Power supply demand portraying method and system for county power grid of power system

Abstract

The invention discloses a power supply demand portraying method for county power grids of a power system. The method comprises the following steps: acquiring and preprocessing data information of each county power grid in a target power system; constructing an index system; performing data clustering analysis based on an adaptive density peak value clustering algorithm; carrying out analysis and type division; and completing a power supply demand portrait of each county power grid in the target power system. The invention also discloses a system for realizing the power supply demand portraying method for the county power grid of the power system. According to the method, the county power grid data information of the target power system is acquired and preprocessed, the corresponding system is constructed, and then data clustering analysis and type division are performed based on the adaptive density peak clustering algorithm, so that the power supply demand portrait of the county power grid of the power system can be realized, the reliability is higher, and the method is more suitable for popularization and application. And the accuracy is better.

Description

Technical Field

[0001] This invention belongs to the field of electrical automation, specifically relating to a method and system for profiling the power supply demand of a county-level power grid. Background Technology

[0002] With economic and technological development and the improvement of people's living standards, electricity has become an indispensable secondary energy source in people's production and daily life, bringing endless convenience. Therefore, ensuring a stable and reliable supply of electricity has become one of the most important tasks of the power system.

[0003] Demand profiling is a crucial foundational task for power systems in formulating generation plans and future development strategies. Therefore, the accuracy and reliability of demand profiling are extremely important for the power system. County-level power grids are a vital component of the power system; thus, demand profiling of county-level power grids is of great significance to the power system as a whole.

[0004] Currently, traditional demand profiling schemes for county-level power grids mainly include schemes based on simple threshold division, schemes based on traditional clustering algorithms, and schemes based on traditional density-peak clustering (DPC). However, schemes based on simple threshold division rely on expert experience for threshold setting, lack data-driven basis, and cannot handle complex relationships between indicators. The results obtained from these schemes are highly subjective and unreliable. Schemes based on traditional clustering algorithms generally assume that the data is spherically distributed, thus they cannot effectively handle the complex distribution characteristics of county-level power grid data. Schemes based on traditional density-peak clustering (DPC) rely on experience for the selection of cutoff distances, lack adaptive mechanisms, require manual participation in decision graph analysis for cluster center identification, are highly subjective, and employ cascading strategies for sample allocation, which easily propagates errors. Therefore, existing demand profiling schemes for county-level power grids all suffer from poor reliability and accuracy. Summary of the Invention

[0005] One of the objectives of this invention is to provide a highly reliable and accurate method for profiling the power supply demand of a county-level power grid.

[0006] The second objective of this invention is to provide a system for profiling the power supply demand of a county-level power grid.

[0007] The method for profiling the power supply demand of a county-level power grid provided by this invention includes the following steps:

[0008] S1. Obtain data information of each county-level power grid in the target power system;

[0009] S2. Preprocess the data obtained in step S1;

[0010] S3. Based on the data obtained in step S2, construct the indicator system;

[0011] S4. Based on the data information obtained in steps S2 and S3, perform data clustering analysis using the adaptive density peak clustering algorithm;

[0012] S5. Analyze and classify the clustering results obtained in step S4;

[0013] S6. Based on the results obtained in step S5, complete the power supply demand profile for each county-level power grid in the target power system.

[0014] Step S1, which involves obtaining data information from various county-level power grids within the target power system, specifically includes the following steps:

[0015] Acquire data information from various county-level power grids within the target power system;

[0016] The data information includes power supply reliability, voltage qualification rate, average transformer capacity per household, main transformer N-1 pass rate, and line N-1 pass rate.

[0017] Step S2, which involves preprocessing the data obtained in step S1, specifically includes the following steps:

[0018] The preprocessing includes the identification and handling of missing data, outlier detection and handling, and data standardization.

[0019] Identifying and handling missing data:

[0020] For the acquired data, if the proportion of missing data in any data set to the total data set is less than a set value, a filling scheme will be used; if the proportion of missing data in any data set to the total data set is not less than a set value, the data will be collected again.

[0021] The filling scheme is to fill using the following formula:

[0022] In the formula Let be the imputed values ​​for the missing indicators of county i in year t; This is the first weighting coefficient set; Let be the index value of the same indicator for county i in year t-1; This is the second weighting coefficient that is set; Let be the index value of the same indicator for county i in year t-2; This is the set third weighting coefficient; Let be the index value of the same indicator for similar counties of county i in year t; similar counties are defined as the counties in the target power system with the closest indicators, based on per capita GDP, the proportion of primary to tertiary industries, and urbanization rate.

[0023] Outlier Detection and Handling:

[0024] The following formula is used for outlier detection and handling:

[0025] In the formula The value of index X after processing; This represents the original value of index X; The index value corresponding to the first quartile of index X; The set adjustment coefficient; Let X be the dispersion value of the index, and ; The index value corresponding to the third quartile of index X;

[0026] Data standardization processing:

[0027] The obtained data metrics were normalized using the Z-score standardization method.

[0028] Step S3, which involves constructing an indicator system based on the data obtained in step S2, specifically includes the following steps:

[0029] Based on the data obtained in step S2, an indicator system is constructed; the indicator system includes a comprehensive index of power grid robustness, an index of balanced power supply quality, an index of investment efficiency assessment, and a degree of coordination in power grid development.

[0030] The comprehensive index of power grid robustness is calculated using the following formula:

[0031] In the formula The comprehensive index of power grid robustness for county i; The first weight for setting the power grid robustness; The pass rate of the main variable N-1 in county i after preprocessing; The second weight for the established power grid robustness; The preprocessed line N-1 throughput of county i; The third weight for the established grid robustness; The power supply reliability of county i after preprocessing;

[0032] The power quality balance index is calculated using the following formula:

[0033] In the formula The power supply quality balance index for county i; The voltage qualification rate of county i after preprocessing;

[0034] The investment efficiency assessment index is calculated using the following formula:

[0035] In the formula The investment efficiency assessment index for county i; The average distribution transformer capacity per household in county i after preprocessing;

[0036] The coordination degree of power grid development is calculated using the following formula:

[0037] In the formula For the coordination of power grid development in county i;

[0038] The pre-processed power supply reliability rate, voltage qualification rate, average distribution transformer capacity per household, main transformer N-1 pass rate and line N-1 pass rate, as well as the calculated comprehensive index of power grid robustness, power supply quality balance index, investment efficiency assessment index and power grid development coordination degree, constitute a profile of the power supply demand characteristics of each county power grid.

