Hierarchical partitioning, energy hub siting and network architecture planning method for regional energy internet

Abstract

The application discloses a layered partition, energy hub site selection and network frame planning method of energy internet, and belongs to the field of energy internet planning, which comprehensively considers centralized energy nodes and distributed energy nodes, solves the problem of huge difference in orders of magnitude between the two through layered planning, promotes energy consumption in place, and optimizes energy hub site selection from local suboptimal optimization to local optimal optimization; the layered planning, greedy strategy heuristic algorithm simplify complex calculation, Cooper iteration and partition energy index smoothing improve solving precision, and the method has fast solving speed and high quality, and performs well on the data set of thousands of nodes of the Dutch Zaanstad city.

Description

Technical Field

[0001] This invention belongs to the field of energy internet planning, and more specifically, relates to a method for hierarchical zoning, energy hub site selection and grid planning of regional energy internet. Background Technology

[0002] The Energy Internet achieves optimal planning and management of distributed energy systems through multi-energy complementarity, energy cascade utilization, and energy storage. Based on its scale, the Energy Internet can be categorized into community-level, regional-level, and global-level systems. The planning problem of regional Energy Internets—that is, how to construct a rational regional Energy Internet system—is the foundation for scheduling optimization and intelligent control, and has received widespread attention from academia and industry. Regional division is a common method for regional Energy Internet planning; deploying energy hubs within zones is beneficial for promoting multi-energy coordination and complementarity and improving energy cascade utilization. However, the regional Energy Internet zoning and site selection problem is quite complex, and a fast, high-quality algorithm is still lacking for large-scale problems. With the rise of new energy sources, distributed energy is widely used, offering advantages such as low loss and flexible operation. However, compared to traditional centralized energy, it is smaller in scale and more numerous, bringing new challenges to the rational planning of centralized-distributed hybrid energy systems. Therefore, designing a fast, high-quality hierarchical zoning, energy hub site selection, and network planning algorithm for large-scale distributed-centralized hybrid energy systems at the urban level to promote multi-energy complementarity and supply-demand balance is of great significance for building a clean and efficient regional Energy Internet. Summary of the Invention

[0003] In response to the above-mentioned deficiencies or improvement needs of existing technologies, this invention provides a method for hierarchical zoning, energy hub site selection, and grid planning of regional energy internet. This method addresses the difficulties in comprehensive planning caused by the failure of existing planning methods to consider the combination of distributed and centralized energy sources, as well as the shortcomings of existing planning methods such as low solution efficiency and small planning scale due to high complexity. This method enables high-quality and high-efficiency energy system planning.

[0004] To achieve the above objectives, according to a first aspect of the present invention, a method for hierarchical zoning, energy hub site selection, and grid planning of a regional energy internet is provided, comprising:

[0005] S1. According to the preset correspondence between the energy supply or demand of energy nodes and the grid hierarchy, each energy node in the target area is assigned to the corresponding level of the grid; among them, the higher the level of the grid, the greater the energy supply or demand of the energy nodes.

[0006] S2, Under the constraint of the number of partitions in each layer of the network, with the goal of minimizing the overall loss of each layer of the network, the first target node of each layer of the network is partitioned sequentially from bottom to top, and the position of the hub node of each partition is determined. Then, each hub node is connected to the first target node of its partition to obtain the overall network.

[0007] The comprehensive loss of each layer of the network includes the line loss between the first target node and its hub node in each partition and the supply and demand balance loss of each partition. The first target node includes the energy node of the network layer and the hub node of the network layer above. Each partition of each layer of the network has one and only one hub node.

[0008] S3, execute S2 q times to obtain q overall network structures, select the overall network structure with the smallest line loss as the network structure to be adjusted, q≥1; when q>1, execute S2 based on the previous overall network structure;

[0009] S4, with the goal of minimizing the line loss between the second target node and the hub node of each partition of each layer of the overall network structure, the position of the hub node of each partition of each layer of the overall network structure is adjusted, and each hub node after the position adjustment is connected to the second target node of its partition to obtain the regional energy internet of the target area.

[0010] The second target node includes energy nodes connected to hub nodes of each zone of each layer of the overall grid, as well as hub nodes of the upper and lower layers of the grid.

[0011] According to a second aspect of the present invention, a method and system for hierarchical zoning, energy hub site selection and grid planning of a regional energy internet is provided, comprising: a computer-readable storage medium and a processor;

[0012] The computer-readable storage medium is used to store executable instructions;

[0013] The processor is configured to read executable instructions stored in the computer-readable storage medium and execute the method as described in the first aspect.

[0014] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects:

[0015] Compared with existing methods, the method provided by this invention comprehensively considers both centralized and distributed energy nodes, and solves the problem of huge order-of-magnitude differences between them through hierarchical planning; it promotes local energy consumption and optimizes the location of energy hubs from local suboptimal to local optimal; it simplifies complex calculations through hierarchical planning and greedy strategy heuristic algorithms, and improves the solution accuracy through Cooper iteration and partitioned energy index smoothing. It has a fast solution speed and high quality, and performs well on the Zaandam, Netherlands dataset with thousands of nodes. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the overall framework of the regional energy internet hierarchical zoning, energy hub site selection and grid planning method provided in the embodiments of the present invention.

[0017] Figure 2 (a) and (b) in the figure are schematic diagrams of the planning steps and adjustment steps provided in the embodiments of the present invention, respectively.

[0018] Figure 3 This is a schematic diagram of the bottom-up planning steps provided in an embodiment of the present invention.

[0019] Figure 4 This is a schematic diagram of the top-down adjustment steps provided in an embodiment of the present invention.

[0020] Figure 5 This is a schematic diagram of the bottom-up adjustment steps provided in an embodiment of the present invention.

