Transformer area topology identification method and device, equipment and storage medium

Abstract

Embodiments of the present application disclose a transformer topology identification method and device, equipment and a storage medium, the method comprising: obtaining total power consumption data of each phase of a transformer in a transformer area and power consumption data of a plurality of users in the transformer area in a preset time period; dividing the total power consumption data of each phase of the transformer and the power consumption data of the plurality of users into a plurality of time segment data sets according to time; for each time segment data set, independently solving a group of candidate user phase sequence attribution relationships based on the principle of energy conservation; aggregating all candidate user phase sequence attribution relationships generated for the time segments to determine the connection probability of each user belonging to each phase; and based on the connection probability and at least one preset confidence threshold, determining the final phase sequence attribution result and the corresponding confidence level for each user. In the above manner, the transformer topology identification combining energy conservation and cross-period consistency analysis is realized, and the accuracy of the identification is significantly improved.

Description

Technical Field

[0001] This invention relates to the field of transformer area topology identification technology, and in particular to a transformer area topology identification method, apparatus, device and storage medium. Background Technology

[0002] The topology of power distribution areas is fundamental data for critical tasks such as distribution network operation monitoring, fault location, and load analysis. Its accuracy directly affects the intelligence level and operation and maintenance efficiency of the power grid. Currently, distribution area topology information mainly relies on manual modeling, on-site inspections, or user application data for construction and maintenance. This results in high labor costs, delayed updates, and insufficient accuracy, making it difficult to meet the automated and precise operation requirements of modern distribution networks. In recent years, some studies have attempted to use the similarity relationships of physical quantities such as voltage and current to assist in identifying distribution area structures. However, these methods generally require high sampling frequency and synchronization accuracy, relying on high-frequency acquisition equipment, which limits their large-scale application in low-frequency metering environments.

[0003] In contrast, the transformer substation identification method based on the principle of energy conservation has stronger engineering adaptability and practical value. This method analyzes the conservation relationship between user electricity consumption and the main power supply of the transformer, and can identify user phase sequence affiliation by relying on low-frequency data in the existing metering system. It has advantages such as wide data sources, no need for additional hardware, and low deployment cost.

[0004] However, the current method has the problem of misjudgment due to small differences in users' electricity consumption during actual identification, and the accuracy of topology identification still needs to be improved. Summary of the Invention

[0005] The main objective of this invention is to provide a method, apparatus, device, and storage medium for identifying transformer substation topology, which can solve the problem that the accuracy of transformer substation topology identification in the prior art still needs to be improved.

[0006] To achieve the above objectives, the first aspect of the present invention provides a method for identifying transformer substation topology, the method comprising: Obtain the total power consumption data of each phase of the transformer in the distribution area and the power consumption data of multiple users in the distribution area within a preset time period; The total power consumption data of each phase of the transformer in the distribution area and the power consumption data of the multiple users are divided into multiple sub-time periods of data. For each of the aforementioned sub-time periods, based on the principle of energy conservation, a set of candidate user phase sequence attribution relationships are independently solved; Aggregate all candidate user phase sequence attribution relationships generated for the sub-time period to determine the connection probability of each user belonging to each phase; Based on the connection probability and at least one preset confidence threshold, the final phase sequence assignment result and corresponding confidence level are determined for each user.

[0007] To achieve the above objectives, a second aspect of the present invention provides a transformer area topology identification device, the device comprising: The data acquisition module is used to acquire the total power consumption data of each phase of the transformer in the distribution area and the power consumption data of multiple users in the distribution area within a preset time period. The data segmentation module is used to divide the total power consumption data of each phase of the transformer in the distribution area and the power consumption data of the multiple users into multiple sub-time period datasets according to time. An independent solution module is used to independently solve a set of candidate user phase sequence attribution relationships for each of the sub-time periods, based on the principle of energy conservation. The probability aggregation module is used to aggregate all candidate user phase sequence attribution relationships generated for the sub-time period to determine the connection probability of each user belonging to each phase. The confidence level decision module is used to determine the final phase sequence assignment result and the corresponding confidence level for each user based on the connection probability and at least one preset confidence level threshold.

[0008] To achieve the above objectives, a third aspect of the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps shown in the first aspect and any feasible implementation.

