Decomposition collaborative-parallel accelerated main matching micro-synchronous asynchronous collaborative power flow calculation method

Abstract

This invention relates to the field of smart grid transmission and distribution coordinated power flow calculation technology, and discloses a decomposed coordinated-parallel accelerated synchronous and asynchronous coordinated power flow calculation method for main-distribution-micro networks. The method includes: a decomposed coordinated power flow calculation framework for main-distribution-micro heterogeneous networks, which divides the main-distribution-micro heterogeneous network into an upper-layer ring network and a lower-layer radial network. The upper and lower layers perform power flow calculations independently, and achieve consistent calculation and solution of global network power flow through interactive iteration of power flow information from coupling interfaces; a synchronous / asynchronous parallel accelerated calculation method based on network scale partitioning; and a synchronous / asynchronous parallel accelerated mechanism for the lower-layer radial network, which partitions the network scale and designs a synchronous / asynchronous parallel accelerated mechanism between a single upper-layer ring network and a massive number of lower-layer radial networks to achieve rapid power flow calculation for large-scale heterogeneous networks. The advantages of this invention are that it can effectively improve the efficiency of power flow calculation for large-scale main-distribution-micro networks, realize rapid iterative calculation of main-distribution-micro power flow, and support the high-frequency safety and stability analysis requirements of the power grid.

Description

Technical Field

[0001] This invention relates to the field of smart grid transmission and distribution coordinated power flow calculation technology, and in particular to a decomposed coordinated-parallel accelerated master-distribution micro-synchronous asynchronous coordinated power flow calculation method. Background Technology

[0002] Currently, the power system is rapidly evolving towards a cleaner, lower-carbon, distributed, and intelligent direction. Distributed resources, represented by photovoltaics, wind power, energy storage, electric vehicles, and flexible loads, are being rapidly integrated into the main grid and distribution microgrids, becoming a key driving force for energy structure transformation and the diversified interactive development of the power system. However, due to the inherent characteristics of distributed resources, such as strong output volatility, poor controllability, and complex control modes, their widespread integration, while increasing the flexibility of the main grid and distribution microgrids, also impacts the safe and stable operation of the power grid. Against this backdrop, power flow calculation, as a core tool for power grid analysis, is becoming increasingly crucial.

[0003] With the continuous integration of distributed resources such as photovoltaics, electric vehicles, distributed energy storage, and new responsive loads, the number of nodes in main distribution microgrids, especially medium- and low-voltage distribution microgrids, is rapidly increasing, exhibiting a large-scale characteristic across multiple voltage levels. Because power flow calculations need to consider power transmission and mutual constraints across voltage levels and regions, the network size not only grows linearly but also tends to expand exponentially with increased node coupling. This large-scale characteristic directly affects the sparse structure and solution complexity of power flow calculations, significantly increasing the size of the Jacobian matrix and the number of non-zero elements, thus increasing the iteration time of power flow calculations. Furthermore, security analysis of main distribution microgrids based on power flow calculations, especially real-time online monitoring and analysis of grid operation status, places high demands on the speed of power flow calculations, which traditional integrated Newton-Lambert method cannot meet. Therefore, there is an urgent need to design a main distribution microgrid power flow calculation method with fast and efficient computational capabilities for large-scale main distribution microgrids.

[0004] Currently, numerous studies have proposed decomposition-cooperative power flow calculation methods for the scalable characteristics of master-distributor micronetworks, such as hierarchical power flow calculation methods based on master-slave splitting and power flow calculation methods based on distributed algorithms. Among these, the hierarchical power flow calculation method was proposed earlier and possesses mature and efficient decomposition-cooperative fast solution capabilities, which has been widely explored and applied by professional researchers. In the decomposition-cooperative master-distributor micronetwork hierarchical power flow calculation process, the lower-level network contains a large number of radial subnetworks that are calculated independently. These subnetworks exhibit significant heterogeneity in terms of node size, topological complexity, and resource access structure, resulting in large differences in power flow solution time between different subnetworks. If a fully synchronous strategy is adopted (i.e., unified interaction after all subnetworks have been calculated), the solution speed of the overall synchronous interaction of multiple subnetworks will be limited by the distribution network with the worst solution speed, leading to a decrease in overall iteration efficiency and affecting the computational scalability of large-scale examples. However, if the subnetworks do not undergo unified iteration and adopt a completely differentiated asynchronous iteration method, the iterative interaction process of multiple subnetworks will be abnormally chaotic, which will not only affect the speed of power flow calculation but may also affect the convergence of master-distributor cooperative power flow. Therefore, the iterative interaction process between upper and lower layer networks lacks an efficient and feasible acceleration management mechanism to further improve the solution efficiency of the master-slave micro-decomposition cooperative power flow calculation method. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a decomposition-cooperative-parallel accelerated master-slave micro-synchronous asynchronous cooperative power flow calculation method.