[0021] Figure 6 This is a schematic diagram of the partitioning location process provided in an embodiment of the present invention; wherein, except for the red dashed box for partitioning location, the dashed boxes represent variables and data (including input data, output data and intermediate results), the solid boxes represent the processing process, and the yellow box diagram is a schematic diagram of the node set, hub location and single-layer network structure.

[0022] Figure 7 The flowchart of the partitioning location problem solving algorithm provided in the embodiment of the present invention is shown.

[0023] Figure 8 The flowchart of the energy hub location initialization algorithm provided in the embodiment of the present invention is shown; the black dashed box represents the input and output, and the colored solid rounded corner box represents the algorithm flow.

[0024] Figure 9 This is a flowchart of the partitioning algorithm provided in an embodiment of the present invention; the black dashed boxes represent inputs and outputs, and the colored solid rounded corner boxes represent the algorithm flow.

[0025] Figure 10 This is a flowchart of the address selection algorithm provided in an embodiment of the present invention; the black dashed boxes represent inputs and outputs, and the colored solid rounded corner boxes represent the algorithm flow.

[0026] Figure 11 This is a flowchart of the partitioned energy index smoothing algorithm provided in an embodiment of the present invention; the black dashed boxes represent inputs and outputs, and the colored solid rounded corner boxes represent the algorithm flow.

[0027] Figure 12 This is a schematic diagram illustrating the gradual flattening of the overall loss as provided in an embodiment of the present invention.

[0028] Figure 13 The flowchart of the address adjustment algorithm provided in the embodiment of the present invention is shown; the dashed box represents the algorithm input and output, and the solid box represents the processing procedure. Detailed Implementation

[0029] 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 merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0030] First, the definitions are as follows:

[0031] Regional Energy Internet: refers to a smart energy network within a certain area, with the distribution network as its core, that achieves deep integration of energy and information, promotes multi-energy coupling and complementarity, and coordinates the development of energy production, distribution, consumption, and storage.

[0032] Clustering partitioning: Using clustering methods, nodes in the regional energy internet are divided into sub-regions to promote local energy consumption and reduce energy network costs.

[0033] Transmission loss: When energy sources such as electricity, heat, and gas are transported along power grids or pipelines, energy loss is unavoidable due to factors such as heat generation and dissipation.

[0034] Supply and demand balance: Within a region, the energy-generating source nodes and the energy-consuming load nodes should maintain a balance between supply and demand as much as possible.

[0035] Energy hub: A device that enables the comprehensive optimization and cascade utilization of multiple energy forms, which can be represented as a multi-input multi-output multi-port network.

[0036] Network planning: Rationally plan and deploy network lines for energy nodes such as users, power / gas stations and energy hubs to achieve efficient and stable energy transmission.

[0037] For large-scale distributed-centralized hybrid energy systems at the urban level, embodiments of the present invention provide a method for hierarchical zoning, energy hub site selection, and grid planning of a regional energy internet, including:

[0038] S1. According to the preset correspondence between the energy supply or demand of energy nodes and the grid hierarchy, each energy node in the target area is assigned to the corresponding level of the grid (i.e., the energy grid); among them, the higher the level of the grid, the greater the energy supply or demand of the energy nodes.

[0039] S2, Under the constraint of the number of partitions in each layer of the network, with the goal of minimizing the overall loss of each layer of the network, the first target node of each layer of the network is partitioned sequentially from bottom to top, and the position of the hub node of each partition is determined. Then, each hub node is connected to the first target node of its partition to obtain the overall network.

[0040] The comprehensive loss of each layer of the network includes the line loss between the first target node and its hub node in each partition and the supply and demand balance loss of each partition. The first target node includes the energy node of the network layer and the hub node of the network layer above. The number of partitions of each layer of the network can be the same or different. Each partition of each layer of the network has one and only one hub node. That is, the number of partitions of the same layer of the network is the same as the number of hub nodes.

[0041] S3, execute S2 q times to obtain q overall network structures, select the overall network structure with the smallest line loss as the network structure to be adjusted, q≥1; when q>1, execute S2 again based on the overall network structure obtained in the previous execution of S2;

[0042] S4, with the goal of minimizing the line loss between the second target node and the hub node of each partition of each layer of the overall network structure, the position of the hub node of each partition of each layer of the overall network structure is adjusted, and each hub node after the position adjustment is connected to the second target node of its partition to obtain the regional energy internet of the target area.

[0043] The second target node includes energy nodes connected to hub nodes of each zone of each layer of the overall grid, as well as hub nodes of the upper and lower layers of the grid.

[0044] As a further preferred embodiment of the present invention, step S2, which involves partitioning the first target nodes of each layer of the network structure and determining the location of the hub nodes of each partition, includes:

[0045] A1. The set of first target nodes is used as a candidate set. A node is randomly selected from the candidate set as the first hub node and added to the hub node set. Then, the node with the largest sum of distances between the candidate set and the nodes in the hub node set is added to the hub node set, until the hub node set has k hub nodes. Here, k is the number of partitions in each layer of the network structure. That is, a node is randomly selected from the candidate set as the first hub node and added to the hub node set, and then removed from the candidate set. The distances between each remaining node in the candidate set and each node in the hub set are calculated sequentially. The node with the largest distance between the candidate set and each node in the hub node set is added to the hub node set and then removed from the candidate set. The distances are calculated again to update the hub node set and the candidate set, until the hub node set has k hub nodes.

[0046] A2, with the goal of minimizing the overall loss, divide the remaining nodes in the candidate set into k partition sets;

[0047] A3 aims to minimize the line loss from the first target node to its hub node in each partition set. It iteratively updates the position of each hub node based on the Cooper method and determines whether the iteration stopping condition has been met. If not, it updates the virtual energy of each hub node and returns to A2. If so, when k has been fully traversed, it takes the partition status corresponding to the number of partitions with the minimum comprehensive loss and the position of the hub node of each partition as the final result. If k has not been fully traversed, it returns to A1 until k has been fully traversed.

