A Distributed Power Supply Cooperative Planning Method and System Considering Island Partitioning
By standardizing data processing and optimizing island partitioning methods, and combining breadth-first search and mixed-integer linear programming, the scientific and stability issues in island partitioning are resolved, achieving efficient coordination between distributed power sources and loads, and improving the operational reliability and economy of the distribution network.
Patent Information
- Application Number
- CN202611411296.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-11
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies lack scientific rigor in islanding, have poor matching effects between distributed power sources and loads, and are difficult to guarantee operational stability in multiple scenarios, making them unsuitable for the operational needs of complex power distribution networks.
By collecting and standardizing core data from the distribution network, a node-branch correlation matrix is constructed. Combining breadth-first search and mixed-integer linear programming, islanding partitioning and distributed power source configuration are optimized. Energy storage control is used to smooth out fluctuations, and multi-objective optimization and simulation verification are carried out.
It enables scientific planning of the distribution network, ensures priority power supply to important loads, improves the fault resistance and power supply continuity of the distribution network, and meets the multiple objectives of economy, reliability and environmental protection.
Smart Images

Figure CN123121803A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of distributed power planning technology, and in particular to a distributed power collaborative planning method and system that considers islanding. Background Technology
[0002] With the increasing penetration of renewable energy and the growing trend of distributed power systems, the traditional centralized power supply model is facing significant changes. The large-scale integration of distributed generation (DG) enables distribution networks to exhibit multi-source collaborative operation characteristics, but it also brings technical challenges such as voltage fluctuations and power flow reversal. Especially in the event of extreme natural disasters or main grid failures, the distribution network may be forced into islanded operation. In such cases, without a reasonable power source layout and coordinated control strategy, power supply reliability will significantly decrease.
[0003] Current methods on the market lack standardized processing of collected data, and outliers and differences in dimensions affect planning accuracy. Furthermore, topology modeling is mostly static, failing to capture real-time dynamic changes in the distribution network topology, and node and load priority assignment lacks standardized guidelines. Islanding assignment often relies on experience or simple algorithms, without combining breadth-first search and mixed-integer linear programming for scientific optimization. This results in either incomplete constraint considerations or an inability to reasonably match distributed generation output with load demand, making it difficult to prioritize power supply to critical loads and potentially leading to redundant islands and excessive network losses. In addition, the selection of candidate nodes for distributed generation is not targeted enough, capacity allocation does not consider multiple objectives such as economy, reliability, and environmental protection, energy storage regulation methods are simplistic and unable to smooth out output fluctuations, and there is a lack of multi-scenario simulation verification. The smoothness of mode switching and operational reliability cannot be effectively guaranteed, resulting in poor overall coordination and difficulty in adapting to the operational needs of complex distribution networks. Summary of the Invention
[0004] To address the shortcomings of existing technologies, such as insufficient scientific rigor in islanding, poor matching between distributed power sources and loads, and difficulty in ensuring operational stability across multiple scenarios, this invention provides a distributed power source collaborative planning method and system that considers islanding. The specific technical solution is as follows: A distributed generation collaborative planning method considering islanding includes the following steps: S1. Collect core data of the distribution network, preprocess the core data to generate a standardized dataset; S2. Based on the node-branch association matrix and the standardized dataset, construct a distribution network topology model, and classify node types and label load priorities according to node specifications; S3. Based on operational requirements, construct constraints including power balance, operating parameters, topology, and load priorities; S4. Using the distributed generation grid connection point as the root node, use breadth-first search to traverse and expand load nodes, verify the constraints point by point, and obtain the initial island boundary; S5. Use mixed-integer linear programming to optimize the number of islands and achieve optimal source-load matching by merging shared load islands and adjusting boundaries; S6. Based on the optimized island partitioning and resource data, screen candidate nodes for photovoltaic, wind power, and energy storage to form a list of candidate nodes for distributed generation; S7. Based on load time-series data, optimize the ratio of photovoltaic, wind power, and energy storage with multiple objectives of economy, reliability, and environmental protection, and configure energy storage V / f control parameters; output the final planning scheme.
[0005] Preferably, step S2 includes the following sub-steps: constructing a binary association matrix with all nodes of the distribution network as rows and all branches as columns, where an element of 1 indicates that the node is directly connected to the corresponding branch, and an element of 0 indicates that it is not directly connected; at the same time, using the branch number as an index, recording the impedance, switch status and rated capacity parameters of each branch respectively, forming an association mapping with the column number of the association matrix; referring to the IEEE 33-node system specification, dividing the nodes into three categories: power supply nodes, load nodes and tie nodes, and assigning them unified numbers; and labeling the load nodes as first-level important loads, second-level relatively important loads and third-level ordinary loads according to the importance of electricity use.
[0006] Preferably, the constraints include power balance constraints, operating parameter constraints, topology constraints, and load priority constraints. The power balance constraint sets the total output of the distributed power sources within the island to meet the load demand within the island, and reserves 10%-20% of hot standby capacity. The operating parameter constraint sets the voltage range of the island operation to be 0.95-1.05 times the rated voltage, and the frequency range to be 50±0.5Hz. The topology constraint sets the island to consist of continuous feeder segments, maintaining a radial topology structure, without ring networks or isolated networks. The load priority constraint sets the power supply priority of important loads to be higher than that of ordinary loads, ensuring the continuity of power supply to important loads during the islanding process, and reserving adjustment capacity for uncontrollable loads.
[0007] Preferably, step S4 includes the following steps: taking all distributed power grid connection points in the planning area as independent root nodes for breadth-first search traversal, with each root node corresponding to a candidate starting point for an initial island; starting from the root node, obtaining the connection relationship between adjacent nodes based on the node-branch association matrix, and expanding layer by layer to adjacent load nodes in descending order of load priority; for each expanded node, including the node in the candidate island set, and verifying in real time whether the set simultaneously satisfies power balance constraints, operating parameter constraints, and topology connectivity constraints; if all constraints are satisfied, continuing to expand to the next layer of nodes; if any constraint is not satisfied, immediately terminating the expansion in that direction, excluding the node from the candidate set, and locking the boundary of the current candidate set as the stopping branch in that direction; after traversing all directions, determining the final candidate set as the initial island, and recording the total load, the type of distributed power source contained in the initial island, and its output potential.
[0008] Preferably, step S5 includes the following steps: Using node affiliation status, branch closure status, branch active and reactive power flow, distributed generation active and reactive power output, node voltage amplitude, and node severable active loads as decision variables, and using the minimum active power loss, the fewest switching operations, and the minimum load shedding penalty as the comprehensive objective function, a mixed-integer linear programming model is constructed. Constraints include power balance constraints, voltage safety constraints, radial topology constraints, distributed generation output limit constraints, hot standby constraints, and primary load uninterruption constraints. When solving the mixed-integer linear programming model, all initial islands are traversed, and load nodes shared by multiple islands are identified. If merging satisfies all constraints, they are merged into a larger island. Source-load matching verification is performed on each island. For islands with insufficient distributed generation output, [the system] directs [the process] to the [relevant] islands. The system expands adjacent low-load-density nodes that have been included in the candidate nodes of distributed generation (DG) but have not yet been assigned to any island. After the expanded node is connected to the new DG, the power balance verification can improve the source-load matching degree of the island to no less than 1.0. For islands with excess DG output, the system identifies DG nodes that can be stripped or reduced within the island. Priority is given to stripping the DG node that is farthest from the load center and has the highest unit transmission cost. Alternatively, the excess DG output can be redistributed to adjacent islands with insufficient output. The system verifies that the source-load matching degree of each island is maintained within the range of 1.0 to 1.25 after stripping or redistribution. After each boundary adjustment, the system calculates the active power loss and the number of switching operations and selects the best option. Finally, the system outputs the optimized number of islands and the precise boundary and source-load matching results of each island.
[0009] Preferably, step S6 includes the following sub-steps: using the optimized island boundaries as spatial screening units, retrieve data on solar radiation, average annual wind speed, available installation area, and load distribution for each island, and screen nodes based on the retrieved data: nodes with abundant solar resources, good lighting conditions, and low access costs are selected as photovoltaic power candidate nodes; nodes with average annual wind speeds reaching the start-up wind speed, good ventilation conditions, and far from obstacles are selected as wind power candidate nodes; nodes within each island that are adjacent to important load centers, have convenient grid access conditions, and sufficient land area are selected as lithium battery energy storage system candidate nodes; perform feasibility verification on the access capacity, line carrying capacity, and equipment installation conditions for each of the initially selected candidate nodes, and evaluate the matching degree between the candidate distributed power type and the island load time sequence characteristics, eliminating nodes that are not feasible to access or have poor synergistic effects, and finally determine the list of distributed power candidate nodes for each island and output it to the capacity configuration module.
[0010] Preferably, nodes are screened based on retrieved data, including: nodes with abundant sunlight resources, good lighting conditions and low access costs are screened as candidate nodes for photovoltaic power generation; nodes with an average annual wind speed that reaches the start-up wind speed, good ventilation conditions and far from obstacles are screened based on retrieved data as candidate nodes for wind power generation; nodes within each island that are close to important load centers, have convenient grid access conditions and sufficient land area are screened as candidate nodes for lithium battery energy storage systems.