[0039] Step S4, which involves performing data clustering analysis based on the data information obtained in steps S2 and S3 using the adaptive density peak clustering algorithm, specifically includes the following steps:

[0040] A. Calculation of cutoff distance:

[0041] The local density index of county i is calculated using the following formula:

[0042] In the formula For county i, the local density index; The number of nearest neighbors, and , The total number of counties; Let k be the set of the k nearest neighbors of county i; Let be the Euclidean distance between county i and county j;

[0043] Based on the local density index, the data space is partitioned using the following formula:

[0044] In the formula The result of data space partitioning; Indicates county-level clusters in developed regions; This indicates a county with a medium level of development. This refers to counties in underdeveloped areas; The local density index is the first quartile of the local density index for all counties. The local density index is the third quartile of the local density index for all counties.

[0045] The cutoff distance for each region is calculated using the following formula:

[0046] In the formula The cutoff distance for county clusters in developed regions; The first scale adjustment factor is set; This is a median function; The cutoff distance for counties with a moderate level of development; The second scale adjustment factor is set; The cutoff distance for counties in underdeveloped areas; The third-scale adjustment factor is set; ;

[0047] The local density of county i is calculated using the following formula:

[0048] In the formula Let be the local density of county i; The cutoff distance is the region to which county i belongs;

[0049] B. Center identification:

[0050] The relative distance to county i is calculated using the following formula:

[0051] In the formula Let i be the relative distance to county i.

[0052] If the local density of county i is the largest, then the relative distance of county i is... for ;

[0053] The kernel density estimation function for the decision values ​​is calculated using the following formula to provide a smooth probability density distribution:

[0054] In the formula For decision value The kernel density estimation function; The decision values ​​already present in the sample Quantity; For bandwidth parameters, and , For decision value Standard deviation; The Gaussian kernel function;

[0055] On the smoothed density curve, the following formula is used to detect candidate center points:

[0056] In the formula The set of candidate center points obtained from the detection;

[0057] Define the statistical significance test rule: if the following conditions are met... If the result is positive, then the statistical significance test is passed; among which, The calculation formula is , For mathematical expectation, Number of clusters The sum of the cluster scatter degrees at time and , This represents the intra-cluster dispersion during clustering. For the first A set of samples for each cluster. For the first The number of samples in each cluster;

[0058] The final center point is selected as the cluster center using the following rules:

[0059] Rule 1: , The mean of the local density. The standard deviation of the local density;

[0060] Rule 2: , The mean of the relative distances. The standard deviation of the relative distance;

[0061] Rule 3: Pass the statistical significance test;

[0062] If county i satisfies rules 1, 2, and 3 simultaneously, then county i will be selected as the final center point.

[0063] C. Global sample allocation:

[0064] The sample allocation problem is conceived as a graph cut optimization problem.

[0065] Graph structure construction:

[0066] In the graph structure, nodes represent all data samples, including selected center points and non-center points; edges connect each sample to its k nearest neighbors, with edge weights... Set as , This is the bandwidth parameter of the Gaussian kernel function;

[0067] Energy function design:

[0068] The following formula is used as the energy function:

[0069] In the formula This is the energy function value; This means assigning sample i to label. The cost, and , For sample i to its label The distance to the corresponding cluster center, For tags The corresponding cluster centers is the distance scale parameter, and its value is the median distance from all samples to the nearest cluster center; The set balance parameters; It is a smoothing term, and ;

[0070] D. Solve:

[0071] The maximum flow / minimum cut algorithm is used to solve the obtained optimization problem;

[0072] Building Stream Networks:

[0073] Create a source node s and a sink node t. For each sample, create a corresponding node and set the edge capacity, as follows:

[0074] In the formula Let be the edge capacity from node s to node i; The cost of maintaining the current label for sample i; Let be the edge capacity from node i to node t; The cost of changing sample i to the new label;

[0075] Initialize the top stream, represented as:

[0076] In the formula The total number of nodes; The height of node s is a function of its height. This represents the current flow value from node s to node v; Let i be the excess flow of node i;

[0077] Active node processing:

[0078] (1) Active node check: Check if there is a node u that satisfies If it exists, proceed to step (2); if it does not exist, proceed to step (4).

[0079] (2) Select an active node: Select a node from the active node queue for processing.

[0080] (3) Allowed arc check: Check if there is an adjacent node v that satisfies: and If it exists, proceed to step (4); if it does not exist, proceed to step (5).

[0081] (4) Flow pushing operation: Transferring excess flow to low-altitude nodes: Increase flow ,Will Reduce flow ,Will Reduce flow ,Will Increase flow Among them, flow The calculation formula is ;

[0082] (5) Remarking operation to resolve deadlock and allow push flow to continue; represented as:

[0083] In the formula This indicates an assignment operation;

[0084] Calculate the maximum flow value:

[0085] When there are no active nodes, the current flow is considered the maximum flow; according to the maximum flow-minimum cut theorem, the maximum flow is set to equal the minimum cut capacity, expressed as:

[0086] In the formula The total flow rate from the source point s is the maximum flow value determined by the maximum flow-minimum cut algorithm.

[0087] Determine the minimum cut:

[0088] In the residual network, a breadth-first search is performed from the source node s to mark all nodes reachable from s, forming a set S, and other nodes form a set T; the edges connecting S and T form a minimum cut, and the minimum cut edge set and the corresponding capacity are output.

[0089] Classification sample labels:

[0090] The following formula is used for optimal label allocation:

[0091] In the formula The optimal cluster label assignment result obtained after graph cut optimization for sample i;

[0092] Results obtained:

[0093] The final clustering result obtained from the clustering algorithm is represented as a cluster assignment vector. .

[0094] Step S5 involves analyzing and classifying the clustering results obtained in step S4, specifically including the following steps:

[0095] (a) Cluster quality assessment:

[0096] Profile coefficient analysis:

[0097] The profile coefficient is calculated using the following formula:

[0098] In the formula The contour coefficient of sample i; Let be the average distance from sample i to the nearest other cluster sample, and , The cluster to which sample i belongs The number of samples; Let be the average distance from sample i to the nearest other cluster sample, and , For other clusters The number of samples;

[0099] The contour coefficient was calculated. average ;

[0100] like If the values ​​are greater than the set value and the average silhouette coefficient of each cluster is greater than the set value, then the clustering is considered to meet the standard; otherwise, the clustering is considered to fail.

