Power distribution network partition and transformer substation power-carbon collaborative planning method and device considering multi-terminal uncertainty

Abstract

The invention relates to a power distribution network partition and transformer substation power-carbon collaborative planning method considering multi-terminal uncertainty. The method comprises the following steps: extracting a representative wind-light output curve and a load curve; constructing an electrical coupling degree index, a load carbon flux index and a load balance degree index, calculating a comprehensive division index, and performing dynamic partitioning; and planning the transformer substation capacity which meets the main transformer N-1 verification, considers the influence of the load coincidence rate and the load prediction coefficient, and can meet the maximum load demand in the network element. The node carbon flux clustering index is combined with the traditional electrical coupling degree and load balance in the partition index, the limitation that traditional partition only depends on the electrical topological relation is broken through, the real-time carbon emission intensity of the load nodes is accurately quantified, and balanced partition of the power distribution network elements is achieved; carbon loss can be obviously reduced, and joint optimization of low carbon emission and high safety in the planning process of the transformer substation is realized; a typical scene expected value curve is used as input, and the influence of uncertainty on power distribution network planning is remarkably reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power distribution network and substation coordination, and in particular to a power distribution network partitioning and substation electric-carbon coordinated planning method and device considering multi-terminal uncertainty. BACKGROUND

[0002] With the construction of new energy high proportion access to new power system, as an important part of the power system, the power distribution network is evolving from the traditional power transmission and distribution channel to the integrated operation platform of power flow, carbon flow and information flow coupling. However, the traditional power distribution network partitioning and substation planning method has significant limitations in dealing with this change, and it is difficult to meet the urgent needs of current low-carbon development, and lacks systematic consideration of carbon emission reduction targets. The core defects mainly manifest in the following aspects: first, the traditional non-partition direct planning lacks coordination, which easily causes overlapping of substation layout or power supply vacuum, causing uneven utilization of equipment and long-distance power supply across regions, which not only increases invalid investment but also greatly increases system carbon loss due to line loss; second, the global unified planning cannot respond to local load characteristics and carbon emission differences, leading to blurred carbon responsibility traceability, making it difficult to deploy precise carbon reduction strategies for high-carbon areas, resulting in insufficient control precision; finally, the traditional partitioning and site selection method also has limitations - partitioning is only based on topological connection or electrical distance to divide the power supply area, and site selection focuses on geographical accessibility and load center to shorten the power supply distance, neither of which considers the overall optimization of carbon emissions, making it difficult to adapt to the planning needs under the low-carbon target.

[0003] How to embed carbon constraints into the whole process of power distribution network "partitioning-site selection-sizing" coordinated planning, while ensuring power supply safety and minimizing carbon emissions, has become a key problem to be solved at present. SUMMARY

[0004] To solve the problems of poor spatial coordination and ignoring the spatial heterogeneity of carbon flow in traditional substation planning methods, the primary purpose of the present application is to provide a power distribution network partitioning considering low-carbon target, which combines node carbon flux clustering index with traditional electrical coupling degree and load balancing in partitioning index, accurately quantifies the real-time carbon emission intensity of load nodes, and realizes the balanced division of power distribution network elements.

[0005] To achieve the above purpose, the present application adopts the following technical solution: a power distribution network partitioning and substation electric-carbon coordinated planning method considering multi-terminal uncertainty, which comprises the following sequential steps:

[0006] (1) In view of the uncertainty of the output fluctuation of wind and light and the load change, a Monte Carlo simulation technology is used to generate an initial scene set, and a K-means clustering algorithm is introduced to optimize and screen the initial scene set, and representative wind and light output curves and load curves are extracted;

[0007] (2) According to the representative wind and light output curves and load curves, an electrical coupling degree index, a load carbon flux index and a load balance degree index are constructed, a comprehensive division index is calculated, dynamic partitioning is carried out, and a final network element partitioning result is obtained;

[0008] (3) According to the final network element partitioning result, when planning a substation, the newly-built substation site is determined according to the principle of minimum line loss, that is, minimum carbon loss; then, considering the load demand and constraint conditions, the substation capacity that meets the main transformer N-1 check and considers the load simultaneous rate and load prediction coefficient influence and can meet the maximum load demand in the network element is planned.