[0011] Preferably, step S7 includes the following sub-steps: Based on the load time-series data of each island, statistically analyze the load peak, load average, and load fluctuation patterns at different times; combine the output characteristics of distributed power sources to obtain the initial capacity of each type of distributed power source; with the minimum investment cost and operating cost of distributed power source planning as the economic objective, the shortest power outage time of the island as the reliability objective, and the lowest carbon emissions as the environmental objective, adjust the capacity ratio of photovoltaic, wind power, and energy storage in each island; the optimized energy storage adopts voltage / frequency dual closed-loop V / f control in island mode, dynamically suppresses wind and solar fluctuations in charge and discharge levels, and introduces droop control to achieve power sharing among multiple energy storage systems; output the planning scheme, which includes the optimized boundaries of each island, the installed capacity of photovoltaic, wind power, and energy storage in each island, and V / f control parameters.
[0012] Preferably, the method also includes step S8, which involves building a simulation platform to simulate three scenarios: normal operation, main network failure, and main network recovery. The platform monitors the operating parameters and collaborative effects in real time and verifies the smoothness of mode switching. If the verification fails, the process is returned for adjustment.
[0013] This invention also provides a distributed power source collaborative planning system considering islanding, applied to the aforementioned distributed power source collaborative planning method considering islanding, comprising: a data acquisition and preprocessing module for acquiring core data of the distribution network, preprocessing the core data to generate a standardized dataset; and a topology modeling and dynamic update module for constructing a distribution network topology model based on the node-branch correlation matrix and the standardized dataset, and referring to EEE... The system includes three modules: a 33-node specification module to classify node types and label load priorities; a constraint construction module to construct constraints including power balance, operating parameters, topology, and load priorities based on operational requirements; an initial islanding module to expand load nodes using breadth-first search with the distributed power grid connection point as the root node, verifying constraints point by point to obtain the initial island boundary; an island optimization and adjustment module to optimize the number of islands and achieve optimal source-load matching by merging shared load islands and adjusting boundaries using mixed-integer linear programming; a candidate node selection module to select candidate nodes for photovoltaic, wind power, and energy storage based on the optimized island partitioning and resource data, forming a list of candidate nodes for distributed power sources; a capacity coordination configuration module to optimize the ratio of photovoltaic, wind power, and energy storage based on load time-series data with multiple objectives of economy, reliability, and environmental protection, and to configure energy storage V / f control parameters; and an output of the final planning scheme. A simulation verification and closed-loop adjustment module is used to build a simulation platform to simulate three scenarios: normal operation, main grid failure, and main grid recovery, monitoring operating parameters and coordination effects in real time, and verifying the smoothness of mode switching. If the verification fails, the system returns to adjustment.
[0014] Compared with existing technologies, the beneficial effects of this invention are as follows: Data is collected collaboratively by multiple devices and processed through standardization, providing accurate data support for planning; a topology model is constructed by combining node-branch correlation matrices, enabling real-time capture of dynamic topology changes; node and load priorities are assigned according to specifications, ensuring the scientific nature of the planning. Breadth-first search and mixed-integer linear programming are used to optimize island partitioning, satisfying multiple constraints while matching output and load, merging shared load islands, and prioritizing power supply to critical loads. Targeted selection of candidate nodes, multi-objective optimization of capacity allocation, and energy storage V / f control to smooth fluctuations achieve optimal economic, reliable, and environmentally friendly multi-objectives; finally, simulation verification and dynamic adjustment in three scenarios ensure the feasibility of the method and the smoothness of mode switching, effectively solving problems such as island adaptation and output fluctuations in distributed power source collaborative operation, and improving the fault resistance and power supply continuity of the distribution network. Attached Figure Description
[0015] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.
[0016] Figure 1 This is a flowchart of the distributed power supply collaborative planning method considering island partitioning in Example 1. Detailed Implementation
[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0018] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.
[0019] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.
[0020] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.
[0021] Example 1 Please refer to Figure 1 , Figure 1 This is a flowchart illustrating a distributed power source collaborative planning method considering islanding. This embodiment provides a distributed power source collaborative planning method considering islanding, which includes the following steps: S1. Collect multi-source heterogeneous data from the power distribution network, preprocess the multi-source heterogeneous data, and generate a standardized dataset.
[0022] Specifically, distribution network automation systems, feeder terminal units, smart meters, and meteorological monitoring equipment are used to collect core data on distribution network nodes, branches, loads, and distributed power sources. In this example, a distribution network automation (DA) system, feeder terminal units (FTUs), smart meters, and meteorological monitoring equipment are used to comprehensively collect core data on the distribution network within the planned area.
[0023] Furthermore, the core data of the distribution network includes distribution network node parameters, branch impedance, line rated capacity, load type and load timing data of all nodes, resource data and output characteristic data of distributed generation, and voltage and frequency constraint data of distribution network operation.
[0024] After collecting core data from the distribution network, the data undergoes preprocessing. Preprocessing includes: using a moving average method to identify outliers in the collected data, and then generating a standardized base dataset through normalization.
[0025] Specifically, the sliding window duration is set to 1 hour, with 12 consecutive data points collected at a frequency of one point every 5 minutes. Outlier detection uses the "three-standard-deviation criterion." Within each sliding window, the mean μ and standard deviation σ of the data within the window are calculated. If the data value at the center point of the window exceeds the range of μ ± 3σ, it is considered an outlier. Physical boundary checks are also used to ensure comprehensive and accurate detection. For data points identified as outliers, linear interpolation is used to fill in the gaps after removal. This involves using the mean of valid data in adjacent windows before and after the outlier to calculate a replacement value according to the time ratio, ensuring the continuity and integrity of the data sequence and avoiding data loss due to direct deletion that could affect the accuracy of subsequent modeling. This moving average method is executed point-by-point along the entire time-series dataset until all data points have undergone anomaly detection and repair.
[0026] The data with different dimensions after filling are uniformly mapped to a standardized range (e.g., voltage, power, and impedance are all converted to per-unit values) to eliminate the impact of dimensional differences on the accuracy of subsequent modeling. Then, the data is archived according to the categories of "topology data - load data - distributed power source data - constraint data" to form a standardized dataset. At the same time, a data interface conforming to the IEC 61850 standard is reserved to support subsequent real-time data updates, historical data queries, and dynamic adjustments, providing an accurate and complete dataset for subsequent islanding, topology modeling, and distributed power source planning.
[0027] S2. Based on the node-branch correlation matrix and combined with the standardized dataset, construct the distribution network topology model, and classify the node types and label the load priorities in accordance with the EEE 33 node specification.
[0028] Specifically, step S2 includes the following sub-steps: S21. Construct a binary association matrix with all nodes of the distribution network as rows and all branches as columns. An element of 1 indicates that the node is directly connected to the corresponding branch, and an element of 0 indicates that it is not directly connected. At the same time, the impedance, switch status and rated capacity parameters of each branch are recorded with the branch number as the index, and an association mapping is formed with the column number of the association matrix.
[0029] Specifically, a node-branch association matrix modeling technique is used to construct the association matrix of the distribution network in the planning area. The number of rows in the matrix corresponds to the number of all nodes in the distribution network, and the number of columns corresponds to the number of all branches. The rule for the matrix element values is as follows: if a node is directly connected to a branch, the corresponding element has a value of 1; if they are not directly connected, the corresponding element has a value of 0.
[0030] Simultaneously, using the branch number as an index, a corresponding attribute record is established for each branch, storing its impedance value (including resistance and reactance), current switch status (1 indicates closed, 0 indicates open), and rated capacity. A one-to-one correspondence is formed between the branch attribute record and the column number of the association matrix—the j-th column of the association matrix represents branch j, and the attribute record similarly stores all its physical parameters using branch j as an index; the two are associated and mapped through the branch number. When complete information about a branch is needed, the branch's topological connection relationship is read from the association matrix, and the corresponding electrical parameters are read from the attribute record using the same branch number. Together, these constitute a complete data description of the branch.
[0031] S22. Referring to the node system specification, nodes are divided into three categories: power supply nodes, load nodes, and tie nodes, and are uniformly numbered; load nodes are labeled as Level 1 important loads, Level 2 relatively important loads, and Level 3 ordinary loads according to the importance of electricity use.
[0032] In this step, referencing the topology construction specifications of the IEEE 33-node standard distribution network test system, a complete distribution network topology model is built with the node-branch association matrix as the core. Visual modeling technology is used to visualize nodes, branches, switches, and other equipment according to their actual geographical locations and electrical connections.
[0033] Based on functional characteristics, distribution network nodes are divided into three categories: The specific classification criteria are as follows: Power Node: A node connected to distributed generation systems (including photovoltaic power generation systems, wind power generation systems, and energy storage systems), possessing the capability to inject active and reactive power. For each power node, record its connected DG type, rated capacity, and current adjustable output range; Load node: A node that consumes only electrical energy, further subdivided into three categories based on the nature of electricity consumption: residential load, commercial load, and industrial load. For each load node, its subcategory and corresponding 24-hour time-series load curve are recorded; Interconnecting nodes: Intermediate nodes that are neither connected to power nor have a load. They are mainly used to connect other nodes. Typical examples include the node where the sectionalizing switch is located and the line T-connection.