[0009] To achieve the above objectives, a fourth aspect of the present invention provides a computer device including a memory and a processor, the memory storing a computer program, which, when executed by the processor, causes the processor to perform the steps shown in the first aspect and any feasible implementation.

[0010] The embodiments of the present invention have the following beneficial effects: This invention provides a method for topology identification of a transformer substation. The method includes: acquiring total power consumption data of each phase of the transformer substation and power consumption data of multiple users within the substation within a preset time period; dividing the total power consumption data of each phase of the transformer substation and the power consumption data of the multiple users into multiple sub-time period datasets; for each sub-time period dataset, independently solving a set of candidate user phase sequence attribution relationships based on the principle of energy conservation; aggregating all candidate user phase sequence attribution relationships generated for the sub-time period to determine the connection probability of each user belonging to each phase; and determining the final phase sequence attribution result and corresponding confidence level for each user based on the connection probability and at least one preset confidence threshold. By dividing the dataset into multiple sub-time period datasets according to time; independently solving a set of candidate user phase sequence attribution relationships for each sub-time period dataset based on the principle of energy conservation; aggregating the candidate user phase sequence attribution relationships to determine the connection probability of each user belonging to each phase; and significantly improving the accuracy of the identification results through stability evaluation of results at multiple time scales. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] in: Figure 1 This is a flowchart of a transformer area topology identification method according to an embodiment of the present invention; Figure 2 This is a schematic diagram of a three-layer structure of a low-voltage distribution network in an embodiment of the present invention; Figure 3 This is a schematic diagram of a time consistency solution process in an embodiment of the present invention; Figure 4 This is a structural block diagram of a transformer area topology identification device according to an embodiment of the present invention; Figure 5 This is a structural block diagram of a computer device in an embodiment of the present invention. Detailed Implementation

[0013] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0014] In the operation and management of modern power distribution networks, accurately understanding the topology of distribution areas, especially the relationship between users and the three phases of transformers, is a prerequisite for achieving advanced applications such as load balancing, rapid fault location, and refined line loss management. Please refer to [link / reference]. Figure 1 , Figure 1 This is a flowchart of a transformer area topology identification method according to an embodiment of the present invention. The application scenario of this method is as follows: Figure 2 The power grid scenario shown. Please refer to [link / reference]. Figure 2 , Figure 2 This is a schematic diagram of a three-layer structure of a low-voltage distribution network according to an embodiment of the present invention. Figure 2 This illustration shows a typical low-voltage distribution network topology used in the embodiments of this application. Power is output from the substation transformer 201, passes through one or more branch boxes 202, and is finally distributed to numerous end users 203. Physically, these users are connected to one of the three phases A, B, and C of the transformer. In practice, the phase sequence information of some users may be known, such as A-phase user 203a, B-phase user 203b, and C-phase user 203c in the figure, but the phase sequence information of a large number of users is unknown or incorrect, such as the unknown phase user 203d shown in the figure. To address this technical problem, embodiments of this application provide a method and apparatus designed to automatically, accurately, and cost-effectively identify the actual phase sequence assignment of the unknown phase user 203d.

[0015] Figure 2 The topology of the low-voltage distribution network is shown. When the topology is unknown, the data-based topology identification of the transformer area mainly needs to solve two problems: (1) identifying which users belong to the transformer area, that is, the identification of the "transformer-user relationship" of the transformer area; (2) determining the specific phase (A, B, C) of the three-phase power connected to each user in the transformer area, that is, the identification of the "phase sequence relationship" of the transformer area.

[0016] When solving the transformer substation topology using energy conservation principles, the problem of identifying substation topology relationships can be described as a multi-subset partitioning problem. First, for identifying "variable user relationships" within a substation, the task is to partition the substations from these users into subsets belonging to the same substation. For a line U with n users, the electricity consumption of each user on the line is:

[0017] It has T substations, and the total electricity consumption of each substation is:

[0018] Then we need to divide these n users into m mutually exclusive non-empty subsets. R i This ensures that the charge is conserved within each subset, i.e.: ; When m=3, the problem degenerates into a transformer area "phase sequence identification" problem, where E1, E2, and E3 represent the total power consumption of phases A, B, and C, respectively. The goal is to divide all users into three mutually exclusive non-empty subsets R1, R2, and R3, such that the total power consumption of each subset is as close as possible to the total power consumption of its corresponding phase.