[0009] Step (1) specifically includes the following sequential steps:

[0010] (1a) The probability characteristics of wind power output are modeled by Weibull distribution, and the probability density function of Weibull distribution is represented as:

[0011] ;

[0012] In the formula, wind speed, is the shape parameter of wind speed distribution, is the scale parameter of wind speed;

[0013] Based on Weibull distribution, combined with a wind speed-power segmented conversion model, the wind speed sequence is mapped to the fan output; when the wind speed is lower than the cut-in wind speed or higher than the cut-out wind speed , the fan output is zero; when the wind speed is between the cut-in wind speed and the rated wind speed , the output increases linearly with the wind speed; after reaching the rated wind speed , the fan output remains at the rated power ;

[0014] A Monte Carlo simulation technology is used to generate a first initial scene set: first, the Weibull distribution parameters are fitted based on historical wind speed data, a large number of wind speed sequences are randomly sampled and converted to output curves to form a first initial scene set;

[0015] The K-means clustering algorithm is introduced to reduce the first initial scene set: randomly initialize the cluster center, calculate the Euclidean distance between each scene and the center, assign the scene to the nearest neighbor cluster, iteratively update the center until convergence, finally take the cluster center as the typical scene, and determine the probability weight according to the proportion of scenes in the cluster, obtain the representative output of the wind power composed of cluster centers and the occurrence probability corresponding to each typical scene, and finally obtain a comprehensive expected output curve of wind power by weighting;

[0016] The probability distribution of photovoltaic output is modeled by Beta distribution, and the probability density function of Beta distribution is ;

[0017] ;

[0018] In the formula, is the period irradiance, and are shape parameters related to light intensity and weather changes, based on Beta distribution, the photovoltaic output is mapped to the actual output power by the irradiance-power conversion model;

[0019] Monte Carlo simulation technology is used to generate the second initial scene set: first, the Beta distribution parameters are fitted based on the historical irradiance data, a large number of irradiance sequences are generated by random sampling, and the photovoltaic output curve is converted to form the second initial scene set;

[0020] Then the K-means clustering algorithm is introduced to optimize and select the second initial scene set, extract representative typical scenes, obtain the representative output of photovoltaic composed of cluster centers and the occurrence probability corresponding to each typical scene, and finally obtain a comprehensive expected output curve of photovoltaic by weighting;

[0021] The probability distribution of electrical load obeys normal distribution, and the probability density function of normal distribution is ;

[0022] ;

[0023] In the formula, is the load power, is the expected value, is the standard deviation;

[0024] Monte Carlo simulation technology is used to generate the third initial scene set: first, the normal distribution parameters are fitted based on the historical load data, and a large number of load power sequences are generated by random sampling to form the third initial scene set;

[0025] Then the K-means clustering algorithm is introduced to optimize and select the third initial scene set, obtain a group of load output curves and their corresponding occurrence probabilities, and finally obtain a load fluctuation curve by weighting.

[0026] The wind power comprehensive expected output curve, the photovoltaic comprehensive expected output curve and the load fluctuation curve constitute the wind-solar power output curve and the load curve.

[0027] Step (2) specifically comprises the following sequential steps:

[0028] (2a) constructing an electrical coupling degree index : based on the power distribution network topology structure and electrical parameters, the electrical coupling degree between nodes is calculated by using a power transmission distribution factor, and a non-directional weight matrix reflecting the electrical coupling strength is constructed : the equivalent impedance between nodes is calculated through an impedance matrix to generate an electrical coupling degree matrix, then the reciprocal of the electrical coupling degree is taken as the weight, the diagonal elements are set to zero, and finally a non-directional weight matrix with the electrical coupling degree as the core is formed ;