[0034] Then, each type of node is uniformly numbered and labeled.
[0035] Based on the requirements for safe operation of the distribution network and the importance of each load node in actual electricity consumption, the priority of each load node is marked and divided into three levels: Level 1 important loads are marked with the highest priority, Level 2 important loads with medium priority, and Level 3 ordinary loads with low priority, thus clarifying the load protection order when islanding. The load priority marking results are stored as tags in the attribute fields of each load node, and are used as the sorting basis for the expansion order during subsequent breadth-first search expansion, that is, nodes with higher priority are expanded into the island candidate set more quickly.
[0036] Furthermore, by employing distribution network situational awareness technology, a real-time monitoring module is embedded in the topology model to access real-time data from feeder terminal units (FTUs) and distribution network automation systems (DAs), thereby capturing dynamic topology changes such as branch switch status changes, node load fluctuations, and equipment failures in real time.
[0037] Specifically, a real-time data interface is embedded in the topology model to receive three types of signals from the feeder terminal units (FTUs) and the distribution automation system (DA): Switch status change signal: When any switchgear undergoes a tripping or closing operation, the FTU uploads the switch number, the changed status, and the operation timestamp. Protection action signal: When a short circuit fault occurs on the line and causes the relay protection device to operate, the fault branch number, fault type and operation time are uploaded; Equipment fault alarm information: When equipment such as transformers and lines experience abnormal operating conditions, upload the fault equipment number, abnormality type, and alarm level.
[0038] When a change in the state of any switch is detected, the system automatically locates the branch number corresponding to that switch and modifies the corresponding elements in the node-branch association matrix in real time according to the switch's on / off state. If the switch changes from closed to open, all non-zero elements corresponding to that branch are set to 0; if the switch changes from open to closed, the connection relationship corresponding to that branch is restored to 1. The connection attributes of that branch in the topology model are updated synchronously.
[0039] Subsequently, an incremental topology verification algorithm is employed, starting from the changeover switch and traversing bidirectionally to adjacent nodes to quickly recalculate the connectivity partitions of the current distribution network without requiring a full network topology reconstruction, thus completing the model update within milliseconds. The core steps of the incremental topology verification algorithm are as follows: 1. Determine the column index j of the change switch in the correlation matrix and the two nodes i1 and i2 connected to it; 2. If the switch changes from closed to open, start from node i1 and traverse all reachable node sets Ω1 along the non-disconnected branch, and start from node i2 and traverse all reachable node sets Ω2 along the non-disconnected branch. If Ω1 and Ω2 originally belonged to the same connected partition, then the switch opening operation splits the partition into two independent partitions. 3. If the switch changes from open to closed, start from nodes i1 and i2 respectively and traverse their respective current connected partitions, merging the two partitions into one large partition; 4. Update the connectivity partition status table of the distribution network without performing a full topology refactoring of the entire network.
[0040] The incremental topology verification algorithm only involves traversing and calculating a local area centered on the change switch. Its complexity is proportional to the local range of change, rather than to the size of the entire network. This allows the model to be refreshed in milliseconds, meeting the time requirements for real-time operation monitoring of the distribution network.
[0041] Finally, a version comparison mechanism is used to compare the updated correlation matrix with the matrix from the previous time step to verify the correctness of the update logic. The comparison includes three dimensions: the change in the number of non-zero elements in the matrix, the change in the number of nodes within each connected partition, and the change in the partition where the power supply node is located. If the comparison result shows that the change caused by the switch is consistent with the traversal calculation result, the update is confirmed to be correct. After confirmation, the new model is pushed to the islanding and collaborative optimization module to ensure that subsequent planning is based on the latest actual operating status of the distribution network and to avoid strategy deviations caused by lagging topology information.
[0042] S3. Based on operational requirements, construct constraints including power balance, operating parameters, topology, and load priority.
[0043] Based on the requirements for distribution network operation safety and the collaborative operation needs of distributed generation, four core constraints for islanding are identified. These constraints include power balance constraints, operating parameter constraints, topology constraints, and load priority constraints. Among them, the power balance constraint is the fundamental survival condition for independent island operation. Its core requirement is that the total available output of all distributed generation sources within any island must meet the electricity demand of all loads within that island, with an additional proportion of hot reserve capacity reserved to cope with uncertainties such as random load fluctuations or sudden drops in distributed generation output.
[0044] Specifically, the preprocessed load time-series data and distributed power generation output data are called up to statistically analyze the load peak, average and fluctuation range for each time period within the island. Combined with the output characteristics of distributed power generation, the core requirements of power balance constraints are determined. It is clarified that the total output of distributed power generation within the island must not be less than the total load demand within the island, and 10%-20% of the total load demand within the island must be reserved for hot standby capacity, with 20% reserved for important load areas and 10% reserved for ordinary load areas.
[0045] The total available output of distributed power sources within the island must be at least 1.1 times that of ordinary load areas or 1.2 times that of critical load areas. The total active power demand within the island for each time period is obtained based on load time-series data. The reserved ratio k is determined based on the proportion of important loads within the isolated island, and the minimum required total DG output is: The specific value for hot standby capacity is as follows: This capacity can be provided by energy storage systems, adjustable distributed generation (DG), or by cutting off some interruptible loads.
[0046] During the capacity configuration phase, the installed capacity of each DG is determined through multi-objective optimization, so that the actual adjustable output can meet the minimum output requirements under typical operating conditions. In the real-time control of islanded operation, the reserve level is dynamically maintained through energy storage V / f control and DG output adjustment to ensure that the system can still operate stably when the load fluctuates or the DG output fluctuates.
[0047] Operating parameter constraints are quality assurance conditions for islanded independent operation. The core requirements are: when the islanded system is operating independently from the main grid, the voltage and frequency must be maintained within the allowable range to ensure the normal operation of various electrical equipment and avoid equipment damage or malfunction of protection devices due to power quality exceeding the limits.
[0048] The specific constraints on operating parameters are determined as follows: referring to the distribution network operation standards and combining the characteristics of islanded independent operation, the constraint ranges for voltage and frequency are determined; the voltage constraint is set to 0.95-1.05 times the rated voltage, with the voltage deviation of important load nodes controlled within ±3% and the voltage deviation of ordinary load nodes controlled within ±5%; the frequency constraint is set to 50±0.5Hz, with instantaneous fluctuations not exceeding ±1Hz and durations not exceeding 2 seconds.
[0049] Topological constraints are the structural conditions for islanding. The core requirement is that the physical connection structure of the islands must meet the basic rules of the radial operation of the distribution network, ensure that the power flow direction is clear and controllable, and avoid circulating currents that cause line overload and protection malfunctions.
[0050] Specifically, an island must consist of continuous feeder segments, with all internal nodes interconnected via branches, and no isolated nodes without a path to the power source within the island (i.e., "islands within islands"). Connectivity verification is performed using a node-branch association matrix. For any candidate island set Ω, starting from any DG grid-connected point within the island, the path corresponding to the non-zero element in the association matrix is traversed. If all nodes are reachable from the DG grid-connected point, connectivity is satisfied; otherwise, unreachable nodes are excluded from the candidate set.
[0051] Isolated networks must strictly maintain a radial topology and are prohibited from forming ring networks (i.e., there are two or more different paths between any two nodes). Furthermore, no "isolated networks" (i.e., passive subnets) that are independent of distributed power supply support are allowed within an isolated network.
[0052] When the main network topology changes (e.g., fault isolation causes some branches to disconnect), the island boundaries must be dynamically adjusted to ensure that the new islands also meet the aforementioned connectivity and radial requirements. Dynamic constraints achieve millisecond-level response through the aforementioned incremental topology verification algorithm.
[0053] Load priority constraints are the strategy-oriented conditions for islanding. The core requirement is that when the islanding power supply is limited, power supply must be guaranteed in order of the importance of the load, and the limited output of distributed power sources should be accurately allocated to the most critical users to maximize the social benefits of islanding operation.
[0054] Based on user electricity consumption characteristics and social impact, load nodes are divided into primary important loads, secondary relatively important loads, and tertiary ordinary loads. It is clear that the power supply priority of important loads is higher than that of ordinary loads, ensuring the power supply continuity of important loads is prioritized during the islanding process, and reserving adjustment capacity for uncontrollable loads.
[0055] S4. Using the distributed power source grid connection point as the root node, a breadth-first search is used to traverse and expand the load nodes, verifying the constraints point by point to obtain the initial island boundary. In this embodiment, the breadth-first search topology traversal technique is used to expand the load nodes with the distributed power source grid connection point as the root node, verifying the constraints point by point to obtain the island boundary, specifically including steps S41 to S44: S41. Take all distributed power grid connection points in the planning area as independent root nodes for breadth-first search traversal, and each root node corresponds to a candidate starting point for an initial island.