[0019] Taking the user phase sequence identification problem as an example, assuming there are n users in a transformer area, the electricity consumption of these users can be represented by a matrix X, where {xm1 xm2 … xmn} is the electricity consumption of user n in the m-th time period (usually the m-th day).

[0020]

[0021] At the same time, a matrix Y can be constructed to represent the phase sequence affiliation of users:

[0022] Therefore, matrix Y should satisfy:

[0023] Here, j∈{1, 2, 3}.

[0024] If the phase-by-phase electrical quantities of the transformer are represented as matrix W:

[0025] And assume that the line loss rate for the corresponding phase in each time period is The total electricity consumption of the corresponding phase during this period is Then the line loss can be expressed as:

[0026] Based on the law of conservation of energy, we have:

[0027] When using dynamic programming to solve this problem, we can treat the user's phase affiliation Y as the variable to be optimized. The algorithm searches for all possible combinations of user phases, with the goal of minimizing the following loss function:

[0028] Thus, once the electricity consumption information of n users over m days and the phase electricity of each transformer are obtained, a solution for the phase sequence attribution relationship of a set of transformer areas can be obtained through dynamic programming algorithm.

[0029] However, within a fixed time period, due to the relatively small differences in users' electricity consumption behavior, the model is prone to confusion when determining phase sequence, leading to a decrease in the accuracy of the identification results. Especially when the transformer area topology is unknown, if the identification results cannot achieve 100% accuracy, all results must be manually verified, rendering "partial correctness" meaningless in practical engineering; that is, incomplete accuracy is equivalent to complete invalidity. Furthermore, simply increasing the amount of data cannot effectively improve the identification performance. Interference from insignificant data will still affect the model's judgment, and increasing the amount of data will not significantly improve the overall identification accuracy; on the contrary, it may increase the computational burden and reduce system efficiency.

[0030] Therefore, in order to improve data utilization and precision of results, we introduce a time consistency analysis strategy based on class ensemble learning to enhance the stability and reliability of the model across multiple time scales.

[0031] Assuming the transformer topology remains unchanged within a certain time period T, ideally, selecting l time subsets {t1, t2, ..., tl} ∈ T from this time period should yield consistent transformer topology allocation results. However, due to algorithm errors and variations in electricity consumption data, the actual results are inconsistent. Therefore, we consider taking the intersection of the topology results corresponding to these l subsets. As l becomes sufficiently large, this intersection can be considered highly reliable, even 100% accurate, for topology determination. Topologies not appearing in the intersection can be assigned a probability value based on their frequency of occurrence in each subset, representing the likelihood that they represent the true topology. The implementation process is as follows... Figure 3 .

[0032] Considering the significant time-varying characteristics of user electricity consumption behavior and to improve data utilization efficiency, this invention employs an overlapping sliding window method to sample sub-time series t, with all sub-time series set to a fixed length. Under this strategy, if a user appears in the solution set of a certain phase calculated from all sub-sequences (e.g., ...), ... Figure 3 If users u1, u3, etc., are identified, then that user is determined to belong to that phase with a 100% probability. For other users, the frequency with which they belong to that phase in each subsequence is counted and used as a probability indicator for assignment. When this probability exceeds a set threshold... ( If the confidence level is set to 50%, it can be determined that the product belongs to that phase; if it is lower than the threshold, the judgment result is retained and marked as "low confidence level", and final confirmation is required through manual review.

[0033] This application provides a technical solution (such as...) Figure 1 The core idea is to divide a long period of electricity consumption data into multiple shorter sub-periods, and independently apply the principle of energy conservation to each sub-period to obtain multiple possible user affiliations. Finally, by aggregating and analyzing the consistency of these affiliations across different time periods, the probability of each user belonging to a certain phase is calculated, and a final judgment with a confidence level is given based on this probability. Methodologically, this scheme draws on the idea of ​​self-gathering aggregation strategies in ensemble learning, which can effectively overcome misjudgments caused by accidental similarities in user electricity consumption behavior or data noise in single long-term data, thereby significantly improving the accuracy and robustness of identification.