[0029] The impedance matrix is obtained by inverting the node conductance matrix : the resistance matrix is first calculated, then the conductance matrix is calculated according to the resistance matrix , ; then the node conductance matrix is constructed, wherein the diagonal elements of the node conductance matrix are the conductance negatives of the corresponding nodes, and the non-diagonal elements are the conductances themselves; finally, the impedance matrix is obtained by inverting the node conductance matrix ;

[0030] (2b) constructing a load carbon flux index : the power flow calculation result is obtained by power flow calculation, the node net load in the power flow calculation result is used, and the carbon trace factor obtained according to the main network carbon emission intensity time curve is used to calculate the average load carbon flux of the node:

[0031] ;

[0032] In the formula, is the average load carbon flux of node i, is the number of time periods;

[0033] An adjacency matrix reflecting the spatial similarity of carbon emissions is constructed , realizing the spatial coupling representation of carbon flow characteristics; is the average load carbon flux of node j;

[0034] The is linearly superimposed on the according to a preset ratio to generate a comprehensive adjacency matrix The preset ratio is 4.2:1.8.

[0035] (2c) Constructing the load balancing index: introducing the load balancing index , the sum of the average of the net loads of all load nodes is obtained to obtain the total load; secondly, the difference square sum of the actual load and the ideal load of each network element is calculated to measure the overall imbalance degree; finally, the imbalance degree is converted into a standardized index between 0 and 1 and the inverse number is taken to represent The greater the index is, the more conducive to the load balancing among the network elements; the ideal load is the total load divided by the number of network elements;

[0036] (2d) Constructing the comprehensive division index: using the community detection algorithm, taking the comprehensive adjacency matrix as the input, constructing the comprehensive division index as follows:

[0037] ;

[0038] In the formula, and are weight coefficients;

[0039] is the modularity of the fusion of the electrical coupling and the carbon flux, and the formula is:

[0040] ;

[0041] In the formula, represents the total edge weight of the network, ; represents the node belonging characteristic value, if the nodes and belong to the same network element, then ; otherwise, ;

[0042] When performing dynamic partitioning: firstly, initial division is performed, and each node is independently defined as an initial community; secondly, iteration merging is performed, a node is randomly selected, and the increment of the comprehensive division index after merging with adjacent nodes is calculated; if there is an adjacent community that can improve the comprehensive division index , then the node is merged into the community with the largest increment; finally, convergence judgment is performed, and the merging operation is repeated for all nodes until the comprehensive division index cannot be further optimized, and the final network element partitioning result is output.

[0043] Step (3) specifically includes the following sequential steps:

[0044] Firstly, based on the node load distribution and equivalent impedance in the network element, a carbon loss minimization objective function is constructed to realize dynamic screening of optimal substation site:

[0045] ;

[0046] In the formula, is the mean value of node net load; represents the equivalent impedance from the node to the candidate substation site; is the loss coefficient; is the total carbon loss; G is a set composed of substation power supply nodes;

[0047] Secondly, based on the maximum active load in the region , combined with the load simultaneous rate and the long-term reservation coefficient , the total apparent power is calculated:

[0048] ;

[0049] ;

[0050] In the formula, N is the number of main transformer stations; S is the rated capacity of a single main transformer; is the power factor;

[0051] According to , the substation capacity configuration that meets the maximum load demand in each network element, considers the influence of the load simultaneous rate and the long-term reservation coefficient , and meets the N-1 safety criterion is obtained.

[0052] Another purpose of the present application is to provide an electronic device comprising:

[0053] a processor; and

[0054] a memory, in which computer program instructions are stored, the computer program instructions, when executed by the processor, causing the processor to perform the power distribution network partitioning and substation electric carbon collaborative planning method considering multi-terminal uncertainty as described above.

[0055] The present application also provides a computer readable storage medium having computer program instructions stored thereon, the computer program instructions, when executed by a processor, causing the processor to perform the power distribution network partitioning and substation electric carbon collaborative planning method considering multi-terminal uncertainty as described above.