[0056] Specifically, the process iterates through the power node records in the standardized dataset, filtering out the node numbers of all nodes connected to distributed power sources, including photovoltaic power generation systems, wind power generation systems, and energy storage systems. If a node is connected to multiple distributed power sources simultaneously, it is still considered a single root node and is not counted repeatedly. A separate traversal operation is performed on each root node. First, the root node is included in the initial island candidate set and marked as a traversed node. Simultaneously, the output data of the distributed power source corresponding to the root node is collected, including rated capacity, currently available output, and output adjustment range. Connection information of surrounding nodes of the root node is also collected. The power balance verification parameters of the candidate set are initialized, including the total distributed power output (initial value equal to the total DG output of the root node) and the total load.
[0057] The root node itself remains powered by default during subsequent checks, does not participate in load shedding, and serves as the starting reference point for island topology connectivity checks. The traversal process of each root node is independent and does not affect each other, and can be executed in parallel to improve computational efficiency.
[0058] S42. Starting from the root node, obtain the connection relationship between adjacent nodes based on the node-branch association matrix, and expand to the adjacent load nodes layer by layer in order of load priority from high to low.
[0059] Create a FIFO (First-In-First-Out) queue and enqueue the root node. Create a set of visited nodes and mark the root node as visited. Create a set of candidate isolated nodes Ω, initially containing only the root node.
[0060] Remove the head node from the queue and query all its neighboring nodes (i.e., other nodes directly connected to it via closed branches) based on the node-branch association matrix. For each unvisited neighboring node, process them in the following priority order: First priority – load level: prioritize expanding nodes with higher load priority. Level 1 critical load nodes take precedence over Level 2 relatively important load nodes, and Level 2 takes precedence over Level 3 ordinary load nodes.
[0061] Second priority – electrical distance: For load nodes of the same priority, they are sorted from closest to furthest from the root node by their electrical distance (cumulative branch impedance), with the closest nodes being expanded first. Tie nodes (without load) are expanded after all load nodes.
[0062] For each adjacent node processed, it is temporarily added to the candidate set Ω, and constraints are immediately validated in real time. If the validation passes, the node is officially added to the candidate set, marked as traversed, and enqueued to await further expansion of its adjacent nodes. If the validation fails, the node is removed from the candidate set, not added to the queue, and further expansion in that direction ceases (i.e., the adjacent nodes of that node are no longer traversed). After processing all nodes in the current layer, the next layer of nodes in the queue is processed until the queue is empty (all expandable nodes have been processed) or all directions are terminated.
[0063] S43. For each node expanded, the node is added to the candidate island set, and the set is checked in real time to see if it simultaneously satisfies the power balance constraint, operating parameter constraint, and topology connectivity constraint. If all constraints are satisfied, the expansion continues to the next level of nodes. If any constraint is not satisfied, the expansion in that direction is terminated immediately, the node is excluded from the candidate set, and the boundary of the current candidate set is locked as the stopping branch in that direction.
[0064] Specifically, the verification content is as follows: 1. Power balance verification: Calculate the total output of all distributed power sources and the total demand of all loads in the current candidate set, and determine whether the condition "total output ≥ load demand + 10%-20% hot standby capacity" is met.
[0065] 2. Verify operating parameters: Simulate the voltage and frequency of the candidate set when it is running independently, and determine whether it meets the constraints of 0.95-1.05 times the rated voltage and 50±0.5Hz; 3. Verify topological connectivity: Ensure that nodes in the candidate set are continuously connected through branches, with no isolated nodes. Specifically, starting from the root node in the candidate set, traverse along the closed branches corresponding to the non-zero elements in the incidence matrix and count the number of reachable nodes. If the number of reachable nodes equals the total number of nodes in the candidate set, connectivity is satisfied; if there are unreachable nodes, remove them from the candidate set.
[0066] If all three constraints are satisfied, the node passes the verification and is formally included in the candidate island set, and expansion continues towards its neighboring nodes. If any constraint is not satisfied, further expansion in that direction is immediately terminated, the node is removed from the candidate set, and the boundary of the current candidate set is determined as the stopping branch in that direction. The boundary determination method is as follows: traverse all nodes in the candidate set, check the neighboring branches of each node, and if the opposite node of a branch is an excluded node, then mark that branch as a stopping branch.
[0067] S44. After traversing all directions, the final candidate set is determined as the initial island, and the total load, the type of distributed power source contained in the initial island, and the output potential are recorded.
[0068] Specifically, when expansion stops in all directions or the queue is empty, the boundary of the current candidate set is determined as the initial island boundary. The boundary locking method is as follows: traverse all nodes in the candidate set, check the adjacent branches of each node, and if the opposite node of a branch is not in the candidate set, then mark the branch as the island boundary branch. By disconnecting all boundary branches, the candidate set can be isolated from the distribution network to form an independent island power supply area.
[0069] Record the following information for each initial island to form an initial island information table: initial island number and set of all node numbers contained therein, set of island boundary branch numbers (all marked disconnected branches), total load within the island, type of distributed power source connected to the island (PV / Wind / Energy Storage) and output potential (rated capacity, maximum / minimum output, current available output).
[0070] S5. Using mixed-integer linear programming, the number of islands is optimized and the source load is optimally matched by merging shared load islands and adjusting the boundaries.
[0071] Specifically, step S5 includes the following sub-steps: S51. Using node affiliation status, branch closure status, branch active and reactive power flow, distributed generation active and reactive power output, node voltage amplitude, and node severable active loads as decision variables, and taking the minimum active power loss, the minimum number of switching operations, and the minimum load shedding penalty as the comprehensive objective function, a mixed integer linear programming model is constructed. The constraints include power balance constraints, voltage safety constraints, radial topology constraints, distributed generation output limit constraints, hot standby constraints, and primary load uninterruption constraints.
[0072] Specifically, mixed-integer linear programming models include the following six types of decision variables: 1. Node ownership state variable : Indicates whether node i is included in a certain island, 1 for inclusion, 0 for non-inclusion; 2. Branch closed-loop state variables : These represent the active power and reactive power transmitted on branch (i,j), respectively; 3. Continuous variables The side road has merit, and the trend is ineffective; 4. Continuous variables : Active and reactive power output of the distributed power source at node i; 5. Continuous variables Active load that can be removed from node i (0≤ ≤ ); 6. Continuous variables Node voltage amplitude (per unit).
[0073] The overall objective function focuses on minimizing active power loss, minimizing the number of switching operations, and minimizing load shedding penalties. The specific expression is as follows: The first term is the active power loss, which is approximated by linearization of the quadratic term. The resistance of branch (i,j); The set represents all branches; the second term represents the number of switch operations. This represents the initial state of the branch, i.e., the initial boundary state of the isolated island. A value of 1 indicates that the state of the branch has changed, and a value of 0 indicates that the state remains unchanged; the third term is the load shedding penalty, and α, β, γ are weighting coefficients; This is the set of all nodes.
[0074] Furthermore, γ is typically the largest, as ensuring power supply continuity is the bottom line for safety in islanded distribution network operation. It is usually set that γ > α and γ > β to ensure that the model will not disconnect critical loads under any circumstances in order to reduce network losses or minimize switch operations. α and β are set according to actual operational preferences. If the power grid company prioritizes economic operation, α is chosen to be larger; if equipment life-cycle costs are high, β is chosen to be larger.
[0075] The importance of the three indicators "minimum network loss", "minimum number of switching", and "minimum load shedding" is compared pairwise using the analytic hierarchy process (AHP). A judgment matrix is constructed, the eigenvector corresponding to the largest eigenvalue is calculated, and a consistency check is performed. Once the check is passed, the normalized weight values of α, β, and γ can be obtained.
[0076] Furthermore, the mixed-integer linear programming model includes six types of constraints: power balance constraints, voltage safety constraints, radial topology constraints, distributed power output limit constraints, hot standby constraints, and uninterrupted primary load constraints.
[0077] 1. Power balance constraint: For any node i, the injected power and the outflow power must satisfy Kirchhoff's current law: in, The trend of contributing to the side roads Provide the active power output of the distributed power source for node i. Let i be the original active load of node i. The active load that can be removed from node i (0 ≤ ≤ ).
[0078] 2. Voltage constraint: For any node i, the voltage amplitude must be maintained within the allowable range. in, This represents the voltage amplitude at node i.
[0079] 3. Topological connectivity and radial orientation: Each island must maintain a radial topology and must not form a ring network. Furthermore, each island must contain at least one distributed power node. in, This represents the total number of isolated islands. This constraint ensures that the number of closed branches equals the total number of included nodes minus the total number of isolated islands.
[0080] Simultaneously, single-item flow auxiliary constraints or spanning tree constraints are used to eliminate loops, ensuring that only one path exists between any two nodes. The logical relationship between the number of closed branches and node affiliation is as follows: .
[0081] 4. Output limit constraints for distributed power sources The active and reactive power outputs of each node's distributed power source must be within their allowable ranges: in, Let be the minimum active power output of the distributed power source at node i. The maximum active power output of the distributed power source at node i; Let be the minimum reactive power output of the distributed power source at node i. This represents the maximum reactive power output of the distributed power source at node i.
