based on real-time power flow calculations, and the K-means clustering algorithm is used for dynamic clustering and partitioning to form several voltage control sub-regions;

[0060] Instruction generation module: Constructs a two-layer collaborative control architecture. The upper layer performs centralized optimization with the goal of minimizing the voltage deviation and network loss of the entire network, generates a global coordination target and issues it down. The lower layer, each voltage control sub-region, performs local rapid adjustment based on the global coordination target. Based on the alternating direction multiplier method, the two-layer collaborative control problem is reconstructed into a distributed optimization problem. Each sub-region solves its local subproblem in parallel and iterates until convergence by exchanging boundary information, generating the optimal scheduling instruction sequence for controllable devices.

[0061] Execution module: issues the optimal sequence of scheduling instructions to the controllable devices for execution, and forms closed-loop control based on execution feedback.

[0062] The technical effects and advantages of this invention are as follows:

[0063] In this invention, a global coordination target is periodically generated by a centralized upper-level controller, aiming to minimize the voltage deviation and network loss across the entire network. Local controllers in each lower-level region then perform rapid local adjustments based on this target. The key optimization process utilizes the ADMM algorithm for distributed solution: the global problem is decomposed into sub-problems that are solved in parallel in each sub-region. Each sub-region only needs to exchange boundary coupling variable information with its neighboring regions, iteratively approximating the global optimum. The advantage of this method is that it retains the centralized global optimization capability, ensuring the overall economical operation and voltage stability of the system; while also achieving second-level response to local disturbances such as wind and solar power fluctuations through distributed parallel computing and local fast control loops. The application of ADMM effectively reduces the burden of centralized computing and communication dependence, enabling the control system to work efficiently and collaboratively even under weak communication conditions, significantly improving the overall robustness and real-time control performance of high-proportion renewable energy distribution networks in the face of complex disturbances.

[0064] In this invention, precise node voltage-active power sensitivity and voltage-reactive power sensitivity matrices are obtained online by inverting the Jacobian matrix based on real-time power flow calculations, which quantifies the dynamic electrical coupling strength between nodes. Subsequently, node feature vectors are constructed using these sensitivity vectors and electrical distances, and online partitioning is performed using the K-means clustering algorithm. The advantage of this method is that it allows the partitioning results to be dynamically updated closely following changes in network operating status, distributed power output, and load, ensuring that nodes within a partition always have a high degree of voltage correlation. Compared to static partitioning, this dynamic adaptive partitioning method makes the scope of reactive power compensation and active power regulation resources more precise, greatly improving the efficiency and accuracy of voltage coordination control within the partition, achieving local reactive power balance, effectively suppressing control resource consumption or regulation blind spots caused by unreasonable partitioning, and laying a precise regional foundation for subsequent hierarchical collaborative control.

[0065] This invention employs synergistic optimization at both the planning and control levels. At the planning level, a multi-objective energy storage optimization configuration model considering the uncertainties of wind and solar power is established and solved using an improved particle swarm optimization (PSO) algorithm. The model simultaneously minimizes economic costs and maximizes power supply reliability. The PSO algorithm searches for the Pareto optimal configuration by iteratively updating particle positions. At the operational level, based on this configuration scheme, a two-layer control architecture finely coordinates the charging and discharging of energy storage with the reactive power output of wind and solar inverters. The effectiveness of this method lies in achieving vertical integration of energy storage configuration and operational control, as well as horizontal synergy of spatiotemporal complementarity between wind, solar, and energy storage resources. Energy storage is not only used for peak shaving and valley filling arbitrage but also plays a crucial role in multiple objectives such as voltage support, smoothing fluctuations, and improving power supply reliability. This synergistic strategy significantly enhances the absorption capacity of new energy sources, reduces the overall system operating cost and network losses, and strengthens the grid's resilience to uncertainties. It transforms the wind-solar-energy storage system from a passively accessed resource into an intelligent asset that actively supports the safe, economical, and reliable operation of the grid, resulting in significant comprehensive benefits. Attached Figure Description

[0066] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts:

[0067] Figure 1 This is a flowchart of the method of the present invention;

[0068] Figure 2 This is a flowchart of the Kmeans algorithm of the present invention. Detailed Implementation

[0069] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0070] like Figure 1 The present invention provides a technical solution: a method for dynamic voltage control of distribution network zones with wind, solar and energy storage coordination, comprising the following steps:

[0071] Step 1: Data Acquisition and Preprocessing

[0072] The system receives raw operational data from key nodes in the power distribution network and wind, solar, and energy storage devices, including electrical measurement data, equipment status data, and environmental data. It synchronizes and aligns the data by establishing a unified timeline, and uses a strategy combining physical thresholding and statistical outlier detection to identify and remove outliers and noise. The Z-score normalization method is then used to normalize the data, outputting a clean and structured dataset to provide a high-quality data foundation for subsequent analysis.