[0006] The objective of this invention is achieved through the following technical solution: a decomposition-cooperative-parallel accelerated master-supplier-micro synchronous-asynchronous cooperative power flow calculation method, which includes,

[0007] The master-supplier micro-heterogeneous network decomposition and collaborative power flow calculation framework divides the master-supplier micro-heterogeneous network into an upper ring network and a lower radial network. The upper and lower networks perform power flow calculations independently, and the consistent calculation and solution of the global network power flow is achieved through interactive iteration of power flow information at the coupling interface.

[0008] A synchronous / asynchronous parallel acceleration method based on network size partitioning is proposed. The lower-layer radial network is partitioned by network size, and a synchronous / asynchronous parallel acceleration mechanism is designed between the upper-layer single ring network and the lower-layer massive radial network to achieve fast power flow calculation of master-supplement micro-large-scale heterogeneous networks.

[0009] Specifically, the interactive iteration process is as follows:

[0010] The upper-layer network calculates the overall power flow solution based on the initial power flow parameters of the interaction interface and passes the obtained interaction interface voltage parameters to the lower-layer network. The lower-layer network updates its overall power flow solution based on the received voltage parameters and feeds back the updated interaction interface power parameters to the upper-layer network for cyclical updates. The iteration continues until the difference in power flow parameters of all coupled interaction interfaces in two adjacent iterations reaches a preset threshold. At this point, the power flow converges, and the internal power flow solution and interaction interface power flow parameters independently calculated by each network at the two levels constitute the global power flow solution of the entire network.

[0011] Specifically, the mathematical model for the interactive iterative process is as follows:

[0012] ;

[0013] ;

[0014] ;

[0015] ;

[0016] In the formula, This represents the voltage vector of all nodes in the upper network except for the boundary nodes at the k-th iteration. This represents the voltage vector of the boundary nodes at the k-th iteration. It is the power vector injected into the boundary nodes during the k-th iteration; The upper-layer network is equivalent to the lower-layer radial network as follows: The formula for calculating power flow after load; To convert the transmission network into an equivalent value in the distribution network The formula for calculating the power flow after the constant pressure source; This represents the voltage vector of all nodes in the lower-level network except for the boundary nodes at the k-th iteration. The iterative formula for the interaction variables; The threshold for determining convergence.

[0017] Specifically, the network size partitioning employs the K-Means clustering algorithm, including:

[0018] For each lower-level radial subnetwork, multiple metrics characterizing its size or complexity are extracted to construct a feature vector. These metrics include at least: number of nodes, number of branches, node connectivity, and maximum network depth.

[0019] The feature vector is constructed as follows:

[0020] ;

[0021] In the formula, Let be the feature vector of the i-th radial subnet; Let be the number of nodes in the i-th subnet; Let be the number of branches in the i-th subnet; Let be the node connectivity of the i-th subnet; The maximum network depth of the i-th subnet;

[0022] Normalize the indicators:

[0023] ;

[0024] In the formula, This is the normalized value of the feature vector of the i-th subnet; , These are the minimum and maximum values ​​of the indicator, respectively.