[0056] It can be seen from the above technical solution that the beneficial effects of the present invention are: First, the present invention proposes a distribution network partitioning that takes low-carbon goals into consideration. In the partitioning index, the node carbon flux clustering index is combined with the traditional electrical coupling degree and load balancing, breaking through the limitation of traditional partitioning that only relies on electrical topology relationships, accurately quantifying the real-time carbon emission intensity of load nodes, and realizing balanced division of distribution network network elements; Second, the present invention optimizes the site with the minimum total carbon loss as the objective function. Compared with the traditional reliance on a single economic site selection or geographical site selection, it can significantly reduce carbon loss and realize the joint optimization of low-carbon emissions and high safety in the substation planning process; Third, the present invention constructs a joint probability distribution model of wind, light and load, and efficiently converts complex multi-dimensional and multi-endpoint uncertainty problems into a representative multi-probability scenario set based on the K-means clustering method; while greatly compressing the data scale and improving computing efficiency, the key probability characteristics of the original uncertainty are retained to the greatest extent; using the typical scenario expected value curve as input, the risk of capacity redundancy or insufficiency caused by deterministic planning is avoided, and the impact of uncertainty on distribution network planning is significantly reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 is a flow chart of the method of the present invention;

[0058] Figure 2 Schematic diagram of the network element division method in the present invention. DETAILED DESCRIPTION

[0059] like Figure 1 As shown, a method for coordinated planning of distribution network zoning and substation electricity and carbon emissions taking into account multi-terminal uncertainties includes the following steps in sequence:

[0060] (1) In view of the uncertainty of wind and solar power output fluctuations and load changes, the Monte Carlo simulation technology is used to generate the initial scenario set, and the K-means clustering algorithm is introduced to optimize and screen the initial scenario set to extract representative wind and solar power output curves and load curves;

[0061] (2) Based on the representative wind and solar power output curve and load curve, the electrical coupling index, load carbon flux index and load balance index are constructed, the comprehensive partitioning index is calculated, dynamic partitioning is performed, and the final network element partitioning result is obtained;

[0062] (3) Based on the final network element zoning results, the site selection for the new substation is determined based on the principle of minimum line loss, i.e., minimum carbon loss, when planning the substation. Then, considering the load demand and constraints, the substation capacity is planned to meet the N-1 verification of the main transformer, the influence of the load simultaneity rate and the load prediction coefficient, and the maximum load demand within the network element.

[0063] Step (1) specifically includes the following steps in order:

[0064] (1a) The probability characteristics of wind power output are modeled by Weibull distribution, and the probability density function of Weibull distribution is represented as:

[0065] ;

[0066] In the formula, is the wind speed, is the shape parameter of the wind speed distribution, is the scale parameter of the wind speed;

[0067] Based on the Weibull distribution, combined with the wind speed-power segmented conversion model, the wind speed sequence is mapped to the fan output; when the wind speed is lower than the cut-in wind speed or higher than the cut-out wind speed , the fan output is zero; when the wind speed is between the cut-in wind speed and the rated wind speed , the output increases linearly with the wind speed; after reaching the rated wind speed , the fan output remains at the rated power ;

[0068] The Monte Carlo simulation technology is used to generate the first initial scene set: first, the Weibull distribution parameters are fitted based on the historical wind speed data, a large number of wind speed sequences are randomly sampled and converted into output curves to form the first initial scene set; Monte Carlo simulation approximates the true probability distribution through large sample sampling, but the number of scenes generated is large, which leads to a decrease in subsequent calculation efficiency, therefore, the K-means clustering algorithm is introduced to reduce the scenes.

[0069] The K-means clustering algorithm is introduced to reduce the first initial scene set: randomly initialize the cluster center, calculate the Euclidean distance between each scene and the centroid, assign the scene to the nearest neighbor cluster, iteratively update the centroid until convergence, finally take the cluster centroid as the typical scene, and determine the probability weight by the proportion of the scene in the cluster, obtain the representative output of wind power composed of cluster centroids and the occurrence probability corresponding to each typical scene, finally weighted to obtain a comprehensive expected output curve of wind power;

[0070] (1b) The probability distribution of photovoltaic power output is modeled by Beta distribution, and the probability density function of Beta distribution is represented as:

[0071] ;