[0048] As a further preferred embodiment of the present invention, in step S2, after connecting each hub node to the first target node of its respective partition, the method further includes:

[0049] With the goal of minimizing the line loss between the first target node and the hub node in each partition of each layer of the network, the positions of the hub nodes in each partition of each layer of the network are adjusted sequentially from top to bottom, and each hub node after the position adjustment is connected to the first target node of its respective partition to update the connection relationship between the hub nodes and the first target node of each partition, thus obtaining the overall network; wherein, the position of each hub node is iteratively updated based on the Cooper method until the iteration stops.

[0050] As a further preferred embodiment of the present invention, the overall loss of each layer of the space frame is:

[0051] L l =M l +β l B l

[0052] Among them, L l M represents the overall loss of the l-th layer of the space frame. l For the line loss of the l-th layer network, Bl The supply and demand balance loss of the l-th layer of the space frame; β l β is the index balance coefficient of the l-th layer space frame. For the first overall space frame, β l As a preset value, for non-first overall space frame, B l ` is the supply and demand balance loss of the l-th layer of the preceding integral space frame (not the first integral space frame), M` l+1 ` represents the line loss of the (l+1)th layer of the preceding integral network frame, which is not the first integral network frame.

[0053] As a further preferred embodiment of the present invention, in step S4, the positions of the hub nodes of each partition of each layer of the grid to be adjusted are alternately adjusted along the first direction and the second direction, and the hub nodes after the position adjustment are connected to the second target node of the partition where they are located, so as to obtain the regional energy internet of the target area.

[0054] The first direction is either from top to bottom or from bottom to top, and the second direction is the other.

[0055] Specifically, such as Figure 1 As shown, the method provided in this embodiment of the invention is divided into the following two stages: (1) Preliminary planning stage: several "planning-adjustment" iterations are performed, and the result with the lowest overall cost is selected as the preliminary regional energy internet; (2) Fine-tuning stage: several "bottom-up adjustment-top-down adjustment" iterations are performed on the preliminary regional energy internet, and the location of energy hubs and grid lines are fine-tuned to converge the final regional energy internet.

[0056] In other words, the optimal preliminary energy network is first selected in the initial planning stage, and then the final energy network is obtained through the fine-tuning stage. This method consists of two basic steps: planning and adjustment. The following description, combined with text, figures, and formulas, will first describe the two basic steps of planning and adjustment in detail, and then further explain the two algorithmic stages of the preliminary planning stage and the fine-tuning stage.

[0057] 1. Basic Steps

[0058] We designed a two-stage regional energy internet planning algorithm, using planning and adjustment as the two basic steps. The planning step is the "bottom-up planning step," while the adjustment step is divided into two types: "bottom-up adjustment step" and "top-down adjustment step," as follows: Figure 2 As shown in (a) and (b) in the figure. We first describe in detail the algorithmic framework of the planning and adjustment steps in sections 1.1 and 1.2. The hierarchical planning algorithm used in both planning and adjustment steps, the partitioned location algorithm used in the planning step, and the location adjustment algorithm used in the adjustment step are explained in detail in sections 1.3-1.5.

[0059] It should be noted that the method provided by this invention is as follows: Figure 1 As shown, it is divided into a preliminary planning stage and a fine-tuning stage. The planning steps (in...) Figure 1 (abbreviated as planning) and adjustment steps (in Figure 1 The two phases, abbreviated as "adjustment" in Chinese, together constitute the preliminary planning phase and the fine-tuning phase.

[0060] 1.1 Planning Steps

[0061] The planning steps aim to design a new interconnected grid for energy nodes, such as Figure 3 As shown. The terminology used is explained as follows: Nodes are divided into energy nodes and hub nodes. Energy nodes are fixed nodes, and the location and energy information (energy demand or supply and specific power) of each energy node are known, such as residential users, small photovoltaic panels, large power plants, etc. According to the size, they can be further subdivided into distributed nodes (residential users, small photovoltaic panels, etc.) and centralized nodes (large power and gas stations, etc.). Hub nodes are nodes to be planned, which are artificially constructed network nodes during the planning process and are responsible for energy management and allocation.

[0062] Each partition has one and only one hub node. All energy nodes in the partition are connected through the hub node to realize energy transmission between them. In other words, the hub node can be regarded as an energy transfer point.

[0063] The node data format is uniformly set to (x 1 x 2 P 1 P 2 P 3 ), where (x 1 x 2 ) represents the two-dimensional position coordinates of the node, (P) 1 P 2 P 3 These represent the node's electrical, thermal, and gas power (using these three energy sources as examples; more energy forms can be represented by P). 4 P 5 ...direct expansion).

[0064] In the initial planning stage, the planning steps directly plan the initial energy nodes to obtain the overall grid structure. In subsequent planning stages, the planning steps refer to some energy information from the previous grid structure to plan the energy nodes and obtain the overall grid structure. The input to the planning steps is the initial nodes or the previous overall grid structure, and the output is the new overall grid structure. The planning steps have a hierarchical characteristic, and their hierarchical direction is "bottom-up". The planning steps are explained step by step below.

[0065] Sub-step 1: Layered planning: The input nodes are divided into layers according to their energy volume. Nodes with large energy supply or demand are assigned to the higher layers, while nodes with small energy supply or demand are assigned to the lower layers. By determining the energy supply or demand range of each layer, each node is assigned to the corresponding layer based on its energy supply or demand. The index balance coefficient for each layer is also determined.