[0101] Davidson-Borgin Index Assessment:

[0102] The Davidson-Borgin index is calculated using the following formula. :

[0103] In the formula The number of clusters; is the average distance from the sample within cluster k to the cluster center; The average distance from the sample within cluster l to the cluster center; Let Euclidean distance be the distance between the centers of cluster k and cluster l.

[0104] The quality of the current clustering results is determined based on the silhouette coefficient and the Davidson-Bolding index.

[0105] Calculate all profile coefficients average And set the contour coefficient threshold. ;like If so, the cluster is considered to have good internal compactness and inter-cluster separation.

[0106] Set the Davidson-Boldt index threshold ;like At that time, it was considered that small intra-cluster distances and large inter-cluster distances indicated good clustering quality;

[0107] Determine the quality of the current clustering result: if the current clustering result satisfies and If the clustering quality is satisfactory, then the current clustering result is deemed acceptable; otherwise, the current clustering result is deemed unacceptable.

[0108] If the current clustering result is of unacceptable quality, the following steps should be taken to adjust it:

[0109] In the conditions for significance testing, add a significance tolerance coefficient. The significance test conditions are updated to... Then perform clustering and quality assessment again; if adjustments are needed... If the clustering results cannot pass the quality assessment, then adjust the balance parameter in the energy function. This is done by changing the weights between data items and smoothing items until the clustering results can pass the quality assessment.

[0110] (II) Type Characteristics Analysis:

[0111] After the clustering quality assessment is passed, the features of each cluster are calculated using the following formula:

[0112] In the formula Let be the mean vector of cluster k; Let be the number of samples in cluster k; Let i be the original feature vector of sample i; Let be the standard deviation of cluster k on the j-th index; Let i be the value of the j-th feature value; Let be the mean of cluster k at the j-th eigenvalue; Let be the coefficient of variation of cluster k on the j-th index;

[0113] The significance of each type is calculated using the following formula:

[0114] In the formula Let k be the significance of cluster k on index j; This represents the overall mean of all counties on indicator j. The total standard deviation of all counties on indicator j;

[0115] (III) Boundary Determination:

[0116] For each cluster type and feature index, the fuzzy membership function is constructed using the following formula:

[0117] In the formula The membership degree of a sample with index value x to type k on index j; The cluster center of type k on index j; Let be the scale parameter of type k on index j, and , This represents the actual observed value of sample i on index j;

[0118] Construct a bi-objective optimization problem:

[0119] Objective 1 - Maximizing total membership degree, represented as:

[0120] In the formula This represents the total membership value; The comprehensive membership degree of sample i to cluster type k. , Let be the significance of cluster k on index j.

[0121] Objective 2 - Minimum boundary overlap, represented as:

[0122] In the formula This refers to the boundary overlap. The membership degree of type k at the lower boundary point of index j; The membership degree of type k at the upper boundary point of index j; The membership threshold is set;

[0123] The constraints are expressed as follows:

[0124] In the formula The lower bound of type k on index j; Let k be the upper bound of index j;

[0125] A multi-objective optimization algorithm is used to solve the constructed bi-objective optimization problem, and the boundary conditions of the county's power supply demand are obtained.

[0126] Step S6, based on the results obtained in step S5, completes the power supply demand profile for each county-level power grid in the target power system, specifically including the following steps:

[0127] Construct a rule base for classifying the types of county-level power grids;

[0128] Based on the constructed county-level power grid type classification rule base, and based on the results obtained in step S5, a profile of the power supply demand of each county-level power grid in the target power system is created.

[0129] This invention also provides a system for realizing the method of profiling the power supply demand of a county-level power grid in the power system, comprising a data acquisition module, a data processing module, a system construction module, a cluster analysis module, an analysis and division module, and a demand profiling module; the data acquisition module, data processing module, system construction module, cluster analysis module, analysis and division module, and demand profiling module are connected in series; the data acquisition module is used to acquire data information of each county-level power grid in the target power system and upload the data information to the data processing module; the data processing module is used to preprocess the acquired data information based on the received data information and upload the data information to the system construction module; the system construction module is used to construct an indicator system based on the received data information and the obtained data information and upload the data information to the cluster analysis module; the cluster analysis module is used to perform data clustering analysis based on the received data information and the obtained data information using an adaptive density peak clustering algorithm and upload the data information to the analysis and division module; the analysis and division module is used to analyze and classify the obtained clustering analysis results based on the received data information and upload the data information to the demand profiling module; the demand profiling module is used to complete the power supply demand profiling of each county-level power grid in the target power system based on the received data information and the obtained results.

[0130] The present invention provides a method and system for profiling the power supply demand of a county-level power grid. By acquiring and preprocessing the county-level power grid data of the target power system and constructing the corresponding system, the present invention performs data clustering analysis and type classification based on an adaptive density peak clustering algorithm. Therefore, the present invention can not only realize the profiling of the power supply demand of a county-level power grid, but also has higher reliability and better accuracy. Attached Figure Description

[0131] Figure 1 This is a schematic diagram of the method flow of the present invention.

[0132] Figure 2 This is a schematic diagram of the functional modules of the system of the present invention. Detailed Implementation

[0133] like Figure 1 The diagram shown is a flowchart of the method of the present invention: The method for profiling the power supply demand of a county-level power grid disclosed in this invention includes the following steps:

[0134] S1. Obtain data information of each county-level power grid in the target power system; specifically including the following steps:

[0135] Acquire data information from various county-level power grids within the target power system;

[0136] The data information includes power supply reliability, voltage qualification rate, average distribution transformer capacity per household, main transformer N-1 pass rate, and line N-1 pass rate. These indicators comprehensively reflect the operation status and development level of the county power grid from the dimensions of power supply reliability, power quality, power supply capacity, and power grid structure strength. The data collection period is annual.

[0137] S2. Preprocess the data obtained in step S1; specifically, this includes the following steps:

[0138] The preprocessing includes the identification and handling of missing data, outlier detection and handling, and data standardization.