[0073] Furthermore, it should be added that: for short-term missing data, interpolation based on time series prediction is used to fill in the missing data; for long-term missing data, the data in that time period is marked as not participating in subsequent model training.

[0074] Furthermore, it is necessary to supplement the following electrical measurement data: node voltage amplitude, phase angle, active power, reactive power, frequency, etc.

[0075] Equipment status data: output power of photovoltaic inverter, output power of wind power converter, energy storage charging and discharging power, SOC (state of charge), reactive power output of SVG, etc.

[0076] Environmental data: meteorological data such as light intensity, wind speed, and temperature.

[0077] Step Two: Operational Status Assessment and Ultra-Short-Term Forecasting

[0078] Based on a clean dataset, key operational indicators such as voltage deviation rate, node voltage over-limit indicators, system network loss, and renewable energy penetration rate are calculated to assess the current operating status of the system. Simultaneously, deep learning-based prediction models, such as LSTM or Transformer, are used to perform ultra-short-term forecasts of wind and solar power output and load power for the next 15 minutes to 4 hours. Combining the current status with the prediction results, the system assesses voltage over-limit and overload risks in the future period, generating risk warning information and providing forward-looking input for subsequent optimized control.

[0079] Step 3: Optimize energy storage configuration considering the uncertainties of wind and solar power.

[0080] 3.1: To achieve the optimal economic efficiency and highest operational reliability throughout the system's entire lifecycle, an energy storage site selection and capacity optimization model is established.

[0081] Economic objectives The expression is as follows:

[0082] .

[0083] in: Initial investment cost for energy storage; Annual replacement cost of energy storage; Annual operation and maintenance costs of energy storage; Costs related to unit start-up and shutdown during time period t; Carbon emission costs during period t; : Cost of electricity for energy storage charging and discharging during time period t; Carbon trading revenue during period t.

[0084] Reliability targets The aim is to quantify the improvement in power supply reliability of the distribution network after configuring energy storage, and its expression is:

[0085] .

[0086] in, The average outage frequency is the average number of outages experienced by users of the distribution network per unit of time. The average power outage duration is the average duration of each power outage event. The average power supply reliability rate of the system refers to the ratio of the actual power supply time to the total time during the statistical period. These are the weighting coefficients for each reliability index, used to adjust the importance of different indices in the optimization process.

[0087] Further details need to be added: In order to ensure that the optimization scheme meets the basic requirements of power grid physics and safe operation, the constraints include power balance constraints, node voltage safety constraints, energy storage charging and discharging power and capacity constraints, energy storage installation location constraints, and state of charge (SOC) constraints.

[0088] 3.2 Solving the problem using an improved Particle Swarm Optimization (PSO) algorithm

[0089] In the scenario of this invention, the position vector of each particle This represents an energy storage configuration scheme (installation location and capacity). The particle's fitness function is the aforementioned multi-objective function. and The weighted sum or Pareto-based evaluation value. The core of the PSO algorithm is the updating of particle position and velocity. In the D-dimensional solution space, the particle iterative update strategy is as follows:

[0090] .

[0091] .

[0092] in, This represents the number of iterations. and These are the position and velocity vectors of particle i, respectively; and These are the individual's historical best position and the population's global historical best position, respectively. , A random number within the range [0,1]. Acceleration factor. , Used to balance individual and group learning abilities, its value range is typically [1.0, 2.0]. In this embodiment, an exemplary value is... Inertia weight To balance global exploration and local development, a linear decreasing strategy is used for updating, as shown in the following formula:

[0093] .

[0094] in, , The upper and lower limits of the inertia weight generally satisfy the following conditions: In this embodiment, the exemplary value is... and ; This represents the maximum number of iterations. The algorithm iteratively optimizes the energy storage configuration (including installation nodes, rated power, and capacity) to the final output.

[0095] Step 4: Dynamic partitioning based on voltage sensitivity

[0096] 4.1 Calculate the voltage sensitivity matrix

[0097] The inverse of the Jacobian matrix J generated from real-time power flow calculations yields the voltage sensitivity matrix, which quantifies the electrical coupling between nodes. The power flow correction equation is as follows:

[0098] .