[0082] 5. Hot standby constraints For each isolated island The total output of distributed power sources must be no less than 1.1 or 1.2 times the load demand. .
[0083] 6. Load Priority For any load point i at any level, its associated state variable Forced to 1, load shedding Forced to 0: =0, =1, ∈ .
[0084] S52. When solving the mixed-integer linear programming model, traverse all initial islands, identify load nodes shared by multiple islands, and merge them into a larger island if all constraints are satisfied after merging.
[0085] Specifically, topology correlation analysis is used to traverse the node sets of all initial islands and identify shared load points among multiple islands, i.e., cases where the same load node is covered by two or more initial islands. The specific method is as follows: construct the node sets of each island, calculate the intersection of each pair of sets, and if a load node exists within a non-empty intersection, then that load node is marked as a shared load node.
[0086] For island pairs with shared load nodes, the feasibility of merging them is assessed. Specifically, the total output of distributed generation and total load demand of the merged island are calculated. Then, it is verified whether the merged island meets power balance constraints and operating parameter constraints. If so, they are merged into a larger island. During merging, priority is given to retaining the core areas of islands with high load priority and sufficient distributed generation output, while redundant branch disconnection points are removed to reduce the cost of switching equipment and control complexity. The merged island serves as input to the optimization model and continues to participate in subsequent boundary adjustment optimizations.
[0087] S53. Perform source-load matching verification on each island. For islands with insufficient distributed generation output, expand to adjacent low-load-density nodes that have been included in the candidate nodes of distributed generation but have not yet been assigned to any island. After the expanded node is connected to the new distributed generation, the source-load matching degree of the island can be improved to no less than 1.0 after power balance verification. For islands with excess distributed generation output, identify the distributed generation nodes that can be stripped or reduced within the island. Prioritize stripping the DG node that is farthest from the load center and has the highest unit transmission cost, or redistribute the excess distributed generation output to adjacent islands with insufficient output. Verify that the source-load matching degree of each island is maintained within the range of 1.0 to 1.25 after stripping or redistribution. Calculate the active power loss and the number of switching operations after each boundary adjustment and select the best option. Finally, output the optimized number of islands and the accurate boundary and source-load matching results of each island.
[0088] Specifically, perform the following boundary adjustments and optimizations: For each isolated island, calculate the source-load matching index: in, Let i be the active power output of the distributed power source at node i. Let i be the original active load of node i. Let be the active load that can be removed from node i.
[0089] like ≥1 and ( If -1)×100% exceeds a preset threshold (e.g., 20%), then it is determined that force is excessive; if If 1 < 1, then the output is insufficient. If 1 ≤ <1.2 or 1≤ If the value is less than 1.25, the match is considered successful.
[0090] For isolated nodes with insufficient power output, expansion is initiated by adding adjacent nodes with low load density that meet the constraints. Specifically, all adjacent external nodes of the insufficient island are identified, prioritizing nodes already listed in the distributed generation candidate node list. External nodes are then attempted to be added one by one, in order of load priority from low to high and electrical distance from near to far. For each added node, the power balance constraints, hot standby constraints, and operating parameter constraints are re-verified. If the source-load matching degree of the island improves to at least 1.0 and all constraints are met after adding the node, the expansion continues to the next node. This process continues until all constraints are met. ≥1 and sufficient hot standby. If extending to a node causes constraint failure or the source load matching degree cannot be improved to 1.0, the extension is terminated and the node is excluded.
[0091] For islanded systems with excess output, identify distributed generation (DG) nodes within the excess island that can be stripped or reduced. In order of increasing electrical distance from the load center and decreasing unit transmission cost, attempt to strip or reduce the available output of each DG node; or redistribute the identified excess DG output to adjacent islands with insufficient output. After stripping or redistribution, re-verify the power balance constraints, hot reserve constraints, and operating parameter constraints of the original island and adjacent islands to ensure that the source-load matching degree ηk of each island remains within the range of 1.0 to 1.25. If the verification passes, execute the stripping or redistribution operation, and record the corresponding switching operations or capacity adjustment schemes.
[0092] After each boundary adjustment, the cost metrics are recalculated. Active power loss: The adjusted active power loss is calculated using the DistFlow branch power flow model. The calculation formula is as follows: in, For the total active power loss of the system, , For the transmission of active and reactive power in branch ij, Let i be the voltage amplitude at node i. Let be the resistor of branch ij. The adjustment scheme must reduce network loss by more than 5% compared to before the adjustment.
[0093] During the boundary adjustment process, a switching operation cost analysis method is adopted to count the number of switching operations required for each island boundary adjustment, and priority is given to the adjustment scheme with fewer switching operations and lower network loss. Through comparison of multiple schemes, the scheme with the largest reduction in network loss and the fewest switching operations is selected as the final boundary adjustment result.
[0094] S6. Based on the optimized island partitioning and combined with resource data, candidate nodes for photovoltaic, wind power, and energy storage are selected to form a list of candidate nodes for distributed power sources.
[0095] Specifically, step S6 includes the following sub-steps: S61. Using the optimized boundaries of each island as the spatial filtering unit, retrieve the solar radiation, average annual wind speed, available installation area, and load distribution data for each island, and filter the nodes based on the retrieved data: 1. Nodes with abundant solar resources, good lighting conditions, and low access costs are selected as candidate nodes for photovoltaic power generation.
[0096] Specifically, retrieve the solar radiation data (annual average total radiation, in kWh / m²) and available installation area data (roof area, vacant land area, in m²) for each node within the isolated island, and screen nodes that simultaneously meet the following conditions as preliminary candidate nodes for photovoltaic power generation: The annual total radiation should not be lower than the local economically feasible threshold for photovoltaic power generation (usually ≥1200 kWh / m² / year), and there should be no long-term obstructions to ensure good lighting conditions; the available roof or ground installation area of the node should not be lower than the minimum requirement for photovoltaic module layout to ensure sufficient installed capacity potential; the electrical distance between the node and the existing grid connection point in the island should be short, and the cost of building new lines should be low. Nodes with an electrical distance of ≤3 branch hops from the existing grid connection point are usually preferred.
[0097] For nodes that meet the above conditions, the potential installed capacity is estimated by "annual power generation per unit area × available area". Nodes with an estimated value of ≥50kW are marked as preliminary photovoltaic candidate nodes.
[0098] 4. Nodes are selected based on the retrieved data. Nodes with an annual average wind speed that meets the start-up wind speed, good ventilation, and are far from obstacles are selected as candidate nodes for wind power generation.
[0099] Specifically, the annual average wind speed should not be lower than the wind turbine start-up speed (usually ≥5m / s), and the turbulence intensity should be low and the wind direction stable to ensure that wind energy resources are exploitable; the node location should have good ventilation, with no obstacles such as mountains or dense building clusters blocking the airflow, and the horizontal distance from the obstacle should not be less than 5 times the height of the obstacle (i.e., the 5H principle); it should be far away from densely populated areas (distance ≥300m) to reduce noise impact, and should not be located in restricted development areas such as ecological protection zones.
[0100] Nodes that meet the above conditions and have an estimated annual equivalent full-load operating hours of ≥2000 hours are marked as preliminary candidate nodes for wind power.
[0101] 5. Nodes within each isolated island that are close to important load centers, have convenient grid access, and have sufficient land area are selected as candidate nodes for lithium battery energy storage systems.
[0102] Specifically, the nodes have sufficient spare capacity or room for modification, and the access voltage level is matched (usually 380V / 10kV), with low access costs; there is enough available land or indoor space near the nodes to install energy storage cabinets, usually requiring an available area of ≥200m², to meet the fire safety distance and safety isolation requirements of lithium battery systems; the nodes are closest to the weighted electrical distance of the primary and secondary important load centers in the island to ensure that the power supply of high-priority loads can be prioritized when energy storage discharges.
[0103] At least 1-2 preliminary candidate nodes for energy storage should be selected for each island, located at the electrical center of the island or near the point of maximum load.
[0104] S62. Perform feasibility verification on the access capacity, line carrying capacity and equipment installation conditions of all candidate nodes in the initial selection, evaluate the matching degree between the candidate distributed power source type and the timing characteristics of the island load, eliminate nodes that are not feasible to access or have poor coordination effect, and finally determine the list of distributed power source candidate nodes for each island and output it to the capacity configuration module.
[0105] Specifically, for each preliminary candidate node, three feasibility checks are performed: Access capacity verification checks the maximum accessible capacity at the location of the node (determined by transformer capacity, rated current of switching equipment, and thermal stability limit of feeder). If this value is less than 1.2 times the rated capacity of the equipment to be installed (with a safety margin), it is deemed unacceptable.
[0106] For line carrying capacity verification, retrieve the current load rate and rated current carrying capacity of the feeder or transformer where the node is located, calculate the maximum transmission power of the line after the addition of the new DG, and verify whether it meets the following condition: maximum transmission power after addition ≤ line rated current carrying capacity × 0.85. If it exceeds this limit, it is judged as failing.