[0139] Identifying and handling missing data:

[0140] For the acquired data, if the proportion of missing data in any data to the total amount of data is less than a set value (preferably 10%), a filling scheme is used for processing; if the proportion of missing data in any data to the total amount of data is not less than the set value, the data is collected again.

[0141] The filling scheme is as follows: The filling is performed using the following formula:

[0142] In the formula Let be the imputed values ​​for the missing indicators of county i in year t; The first weighting coefficient is set (preferably 0.5); Let be the index value of the same indicator for county i in year t-1; The second weighting coefficient is set (preferably 0.3); Let be the index value of the same indicator for county i in year t-2; The third weighting coefficient is set (preferably 0.2); Let be the index value of the same indicator for similar counties of county i in year t; similar counties are defined as the counties in the target power system with the closest indicators, based on per capita GDP, the proportion of primary to tertiary industries, and urbanization rate.

[0143] Outlier Detection and Handling:

[0144] The following formula is used for outlier detection and handling:

[0145] In the formula The value of index X after processing; This represents the original value of index X; The index value corresponding to the first quartile of index X; The set adjustment coefficient; Let X be the dispersion value of the index, and ; The value of the index is the third and fourth quartiles of index X. This processing method preserves the statistical characteristics of the data and avoids outliers from having an excessive impact on subsequent cluster analysis. After outlier processing, the data quality is significantly improved. The next step is to perform standardization to eliminate the influence of units.

[0146] Data standardization processing:

[0147] The obtained data metrics were normalized using the Z-score standardization method.

[0148] S3. Based on the data obtained in step S2, construct the indicator system; specifically, this includes the following steps:

[0149] Based on the data obtained in step S2, an indicator system is constructed; the indicator system includes a comprehensive index of power grid robustness, an index of balanced power supply quality, an index of investment efficiency assessment, and a degree of coordination in power grid development.

[0150] The comprehensive index of power grid robustness is calculated using the following formula:

[0151] In the formula The comprehensive index of power grid robustness for county i; The first weight for setting the power grid robustness; The pass rate of the main variable N-1 in county i after preprocessing; The second weight for the established power grid robustness; The preprocessed line N-1 throughput of county i; The third weight for the established grid robustness; The power supply reliability rate of county i after preprocessing; the comprehensive power grid robustness index aims to comprehensively evaluate the overall risk resistance and structural integrity of the power grid.

[0152] The power quality balance index is calculated using the following formula:

[0153] In the formula The power supply quality balance index for county i; The voltage qualification rate of county i after preprocessing; the power supply quality balance index is used to evaluate the synergistic development relationship between power supply reliability and voltage qualification rate.

[0154] The investment efficiency assessment index is calculated using the following formula:

[0155] In the formula The investment efficiency assessment index for county i; The average distribution transformer capacity per household in county i after preprocessing; the investment efficiency assessment index aims to address the misleading nature of solely assessing the power grid development level based on the average distribution transformer capacity per household.

[0156] The coordination degree of power grid development is calculated using the following formula:

[0157] In the formula The power grid development coordination degree of county i; the power grid development coordination degree is used to assess the level of coordinated development among various core indicators and to identify counties with uncoordinated development.

[0158] The pre-processed power supply reliability rate, voltage qualification rate, average distribution transformer capacity per household, main transformer N-1 pass rate and line N-1 pass rate, as well as the calculated comprehensive index of power grid robustness, power supply quality balance index, investment efficiency assessment index and power grid development coordination degree, constitute a profile of the power supply demand characteristics of each county power grid.

[0159] S4. Based on the data information obtained in steps S2 and S3, perform data clustering analysis using the adaptive density peak clustering algorithm; specifically including the following steps:

[0160] After constructing the complete nine-dimensional feature vector, a suitable clustering algorithm is needed to discover the inherent group distribution patterns in the data. The goal of cluster analysis is to automatically group similar counties into the same category and divide counties with significant differences into different categories, thereby identifying county groups with similar power supply demand characteristics.

[0161] The present invention makes three core improvements to the density peak clustering algorithm: first, it adaptively determines the parameters through a multi-scale cutoff distance mechanism; second, it automatically identifies the cluster centers based on statistical methods; and finally, it achieves the globally optimal sample allocation through graph optimization.

[0162] A. Calculation of cutoff distance:

[0163] The local density index of county i is calculated using the following formula:

[0164] In the formula For county i, the local density index; The number of nearest neighbors, and , The total number of counties; Let k be the set of the k nearest neighbors of county i; The Euclidean distance between county i and county j is used; the purpose of this step is to identify dense and sparse regions in the data space, providing a basis for subsequent multi-scale processing.

[0165] Based on the local density index, the data space is partitioned using the following formula:

[0166] In the formula The result of data space partitioning; Indicates county-level clusters in developed regions; This indicates a county with a medium level of development. This refers to counties in underdeveloped areas; The local density index is the first quartile of the local density index for all counties. The local density index is the third quartile of the local density index for all counties; this division ensures that appropriate cutoff distance parameters can be obtained for different density areas.

[0167] The cutoff distance for each region is calculated using the following formula:

[0168] In the formula The cutoff distance for county clusters in developed regions; The first scale adjustment factor is set; This is a median function; The cutoff distance for counties with a moderate level of development; The second scale adjustment factor is set; The cutoff distance for counties in underdeveloped areas; The third-scale adjustment factor is set; The core idea of ​​this step is to use a smaller cutoff distance in dense regions to capture detailed features, and a larger cutoff distance in sparse regions to avoid over-segmentation.

[0169] The local density of county i is calculated using the following formula:

[0170] In the formula Let be the local density of county i; The cutoff distance is the region to which county i belongs;

[0171] B. Center identification:

[0172] Relative distance is another key indicator for identifying cluster centers. It measures the degree of separation between a sample and samples with higher density. Cluster centers typically have high local density and large relative distances. The relative distance of county i is calculated using the following formula:

[0173] In the formula Let i be the relative distance to county i.