[0099] After transformation, we get:

[0100] .

[0101] In the formula: , The changes in active and reactive power are injected into node j, respectively. , These represent the changes in the phase angle and magnitude of the voltage at node i, respectively. , These are the voltage active power sensitivity index and voltage reactive power sensitivity index between nodes i and j, respectively.

[0102] Consider the voltage change at node i caused by changing the photovoltaic output at node j:

[0103] .

[0104] In the formula: , These represent the changes in active and reactive power output of the photovoltaic system at node j, respectively.

[0105] 4.2 Constructing Voltage Variation Indicators and Zoning Criteria

[0106] In reactive voltage partitioning, define voltage variation indicators. Let t be the expected change in voltage at node i caused by the photovoltaic system at node j adjusting to its maximum reactive power capacity during time period t.

[0107] .

[0108] in, ), Let J be the maximum reactive power regulation capacity of the photovoltaic system at time τ within time period t. for Expected value; The apparent capacity of photovoltaics at node j; Let j be the active power output of the photovoltaic system at time τ. This is the expectation operator. The voltage change threshold for the reactive voltage zone to which node j belongs in time period t is set as the average of the expected changes across all nodes caused by the photovoltaic system at node j being adjusted to its maximum reactive capacity in time period t. Its expression is:

[0109] .

[0110] In the formula: The total number of nodes; This is the partition size coefficient.

[0111] The criterion is: if If node i is selected, then node j will be assigned to the reactive voltage control sub-region centered on node j. Active voltage zoning can be based on the voltage active sensitivity index. Analogy construction.

[0112] like Figure 2 As shown in section 4.3, dynamic clustering and partitioning are performed using the K-means algorithm.

[0113] To obtain more accurate and adaptive partitions, each node i is represented as a feature quantity. , : The voltage active power sensitivity vector of node i with respect to all other nodes. : The voltage reactive power sensitivity vector of node i with respect to all other nodes. : The electrical distance vector of node i (used to characterize its topological or impedance relationship with other nodes). Based on all node feature vectors calculated in steps 4.1 and 4.2, dynamic partitioning is performed using the K-means clustering algorithm. As a representative of unsupervised clustering algorithms, the K-means clustering algorithm mainly divides N node samples into K disjoint clusters (i.e., voltage-controlled sub-regions), and each node is partitioned. The node sample data is as follows: , where each sample Each is an n-dimensional vector representing the electrical characteristics of a node. The algorithm formula uses... The objective function is to cluster N node samples into K clusters.

[0114] in, This represents the set of node samples in cluster i. The eigenvector of the j-th node. The center of cluster i is typically the average of the feature vectors of the node samples within that cluster. .

[0115] The main idea of ​​the K-means algorithm is as follows: From N node samples, randomly select K samples (K being the predetermined number of partitions) as the initial centers of K clusters; for the remaining samples, search and calculate the Euclidean distance between the sample and each cluster center, selecting the cluster with the smallest distance to join, forming a new cluster partition; for each newly formed cluster, according to... Calculate new cluster centers; then, based on the new cluster centers, re-retrieve all samples and assign them to the nearest cluster; repeat the process of updating cluster centers and the cluster assignment method for each sample until the cluster assignment result no longer changes or the maximum number of iterations is reached, at which point the assignment ends, yielding the optimal cluster assignment method, i.e., the final dynamic voltage partitioning result. The number of partitions K can be determined based on system size, complexity, or heuristic methods such as the Elbow Method. The partitioning period can be updated based on changes in operating status, network topology changes, or timed triggering.

[0116] Step 5: Generation of Two-Layer Cooperative Voltage Control Strategy

[0117] Upper-level centralized control layer: A global optimization model is established with the objectives of minimizing the sum of squares of voltage deviations across the entire network and minimizing total active power loss. Decision variables include the power exchange between each sub-region and the backbone network, and the overall output of key controllable resources (such as centralized energy storage and SVG). The optimization cycle is relatively long (e.g., 5-15 minutes). After solving, power exchange targets or key node voltage reference values ​​are distributed to each lower-level sub-region.