[0107] Equipment installation, adjustment, and verification are conducted on-site or based on GIS data to assess the installation space, transportation channels, construction conditions, and environmental impact of candidate nodes. If there are insurmountable installation obstacles (such as insufficient space, poor geological conditions, or environmental restrictions), the application is deemed unsuccessful.
[0108] Furthermore, a collaborative effect evaluation method is adopted to analyze the matching degree between the distributed power supply type of the candidate nodes and the characteristics of the islanded load, and to eliminate candidate nodes with poor collaborative effect and excessive access cost; finally, the candidate nodes and types of distributed power supply for each island are determined, and the candidate node number, type, selection basis and core parameters are recorded in detail to provide accurate support for the collaborative configuration of distributed power supply capacity.
[0109] Specifically, the time-series output curves of candidate distributed generation (DG) sources are coupled and correlated with the time-series load curves of isolated power sources. The correlation coefficient method or complementary coefficient method is used to extract the daily peak and valley periods and volatility of the islanded load. The time-series overlap and power matching degree are evaluated against typical output characteristics such as high midday power generation for photovoltaics and high output at night or in winter for wind power. If the peak output of DG is significantly misaligned with the peak load, or if the direction of output fluctuation is opposite to the trend of load demand change, it is judged as having poor synergy and is eliminated. Simultaneously, the ability of energy storage to smooth and fill peak-valley differences in load is assessed. A comprehensive quantitative matching score is used to eliminate nodes with scores below the threshold or those requiring significant line modifications or with poor economic viability. Finally, high-quality candidate nodes with tight source-load time-series coupling are selected.
[0110] After two rounds of screening, the final candidate node list for each island was determined, specifically including: photovoltaic power candidate nodes and recommended installed capacity range (based on 70% to 90% of the upper limit of the access capacity), wind power candidate nodes and recommended installed capacity range, and lithium battery energy storage system candidate nodes and recommended power / energy capacity range.
[0111] The system records the number, type, selection criteria (feasibility verification passed items, matching degree score) and core parameters of each candidate node in detail, and outputs the list of candidate nodes for distributed power sources to the capacity configuration module, providing accurate location constraints and boundary conditions for subsequent capacity collaborative configuration.
[0112] S7. Based on load time series data, optimize the ratio of photovoltaic, wind power and energy storage with multiple objectives of economy, reliability and environmental protection, and configure the energy storage V / f control parameters; output the final planning scheme.
[0113] Specifically, step S7 includes the following sub-steps: S71. Based on the load time-series data of each island, statistically analyze the load peak, load average and load fluctuation patterns at different time periods, and combine the output characteristics of distributed power sources to obtain the initial capacity of each type of distributed power source.
[0114] Specifically, load time-series analysis is used to statistically analyze the hourly load values, peak load, average load, and load fluctuation amplitude for each island over a 24-hour period, dividing the load into peak, average, and valley periods. Combining the output characteristics of distributed generation, the initial capacity of each type of distributed generation is determined using the following method: The initial capacity of the photovoltaic power source is calculated based on the average daily sunshine duration of the island and the conversion efficiency of the photovoltaic modules, ensuring that the average daily output can cover more than 30% of the load during the off-peak hours of the island. The specific calculation formula is as follows: in, This represents the island's average daily total electricity consumption. The proportion of electricity that photovoltaic systems are required to supply (usually taken as 30% or more). It is the annual peak sunshine hours (i.e., the daily equivalent full-load hours, based on historical meteorological data statistics). For photovoltaic module conversion efficiency, The overall system efficiency (including inverter, line, and other losses).
[0115] For example, if an isolated island has an average daily load of 300 kWh, taking a proportion of 30%, an average daily sunshine duration of 4 hours, a module efficiency of 20%, and a system efficiency of 0.8, then... =300×0.3 / (4×0.2×0.8)≈140.6kW. After calculation, the feasibility of the installation area still needs to be verified, and a 10% margin is reserved. Finally, it will be used as the initial value for multi-objective optimization and participate in the ratio adjustment.
[0116] The initial capacity of the wind power source is calculated based on the island's average annual wind speed and the rated power of the turbines, ensuring that the output during off-peak hours is not less than a certain percentage of the island's load. The average load during off-peak hours on the island is statistically analyzed. And set the minimum guarantee ratio that wind power must bear. Obtain the annual average wind speed of the isolated area. By combining the power characteristic curve of the selected wind turbine, the output coefficient k corresponding to that wind speed is found. Based on this, the installed wind power capacity is calculated. : in, Take a value of 0.15~0.25. For overall efficiency. If a precise power curve is unavailable, it can be simplified to... , This refers to the rated wind speed of the fan.
[0117] For example, if an isolated island has a low-load of 150kW and an average annual wind speed of 6m / s, and the selected wind turbine has a rated wind speed of 12m / s, then k= =0.125, taking a proportion of 20% and an efficiency of 0.85, then PWT = 150 × 0.2 / (0.125 × 0.85) ≈ 282 kW. After calculation, verify the installation site area and use it as the initial value for subsequent multi-objective optimization.
[0118] The initial capacity of the energy storage system is estimated based on the peak-valley characteristics of the islanded load and the output fluctuations of distributed power sources. The power difference between the typical daily load curve of the islanded load and the predicted output curves of photovoltaic and wind power is statistically analyzed to preliminarily determine the peak-shaving and valley-filling power requirements and corresponding energy requirements that the energy storage system needs to undertake, which serve as the initial input values for subsequent multi-objective optimization.
[0119] S72. With the goal of minimizing investment and operating costs in distributed power planning, the goal of minimizing power outage time in isolated islands, and the goal of minimizing carbon emissions, adjust the capacity ratio of photovoltaic, wind power, and energy storage in each isolated island.
[0120] Specifically, starting from the initial capacity determined in step S71, the capacity allocation is iteratively optimized using a multi-objective optimization algorithm.
[0121] The economic objective uses the annualized cost-per-year method, aiming to minimize the sum of the annualized total investment cost and the annual operation and maintenance cost. The investment cost is converted into an annuity over the equipment's lifespan, and the operating cost includes fuel costs, maintenance costs, and network loss conversion costs. The reliability objective uses the expected power outage time, calculated through time-series simulations of the annual cumulative power outage hours for all load nodes within the island, with the weighting factor for important load outage times doubled. The environmental objective uses the annual total carbon emissions, calculated based on the emission reduction coefficients of various types of distributed generation (DG) replacing coal-fired power generation. The emission reduction coefficients for photovoltaic and wind power are converted according to the local grid emission factor, and energy storage only considers the indirect impact of charging and discharging losses.
[0122] Multi-objective optimization employs a linear weighting method, which normalizes the three objectives and assigns weights to them to construct a comprehensive objective function.
[0123] The expression for the overall objective function is: in, , , These are the weighting coefficients. The value after normalization of the economic objective. The value after normalization of the reliability target. The value is normalized to reflect environmental protection goals.
[0124] Furthermore, ω1, ω2, and ω3 are the weight coefficients for the economic, reliability, and environmental protection objectives. A pairwise comparison judgment matrix is constructed using the analytic hierarchy process (AHP), and normalized weights are derived after consistency verification. Alternatively, the entropy weight method can be used to objectively assign weights based on the degree of variation in each indicator's data, or a combination of expert experience can be used to comprehensively set the weights. If the local power grid prioritizes power supply reliability, then ω2 takes a larger value; if it prioritizes low-carbon operation, then ω3 takes a larger value. After weight calibration, the results are input into the comprehensive objective function to guide the genetic algorithm in iterative optimization, while retaining the Pareto non-dominated solution set for decision-makers to flexibly adjust the weights and reselect.
[0125] A genetic algorithm is used to encode and iterate the capacity allocation of photovoltaic, wind power, and energy storage. The population size is set to 50, the crossover probability is 0.8, the mutation probability is 0.1, and the algorithm converges after 100 generations. Each generation of schemes must satisfy power balance, voltage frequency, and hot reserve constraints. The final output is the capacity allocation scheme with the minimum weighted comprehensive cost. At the same time, the Pareto non-dominated solution set is retained so that decision-makers can adjust the weights and reselect according to actual policy preferences.
[0126] S73. The optimized energy storage adopts voltage / frequency dual closed-loop V / f control in islanded mode, dynamically suppresses wind and solar fluctuations in charge and discharge levels, and introduces droop control to achieve power sharing among multiple energy storage units.
[0127] Specifically, in islanded mode, the energy storage system uses voltage / frequency (V / f) control to serve as the main power source for the island, providing a stable voltage and frequency reference (rated voltage, 50Hz) for the entire island. Other distributed power sources (photovoltaics, wind power) follow this reference and operate in grid-connected mode.
[0128] The energy storage V / f control adopts a dual closed-loop structure of voltage-frequency outer loop + current inner loop: Outer loop (voltage / frequency regulation): The deviation between the island's real-time voltage and the reference voltage, and the deviation between the real-time frequency and the reference frequency, are respectively processed by the PI regulator to generate active current commands and reactive current commands.
[0129] Inner loop (current tracking): Quickly tracks active and reactive current commands, outputs PWM drive signals to control the switching transistors of the energy storage converter, dynamically adjusts the energy storage charging and discharging power, and achieves a response time in milliseconds.