[0072] In the formula, is the period irradiance, and respectively, are shape parameters related to the light intensity and weather change, based on Beta distribution, the photovoltaic output is mapped to the actual output power through the irradiance-power conversion model;

[0073] The second initial scene set is generated by using the Monte Carlo simulation technology: firstly, Beta distribution parameters are fitted based on historical irradiance data, a large number of irradiance sequences are generated by random sampling and are converted into photovoltaic output curves to form the second initial scene set;

[0074] Then, the K-means clustering algorithm is introduced to optimize and screen the second initial scene set, representative typical scenes are extracted, and a photovoltaic representative output composed of cluster centers and the occurrence probability corresponding to each typical scene are obtained, and finally a comprehensive expected photovoltaic output curve is weighted;

[0075] The influence of typical photovoltaic scene quantization on light fluctuation on carbon emission and equipment capacity is provided, and scene support is provided for photovoltaic power system planning and operation.

[0076] (1c) the probability distribution of the electrical load obeys a normal distribution, and the probability density function of the normal distribution is represented as:

[0077] ;

[0078] In the formula, P is the load power, is the expected value, is the standard deviation;

[0079] The third initial scene set is generated by using the Monte Carlo simulation technology: firstly, normal distribution parameters are fitted based on historical load data, a large number of load power sequences are generated by random sampling to form the third initial scene set;

[0080] Then, the K-means clustering algorithm is introduced to optimize and screen the third initial scene set, a group of load output curves and the corresponding occurrence probability are obtained, and finally a load fluctuation curve is weighted; the typical load scene provides reliable input for distribution network partitioning and substation planning, and a multi-end uncertainty joint analysis framework is constructed combined with the typical wind power and photovoltaic scenes.

[0081] The wind power comprehensive expected output curve, the photovoltaic comprehensive expected output curve and the load fluctuation curve constitute the wind and light output curve and the load curve.

[0082] As shown in Figure 2 , step (2) specifically includes the following sequential steps:

[0083] (2a) constructing an electrical coupling degree index Based on the power transmission distribution factor, the electrical coupling degree between nodes is calculated according to the power grid topology and electrical parameters, and an undirected weight reflecting the electrical coupling strength is constructed Matrix: first, the equivalent impedance between nodes is calculated by the impedance matrix, and the electrical coupling degree matrix is generated, then the reciprocal of the electrical coupling degree is taken as the weight, and the diagonal elements are set to zero, and finally the undirected weight matrix with the electrical coupling degree as the core is formed ;

[0084] The impedance matrix is obtained by inverting the node conductance matrix First, the resistance matrix is calculated, and then the conductance matrix is calculated according to the resistance matrix , ; then the node conductance matrix is constructed, wherein the diagonal elements of the node conductance matrix are the conductance negative sums of the corresponding nodes, and the non-diagonal elements are the conductances themselves; finally, the impedance matrix is obtained by inverting the ;

[0085] (2b) Constructing the load carbon flux index : the load flow calculation result is obtained by load flow calculation, the node net load in the load flow calculation result is obtained, and the carbon trace factor obtained according to the main grid carbon emission intensity time curve is obtained, and the node average load carbon flux is calculated:

[0086] ;

[0087] In the formula, the average load carbon flux of node i is ; the number of time periods is , and =96h;

[0088] The load flow calculation obtains the active power, reactive power, node net load, branch load flow distribution, unit injection matrix, renewable energy distribution matrix, node active flux matrix, etc.