[0118] Lower-level distributed control layer: Each voltage control sub-region establishes a local controller to receive commands from the upper layer and collect real-time information on voltage, wind and solar power output, and load of all nodes within its sub-region. The control objective is to quickly eliminate voltage deviations within the sub-region and minimize regulation losses while meeting the requirements of the upper layer. This is achieved through coordinated control of the reactive power output (Q control) of photovoltaic inverters and wind power converters, as well as the charging and discharging power (P control) of distributed energy storage, enabling rapid responses at the second or minute level.

[0119] Step Six: Distributed Scrolling Optimization Based on ADMM

[0120] The two-layer cooperative voltage control problem in step five is reconstructed into an optimization problem that can be solved in a distributed manner across sub-regions. The Alternating Directional Multiplier Method (ADMM) is adopted as the distributed solution framework. Specifically, the global objective function is decomposed into sub-problems for each sub-region. Each sub-problem only requires data from its own sub-region and information on boundary coupling variables of adjacent sub-regions (such as boundary node voltages and tie-line power). Sub-problem s is solved in the k-th iteration as an optimization problem of the following form:

[0121]

[0122] in, Let be the local objective function of subregion s. For its decision variables, For globally consistent variables, As dual variables, For penalty parameters, The local constraint matrix is ​​used to extract coupling variables. Boundary information is exchanged after parallel solving of each sub-region, and the process iterates until convergence. This process is based on ultra-short-term prediction data and is performed in rolling cycles with shorter periods (e.g., 5 minutes) to dynamically generate the optimal scheduling instruction sequence for each controllable device.

[0123] Step 7: Control command execution and closed-loop feedback

[0124] The optimized control commands are converted into executable control signals and sent to controllable devices such as photovoltaic inverters, wind power converters, and energy storage systems via standard communication protocols. The system monitors the command reception status, the deviation between actual device output and setpoints, and the device operating status in real time. The actual execution results are fed back to the upper-level control module, forming a closed-loop control system. When the execution deviation exceeds a preset threshold or a device malfunctions, a control strategy readjustment mechanism is triggered. Simultaneously, command safety boundaries and protection logic are set to prevent misoperation from causing device damage or grid accidents.

[0125] Step 8: Simulation Verification and Practical Application

[0126] 8.1 Digital Simulation Verification

[0127] Based on a typical low-voltage distribution network topology (such as the IEEE 33-bus system), a test model with a high proportion of random fluctuations in wind and solar power and energy storage configuration was built in a simulation platform. The complete control strategy proposed in this invention was deployed, and long-term series simulations were performed. Test indicators included: 24-hour voltage qualification rate, voltage fluctuation rate, system network loss, renewable energy absorption rate, control response time, etc., to comprehensively evaluate the effectiveness and robustness of the strategy.

[0128] 8.2 Practical Demonstration Application

[0129] A distribution substation with a high proportion of distributed photovoltaic (PV) grid connection and voltage quality issues was selected as a demonstration project. Local controllers and communication units were deployed, the control interfaces of existing equipment were modified, and a voltage control software system integrating the method of this invention was deployed. A field trial run lasting several months was conducted, comparing key performance indicators (such as voltage compliance rate, number of over-limit occurrences, and PV curtailment rate) before and after the implementation of the control strategy. The ultimate goal was to achieve a 100% 24-hour voltage compliance rate in the demonstration area, thus forming a replicable and scalable engineering solution.

[0130] This embodiment also provides a technical solution: a wind-solar-storage coordinated distribution network zone dynamic voltage control system, comprising:

[0131] Data acquisition module: Collects and preprocesses operational data from key nodes in the power distribution network and wind, solar, and energy storage equipment;

[0132] Prediction module: Based on the preprocessed data, assess the current operating status of the system and make ultra-short-term predictions of wind and solar power output and load power;

[0133] Partitioning Module: With the goal of achieving optimal economic efficiency and highest operational reliability throughout the system's lifecycle, an energy storage optimization configuration model is established and solved; the voltage sensitivity matrix is ​​obtained based on real-time power flow calculations, and the K-means clustering algorithm is used for dynamic clustering and partitioning to form several voltage control sub-regions;

[0134] Instruction generation module: Constructs a two-layer collaborative control architecture. The upper layer performs centralized optimization with the goal of minimizing the voltage deviation and network loss of the entire network, generates a global coordination target and issues it down. The lower layer, each voltage control sub-region, performs local rapid adjustment based on the global coordination target. Based on the alternating direction multiplier method, the two-layer collaborative control problem is reconstructed into a distributed optimization problem. Each sub-region solves its local subproblem in parallel and iterates until convergence by exchanging boundary information, generating the optimal scheduling instruction sequence for controllable devices.