[0174] If the local density of county i is the largest, then the relative distance of county i is... for ;

[0175] The kernel density estimation function for the decision values ​​is calculated using the following formula to provide a smooth probability density distribution:

[0176] In the formula For decision value The kernel density estimation function; The decision values ​​already present in the sample Quantity; For bandwidth parameters, and , For decision value Standard deviation; The Gaussian kernel function;

[0177] On the smoothed density curve, the following formula is used to detect candidate center points:

[0178] In the formula The set of candidate center points obtained from the detection;

[0179] Define the statistical significance test rule: if the following conditions are met... If the result is positive, then the statistical significance test is passed; among which, The calculation formula is , For mathematical expectation, Number of clusters The sum of the cluster scatter degrees at time and , This represents the intra-cluster dispersion during clustering. For the first A set of samples for each cluster. For the first The number of samples in each cluster;

[0180] The final center point is selected as the cluster center using the following rules:

[0181] Rule 1: , The mean of the local density. The standard deviation of the local density;

[0182] Rule 2: , The mean of the relative distances. The standard deviation of the relative distance;

[0183] Rule 3: Pass the statistical significance test;

[0184] If county i satisfies rules 1, 2, and 3 simultaneously, then county i will be selected as the final center point.

[0185] C. Global sample allocation:

[0186] After determining the cluster centers, traditional DPC uses a cascading allocation strategy, but this strategy is prone to propagating errors, and the incorrect allocation of a single sample will affect all subsequent related samples. To solve this problem, this invention constructs the sample allocation problem as a graph cut optimization problem.

[0187] Graph structure construction:

[0188] In the graph structure, nodes represent all data samples, including selected center points and non-center points; edges connect each sample to its k nearest neighbors, with edge weights... Set as , This is the bandwidth parameter of the Gaussian kernel function;

[0189] Energy function design:

[0190] The following formula is used as the energy function:

[0191] In the formula This is the energy function value; This means assigning sample i to label. The cost, and , For sample i to its label The distance to the corresponding cluster center, For tags The corresponding cluster centers is the distance scale parameter, and its value is the median distance from all samples to the nearest cluster center; The set balance parameters; It is a smoothing term, and ;

[0192] D. Solve:

[0193] The maximum flow / minimum cut algorithm is used to solve the obtained optimization problem;

[0194] Building Stream Networks:

[0195] Create a source node s and a sink node t. For each sample, create a corresponding node and set the edge capacity, as follows:

[0196] In the formula Let be the edge capacity from node s to node i; The cost of maintaining the current label for sample i; Let be the edge capacity from node i to node t; The cost of changing sample i to the new label;

[0197] Initialize the top stream, represented as:

[0198] In the formula The total number of nodes; The height of node s is a function of its height. This represents the current flow value from node s to node v; Let i be the excess flow of node i;

[0199] Active node processing:

[0200] (1) Active node check: Check if there is a node u that satisfies If it exists, proceed to step (2); if it does not exist, proceed to step (4).

[0201] (2) Select an active node: Select a node from the active node queue for processing.

[0202] (3) Allowed arc check: Check if there is an adjacent node v that satisfies: and If it exists, proceed to step (4); if it does not exist, proceed to step (5).

[0203] (4) Flow pushing operation: Transferring excess flow to low-altitude nodes: Increase flow ,Will Reduce flow ,Will Reduce flow ,Will Increase flow Among them, flow The calculation formula is ;

[0204] (5) Remarking operation to resolve deadlock and allow push flow to continue; represented as:

[0205] In the formula This indicates an assignment operation;

[0206] Calculate the maximum flow value:

[0207] When there are no active nodes, the current flow is considered the maximum flow; according to the maximum flow-minimum cut theorem, the maximum flow is set to equal the minimum cut capacity, expressed as:

[0208] In the formula The total flow rate from the source point s is the maximum flow value determined by the maximum flow-minimum cut algorithm.

[0209] Determine the minimum cut:

[0210] In the residual network, a breadth-first search is performed from the source node s to mark all nodes reachable from s, forming a set S, and other nodes form a set T; the edges connecting S and T form a minimum cut, and the minimum cut edge set and the corresponding capacity are output.

[0211] Classification sample labels:

[0212] The following formula is used for optimal label allocation:

[0213] In the formula The optimal cluster label assignment result obtained after graph cut optimization for sample i;

[0214] Results obtained:

[0215] The final clustering result obtained from the clustering algorithm is represented as a cluster assignment vector. ;

[0216] S5. Analyze and classify the clustering results obtained in step S4; specifically, this includes the following steps:

[0217] (a) Cluster quality assessment:

[0218] Profile coefficient analysis:

[0219] The profile coefficient is calculated using the following formula:

[0220] In the formula The contour coefficient of sample i; Let be the average distance from sample i to the nearest other cluster sample, and , Let i be the cluster to which sample i belongs The number of samples; Let be the average distance from sample i to the nearest other cluster sample, and , For other clusters The number of samples;

[0221] The contour coefficient was calculated. average ;

[0222] like If the average profile coefficient of each cluster is greater than a set value (preferably 0.5) and the average profile coefficient of each cluster is greater than a set value (preferably 0.4), then the clustering is deemed to have met the standard; otherwise, the clustering is deemed to have failed.

[0223] Davidson-Borgin Index Assessment:

[0224] The Davidson-Borgin index is calculated using the following formula. :

[0225] In the formula The number of clusters; is the average distance from the sample within cluster k to the cluster center; The average distance from the sample within cluster l to the cluster center; Let Euclidean distance be the distance between the centers of cluster k and cluster l.

[0226] The quality of the current clustering results is determined based on the silhouette coefficient and the Davidson-Bolding index.

[0227] Profile coefficient average It should be as close as possible to the maximum value of 1; therefore, a threshold is set. ,like If so, the cluster is considered to have good internal compactness and inter-cluster separation.

[0228] Davidson-Borgin Din Index It should be as close as possible to the minimum value of 0; therefore, a threshold is set. ,like At that time, it was considered that small intra-cluster distances and large inter-cluster distances indicated good clustering quality;

[0229] Therefore, the determination method is as follows: if the current clustering result is... and If the clustering is successful, the clustering quality is deemed acceptable; otherwise, it is deemed unacceptable.

[0230] If the clustering quality assessment result is unsatisfactory, an iterative optimization mechanism for parameters and processes will be triggered, with the specific steps as follows:

[0231] The judgment criteria in statistical significance testing can be appropriately relaxed or tightened. Specifically, an adjustable significance tolerance coefficient can be added to the criteria for significance testing. The significance test conditions are updated to... Then perform clustering and quality assessment again; if adjustments are needed... If the clustering results cannot pass the quality assessment, then adjust the balance parameter in the energy function. This is done by changing the weights between data items and smoothing items until the clustering results can pass the quality assessment.