[0066] For a hierarchical planning process divided into L layers (L is pre-selected as a hyperparameter, usually an empirical value; for example, for 2000 energy nodes, the preferred range of L is 5-10), its input is the initial node set or the previous network structure (i.e., the overall network structure obtained from the previous planning), and its output is: 1) the original node set of each layer (Layer1, Layer2, ..., Layer...). L Layer i 1) The original node set of the i-th layer (which can be an empty set; if there are no energy nodes in this layer, the hub node of the next layer is directly used as the energy node of this layer), which are disjoint and their union is the initial node set; 2) The index balance coefficients of each layer (β1, β2, ..., β... L This is used for multi-objective balancing during the zoning site selection process (in the previous network structure, it was only used to calculate the index balance coefficients for each layer; see Section 1.3 for details on the hierarchical planning algorithm). Its calculation and uses are detailed in Sections 1.3 and 1.4, respectively. See Section 1.3 for details on hierarchical planning.

[0067] Sub-step 2: Zonal Site Selection: Zonal site selection is performed for each level from bottom to top. The node set of that level is divided into zones, and energy hubs are deployed within each zone to construct a single-layer network structure that is interconnected within the zone. Except for the bottom layer, the original nodes of each layer are first merged with the hub nodes obtained after zonal site selection of the next layer to obtain the node set of that layer. The zonal site selection process is carried out under the guidance of the index balance coefficient β of that layer to obtain the single-layer network structure of that layer. For details of the zonal site selection process, please refer to section 1.4.

[0068] Sub-step 3: Space Frame Integration: The obtained single-layer space frames are merged and connected according to the corresponding hub relationships to obtain the overall space frame. It should be noted that... Figure 3 For readability reasons, the integration process of single-layer grids is explicitly represented in the code. However, in the actual algorithm, explicit grid integration is not required. As single-layer grids are generated from bottom to top, the overall grid is naturally obtained.

[0069] 1.2 Adjustment Steps

[0070] To significantly improve algorithm efficiency, the planning step employs a bottom-up "greedy strategy," utilizing only information from the current layer's energy nodes and lower-layer hubs, neglecting information from upper-layer hubs. Consequently, the resulting overall network structure does not have optimal line losses, leaving room for optimization. To reduce line losses, the adjustment step comprehensively utilizes information from both upper and lower-layer hubs. By adjusting the positions of energy hubs layer by layer, it modifies the overall network structure obtained from the planning step, bringing its overall line losses to a local optimum. The input to the adjustment step is the overall network structure obtained from the planning step, i.e., the network structure to be adjusted, and the output is the adjusted network structure.

[0071] The adjustment process involves a layer-by-layer adjustment of the energy hub's location. Based on the order of the adjustments, it is divided into "top-down adjustment steps" and "bottom-up adjustment steps," as follows: Figure 4 , Figure 5 As shown, by alternating between the two processes and gradually optimizing, the overall network line loss can be ensured to converge to a local optimum. The following clarification is provided regarding the terminology used: When adjusting the i-th layer hub, the positions of the other layer hubs are fixed and cannot be adjusted; these are called "fixed hubs," and the i-th layer hub is called the "hub to be adjusted." Unlike the planning step, which only uses the previous network information to calculate the balance coefficients of each layer for replanning, the adjustment step aims to optimize the network based on the network to be adjusted. Similar to the planning step, the adjustment step can also be divided into three sub-steps: layered planning, site selection adjustment, and network integration. The "top-down" / "bottom-up" directionality is mainly reflected in the layered planning part, and each sub-step is explained below.

[0072] Sub-step 1, Layered Planning: Layered planning aims to decompose the network structure to be adjusted into cross-layer network structures sequentially ("from top to bottom" adjustments to layers L and L-1, L-1 and L-2, etc., and "from bottom to top" adjustments to layers 1 and 2, 2 and 3, etc.), providing node information for this layer and hub information for the upper and lower layers for layer-by-layer site selection adjustments. Unlike the planning step, the adjustment step does not require partitioning operations, and for the single objective of reducing line losses, the indicator balance coefficient β is no longer needed.

[0073] Sub-step 2: Site Selection Adjustment: Site selection adjustment aims to adjust the location of the current-layer hubs based on the current-layer node information and the upper and lower-layer hub information (collectively referred to as zoning and network information). The zoning and network information is explained as follows depending on the adjustment direction: 1) "Top-down adjustment step": For the hub to be adjusted in layer l, the energy node information of layer l and the hub node information of layer l-1 are given by the hierarchical planning, and the hub node information of layer l+1 is given by the upper-layer site selection adjustment results (assuming the hub information of layer l and layer l+1 is empty), such as... Figure 4As shown; 2) "Bottom-up adjustment steps", for the hub to be adjusted in layer l, the energy node information of layer l and the hub node information of layer l+1 are given by the hierarchical planning, and the hub node information of layer l-1 is given by the upper-level site selection adjustment results (note that the hub information of layer 0 and layer L+1 is empty), such as Figure 5 As shown. Details of the site selection adjustment process can be found in Section 1.5.

[0074] Sub-step 3: Network integration: The obtained cross-layer network structures are merged and connected according to the corresponding hub relationships to obtain the adjusted network structure. It should be noted that in the actual algorithm, explicit network integration is not required. The adjusted network structure is naturally obtained by adjusting the location from top to bottom / bottom to top.

[0075] 1.3 Hierarchical Programming Algorithm

[0076] As described in sections 1.1 and 1.2, both the planning and adjustment steps involve hierarchical programming algorithms to process the problem hierarchically, but their functions and outputs differ slightly. In the actual algorithm, the overall network structure is stored hierarchically. The hierarchical programming algorithm in the adjustment step of section 1.2 only requires simple processing during traversal and is relatively easy to implement. This section focuses on explaining the details of the hierarchical programming algorithm in the planning step of section 1.1.