[0025] Randomly select k initial cluster centers for iteration. For each subgroup, calculate its distance to each cluster center:

[0026] ;

[0027] In the formula, Let be the weight of the j-th indicator; This is the normalized value of the j-th indicator in the i-th subnet; The value of the j-th index for the k-th cluster center;

[0028] Each distribution network is assigned to the cluster containing its nearest cluster center. The mean of each cluster is calculated, and the cluster center is updated to the mean of that cluster.

[0029] ;

[0030] In the formula, Let k be the distribution network set of the k-th cluster; This represents the number of subnets in the cluster;

[0031] Using the K-Means clustering algorithm, the lower-level subnet is divided into K network groups, each with similar size and connectivity. The resulting network partitioning is as follows:

[0032] ;

[0033] In the formula, This represents the sub-network group after the kth clustering.

[0034] Specifically, the synchronous / asynchronous parallel acceleration mechanism is as follows:

[0035] For each radial subnetwork within the subnetwork group obtained by the same cluster, a synchronous interaction method is adopted. After all subnetworks in the group have completed the power flow calculation of the current iteration, their interaction parameters are uniformly passed to the upper layer network.

[0036] For sub-network groups obtained from different clusters, an asynchronous interaction method is adopted. After each sub-network group completes the power flow calculation of all its internal sub-networks, it does not need to wait for other network groups and immediately passes its interaction parameters to the upper-layer network for updating.

[0037] The present invention has the following advantages:

[0038] This invention significantly improves the iterative efficiency of power flow calculation for large-scale heterogeneous power grids by decoupling the main and distribution microgrids in layers and adopting a parallel interaction strategy of intra-group synchronization and inter-group asynchronous interaction based on network-scale clustering, while ensuring computational convergence. This enables it to better support the real-time requirements of high-frequency security and stability analysis of the power grid. Attached Figure Description

[0039] Figure 1 This is a schematic diagram of the power flow calculation method of the present invention;

[0040] Figure 2 A schematic diagram of the master-slave micro-heterogeneous network decomposition and collaborative power flow calculation framework;

[0041] Figure 3 This is a schematic diagram of the power system topology. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only for explaining the invention and are not intended to limit the invention; that is, the described embodiments are merely some embodiments of the invention, and not all embodiments. The components of the embodiments of the invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0043] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0044] It should be noted that relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0045] The present invention will be further described below with reference to the accompanying drawings, but the scope of protection of the present invention is not limited to the following description.

[0046] like Figures 1 to 3 As shown, a decomposition-cooperative-parallel accelerated master-slave synchronous-asynchronous cooperative power flow calculation method is presented, which includes:

[0047] The master-supplier micro-heterogeneous network decomposition and collaborative power flow calculation framework divides the master-supplier micro-heterogeneous network into an upper ring network and a lower radial network. The upper and lower networks perform power flow calculations independently, and the consistent calculation and solution of the global network power flow is achieved through interactive iteration of power flow information at the coupling interface.

[0048] The primary and secondary heterogeneous networks are divided into an upper ring network of 220kV-110kV and a lower radial network of 10kV-0.38kV.

[0049] The interactive iteration process is as follows:

[0050] The upper-layer network calculates the overall power flow solution based on the initial power flow parameters of the interaction interface and passes the obtained voltage parameters of the interaction interface to the lower-layer network. The lower-layer network updates its overall power flow solution based on the received voltage parameters and feeds back the updated power flow parameters of the interaction interface to the upper-layer network for cyclical updates. The iteration continues until the difference in power flow parameters of all coupled interaction interfaces in two adjacent iterations reaches a preset threshold. At this point, the power flow converges and the interaction iteration ends. The upper-layer network is a single ring structure, and the lower-layer network is a group of radial sub-networks directly connected to the upper-layer network. Therefore, there are multiple coupled interaction interfaces between the upper and lower-layer networks. The power flow parameters of all interfaces in two adjacent iterations must reach a preset threshold to determine the power flow convergence of the main-distribution micro-network. The internal power flow solution and the power flow parameters of the interaction interface calculated independently by each network at both levels constitute the global power flow solution of the overall network.