[0232] (II) Type Characteristics Analysis:

[0233] After the clustering quality assessment is passed, the features of each cluster are calculated using the following formula:

[0234] In the formula Let be the mean vector of cluster k; Let be the number of samples in cluster k; Let i be the original feature vector of sample i; Let be the standard deviation of cluster k on the j-th index; Let i be the value of the j-th feature value; Let be the mean of cluster k at the j-th eigenvalue; Let be the coefficient of variation of cluster k on the j-th index;

[0235] The significance of each type is calculated using the following formula:

[0236] In the formula Let k be the significance of cluster k on index j; This represents the overall mean of all counties on indicator j. The total standard deviation of all counties on indicator j;

[0237] (III) Boundary Determination:

[0238] For each cluster type and feature index, the fuzzy membership function is constructed using the following formula:

[0239] In the formula The membership degree of a sample with index value x to type k on index j; The cluster center of type k on index j; Let be the scale parameter of type k on index j, and , This represents the actual observed value of sample i on index j;

[0240] Construct a bi-objective optimization problem:

[0241] Objective 1 - Maximizing total membership degree, represented as:

[0242] In the formula This represents the total membership value; The comprehensive membership degree of sample i to cluster type k. , Let be the significance of cluster k on index j.

[0243] Objective 2 - Minimum boundary overlap, represented as:

[0244] In the formula This refers to the boundary overlap. The membership degree of type k at the lower boundary point of index j; The membership degree of type k at the upper boundary point of index j; The membership threshold is set;

[0245] The constraints are expressed as follows:

[0246] In the formula The lower bound of type k on index j; Let k be the upper bound of index j;

[0247] A multi-objective optimization algorithm is used to solve the constructed bi-objective optimization problem, and the boundary conditions of the county's power supply demand are obtained.

[0248] S6. Based on the results obtained in step S5, complete the power supply demand profile for each county-level power grid in the target power system; specifically including the following steps:

[0249] Construct a rule base for classifying the types of county-level power grids;

[0250] Based on the constructed county-level power grid type classification rule base, and based on the results obtained in step S5, the power supply demand of each county-level power grid in the target power system is profiled.

[0251] In practice, the rule base that can be constructed can be:

[0252] 1) Profiling of power supply demand in counties surrounding major cities

[0253] Power supply reliability 99.96%; Voltage qualification rate 99.88%; Average transformer capacity per household 3.0 kVA / household; Main transformer N-1 throughput 85%; Line N-1 throughput 80%; Power Grid Robustness Index 0.82; Power quality balance index 0.75; Investment Efficiency Assessment Index 0.70; Coordination degree of power grid development 0.75;

[0254] 2) Professional Functional County Power Supply Demand Profile

[0255] Power supply reliability [99.90%, 99.96%); Voltage qualification rate [99.80%, 99.88%); Average transformer capacity per household [2.5, 3.0) kVA / household; Main transformer N-1 throughput [75%, 85%); Line N-1 throughput [70%, 80%); Power Grid Robustness Index [0.70, 0.82); Power quality balance index [0.65, 0.75); Investment efficiency assessment index [0.60, 0.70); Coordination degree of power grid development [0.65, 0.75);

[0256] 3) Profiling of electricity demand in counties of major agricultural production areas

[0257] Power supply reliability [99.85%, 99.90%); Voltage qualification rate [99.75%, 99.80%); Average transformer capacity per household [2.0, 2.5) kVA / household; Main transformer N-1 throughput [65%, 75%); Line N-1 throughput [60%, 70%); Power Grid Robustness Index [0.60, 0.70); Power quality balance index [0.60, 0.65); Investment efficiency assessment index [0.50, 0.60); Coordination degree of power grid development [0.60, 0.65);

[0258] 4) Profiling of Power Supply Demand in Key Ecological Function Zones

[0259] Power supply reliability 99.85%; Voltage qualification rate 99.75%; Average transformer capacity per household 2.0 kVA / household; Main transformer N-1 throughput 65%; Line N-1 throughput 60%; Power Grid Robustness Index 0.60; Power quality balance index 0.60; Investment Efficiency Assessment Index 0.50; Coordination degree of power grid development 0.60;

[0260] Ultimately, a profile of the power grid in each county is obtained; based on the corresponding profile, a corresponding resource allocation plan can be set, with the priority order as follows: key ecological function zones > major agricultural production areas > specialized functions > areas surrounding large cities, to ensure that areas with weak infrastructure are identified first.

[0261] like Figure 2The diagram shows the functional modules of the system of the present invention: The system disclosed in this invention for realizing the power supply demand profiling method of the county-level power grid in the power system includes a data acquisition module, a data processing module, a system construction module, a cluster analysis module, an analysis and division module, and a demand profiling module; these modules are connected in series. The data acquisition module acquires data information from various county-level power grids in the target power system and uploads the data information to the data processing module. The data processing module preprocesses the acquired data information based on the received data information and uploads the data information to the system construction module. The system is divided into three modules: a system construction module and a clustering analysis module. The system construction module constructs an indicator system based on the received data and uploads the data to the clustering analysis module. The clustering analysis module performs data clustering analysis based on the received data and the adaptive density peak clustering algorithm and uploads the data to the analysis and partitioning module. The analysis and partitioning module analyzes and classifies the clustering analysis results based on the received data and uploads the data to the demand profiling module. The demand profiling module completes the power supply demand profile for each county power grid in the target power system based on the received data and the results.

Claims

1. A power system county grid power supply demand profiling method, comprising the following steps: S1. Obtain data information of each county grid in the target power system; S2. Preprocess the data information obtained in step S1; S3. Construct an index system based on the data information obtained in step S2; S4. Perform data clustering analysis based on the data information obtained in steps S2 and S3 using an adaptive density peak clustering algorithm; S5. Analyze and classify the clustering analysis results obtained in step S4; S6. Complete the power supply demand profiling of each county grid in the target power system based on the results obtained in step S5.