[0077] In the planning process, a hierarchical planning method is adopted to divide the overall network planning into layers. This is to: 1) plan distributed energy nodes and centralized energy nodes at different levels, solve the difference in energy volume between the two, and solve the distributed-centralized energy hybrid planning problem that is difficult to handle by traditional methods; 2) facilitate the use of a "greedy strategy" to carry out bottom-up layer-by-layer planning, which greatly reduces the complexity of the problem, improves the efficiency of the algorithm, and increases the scale of the problem solution; 3) in the layer-by-layer planning, the line wheel loss estimation is often more accurate than the pre-calculated line loss estimation matrix between all nodes.

[0078] In the adjustment process, a hierarchical planning method is adopted to divide the overall grid planning into layers. This is to decompose the grid to be adjusted into cross-layer grids in sequence, providing node information of the current layer and hub information of the upper and lower layers for layer-by-layer site selection and adjustment. There is no need to perform zoning operations or the indicator balance coefficient β is no longer required.

[0079] The output of the hierarchical programming algorithm is the original set of nodes (Layer1, Layer2, ..., Layer) for each layer. L ) and the index balance coefficients (β1, β2, ..., β) of each layer L The following explains how to solve both problems:

[0080] 1) The original node set of each layer (Layer1, Layer2, ..., Layer) LLayer i Let be the original set of nodes at level i (which can be empty), all of which are disjoint and whose union is the initial node set. Distributed nodes are assigned to lower-level node sets, and centralized nodes are assigned to higher-level node sets. If The i-th layer obtained from the planning steps only contains hub nodes. Its purpose is to establish a fine-grained network structure with multiple hubs. For a larger number of distributed nodes, a network structure with more layers can be used (i.e., a larger L with more nodes). Constructing more sophisticated and complex networks improves local energy consumption and avoids long-distance transportation. The existing node set partitioning method needs to be manually designed before algorithm implementation and flexibly adjusted according to actual needs and experience. For example, a "distributed node-" approach could be adopted. - -Centralized Nodes- The structure is a 5-layer structure. Alternatively, it can be divided into multiple layers of energy network based on the energy volume of each energy node; for example: distributed nodes of a first preset volume - distributed nodes of a second preset volume - - Centralized nodes of the third preset size - Centralized nodes of the fourth preset size

[0081] 2) Balance coefficients of each level index (β1, β2, ..., β) L β is used as a weighting coefficient in the partitioned location process to guide the balance between the two objective functions. In the partitioned location algorithm, we consider two losses: line loss M and supply-demand balance B (definition and calculation are detailed in Section 1.4). These two losses are weighted by the index balance coefficient β to obtain the comprehensive objective L. For example, in the l-th layer, we have L. l =M l +β l B l Let l = 1, 2, ..., L. In this bi-objective optimization problem, the supply and demand balance objective aims to promote local energy consumption to reduce high-rise line losses. The index balance coefficient β... l This will affect the solution to the problem; therefore, the hierarchical programming algorithm provides an appropriate β in two cases. l ① The input is the initial node set: β l ① Use a preset initial value, such as 0.5; ② The input is the previous grid structure (i.e., the overall grid structure obtained from the previous planning): use the previous grid structure to estimate β. l Based on the aforementioned meaning of the supply and demand balance target, let L l =M l +β l B l =M l +M l+1 It is estimated that (for the top β) L , ).

[0082] 1.4 Partition Addressing Algorithm

[0083] The partitioned site selection algorithm aims to partition the node set of this layer and deploy energy hubs within each partition, constructing a single-layer network structure with connectivity within the partition. For the l-th layer, the input to its partitioned site selection process is the index balance coefficient β. l and the set of nodes in this layer. l ′(Layer l +lower-level hub), the output is a single-layer network structure connected within the l-th layer partition. Here, the node at that layer refers to the first target node at that layer. Specifically, such as... Figure 6 As shown, the idea behind the partitioned site selection process is to model the single-layer network planning problem as a bi-objective optimization problem (the node set undergoes necessary preprocessing to unify units). The partitioned site selection algorithm is used to divide the node set into a specified number of partitions, and an energy hub is deployed within each partition, with its optimal location selected. Connecting the nodes within each partition to the corresponding energy hub yields the single-layer network structure. The following sections will first provide supplementary explanations of the symbols and terminology used in this subsection, then briefly describe the optimization problem modeling, and finally detail the solution process for the partitioned site selection algorithm.

[0084] Suppose there are n nodes in the current layer (referring to the first target node), and m energy forms (m=3 when considering electricity, gas, and heat). Consider dividing the n nodes into k partitions and deploying energy hubs in each partition. Let the coordinates of the i-th first target node be denoted as [i, y]. The set of energy or load nodes in the j-th partition is denoted as C. j The corresponding energy hub location is denoted as The energy demand / production power of node i is denoted as . Each zone has one and only one energy hub, meaning the number of energy hubs is the same as the number of zones.

[0085] We consider the line loss index M and the supply-demand balance index B as two optimization objectives. Both can be decomposed into the sum of the corresponding indices for each region, and both should be as small as possible.

[0086]

[0087]

[0088] The following is about M j and B j The calculations will be explained separately.

[0089] 1) For the line loss (i.e., line transport loss) index M of the j-th partition j It is equal to the line loss m between each first target node i (hereinafter referred to as node i) in the j-th partition and the energy hub. i sum:

[0090]

[0091] Based on the derivation, the line loss m between node i and energy hub o is... i The combined load moment can be calculated from the combined load moment of the two, where "combined" means the sum of the load moments of each energy form r:

[0092]

[0093] Where D io pass Calculate the Euclidean distance, where o is the energy hub in energy partition j. Defined as:

[0094]

[0095] Where, η r It is the transmission efficiency of energy form r.

[0096] 2) For the supply and demand balance index B of the j-th partition j We use the absolute value of the total energy in a region to measure the supply and demand balance in a zone.

[0097] For partition j, the total energy of form r in the partition is defined as:

[0098]

[0099] The overall supply and demand balance of partition j is defined as follows:

[0100]

[0101] B j The smaller the value, the better the source load balancing in partition j.