[0089] An adjacency matrix reflecting the spatial similarity of carbon emissions is constructed , realizing the spatial coupling representation of carbon flow characteristics The average load carbon flux of node j is

[0090] The is linearly superimposed with the according to a preset ratio to generate a comprehensive adjacency matrix ; the preset ratio is 4.2:1.8;

[0091] (2c) Constructing the load balance index: introducing the load balance index , the sum of the average of the net load of all load nodes is obtained to obtain the total load; secondly, the difference square sum of the actual load and the ideal load of each network element is calculated to measure the overall imbalance degree, the greater the value, the more uneven the load distribution; finally, the imbalance degree is converted into a standardized index between 0 and 1 and the opposite number is taken to represent The greater the index, the more conducive to the load balance between network elements; the ideal load is the total load of the system divided by the number of network elements;

[0092] (2d) Constructing a comprehensive division index: using a community detection algorithm, taking the comprehensive adjacency matrix as input, constructing a comprehensive division index as follows:

[0093] ;

[0094] , and and are weight coefficients;

[0095] is the modularity of the fusion of electrical coupling and carbon flux, and the formula is:

[0096] ;

[0097] , wherein represents the total edge weight of the network, ; represents the node attribution eigenvalue, if the node belongs to the same network element as , then ; otherwise, ;

[0098] When performing dynamic partitioning: first, perform initial division, and independently define each node as an initial community; second, perform iterative merging, randomly select a node, calculate the increment of the comprehensive division index after merging with adjacent nodes; if there is an adjacent community that improves the comprehensive division index , then merge the node into the community with the largest increment; finally, perform convergence judgment, repeat the traversal of all nodes for merging operation until the comprehensive division index cannot be further optimized, and output the final network element partitioning result.

[0099] Step (3) specifically includes the following sequential steps:

[0100] First, based on the node load distribution and equivalent impedance within the network element, a carbon loss minimization objective function is constructed to realize dynamic screening of the optimal substation site:

[0101] ;

[0102] Where, is the mean net load of the node; Represents the equivalent impedance from the node to the candidate site; is the loss coefficient; is the total carbon loss; G is the set of power supply nodes of the substation;

[0103] Secondly, based on the maximum active load in the area , combined with the load simultaneity rate and the forward reservation coefficient , calculate the total apparent power :

[0104] ;

[0105] ;

[0106] Where, N is the number of main transformers; S is the rated capacity of a single main transformer; is the power factor;

[0107] according to Get the maximum load demand of each network element and consider the load simultaneity rate and the forward reservation coefficient The substation capacity configuration should be configured to minimize the impact of power outages and comply with the N-1 safety criterion; ensure that the short-term overload capacity of the remaining main transformers does not exceed 130% when a single main transformer exits, and that the normal operating load rate is stable within the 70% economic range.

[0108] In summary, the present invention proposes a distribution network partitioning that takes low-carbon goals into consideration. In the partitioning index, the node carbon flux clustering index is combined with the traditional electrical coupling degree and load balancing, breaking through the limitation of traditional partitioning that only relies on electrical topology relationships, accurately quantifying the real-time carbon emission intensity of load nodes, and realizing balanced division of distribution network network elements; the present invention optimizes the site with the minimum total carbon loss as the objective function. Compared with the traditional reliance on a single economic site selection or geographical site selection, it can significantly reduce carbon loss and realize the joint optimization of low-carbon emissions and high safety in the substation planning process; according to the joint probability distribution of wind, light and load, the present invention efficiently converts complex multi-dimensional and multi-endpoint uncertainty problems into a representative multi-probability scenario set based on the K-means clustering method; while greatly compressing the data scale and improving computing efficiency, the key probability characteristics of the original uncertainty are retained to the greatest extent; using the expected value curve of the typical scenario as input, the risk of capacity redundancy or insufficiency caused by deterministic planning is avoided, and the impact of uncertainty on distribution network planning is significantly reduced.

[0109] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above-mentioned embodiments, and the above-mentioned embodiments and descriptions in the specification are only the principles of the present application. Various changes and improvements can be made without departing from the spirit and scope of the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A power distribution network partitioning and substation electric-carbon collaborative planning method considering multi-terminal uncertainty, characterized in that: The method comprises the following steps in sequence: (1) In view of the uncertainty of wind, light output fluctuation and load change, a Monte Carlo simulation technology is used to generate an initial scene set, and a K-means clustering algorithm is introduced to optimize and screen the initial scene set, and representative wind and light output curves and load curves are extracted; (2) According to the representative wind and light output curves and load curves, an electrical coupling degree index, a load carbon flux index and a load balance degree index are constructed, a comprehensive division index is calculated, dynamic partitioning is carried out, and a final network element partitioning result is obtained; (3) According to the final network element partitioning result, when planning a substation, the minimum line loss, that is, the minimum carbon loss principle is used to determine the site selection of a newly-built substation; then, considering the load demand and constraint conditions, the substation capacity that meets the main transformer N-1 check and considers the load simultaneous rate and load prediction coefficient influence and can meet the maximum load demand in the network element is planned.