2. The power system county grid power supply demand profiling method of claim 1, wherein The data information of each county grid in the target power system obtained in step S1 includes the following steps: Obtain data information of each county grid in the target power system; The data information includes power supply reliability, voltage qualification rate, per-household distribution transformer capacity, main transformer N-1 passing rate, and line N-1 passing rate.

3. The power system county grid power supply demand profiling method of claim 2, wherein The preprocessing of the data information obtained in step S1 includes the following steps: The preprocessing includes missing data judgment and processing, outlier detection and processing, and data standardization processing; Missing data judgment and processing: For the obtained data, if the proportion of missing data in any data to the total data is less than a certain value, use the filling scheme to process; if the proportion of missing data in any data to the total data is not less than a certain value, re-collect; The filling scheme is filled using the following formula: In the formula is the filling value of the missing index of county i in the tth year; is the first weight coefficient set; is the index value of the same index of county i in the t-1th year; is the second weight coefficient set; is the index value of the same index of county i in the t-2th year; is the third weight coefficient set; is the index value of the same index of the similar county of county i in the tth year; the definition of the similar county is that the index closest to the target power system is found in the target power system with per capita GDP, the proportion of the first to the third industry, and the urbanization rate as the index. Outlier detection and processing: Use the following formula for outlier detection and processing: In the formula is the value of the processed index X; is the original value of the index X; is the index value corresponding to the first quartile of the index X; is the set adjustment coefficient; is the dispersion degree value of the index X, and ; is the index value corresponding to the third quartile of the index X; Data standardization processing: The obtained data indicators are normalized using the Z-score standardization method.

4. The power system county grid power supply demand profiling method of claim 3, wherein The construction of the index system based on the data information obtained in step S2 includes the following steps: Construct an index system based on the data information obtained in step S2; the index system includes a power grid robustness comprehensive index, a power supply quality balance index, an investment efficiency evaluation index, and a power grid development coordination degree; The power grid robustness comprehensive index is calculated using the following formula: In the formula is the grid robustness comprehensive index of the county i; is the first weight of the set grid robustness; is the pretreated main transformer N-1 passing rate of the county i; is the second weight of the set grid robustness; is the pretreated line N-1 passing rate of the county i; is the third weight of the set grid robustness; is the pretreated power supply reliability rate of the county i; The power supply quality balance index is calculated using the following formula: In the formula is the power supply quality balance index of the county i; is the voltage pass rate of the county i after preprocessing; The investment efficiency evaluation index is calculated using the following formula: In the formula is the investment efficiency evaluation index of county i; is the household average distribution transformer capacity of county i after preprocessing; The power grid development coordination degree is calculated using the following formula: In the formula is the grid development coordination degree of the county i; The preprocessed power supply reliability, voltage qualification rate, per-household distribution transformer capacity, main transformer N-1 passing rate, and line N-1 passing rate, as well as the calculated power grid robustness comprehensive index, power supply quality balance index, investment efficiency evaluation index, and power grid development coordination degree, constitute the power supply demand characteristic profiling of each county grid.

5. The power system county grid power supply demand profiling method of claim 4, wherein The data clustering analysis based on the data information obtained in steps S2 and S3 using an adaptive density peak clustering algorithm includes the following steps: A. Truncated distance calculation: The local density index of county i is calculated using the following formula: wherein is the local density index for county i; is the number of neighbors, and , is the total number of counties; is the set of k nearest neighbors for county i; is the Euclidean distance between county i and county j; Based on the local density index, the data space is divided using the following formula: In the formula is the data space division result; represents the developed area county cluster; represents the medium development level county; represents the underdeveloped area county; is the first quartile of the local density index of all counties corresponding to the local density index; is the third quartile of the local density index of all counties corresponding to the local density index; The truncated distance of each region is calculated using the following formula: wherein is the truncation distance for developed areas county clusters; is the set first scale adjustment factor; is the median function; is the truncation distance for medium development level county; is the set second scale adjustment factor; is the truncation distance for underdeveloped areas county; is the set third scale adjustment factor; ; The local density of the county i is calculated by using the following formula: wherein is the local density of the county i; is the cutoff distance of the region to which the county i belongs. B. Center identification: The relative distance of the county i is calculated by using the following formula: In the formula is the relative distance of the county i; If the local density of the county i is the maximum, the relative distance of the county i is ; The kernel density estimation function of the decision value is calculated by using the following formula to provide a smooth probability density distribution: wherein is the kernel density estimation function is the decision value is the number of decision values already present in the sample; is the bandwidth parameter, and , is the standard deviation of the decision values ; is the Gaussian kernel function; On the smoothed density curve, the candidate center point is detected by using the following formula: In the formula is a set of candidate center points obtained by detection; Set a statistical significance test rule: if , then determine that the statistical significance test is passed; wherein, The calculation formula of , is the mathematical expectation, is the number of clusters is the sum of intra-cluster dispersion at time t, and , is the intra-cluster dispersion at time t, is the sample set of the th cluster, is the number of samples in the th cluster; The final center point is selected as the clustering center by using the following rules: Rule 1: , is the mean of the local density, is the standard deviation of the local density; Rule 2: , is the mean of the relative distances, is the standard deviation of the relative distances; Rule 3: Passing the statistical significance test; If the county i meets rules 1, 2 and 3 at the same time, the county i is selected as the final center point; C. Global sample allocation: The sample allocation problem is constructed as a graph cut optimization problem; Graph structure construction: Setting the nodes in the graph structure to be all data samples, the nodes including the already selected center points and non-center points; the edges in the graph structure are the edges connecting each sample with its k nearest neighbors, the weight of the edge Setting the nodes in the graph structure to be all data samples, the nodes including the already selected center points and non-center points; the edges in the graph structure are the edges connecting each sample with its k nearest neighbors, the weight of the edge , is a bandwidth parameter of the Gaussian kernel function; Energy function design: The energy function is used as follows: In the formula This is the energy function value; This means assigning sample i to label. The cost, and , For sample i to its label The distance to the corresponding cluster center, For tags The corresponding cluster centers is the distance scale parameter, and its value is the median distance from all samples to the nearest cluster center; The set balance parameters; It is a smoothing term, and ; D. Solution: The maximum flow / minimum cut algorithm is used to solve the obtained optimization problem; Flow network construction: A source point s and a sink point t are created, and each sample creates a corresponding node and sets the edge capacity, denoted as: wherein is the edge capacity from node s to node i; is the cost of keeping the current label for sample i; is the edge capacity from node i to node t; is the cost of changing to a new label for sample i; Initialize the top flow, denoted as: wherein is the total number of nodes; is the height function of node s; is the current flow value from node s to node v; is the excess flow of node i; Active node processing: (1) Active node check: check if there exists a node u such that If yes, go to step (2), if no, go to step (4); (2) Select an active node from the active node queue for processing (3) Allowable arc check: check whether there exists an adjacent node v satisfying: and : if exists, go to step (4), if not, go to step (5); (4) Push operation: transfer excess flow to low-height nodes: transfer increase flow , to reduce flow , to reduce flow , to increase flow ; wherein the calculation formula of flow is ; (5) Re-label operation to solve the deadlock situation and continue the push flow; denoted as: In the formula represents an assignment operation; Calculate the maximum flow value: When there is no active node, the current flow is considered as the maximum flow; according to the maximum flow minimum cut theorem, the maximum flow is set equal to the minimum cut capacity, denoted as: In the formula is the total flow from the source point s; Determine the minimum cut: In the residual network, perform a breadth-first search from the source point s to mark all nodes reachable from s, forming set S, and other nodes form set T; the edges connecting S and T form the minimum cut, and the minimum cut edge set and corresponding capacity are output; Classify sample labels: The optimal label allocation is performed by using the following formula: In the formula is the optimal cluster label assignment result of sample i after graph cut optimization; Get the result: The clustering result of the clustering algorithm is finally obtained, represented as a cluster assignment vector .