[0051] The mathematical model for the interactive iterative process is as follows:

[0052] ;

[0053] ;

[0054] ;

[0055] ;

[0056] In the formula, This represents the voltage vector of all nodes in the upper network except for the boundary nodes at the k-th iteration. This represents the voltage vector of the boundary nodes at the k-th iteration. It is the power vector injected by the boundary node at the k-th iteration, and it is the voltage vector of other nodes in the lower layer network except the root node at the k-th iteration; The upper-layer network is equivalent to the lower-layer radial network as follows: The formula for calculating power flow after load change, at this time Given quantities , The quantity to be determined; To convert the transmission network into an equivalent value in the distribution network The formula for calculating the power flow after the constant pressure source, at this time Given quantities , The quantity to be determined; This represents the voltage vector of all nodes in the lower-level network except for the boundary nodes at the k-th iteration. The iterative formula for the interaction variables; The threshold for determining convergence.

[0057] A synchronous / asynchronous parallel acceleration method based on network size partitioning is proposed. The lower-layer radial network is partitioned by network size, and a synchronous / asynchronous parallel acceleration mechanism is designed between the upper-layer single ring network and the lower-layer massive radial network to achieve fast power flow calculation of master-supplement micro-large-scale heterogeneous networks.

[0058] The network size partitioning employs the K-Means clustering algorithm, including:

[0059] For each lower-level radial subnetwork, multiple indicators representing its size or complexity are extracted to construct a feature vector. The indicators include at least the number of nodes, the number of branches, the node connectivity, and the maximum network depth. Each indicator is assigned an appropriate weight. By clustering and dividing a large number of lower-level radial subnetworks, the distribution network nodes are divided into several subnetworks according to their size and connectivity, so as to facilitate subsequent efficient interactive iterative parallel computing.

[0060] The feature vector is constructed as follows:

[0061] ;

[0062] In the formula, Let be the feature vector of the i-th radial subnet; The number of nodes in the i-th subnet is a metric that reflects the scale of the distribution network. The more nodes there are, the higher the network complexity. The number of branches in the i-th subnet is a metric that measures the connectivity and topological complexity within the subnet. Let represent the node connectivity of the i-th subnet. This metric measures the average connectivity between nodes in the distribution network and reflects the sparsity of the network. The maximum network depth of the i-th subnet is an indicator that represents the maximum distance between the farthest node in the subnet and the root node, reflecting the scalability and depth of the network.

[0063] To perform effective K-Means clustering, these metrics need to be normalized to eliminate the influence between different units of measurement. The four metrics for each subnet are normalized to ensure that each metric is on the same unit of measurement. The normalization process for the metrics is as follows:

[0064] ;

[0065] In the formula, This is the normalized value of the feature vector of the i-th subnet; , These are the minimum and maximum values ​​of the indicator, respectively.

[0066] Randomly select k initial cluster centers to ensure the representativeness of the initial cluster points, and then iterate. For each sub-cluster, calculate its distance to each cluster center:

[0067] ;

[0068] In the formula, Let be the weight of the j-th indicator; This is the normalized value of the j-th indicator in the i-th subnet; The value of the j-th index for the k-th cluster center;

[0069] Each distribution network is assigned to the cluster containing its nearest cluster center. The mean of each cluster is calculated, and the cluster center is updated to the mean of that cluster.

[0070] ;

[0071] In the formula, Let k be the distribution network set of the k-th cluster; This represents the number of subnets in the cluster;

[0072] If the change in the cluster center position is less than a certain set threshold, the iteration stops; otherwise, return to the previous step and continue iterating.

[0073] Using the K-Means clustering algorithm, the lower-level subnet is divided into K network groups, each with similar size and connectivity. The resulting network partitioning is as follows:

[0074] ;

[0075] In the formula, This represents the sub-network group after the kth clustering.