[0135] Execution module: issues the optimal sequence of scheduling instructions to the controllable devices for execution, and forms closed-loop control based on execution feedback.

[0136] The technical scope of this invention is not limited to the content described above. Those skilled in the art can make various modifications and variations to the above embodiments without departing from the technical concept of this invention, and all such modifications and variations should fall within the protection scope of this invention.

[0137] The technical scope of this invention is not limited to the content described above. Those skilled in the art can make various modifications and variations to the above embodiments without departing from the technical concept of this invention, and all such modifications and variations should fall within the protection scope of this invention.

Claims

1. A method for dynamic voltage control of distribution network zones integrating wind, solar, and energy storage, characterized in that, Includes the following steps: Collect and preprocess operational data from key nodes of the power distribution network and wind, solar and energy storage equipment; Based on the preprocessed data, the current operating status of the system is assessed, and ultra-short-term forecasts of wind and solar power output and load power are made. With the goal of achieving optimal economic efficiency and highest operational reliability throughout the system's life cycle, an energy storage optimization configuration model is established and solved. The voltage sensitivity matrix is ​​obtained based on real-time power flow calculation, and the K-means clustering algorithm is used for dynamic clustering and partitioning to form several voltage control sub-regions; A two-layer collaborative control architecture is constructed. The upper layer performs centralized optimization with the goal of minimizing the voltage deviation and network loss of the entire network, generates a global coordination target and distributes it. The lower layer, each voltage control sub-region, performs local rapid adjustment based on the global coordination target. Based on the alternating direction multiplier method, the two-layer cooperative control problem is reconstructed into a distributed optimization problem. Each sub-region solves its local sub-problem in parallel and iterates until convergence by exchanging boundary information, generating the optimal scheduling instruction sequence for the controllable equipment. The optimal scheduling instruction sequence is sent to the controllable device for execution, and closed-loop control is formed based on the execution feedback.

2. The method for dynamic voltage control of distribution network zones with wind, solar, and energy storage synergy as described in claim 1, characterized in that: The process of obtaining the voltage sensitivity matrix based on real-time power flow calculation and performing dynamic clustering partitioning using the K-means clustering algorithm includes: Inverting the Jacobian matrix J generated from real-time power flow calculations yields the voltage sensitivity matrix J. -1 This includes voltage active power sensitivity indicators. and voltage reactive power sensitivity index ; Construct a feature vector for each node i ,in: and Let be the voltage active and reactive power sensitivity vectors of node i with respect to all other nodes, respectively. : The electrical distance vector of node i; The K-means clustering algorithm is used to dynamically cluster the feature vectors of all nodes, dividing the distribution network nodes into K voltage control sub-regions. The objective function of the K-means algorithm is: , The set of node samples in cluster i. : The feature vector of the j-th node The center of cluster i.

3. The method for dynamic voltage control of distribution network zones with wind, solar, and energy storage synergy as described in claim 2, characterized in that: The dynamic clustering using the K-means clustering algorithm includes: Select K node samples as the initial cluster centers; Calculate the Euclidean distance between the remaining samples and the center of each cluster, and assign each sample to the cluster with the closest distance. For each cluster, the average of the feature vectors of all samples within it is used as the new center. All samples were re-divided based on the new center; Repeat the process of updating the center and repartitioning until the cluster partition remains unchanged or the maximum number of iterations is reached, to obtain the final partition result.

4. The method for dynamic voltage control of distribution network zones with wind, solar, and energy storage synergy as described in claim 2, characterized in that: Prior to the dynamic clustering partitioning, a reactive power voltage partitioning criterion is also included: Define the expected voltage change at node i caused by the photovoltaic system at node j adjusting to its maximum reactive power capacity. ,in , The maximum reactive power regulation capacity of photovoltaic system at node j at time τ within the time period t. : Expected value; Node j: Apparent photovoltaic capacity : Photovoltaic active power output at node j at time τ Expectation operator; Set voltage change threshold ,in: Total number of nodes : Partition size coefficient, if If so, then node i will be assigned to the reactive voltage control sub-region centered on node j.