2. The power distribution network partitioning and substation electric-carbon coordinated planning method considering multi-terminal uncertainty according to claim 1, characterized in that: Step (1) specifically comprises the following steps in sequence: (1a) The probabilistic characteristics of wind power output are modeled by a Weibull distribution, whose probability density function is denoted as: ； wherein wind speed, is a shape parameter of the wind speed distribution, is a scale parameter of the wind speed; Based on Weibull distribution, the wind speed sequence is mapped to the wind turbine output by combining the wind speed-power piecewise conversion model. When the wind speed is lower than the cut-in wind speed or higher than the cut-out wind speed , the wind turbine output is zero; when the wind speed is between the cut-in wind speed and the rated wind speed , the output increases linearly with the wind speed; after reaching the rated wind speed , the wind turbine output remains at the rated power . A first initial scene set is generated by using the Monte Carlo simulation technology: first, the Weibull distribution parameters are fitted based on the historical wind speed data, a large number of wind speed sequences are randomly sampled and converted into output curves to form the first initial scene set; The K-means clustering algorithm is introduced to reduce the first initial scene set: the cluster centers are randomly initialized, the Euclidean distances of each scene and the centroid are calculated, the scenes are assigned to the nearest neighbor cluster, and the centroid is iteratively updated until convergence, finally the cluster centroid is taken as the typical scene, and the cluster scene ratio is used to determine the probability weight, a wind power representative output composed of the cluster centroid and the occurrence probability corresponding to each typical scene are obtained, and finally a comprehensive expected wind power output curve is weighted out; (1 b) The probability distribution of the photovoltaic power output is modeled by a Beta distribution, whose probability density function is denoted by: ； wherein is the time period irradiance, and are shape parameters related to the light intensity and weather variations, respectively, based on a Beta distribution, the photovoltaic output is mapped to the actual output power by an irradiance-power conversion model; A second initial scene set is generated by using the Monte Carlo simulation technology: first, the Beta distribution parameters are fitted based on the historical irradiance data, a large number of irradiance sequences are randomly sampled and converted into photovoltaic output curves to form the second initial scene set; Then, the K-means clustering algorithm is introduced to optimize and screen the second initial scene set, and representative typical scenes are extracted, a photovoltaic representative output composed of the cluster centroid and the occurrence probability corresponding to each typical scene are obtained, and finally a comprehensive expected photovoltaic output curve is weighted out; (1c) the probability distribution of the electrical load obeys a normal distribution, a probability density function of the normal distribution is represented as: ； wherein Pload is the load power, Pdesired is the desired value, σ is the standard deviation; A third initial scene set is generated by using the Monte Carlo simulation technology: first, the normal distribution parameters are fitted based on the historical load data, a large number of load power sequences are randomly sampled to form the third initial scene set; Then, the K-means clustering algorithm is introduced to optimize and screen the third initial scene set, a group of load output curves and their corresponding occurrence probabilities are obtained, and finally a load fluctuation curve is weighted out; The wind power comprehensive expected output curve, the photovoltaic comprehensive expected output curve and the load fluctuation curve constitute the wind and light output curve and the load curve.