[0076] In the decomposition and collaborative interactive iterative master-distributor micro-power flow calculation process, the lower-level network contains a large number of radial subnetworks that are calculated independently. These subnetworks exhibit significant heterogeneity in terms of node size, topological complexity, and resource access structure, resulting in large differences in power flow solution time among different subnetworks. If a fully synchronous strategy is adopted (i.e., unified interaction after all subnetworks have been calculated), the solution speed of the overall synchronous interaction of multiple subnetworks will be limited by the distribution network with the worst solution speed, leading to a decrease in overall iteration efficiency and affecting the computational scalability of large-scale cases. However, if the subnetworks do not undergo unified iteration and adopt a completely differentiated asynchronous iteration method, the iterative interaction process of multiple subnetworks will be abnormally chaotic, affecting not only the speed of power flow calculation but also the convergence of master-distributor collaborative power flow. To further improve the efficiency of master-distributor micro-power flow calculation, this invention classifies a large number of radial subnetworks in the lower layer according to their size using a network size partitioning method based on the K-Means clustering algorithm. Relying on the parallel acceleration mechanism of synchronous / asynchronous interaction between the upper and lower layers, it achieves orderly and efficient interactive iteration between the upper and lower layers, thereby improving the overall efficiency of network power flow calculation.

[0077] To further improve the iterative efficiency of power flow calculation in the entire main-distribution-micronetwork, this invention proposes a parallel acceleration mechanism for synchronous / asynchronous interaction between upper and lower network layers. This mechanism, within a decomposed collaborative computing framework, efficiently parallelizes the iterative interaction process between the main network and the lower-level distribution network to achieve rapid power flow solutions for large-scale power grids. It includes:

[0078] Due to the same sub-network group The computation speed of each radial network within the subnet is basically the same. Therefore, the subnets within the subnet group interact synchronously with the upper-layer network. That is, after all the subnets within the network group have completed the power flow calculation, they are synchronously transmitted to the upper-layer main network for iterative calculation. The synchronous iteration with small differences in computation speed within the subnets will not affect the computation iteration efficiency of the lower-layer network. At the same time, it can improve the computation iteration efficiency of the main network, thereby improving the overall power flow calculation efficiency.

[0079] Because the computation speeds of different sub-network groups vary significantly, different sub-network groups Internal subnets , After completing the calculation of the internal subnet, there is no need to wait for other network groups. It can instantly interact and update with the upper-layer network. The asynchronous iteration of the subnets and the main network within different network groups improves the efficiency of interaction and iteration between the upper and lower layers, thereby improving the efficiency of the main-partition micro-power flow calculation.

[0080] To further verify the effectiveness of the master-distributor micro-synchronous / asynchronous power flow calculation method provided by this invention in improving the calculation speed of master-distributor micro-decomposition cooperative power flow, an evaluation index, speed ratio, is defined. ,in The calculation time of the power flow calculation method proposed in this invention is [missing information]. The calculation time of the traditional decomposition and cooperative power flow calculation method master-slave split method was calculated, and the simulation results were compared and verified in four power supply and distribution systems. The speedup results under different operating conditions of different systems are shown in Table 1 below.

[0081] Table 1 Comparison of speedup ratios under different operating conditions for different systems

[0082]

[0083] As can be seen from the data in the table, the decomposition-coordination-parallel acceleration master-slave micro synchronous / asynchronous power flow calculation method proposed in this invention has a significant acceleration effect under different operating states of different systems, and the acceleration effect is more obvious as the system scale increases.

[0084] The above description is merely a preferred embodiment of the present invention and does not constitute any limitation on the present invention. Any person skilled in the art can make many possible variations and modifications to the technical solution of the present invention, or modify it into equivalent embodiments, without departing from the scope of the present invention. Therefore, any modifications, equivalent changes, and alterations made to the above embodiments based on the technology of the present invention without departing from the scope of the present invention are within the protection scope of the present invention.

Claims

1. A decomposition-coordinated-parallel accelerated master-supplier-micro synchronous-asynchronous cooperative power flow calculation method, characterized by: The method includes, The master-slave micro-heterogeneous network decomposition and collaborative power flow calculation framework includes dividing the master-slave micro-heterogeneous network into an upper ring network and a lower radial network. The upper and lower networks perform power flow calculations independently, and the consistent calculation and solution of the global network power flow is achieved through the interactive iteration of power flow information of the coupling interface. A synchronous / asynchronous parallel acceleration method based on network scale partitioning is proposed. This method involves partitioning the lower-level radial network by network scale and designing a synchronous / asynchronous parallel acceleration mechanism between a single upper-level ring network and a massive lower-level radial network to achieve fast power flow calculation in a primary-secondary-micro-scale heterogeneous network.