[0130] Furthermore, droop control characteristics are introduced as an auxiliary adjustment. When the voltage or frequency deviation exceeds the limit, the power output is automatically adjusted according to the preset droop coefficient. When the voltage deviation exceeds ±3% or the frequency deviation exceeds ±0.2Hz, droop auxiliary adjustment is triggered; the droop coefficient is set at 5% to 10% of the rated energy storage capacity.
[0131] Based on the load curve of S71 and the predicted output curves of photovoltaic and wind power, a charging and discharging strategy for energy storage is formulated: When the output of distributed power sources increases suddenly (such as a midday surge in solar power generation), the energy storage charges to absorb the surplus power; when the output of distributed power sources decreases suddenly (such as the shutdown of solar power at night) or the load increases suddenly, the energy storage discharges to make up for the shortfall. Ensure that the island voltage is stable at 0.95~1.05 times the rated voltage, the frequency is stable within the range of 50±0.5Hz, and the transient adjustment time does not exceed 2 seconds.
[0132] S74. Output the planning scheme, which includes the optimized boundaries (node set) of each island, the installed capacity of photovoltaic, wind power, and energy storage of each island, and the V / f control parameters.
[0133] Furthermore, the distributed power supply collaborative planning method considering island partitioning also includes step S8, which involves building a simulation platform to simulate three scenarios: normal operation, main network failure, and main network recovery. The platform monitors the operating parameters and collaborative effects in real time and verifies the smoothness of mode switching. If the verification fails, the platform is returned for adjustment.
[0134] Specifically, the simulation verification was based on a mathematical simulation software platform to build a distribution network model containing distributed power sources and energy storage, with a sampling step size of 20μs and a simulation duration of 10s.
[0135] The model incorporates the following key modules: distribution network topology model, distributed power generation model, energy storage system model, load model, and synchronous grid connection control module.
[0136] In normal operation, the main grid power supply is set to be normal, with a voltage of 1.0 pu and a frequency of 50 Hz. The DG operates in grid-connected mode at optimized capacity, and the energy storage is in standby peak-shaving mode. Monitoring shows that the voltage deviation of each node is ≤ ±3%, the frequency deviation is ≤ ±0.2 Hz, the line load is ≤ 80%, the tie-line power fluctuation does not exceed ±5%, and the energy storage system is in grid-connected peak-shaving mode and can participate in main grid peak-shaving. If any indicator exceeds the standard, the system is judged as unqualified and needs to be returned for adjustment.
[0137] In the main grid fault scenario, a three-phase short-circuit fault on the main grid incoming line is simulated at 2 seconds, with a fault resistance of 0.01Ω. After 2.5 seconds, the protection trips the main grid switch, and each island switches to independent operation via energy storage V / f. Monitoring parameters include island voltage (0.95~1.05 pu), frequency (50±0.5Hz), energy storage V / f response time (≤3ms), DG output matching with load (≥90%), no interruption of primary load, and dynamic response of energy storage charging and discharging power. If a power deficit persists for more than 0.2 seconds leading to low-frequency load shedding, or if voltage / frequency exceeds limits, the operation is deemed unstable.
[0138] In the main grid recovery scenario, the fault is cleared after 5 seconds, and synchronous grid connection technology is used to achieve a smooth handover between the islanded system and the main grid. Before handover, the voltage difference, frequency difference, and phase difference between the islanded system and the main grid are verified: voltage difference ≤ 5%, frequency difference ≤ ±0.2Hz, and phase difference ≤ 10°. After meeting the conditions, the system is closed and connected to the grid after 6 seconds. During the handover process, the inrush current is monitored to be ≤ 1.2 times the rated value, the voltage sag is ≤ 10%, and the transient adjustment time is ≤ 2 seconds. If any condition is not met, the handover is considered a failure and adjustment is required.
[0139] The objective method for judging operational instability adopts the sliding window standard deviation method, with a sliding window of 100ms, to calculate the standard deviation of voltage and frequency in real time. If the standard deviation of voltage exceeds 0.02pu or the standard deviation of frequency exceeds 0.15Hz for three consecutive windows, it is judged as continuous oscillation instability. At the same time, it is combined with the rate of change limit detection. When the rate of change of voltage |dV / dt|>0.2pu / s or the rate of change of frequency |df / dt|>1.5Hz / s, an instability alarm is triggered. In addition, the small signal stability criterion is used to calculate the eigenvalues of the state matrix of the islanded system. If there are eigenvalues with positive real parts, it is judged as small signal instability.
[0140] The adjustment process proceeds hierarchically, adjusting the V / f control parameters. If this is ineffective, the process returns to modify the energy storage capacity or DG ratio. If this is still ineffective, the process returns to re-optimize the islanding boundary. After each adjustment, the corresponding steps and simulation verification are re-executed until all indicators are satisfactory.
[0141] After simulation verification, the following indicators are used to quantify and evaluate the collaborative effect: Voltage qualification rate: The percentage of time the voltage at the monitoring point is within 0.95~1.05 pu, with a target of ≥99.5%; Frequency compliance rate: The percentage of time spent at frequencies within 50±0.5Hz, with a target of ≥99%; DG penetration rate: The proportion of DG power generation to total load electricity consumption, with a target of ≥40%; Source-load matching degree: The correlation coefficient between DG output and load demand within 1 hour, with a target of ≥0.85; Carbon emission reduction rate: Compared with pure grid power supply, the annual carbon emission reduction rate is ≥25%.
[0142] Once all five evaluation indicators reach their target values, the planning scheme is validated, and the final planning scheme is then output.
[0143] If any evaluation indicator fails to meet the standard, the process returns to the previous step and adjustments are performed at each level. The V / f control parameters are adjusted, and if this is ineffective, the process returns to modify the energy storage capacity or DG ratio. If this is still ineffective, the process returns to re-optimize the island boundary. This continues until all five evaluation indicators meet the standard simultaneously.
[0144] In summary, this invention provides precise data support for planning by collecting data collaboratively from multiple devices and processing it in a standardized manner. By constructing a topology model using a node-branch correlation matrix, it can capture real-time dynamic changes in the topology and prioritize nodes and loads according to specifications, ensuring the scientific nature of the planning. It employs breadth-first search and mixed-integer linear programming to optimize island partitioning, satisfying multiple constraints while matching output and load, merging shared load islands, and prioritizing power supply to critical loads. Through targeted selection of candidate nodes, multi-objective optimization of capacity allocation, and energy storage V / f control to smooth fluctuations, it achieves optimal balance between economy, reliability, and environmental protection. Finally, simulation verification and dynamic adjustments in three scenarios ensure the feasibility of the method and the smoothness of mode switching, effectively solving problems such as island adaptation and output fluctuations in distributed power generation collaborative operation, and improving the fault tolerance and power supply continuity of the distribution network.
[0145] Example 2 This embodiment provides a distributed power supply cooperative planning system considering islanding, applied to the distributed power supply cooperative planning method considering islanding in Embodiment 1, including: The data acquisition and preprocessing module is used to acquire core data of the distribution network, preprocess the core data of the distribution network, and generate standardized datasets. The topology modeling and dynamic update module is used to construct a distribution network topology model based on the node-branch association matrix and a standardized dataset, and to classify node types and label load priorities in accordance with the EEE 33 node specification. The constraint building module, in conjunction with operational requirements, constructs constraints including power balance, operating parameters, topology, and load priority. The initial islanding module uses the distributed power grid connection point as the root node, adopts breadth-first search to traverse the extended load nodes, verifies the constraints point by point, and obtains the initial islanding boundary. The island optimization and adjustment module adopts mixed integer linear programming to optimize the number of islands and achieve optimal source-load matching by merging shared load islands and adjusting boundaries. The candidate node screening module, based on the optimized island partitioning and combined with resource data, screens candidate nodes for photovoltaic, wind power, and energy storage to form a list of candidate nodes for distributed power sources. The capacity collaborative configuration module, based on the load matching method and multi-objective optimization strategy, determines the initial capacity by combining load time series data, adjusts the ratio of photovoltaic, wind power and energy storage, and uses V / f control for energy storage to smooth fluctuations, achieves multi-objective optimization, and outputs the multi-objective optimal solution as the capacity configuration scheme. The simulation verification and closed-loop adjustment module is used to build a simulation platform to simulate three scenarios: normal operation, main network failure, and main network recovery. It monitors the operating parameters and collaborative effects in real time and verifies the smoothness of mode switching. If the verification fails, it returns to the adjustment stage.
[0146] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.
Claims
1. A distributed power source cooperative planning method considering islanding, characterized in that, Includes the following steps: S1. Collect core data of the distribution network, preprocess the core data of the distribution network, and generate a standardized dataset; S2. Based on the node-branch correlation matrix and combined with standardized datasets, construct a distribution network topology model, and classify node types and label load priorities according to node specifications; S3. Based on operational requirements, construct constraints including power balance, operating parameters, topology, and load priority; S4. Using the distributed power grid connection point as the root node, a breadth-first search is used to traverse the extended load nodes, and the constraints are checked point by point to obtain the initial island boundary. S5. Using mixed-integer linear programming, the number of islands is optimized and the source load is optimally matched by merging shared load islands and adjusting the boundaries; S6. Based on the optimized island partitioning and combined with resource data, candidate nodes for photovoltaic, wind power, and energy storage are selected to form a list of candidate nodes for distributed power sources; S7. Based on load time series data, optimize the ratio of photovoltaic, wind power and energy storage with multiple objectives of economy, reliability and environmental protection, and configure the energy storage V / f control parameters; output the final planning scheme.