[0102] The bi-objective optimization problem is now transformed into a single-objective optimization problem and formalized. By linearly weighting the line loss index and the supply-demand balance index according to the balance coefficient β obtained from hierarchical planning, the comprehensive objective L is expressed as:

[0103]

[0104] L j =M j +βB j

[0105] Where L j This is the overall loss for partition j. All partition node sets are labeled as vector C, and the coordinates of all energy hubs are labeled as vector X:

[0106] C = [C1, C2, ..., Ck ]

[0107]

[0108] We formalize the optimization problem into the following comprehensive objective minimization problem:

[0109]

[0110] Expanding the objective function, the complete objective function can be expressed as:

[0111]

[0112] The above describes the optimization problem modeling process. It can be proven that this problem is NP-hard and can be solved using existing methods such as integer programming and mixed integer programming. However, for large-scale energy internet planning problems, the direct solution time exceeds acceptable limits. Therefore, this invention designs a partitioned location algorithm, which, as a fast and high-quality heuristic algorithm, can significantly improve solution efficiency while maintaining good solution performance. The main process and sub-processes of the partitioned location algorithm are as follows: Figure 7 As shown, the energy hub location initialization, zoning, site selection, and zoning energy index smoothing algorithms are respectively as follows: Figure 8-11 As shown, the main process of the partitioning addressing algorithm is divided into an outer loop and an inner loop. Based on the objective function described above, the algorithm needs to optimize the parameters k, C, and X. The outer loop is responsible for optimizing parameter k, and the inner loop is responsible for optimizing parameters C and X. The inner and outer loops are explained below.

[0113] 1) Inner Loop: Given a specified number of partitions k, optimize C and X, i.e., solve for the optimal partitioning and optimal energy hub location. The inner loop is an iterative algorithm. In each iteration, it sequentially executes the partitioning, location selection, and partition energy exponential smoothing steps. Before executing the partition energy exponential smoothing step, it checks if the inner loop stopping condition is met. If it is, the loop exits. The inner loop stopping condition is that the iteration reaches the specified maximum number of rounds, or the change in the overall loss L after that round of iterations is less than a specified threshold (to exit the loop early).

[0114] 2) Outer loop: within the specified [K] min K max The inner loop iterates through the number of partitions k, obtaining the minimum overall loss for that number of partitions. After the iteration, if the overall loss first decreases and then increases with increasing k, then the number of partitions k′ with the minimum overall loss is selected as the optimal number of partitions; if the overall loss decreases with increasing k, then the number of partitions k′ with the minimum overall loss is selected (e.g., ...). Figure 12 The number of partitions k′ shown is the optimal number of partitions. Figure 12In the figure, the overall loss always shows a decreasing trend as k increases. When the number of partitions k = 5, the decrease in overall loss tends to level off. Therefore, k′ = 5 is chosen as the optimal number of partitions.

[0115] The following sections will provide a detailed introduction to the steps of energy hub initialization, zoning, site selection, and zoning energy index smoothing in the internal circulation.

[0116] 1) Energy Hub Initialization: Energy hub initialization aims to select suitable node locations from the node set as the initial energy hub locations. First, the candidate node set is the entire node set (i.e., the entire energy node set). A node location is randomly selected from the candidate node set as the first energy hub location, added to the initial hub set, and then removed from the candidate node set. To select the next hub location, the distances between each node in the candidate node set and each hub in the hub set are calculated sequentially. The hub node with the closest distance to each candidate node is determined. Then, the distance between each candidate node and its closest hub node is calculated again. The candidate node with the largest distance is selected as the next hub node, added to the hub set, and then removed from the candidate node set. The distance matrix is ​​then recalculated, and this process is repeated until the locations of k hubs are initialized.

[0117] That is, assuming there are 2 nodes {i1, i2} in the current candidate set and 3 nodes {o1, o2, o3} in the current hub set, for node i1, calculate its distances {d1, d2, d3} to each node in the hub set, and select the distance with the smallest value among {d1, d2, d3} as i1's "target distance"; similarly, for node i2, calculate its distances {d1`, d2`, d3`} to each node in the hub set, and select the distance with the smallest value among {d1`, d2`, d3`} as i2's "target distance"; then, compare the "target distance" of i1 with the "target distance" of i2. If the "target distance" of i1 is greater than the "target distance" of i2, then i1 is selected as the next hub node, otherwise i2 is selected as the next hub node.

[0118] 2) Partitioning: The partitioning step aims to divide the node set into k partitions using the hub set information. Initially, all partitions are empty sets. Starting from i=1, the node set is traversed. To determine which partition to assign node i to, the comprehensive loss {l1, l2, ..., l...} for each partition is calculated. j , ..., l k}, the comprehensive loss for the j-th partition is l j The calculation is as follows:

[0119] l j =m j +βb j

[0120] Where β is the index balance coefficient, mj Here is the line loss metric from the node to the j-th partition:

[0121]

[0122] b j Let be the supply and demand balance index from node to the j-th partition (defined here as hinge supply and demand balance loss). (See definition above):

[0123]

[0124] Assign node i to the partition with the minimum overall loss and update the partition. The energy form of energy hub j is virtual energy r (see the partitioned energy index smoothing section below). Then, the same partitioning operation is performed on the next node until all nodes are partitioned.

[0125] 3) Location Selection: The location selection step aims to relocate the hub location for each partition. Each partition is traversed, and the Cooper method is used to iterate the energy hub location for each partition. Iteration stops when the Cooper iteration stopping condition is met. The same method is used to relocate energy hubs for all partitions. The Cooper iteration stopping condition is that the iteration reaches a specified number of rounds, or the change in hub location before and after a round of iterations is less than a specified threshold (early exiting the iteration). For the j-th partition, its Cooper iteration formula is derived as follows:

[0126]

[0127] in, Let represent the coordinates of the j-th partition hub Cooper during the t-th iteration. The coordinates of the (t+1)-th iteration are obtained by iterating from the coordinates of the t-th iteration.