2. The decomposition-cooperative-parallel accelerated master-supplier-micro synchronous-asynchronous cooperative power flow calculation method according to claim 1, characterized in that: The interactive iteration process is as follows: The upper-layer network calculates the overall power flow solution based on the initial power flow parameters of the interaction interface and passes the obtained interaction interface voltage parameters to the lower-layer network. The lower-layer network updates its overall power flow solution based on the received voltage parameters and feeds back the updated interaction interface power parameters to the upper-layer network for cyclical updates. The iteration continues until the difference in power flow parameters of all coupled interaction interfaces in two adjacent iterations reaches a preset threshold. At this point, the power flow converges, and the internal power flow solution and interaction interface power flow parameters independently calculated by each network at the two levels constitute the global power flow solution of the entire network.

3. The decomposition-cooperative-parallel accelerated master-supplier-micro synchronous-asynchronous cooperative power flow calculation method according to claim 2, characterized in that: The mathematical model for the interactive iterative process is as follows: ； ； ； ； In the formula, This represents the voltage vector of all nodes in the upper network except for the boundary nodes at the k-th iteration. This represents the voltage vector of the boundary nodes at the k-th iteration. It is the power vector injected into the boundary nodes during the k-th iteration; The upper-layer network is equivalent to the lower-layer radial network as follows: The formula for calculating power flow after load; To convert the transmission network into an equivalent value in the distribution network The formula for calculating the power flow after the constant pressure source; This represents the voltage vector of all nodes in the lower-level network except for the boundary nodes at the k-th iteration. The iterative formula for the interaction variables; The threshold for determining convergence.

4. The decomposition-cooperative-parallel accelerated master-supplier-micro synchronous-asynchronous cooperative power flow calculation method according to claim 1, characterized in that: The network size partitioning employs the K-Means clustering algorithm, including: For each lower-level radial subnetwork, multiple metrics characterizing its size or complexity are extracted to construct a feature vector. These metrics include at least: number of nodes, number of branches, node connectivity, and maximum network depth. The feature vector is constructed as follows: ； In the formula, Let be the feature vector of the i-th radial subnet; Let be the number of nodes in the i-th subnet; Let be the number of branches in the i-th subnet; Let be the node connectivity of the i-th subnet; The maximum network depth of the i-th subnet; Normalize the indicators: ； In the formula, This is the normalized value of the feature vector of the i-th subnet; , These are the minimum and maximum values ​​of the indicator, respectively. Randomly select k initial cluster centers for iteration. For each subgroup, calculate its distance to each cluster center: ； In the formula, Let be the weight of the j-th indicator; This is the normalized value of the j-th indicator in the i-th subnet; The value of the j-th index for the k-th cluster center; Each distribution network is assigned to the cluster containing its nearest cluster center. The mean of each cluster is calculated, and the cluster center is updated to the mean of that cluster. ； In the formula, Let k be the distribution network set of the k-th cluster; This represents the number of subnets in the cluster; Using the K-Means clustering algorithm, the lower-level subnet is divided into K network groups, each with similar size and connectivity. The resulting network partitioning is as follows: ； In the formula, This represents the sub-network group after the kth clustering.

5. The decomposition-cooperative-parallel accelerated master-supplier-micro synchronous-asynchronous cooperative power flow calculation method according to claim 4, characterized in that: The synchronous / asynchronous parallel acceleration mechanism is as follows: For each radial subnetwork within the subnetwork group obtained by the same cluster, a synchronous interaction method is adopted. After all subnetworks in the group have completed the power flow calculation of the current iteration, their interaction parameters are uniformly passed to the upper layer network. For sub-network groups obtained from different clusters, an asynchronous interaction method is adopted. After each sub-network group completes the power flow calculation of all its internal sub-networks, it does not need to wait for other network groups and immediately passes its interaction parameters to the upper-layer network for updating.

Images