3. The power distribution network partitioning and substation electric-carbon coordinated planning method considering multi-terminal uncertainty according to claim 1, characterized in that: Step (2) specifically comprises the following steps in sequence: (2a) Constructing electrical coupling degree index : Based on the distribution network topology and electrical parameters, the electrical coupling degree between nodes is calculated by power transmission distribution factor, and the undirected weight reflecting the electrical coupling strength is constructed Matrix: First, the equivalent impedance between nodes is calculated by impedance matrix to generate the electrical coupling degree matrix, then the reciprocal of the electrical coupling degree is taken as the weight, the diagonal elements are set to zero, and finally the undirected weight matrix with electrical coupling degree as the core is formed ; The impedance matrix is ​​obtained by the node conductance matrix Inverse: First calculate the resistance matrix , then according to the resistance matrix Calculate the conductance matrix , ; Then construct the node conductance matrix , where the node conductance matrix The diagonal elements of are the negative sum of the conductances of the corresponding nodes, and the off-diagonal elements are the conductances themselves; finally, by finding The inverse matrix of is the impedance matrix; (2b) Constructing the load carbon flux index : obtaining a power flow calculation result through power flow calculation, according to a node net load in the power flow calculation result , and according to a carbon trace factor obtained according to a main grid carbon emission intensity time sequence curve , the node average load carbon flux is obtained: ； wherein is the average load carbon flux for node i, is the number of time periods; An adjacency matrix reflecting the similarity of carbon emission space is constructed , realizing the spatial coupling representation of carbon flow characteristics; The average load carbon flux of node j; Will With Linearly superimposed according to a preset ratio, a comprehensive adjacency matrix is generated ; the preset ratio is 4.2:1.8; (2c) constructing a load balance index: introducing a load balance index , calculating the sum of the average of the net load of all load nodes to obtain the total load; secondly, calculating the difference square sum of the actual load and the ideal load of each network element to measure the overall imbalance degree; finally, converting the imbalance degree into a standardized index between 0 and 1 and taking the opposite number to represent The larger the index is, the more conducive to the load balance among network elements; the ideal load is the total load of the system divided by the number of network elements; (2d) Constructing a comprehensive partitioning index: Using a community detection algorithm to comprehensively analyze the adjacency matrix As input, construct a comprehensive partitioning index as follows: ； wherein and are weight coefficients; To fuse the electrical coupling with the carbon flux modularity, the formula is: ； In the formula, represents the total edge weight of the network, ; represents the node attribute value, if the node belongs to the same network element as , then ; otherwise, ; In the dynamic partitioning: firstly, initial partitioning is performed, and each node is defined as an initial community independently; secondly, iterative merging is performed, a node is randomly selected, and a comprehensive partitioning index of the node after merging with adjacent nodes is calculated ; if there is an adjacent community that can improve the comprehensive partitioning index , the node is merged into the community with the largest increment; finally, convergence judgment is performed, and the merging operation is repeated for all nodes until the comprehensive partitioning index cannot be further optimized, and the final network element partitioning result is output.

4. The power distribution network partitioning and substation electric-carbon coordinated planning method considering multi-terminal uncertainty according to claim 1, characterized in that: Step (3) specifically comprises the following steps in sequence: Firstly, based on the node load distribution and equivalent impedance in the network element, a carbon loss minimization objective function is constructed to realize dynamic screening of the optimal substation site: ； wherein is the average of the net load of the nodes; denotes the equivalent impedance of the nodes to the candidate station site; is the loss coefficient; is the total carbon loss; G is the set of substation supply nodes; Secondly, based on the maximum active load in the region , combined with the load simultaneous rate and the long-term reservation factor , the total apparent power is calculated ； ； In the formula, N is the number of main transformers; S is the rated capacity of a single main transformer; is the power factor; According to The substation capacity configuration that meets the maximum load demand in each network element, and considers the simultaneous load rate The influence of the long-term reservation coefficient , and conforms to the N-1 safety criterion.

5. An electronic device comprising: a processor; and A memory, in which computer program instructions are stored, the computer program instructions, when executed by the processor, causing the processor to perform the power distribution network partitioning and substation electric-carbon collaborative planning method considering multi-terminal uncertainty as claimed in any one of claims 1-4. 6.A computer readable storage medium, having stored thereon computer program instructions, the computer program instructions, when executed by a processor, causing the processor to perform the power distribution network partitioning and substation electric-carbon collaborative planning method considering multi-terminal uncertainty as claimed in any one of claims 1-4.

Images