[0128] The derivation process is as follows: For the objective function proposed above:

[0129]

[0130] Its for For convex optimization problems, respectively... Taking the derivative and setting it to 0, we get:

[0131]

[0132]

[0133] Simplifying, we get:

[0134]

[0135]

[0136] Based on the principle of fixed-point iteration, it can be transformed into an iterative formula, namely:

[0137]

[0138] The initial value for the iteration is:

[0139]

[0140]

[0141] 4) Partition Energy Exponential Smoothing: The site selection step re-selects the location coordinates for the energy hub, but the hub itself has no energy power. The inner loop is essentially a local optimization algorithm. Due to the local smoothness of the solution space, partitions are similar before and after each iteration of the inner loop. Therefore, the historical energy state of partition j can serve as prior information, guiding which energy nodes can be assigned to partition j to help balance the source load of the partition. For example, if the historical energy state of a partition is always insufficient, the next iteration should aim to add energy supply nodes. This historical information can be achieved by setting a certain initial energy value when there are no nodes in the partition, storing it as "virtual energy information" for the partition hub. Typically, this historical information is not suitable for using the partition energy situation of the previous iteration, which will lead to poor algorithm convergence or non-ideal solutions. We use partition energy exponential smoothing to achieve this memory:

[0142]

[0143]

[0144] in, This refers to the virtual energy information of hub j during the t-th round of internal circulation, i.e., the energy memory of the partition.

[0145] 1.5 Location Adjustment Algorithm

[0146] The location adjustment algorithm adjusts the position of the current layer hub node based on the partition and network structure information (specifically, including the current layer energy node, the upper layer hub, and the lower layer hub). The process is similar to the location selection steps in the partitioned location algorithm, as follows: Figure 13 As shown. This corresponds to the partition set in the location selection step. Unlike the partitioning step, the location adjustment step uses all nodes connected to hub node j (including the current layer energy node, the upper layer hub, and the lower layer hub) as the node set C of partition j. jThe Cooper algorithm is then used for site selection. Since the site selection problem is convex, it can be proven that after site selection adjustment, if the line loss of the original network structure has not yet reached its minimum, the line loss of the cross-layer network structure will definitely be less than that of the original network structure, ensuring that the algorithm converges to the optimal network structure under this partitioning condition.

[0147] 2. Algorithm Stage

[0148] The above section outlines the algorithm's two basic steps: planning and tuning, along with detailed explanations of each sub-process. Our complete algorithm uses planning and tuning as its two fundamental steps, divided into two phases: preliminary planning and fine-tuning. Figure 1 , 2 As shown in the diagram. First, through the preliminary planning stage, several preliminary network structures are obtained as candidates. From these, the preliminary network structure with the lowest overall line loss is selected. Then, through the fine-tuning stage, the network structure is further adjusted to gradually reduce line losses until the network structure converges and stabilizes. The following explains how to construct the preliminary planning and fine-tuning stages through planning and adjustment steps.

[0149] 2.1 Preliminary Planning Stage

[0150] The preliminary planning phase aims to obtain a series of preliminary network structures as candidates, and then select the optimal network structure (i.e., the one with the lowest overall line loss) for the fine-tuning phase. Preliminary planning uses a basic planning process consisting of "one bottom-up planning step + one bottom-up adjustment step." One basic planning process yields one preliminary network structure. Executing q basic planning processes sequentially yields q preliminary network structures. Finally, the network structure with the lowest overall line loss is selected from these q preliminary network structures as the optimal initial network structure, which serves as the input for the fine-tuning phase (q is a hyperparameter, adjusted according to actual needs; an appropriate q value ensures a good preliminary network structure without excessive computation).

[0151] Specifically, a basic planning process consists of "one bottom-up planning step + one bottom-up adjustment step." The planning step is responsible for replanning the grid structure for all energy nodes, while the adjustment step makes necessary adjustments to the resulting grid structure. Since the preliminary planning only needs to obtain a preliminary grid structure, one adjustment step is sufficient. Therefore, "one bottom-up planning step + one bottom-up adjustment step" is considered a basic planning process. When executing the basic planning process sequentially, because the planning process only uses the previous grid structure information to estimate the balance coefficient β of each layer and replans the grid structure for the energy nodes, the grid structure obtained in each basic planning process has a certain degree of randomness, and their quality is not necessarily related. Therefore, it is necessary to ultimately select the optimal grid structure from these initial grid structures. However, at the same time, because the planning step estimates the balance coefficient β of each layer, as the preliminary planning progresses, the estimated value of β for each layer gradually converges. The β values ​​in later basic planning processes are more accurate, and the resulting preliminary grid structures often have better performance.

[0152] 2.2 Fine-tuning stage

[0153] The fine-tuning phase aims to fine-tune the selected optimal initial network layer by layer, relocating the hub nodes at each layer to converge to the optimal network for that connection relationship. The fine-tuning phase uses a basic fine-tuning process of "one bottom-up adjustment step + one top-down adjustment step," executing this process sequentially until the network converges. The algorithm's convergence is theoretically guaranteed. The bottom-up and top-down adjustments are alternated because the result of one adjustment step cannot be further optimized through adjustments in the same direction. Alternating optimization in different directions ensures that the overall line loss decreases with each result (even when the network is still not optimal), ultimately converging to the network with the minimum line loss for that connection relationship (i.e., the optimal hub location at each layer).

[0154] This invention provides a method and system for hierarchical zoning, energy hub site selection, and grid planning of a regional energy internet, comprising: a computer-readable storage medium and a processor;

[0155] The computer-readable storage medium is used to store executable instructions;

[0156] The processor is configured to read executable instructions stored in the computer-readable storage medium and execute the method as described in any of the above embodiments.