6. The power system county grid power demand profiling method of claim 5, wherein Step S5 analyzes and classifies the clustering analysis result obtained in step S4, specifically including the following steps: (I) Cluster quality evaluation: Contour coefficient analysis: The contour coefficient is calculated by using the following formula: wherein is the profile coefficient for sample i; is the average distance of sample i to the nearest other cluster sample, and , is the number of samples in the cluster to which sample i belongs; is the average distance of sample i to the nearest other cluster sample, and , is the number of samples in the other cluster ; The profile coefficient is calculated the average value ; If If the average silhouette coefficient of each cluster is greater than a set value, then the clustering is determined to be satisfactory; otherwise, the clustering is determined to be unsatisfactory. Davison-Bowden index evaluation: The Davison-Bunting index is calculated using the following formula : wherein is the number of clusters; is the average distance of samples within cluster k to the cluster center; is the average distance of samples within cluster l to the cluster center; is the Euclidean distance between the cluster k and cluster l centers; According to the contour coefficient and the Davison-Bowden index, the quality of the current clustering result is judged: Compute the average of all profile coefficients and set a profile coefficient threshold ; if , then the cluster is considered to have good internal compactness and separation from other clusters.​ Setting davies-bouldin index threshold ; if the cluster quality is good when the intra-cluster distance is small and the inter-cluster distance is large. determining the quality of the current clustering result: if the current clustering result satisfies and , it is determined that the clustering quality of the current clustering result is qualified; otherwise, it is determined that the clustering quality of the current clustering result is unqualified; If the clustering quality of the current clustering result is unqualified, the following steps are used for adjustment: increasing the significance tolerance factor in the condition of the significance test updating the condition of the significance test to and performing clustering and quality determination again; if adjusting the balance parameter in the energy function fails to make the clustering result pass the quality determination, adjusting the weight between the data items and the smoothing items to change the weight between the data items and the smoothing items until the clustering result can pass the quality determination; (II) Type feature analysis: After the cluster quality evaluation, the following formula is used to calculate the features of each cluster: where is the mean vector of cluster k; is the number of samples of cluster k; is the original feature vector of sample i; is the standard deviation of cluster k on the jth index; is the value of sample i on the jth feature value; is the mean of cluster k on the jth feature value; is the coefficient of variation of cluster k on the jth index; The significance of each type is calculated by using the following formula: wherein is the prominence of cluster k on index j; is the total mean of all counties on index j; is the total standard deviation of all counties on index j; (III) Boundary determination: For each cluster type and feature index, a fuzzy membership function is constructed by using the following formula: wherein is the membership of sample i with index value x to type k on index j; is the cluster center of type k on index j; is the scale parameter of type k on index j, and , is the actual observed value of sample i on index j; Build a double-objective optimization problem: Objective 1 - Maximum total membership, denoted as: wherein is the total membership value; is the overall membership of sample i to cluster type k, , is the prominence of cluster k on index j; Objective 2 - Minimum boundary overlap, denoted as: wherein is the boundary overlap degree; is the membership degree of the lower boundary point of type k on index j; is the membership degree of the upper boundary point of type k on index j; is the set membership threshold value; Constraint condition, denoted as: wherein is the lower bound of the type k on the index j; is the upper bound of the type k on the index j; A multi-objective optimization algorithm is used to solve the constructed double-objective optimization problem to obtain the boundary conditions of the county power supply demand.

7. The power system county grid power demand profiling method of claim 6, wherein Step S6 completes the power supply demand portrait of each county power grid in the target power system according to the result obtained in step S5, specifically including the following steps: Build a type classification rule library for county power grids; According to the constructed type classification rule library for county power grids, the power supply demand of each county power grid in the target power system is profiled according to the result obtained in step S5.

8. A system for implementing the power demand profiling method of any one of claims 1 to 7, characterized by The system comprises a data acquisition module, a data processing module, a system construction module, a clustering analysis module, an analysis and division module, and a demand portrait module; the data acquisition module, the data processing module, the system construction module, the clustering analysis module, the analysis and division module, and the demand portrait module are sequentially connected; the data acquisition module is used for acquiring data information of each county power grid in a target power system and uploading the data information to the data processing module; The data processing module is used for pre-processing the acquired data information according to the received data information and uploading the data information to the system construction module; the system construction module is used for constructing an index system according to the received data information and the obtained data information and uploading the data information to the clustering analysis module; The clustering analysis module is used for performing data clustering analysis based on a self-adaptive density peak value clustering algorithm according to the received data information and the obtained data information and uploading the data information to the analysis and division module; the analysis and division module is used for analyzing and classifying the obtained clustering analysis result according to the received data information and uploading the data information to the demand portrait module; and the demand portrait module is used for completing a power supply demand portrait of each county power grid in the target power system according to the received data information and the obtained result.

Images