2. The distributed power supply cooperative planning method considering islanding as described in claim 1, characterized in that, include: Step S2 includes the following sub-steps: A binary association matrix is constructed with all nodes of the distribution network as rows and all branches as columns. An element of 1 indicates that the node is directly connected to the corresponding branch, and an element of 0 indicates that it is not directly connected. At the same time, the impedance, switch status and rated capacity parameters of each branch are recorded with the branch number as the index, and an association mapping is formed with the column number of the association matrix. Referring to the IEEE 33-node system specification, nodes are divided into three categories: power supply nodes, load nodes, and tie nodes, and are uniformly numbered; based on the importance of power consumption, load nodes are labeled as Level 1 Important Loads, Level 2 Relatively Important Loads, and Level 3 Ordinary Loads.
3. The distributed power supply cooperative planning method considering islanding as described in claim 1, characterized in that, The constraints include power balance constraints, operating parameter constraints, topology constraints, and load priority constraints. The power balance constraint sets the total output of distributed power sources within the island to meet the load demand within the island, and reserves 10%-20% of hot standby capacity. The operating parameter constraint sets the voltage range of island operation to be 0.95-1.05 times the rated voltage, and the frequency range to be 50±0.5Hz. The topology constraint sets the island to consist of continuous feeder segments, maintaining a radial topology structure, with no ring networks or isolated networks. The load priority constraint sets the power supply priority of important loads to be higher than that of ordinary loads, ensuring the continuity of power supply to important loads during island division, and reserving adjustment capacity for uncontrollable loads.
4. The distributed power supply cooperative planning method considering islanding as described in claim 1, characterized in that, Step S4 includes the following steps: All distributed power grid connection points within the planning area are used as independent root nodes for breadth-first search traversal, with each root node corresponding to a candidate starting point for an initial island. Starting from the root node, the connection relationship between adjacent nodes is obtained based on the node-branch association matrix, and the connection is extended layer by layer to the adjacent load nodes in order of load priority from high to low. For each node expanded, the node is added to the candidate island set, and the set is checked in real time to see if it simultaneously satisfies power balance constraints, operating parameter constraints, and topology connectivity constraints. If all constraints are satisfied, the expansion continues to the next level of nodes. If any constraint is not satisfied, the expansion in that direction is terminated immediately, the node is excluded from the candidate set, and the boundary of the current candidate set is locked as the stopping branch in that direction. After traversing all directions, the final candidate set is determined as the initial island, and the total load, the type of distributed power source contained in the initial island, and the output potential are recorded.
5. The distributed power supply cooperative planning method considering islanding as described in claim 1, characterized in that, Step S5 includes the following steps: Using node ownership status, branch closure status, branch active and reactive power flow, distributed generation active and reactive power output, node voltage amplitude, and node severable active load as decision variables, and taking the minimum active power loss, the minimum number of switching operations, and the minimum load shedding penalty as the comprehensive objective function, a mixed integer linear programming model is constructed. The constraints include power balance constraints, voltage safety constraints, radial topology constraints, distributed generation output limit constraints, hot standby constraints, and primary load uninterruption constraints. When solving the mixed-integer linear programming model, all initial islands are traversed, load nodes shared by multiple islands are identified, and if all constraints are satisfied after merging, they are merged into a larger island. For each island, source-load matching verification is performed. For islands with insufficient distributed generation output, expansion is performed to adjacent low-load-density nodes that have been included in the candidate nodes of distributed generation but have not yet been assigned to any island. After the expanded node connects to the new distributed generation, the source-load matching degree of the island can be improved to no less than 1.0 after power balance verification. For islands with excess distributed generation output, distributed generation nodes that can be stripped or reduced within the island are identified. Priority is given to stripping the DG node that is farthest from the load center and has the highest unit transmission cost, or the excess distributed generation output is redistributed to adjacent islands with insufficient output. The source-load matching degree of each island after stripping or redistribution is verified to be maintained within the range of 1.0 to 1.
25. After each boundary adjustment, the active power loss and the number of switching operations are calculated and the best is selected. Finally, the optimized number of islands and the accurate boundary and source-load matching results of each island are output.
6. The distributed power supply cooperative planning method considering islanding as described in claim 1, characterized in that, Step S6 includes the following sub-steps: Using the optimized boundaries of each isolated island as the spatial screening unit, data on solar radiation, average annual wind speed, available installation area, and load distribution within each island were retrieved. Based on the retrieved data, nodes were screened: nodes with abundant solar resources, good lighting conditions, and low access costs were selected as candidate nodes for photovoltaic power generation; nodes with average annual wind speeds reaching the start-up wind speed, good ventilation conditions, and far from obstacles were selected as candidate nodes for wind power generation; and nodes within each island that are adjacent to important load centers, have convenient grid access conditions, and sufficient land area were selected as candidate nodes for lithium battery energy storage systems. The feasibility of access capacity, line carrying capacity and equipment installation conditions of all candidate nodes in the initial selection is verified one by one. The matching degree between the candidate distributed power source type and the timing characteristics of the island load is evaluated. Nodes that are not feasible to access or have poor coordination effect are eliminated. Finally, the list of candidate distributed power source nodes for each island is determined and output to the capacity configuration module.
7. The distributed power supply cooperative planning method considering islanding as described in claim 6, characterized in that, Based on the retrieved data, nodes were screened as follows: nodes with abundant sunlight resources, good lighting conditions, and low access costs were selected as candidate nodes for photovoltaic power generation; nodes with an average annual wind speed that meets the start-up wind speed, good ventilation conditions, and are far from obstacles were selected as candidate nodes for wind power generation; and nodes within each island that are close to important load centers, have convenient grid access conditions, and sufficient land area were selected as candidate nodes for lithium battery energy storage systems.
8. The distributed power supply cooperative planning method considering islanding as described in claim 1, characterized in that, Step S7 includes the following sub-steps: Based on the load time-series data of each island, the peak load, average load and load fluctuation patterns of different time periods are statistically analyzed. Combined with the output characteristics of distributed power sources, the initial capacity of each type of distributed power source is obtained. With the goal of minimizing investment and operating costs in distributed power planning, the goal of minimizing power outage time in isolated islands, and the goal of minimizing carbon emissions, the capacity ratio of photovoltaic, wind power, and energy storage in each isolated island is adjusted. The optimized energy storage system adopts dual closed-loop V / f control of voltage and frequency in islanded mode, dynamically suppresses wind and solar fluctuations in charge and discharge levels, and introduces droop control to achieve power sharing among multiple energy storage units. Output the planning scheme, which includes the optimized boundaries of each island, the installed capacity of photovoltaic, wind power, and energy storage on each island, and the V / f control parameters.
9. The distributed power supply cooperative planning method considering islanding as described in claim 1, characterized in that, It also includes step S8, which involves building a simulation platform to simulate three scenarios: normal operation, main network failure, and main network recovery. The platform monitors the operating parameters and collaborative effects in real time and verifies the smoothness of mode switching. If the verification fails, the platform is returned for adjustment.
10. A distributed power supply cooperative planning system considering islanding, characterized in that, The distributed power supply cooperative planning method considering islanding as described in any one of claims 1 to 9 includes: The data acquisition and preprocessing module is used to acquire core data of the distribution network, preprocess the core data of the distribution network, and generate standardized datasets. The topology modeling and dynamic update module is used to construct a distribution network topology model based on the node-branch correlation matrix and a standardized dataset, and to classify node types and label load priorities according to node specifications. The constraint building module, in conjunction with operational requirements, constructs constraints including power balance, operating parameters, topology, and load priority. The initial islanding module uses the distributed power grid connection point as the root node, adopts breadth-first search to traverse the extended load nodes, verifies the constraints point by point, and obtains the initial islanding boundary. The island optimization and adjustment module adopts mixed integer linear programming to optimize the number of islands and achieve optimal source-load matching by merging shared load islands and adjusting boundaries. The candidate node screening module, based on the optimized island partitioning and combined with resource data, screens candidate nodes for photovoltaic, wind power, and energy storage to form a list of candidate nodes for distributed power sources. The capacity coordination configuration module optimizes the ratio of photovoltaic, wind power, and energy storage based on load time-series data, taking into account multiple objectives such as economy, reliability, and environmental protection, and configures the energy storage V / f control parameters; it also outputs the final planning scheme. The simulation verification and closed-loop adjustment module is used to build a simulation platform to simulate three scenarios: normal operation, main network failure, and main network recovery. It monitors the operating parameters and collaborative effects in real time and verifies the smoothness of mode switching. If the verification fails, it returns to the adjustment stage.