5. The method for dynamic voltage control of distribution network zones with wind, solar, and energy storage synergy as described in claim 1, characterized in that: The construction of the two-layer collaborative control architecture includes: The upper centralized control layer establishes a global optimization model with the goal of minimizing the sum of squares of voltage deviations across the entire network and minimizing the total active power loss of the system. The decision variables include the exchange power between each sub-region and the backbone network and the overall output of key controllable resources. The power exchange target or key node voltage reference value is solved and issued in the first optimization cycle. The lower-level distributed control layer establishes a local controller in each voltage control sub-region to receive instructions from the upper layer and collect information on voltage, wind and solar power output, and load in the sub-region. With the goal of quickly eliminating voltage deviations within the sub-region and minimizing regulation losses, it coordinates and controls the reactive power output of photovoltaic inverters and wind power converters, as well as the charging and discharging power of distributed energy storage.

6. The method for dynamic voltage control of distribution network zones with wind, solar, and energy storage synergy as described in claim 5, characterized in that: The distributed optimization problem based on the alternating direction multiplier method includes: The global objective function is decomposed into sub-problems in each sub-region, and the decision variables for each sub-problem are... It only includes scheduling instructions for controllable devices within this sub-region; In the k-th iteration, subregion s solves the local optimization subproblem: ,in The local objective function of subregion s. Its decision variables, Globally consistent variables Dual variable, Penalty parameter, Local constraint matrix; After solving each sub-region, the boundary coupling variable information is exchanged with the adjacent sub-regions, z and u are updated, and the process is iterated until convergence, generating the optimal scheduling instruction sequence for each controllable device.

7. The method for dynamic voltage control of distribution network zones with wind, solar, and energy storage synergy as described in claim 1, characterized in that: The establishment and solution of the energy storage optimization configuration model includes: Establish a multi-objective optimization model, with the objective function including economic objectives. and reliability objectives ,in: Initial investment cost for energy storage; Annual replacement cost of energy storage; Annual operation and maintenance costs of energy storage; Costs related to unit start-up and shutdown during time period t; Carbon emission costs during period t; : Cost of electricity for energy storage charging and discharging during time period t; Carbon trading revenue during period t Average system outage frequency Average system outage duration System average power supply reliability These are the weighting coefficients for each reliability index.

8. The method for dynamic voltage control of distribution network zones with wind, solar, and energy storage synergy as described in claim 7, characterized in that: The energy storage optimization configuration model is solved using an improved particle swarm optimization algorithm. Position vector of each particle This represents a set of energy storage configuration solutions; The fitness function is and The weighted sum or Pareto ranking evaluation value; The particle update formula is: , ， in: Number of iterations; and Let be the position and velocity vectors of particle i, respectively. and These are the individual's historical best position and the population's global historical best position, respectively. , A random number within the range [0,1], acceleration factor , It is used to balance individual and group learning abilities.

9. The method for dynamic voltage control of distribution network zones with wind, solar, and energy storage synergy according to claim 1, characterized in that: The step of issuing scheduling instructions to controllable devices for execution and forming closed-loop control includes: The optimal scheduling instruction sequence is converted into control signals and sent to photovoltaic inverters, wind power converters and energy storage systems through standard communication protocols; Real-time monitoring of command reception status, deviation between actual device output and set value, and device operating status; The actual execution results are fed back to the upper control layer, forming a closed loop; When the deviation exceeds the threshold or the equipment malfunctions, the control strategy is readjusted.

10. A dynamic voltage control system for a distribution network zoned by wind, solar, and energy storage, characterized in that, include: Data acquisition module: Collects and preprocesses operational data from key nodes in the power distribution network and wind, solar, and energy storage equipment; Prediction module: Based on the preprocessed data, assess the current operating status of the system and make ultra-short-term predictions of wind and solar power output and load power; Partitioning Module: With the goal of achieving optimal economic efficiency and highest operational reliability throughout the system's lifecycle, an energy storage optimization configuration model is established and solved; the voltage sensitivity matrix is ​​obtained based on real-time power flow calculations, and the K-means clustering algorithm is used for dynamic clustering and partitioning to form several voltage control sub-regions; Instruction generation module: Constructs a two-layer collaborative control architecture. The upper layer performs centralized optimization with the goal of minimizing the voltage deviation and network loss of the entire network, generates a global coordination target and issues it down. The lower layer, each voltage control sub-region, performs local rapid adjustment based on the global coordination target. Based on the alternating direction multiplier method, the two-layer collaborative control problem is reconstructed into a distributed optimization problem. Each sub-region solves its local subproblem in parallel and iterates until convergence by exchanging boundary information, generating the optimal scheduling instruction sequence for controllable devices. Execution module: issues the optimal scheduling instruction sequence to the controllable device for execution, and forms closed-loop control based on execution feedback.

Images