[0157] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for hierarchical zoning, energy hub site selection, and grid planning of a regional energy internet, characterized in that: include: S1. According to the preset correspondence between the energy supply or demand of energy nodes and the grid hierarchy, each energy node in the target area is assigned to the corresponding level of the grid; among them, the higher the level of the grid, the greater the energy supply or demand of the energy nodes. S2, Under the constraint of the number of partitions in each layer of the network, with the goal of minimizing the overall loss of each layer of the network, the first target node of each layer of the network is partitioned sequentially from bottom to top, and the position of the hub node of each partition is determined. Then, each hub node is connected to the first target node of its partition to obtain the overall network. The comprehensive loss of each layer of the network includes the line loss between the first target node and its hub node in each partition and the supply and demand balance loss of each partition. The first target node includes the energy node of the network layer and the hub node of the network layer above. Each partition of each layer of the network has one and only one hub node. S3, execute sequentially q S2 was obtained q From the overall network structure, the network structure with the lowest line loss is selected as the network structure to be adjusted. q ≥1; when q When the value is >1, execute S2 based on the previous overall grid structure; S4, with the goal of minimizing the line loss between the second target node and the hub node of each partition of each layer of the overall network structure, the position of the hub node of each partition of each layer of the overall network structure is adjusted, and each hub node after the position adjustment is connected to the second target node of its partition to obtain the regional energy internet of the target area. The second target node includes energy nodes connected to hub nodes of each zone of each layer of the overall grid and hub nodes of the upper and lower layers of the grid. In step S2, the step of partitioning the first target node of each layer of the network structure and determining the location of the hub node of each partition includes: A1. Take the set of first target nodes as a candidate set, randomly select a node as the first hub node and put it into the hub set, and determine the target distance of each remaining node in the candidate set. Put the remaining node with the largest target distance into the hub set. The target distance is the minimum distance between each remaining node and each node in the hub set, until the number of hub nodes in the hub set is equal to the number of partitions of the network structure at this layer. A2, with the objective of minimizing the overall loss, divides the remaining nodes in the candidate set into... k A set of partitions; A3 aims to minimize the line loss from the first target node to its hub node in each partition set. It iteratively updates the position of each hub node using the Cooper method and determines if the iteration stops. If not, it updates the virtual energy of each hub node and returns to A2; otherwise, it proceeds... k When the entire traversal has been completed, the partition details corresponding to the number of partitions with the minimum overall loss and the location of the hub node in each partition are taken as the final result. k If the traversal is not complete, return to A1, until... k Complete traversal; The iterative update of the position of each hub node based on the Cooper method includes: The positions of each hub node are iteratively updated according to the Cooper iteration formula, which is: ； ； in, Indicates the first The hub node of the partition During the round of iteration, the coordinates are obtained through the first iteration. The coordinate iteration of the wheel yields the first... Coordinates during the cycle , This is the first r The transmission efficiency of this type of energy Indicates the first The first target node The power produced or demanded by a certain form of energy For the first The coordinates of the first target node, For the total number of energy forms, For the first The first target node set in the partition; The virtual energy of the hub node is: in, t For the number of iterations, For the first t At the +1st iteration, the... j The hub node of the first partition r Virtual energy of various types of energy For the first t During the nth iteration j The hub node of the first partition r Virtual energy of various types of energy For the first t During the nth iteration j In the partition, the th r The total energy of all types of energy This is the predetermined smoothing coefficient for the energy index of the partition. .

2. The method as described in claim 1, characterized in that, In step S2, after connecting each hub node to the first target node of its respective partition, the process further includes: With the goal of minimizing the line loss between the first target node and the hub node in each partition of each layer of the network, the positions of the hub nodes in each partition of each layer of the network are adjusted sequentially from top to bottom, and each hub node after the position adjustment is connected to the first target node of its partition to obtain the overall network; wherein, the position of each hub node is iteratively updated based on the Cooper method until the iteration stops.

3. The method as described in claim 1, characterized in that, The iteration stopping condition is: the iteration reaches the required number of iterations or the change in the overall loss after this round of iterations is less than a preset threshold.

4. The method as described in claim 1 or 2, characterized in that, The load moment between the first target node and its hub node in each partition is taken as the line loss between them. The supply and demand balance loss between the first target node and its hub node in each partition is: in, For the first j In the partition, the th r The total energy of all types of energy For nodes No. Types of energy demand or production capacity, For the first j The hub node of the first partition r Virtual energy of various types of energy .

5. The method as described in claim 1 or 2, characterized in that, The overall loss of each layer of the space frame is: L l = M l + β l B l in, L l For the first l The overall loss of the multi-layered space frame, M l For the first l Line loss of the multi-layer network structure B l For the first l Supply and demand imbalance loss of the multi-layered space frame; β l For the first l The index balance coefficient of the layered space frame, for the first overall space frame, β l As a preset value, for non-first overall space frame, β l = , The first integral space frame preceding the first integral space frame (not the first integral space frame) l The supply and demand imbalance of the multi-layered space frame, The first integral space frame preceding the first integral space frame (not the first integral space frame) l +1 layer of network structure line loss.

6. The method as described in claim 1, characterized in that, In step S4, the positions of the hub nodes of each section of each layer of the grid to be adjusted are alternately adjusted along the first direction and the second direction, and the adjusted hub nodes are connected to the second target node of their respective sections to obtain the regional energy internet of the target area. The first direction is either from top to bottom or from bottom to top, and the second direction is the other.

7. A system for hierarchical zoning, energy hub site selection, and grid planning of a regional energy internet, characterized in that: include: Computer-readable storage media and processors; The computer-readable storage medium is used to store executable instructions; The processor is configured to read executable instructions stored in the computer-readable storage medium and execute the method as described in any one of claims 1-6.

Images