[0034] Please see Figure 3 , Figure 3 This is a schematic diagram of a time consistency solution process according to an embodiment of the present invention. Figure 3 The principle of cross-time period consistency analysis in the core technical logic of this application is illustrated schematically. A complete time series T containing electricity consumption data of transformers and all users is processed by data segmentation to generate multiple sub-time series, such as sub-time series t1, t2, and t3 (where data segmentation adopts an overlapping sliding window method). Each sub-time series is fed into an independent solver, as shown in solver 301, solver 302, and solver 303 in the figure. Accordingly, each solver outputs a set of candidate solutions that are optimal for that time period based on the principle of energy conservation, as shown in solution sets 304, 305, and 306 in the figure. Due to the differences in data subsets, these solution sets may differ. Finally, the aggregation analysis result is obtained through aggregation analysis. The aggregation analysis result collects all solution sets and performs statistical analysis to calculate the final connection probability of each user's affiliation, such as... Figure 3 The aggregation analysis results shown are as follows: the connection probability of users u1 and u3 with phase A is 100%, meaning that users u1 and u3 are 100% located in phase A; the connection probability of users u2, u6, and u4 with phase A is 66.6%, meaning that users u2, u6, and u4 are 66.6% located in phase A; the connection probability of users u7 and u8 with phase A is 33.3%, meaning that users u7 and u8 are 33.3% located in phase A.

[0035] The technical solution of this application will be described in detail below through specific embodiments. Embodiment 1 describes in detail the basic implementation of a transformer area topology identification method based on energy conservation and cross-time consistency analysis. In this embodiment, data segmentation adopts an overlapping sliding window method, independent solution adopts a dynamic programming algorithm, and confidence judgment adopts a double-layer threshold logic.

[0036] Please continue reading. Figure 1 , Figure 1 This is a flowchart of a transformer area topology identification method according to an embodiment of the present invention. This method can be applied to either a terminal or a server. The terminal can be a desktop terminal or a mobile terminal; a mobile terminal can be at least one of a mobile phone, tablet computer, or laptop computer. The server can be a standalone server or a server cluster composed of multiple servers. This embodiment uses a terminal application as an example. Figure 1 The method shown includes the following steps: 101. Obtain the total power consumption data of each phase of the transformer in the distribution area and the power consumption data of multiple users in the distribution area within a preset time period; First, perform data acquisition step 101. This involves acquiring electricity consumption data for a preset time period from the power company's metering automation system, electricity information collection system, or other relevant databases. This includes the total electricity consumption data for each phase of the transformer in the distribution area, as well as the electricity consumption data of multiple users within the distribution area. Assume there are N users to be identified in the target distribution area (e.g., N=40), and the preset time period is a continuous D days (e.g., D=14 days). The acquired data includes: 1. Total electricity consumption data W for each phase of the transformer in the distribution area: This is time-series data, recording the total electrical energy output by phases A, B, and C of the transformer at each acquisition point within the preset time period. If the data acquisition frequency is once daily, it can be represented as a D-row, 3-column matrix W, where the elements... This represents the total electricity consumption of phase p (p∈{A, B, C}) on day d. 2. Electricity consumption data X for multiple users within the distribution area: This is a time-series data set containing N users. Also using a daily collection period, it can be represented as a D x N matrix X, where the elements... This represents the total electricity consumption of user u on day d.

[0037] It is understood that the electricity consumption data is low-frequency metering data, and its collection period can be determined according to the configuration of the existing metering system, such as 15 minutes, 1 hour, or 1 day. This embodiment uses daily electricity consumption data as an example for illustration, but this does not constitute a limitation. Using low-frequency data allows this method to directly utilize most existing metering facilities without the need for additional investment in expensive hardware, thus possessing high economic efficiency and scalability.

[0038] 102. Divide the total power consumption data of each phase of the transformer in the distribution area and the power consumption data of the multiple users into multiple sub-time periods according to time; Next, data segmentation step 102 is performed to obtain the complete time period data (matrices W and X), which is then segmented into multiple sub-time period datasets. In one feasible implementation, step 102 includes: using overlapping sliding windows or non-overlapping windows to segment the total power consumption data of each phase of the transformer substation and the power consumption data of the multiple users to generate the multiple sub-time period datasets.

[0039] In this embodiment, an overlapping sliding window method is used as an example. The specific operation is as follows: Set the window length L (e.g., L = 8 days) and the sliding step S (e.g., S = 1 day). The window length L should be chosen to ensure that the data within that time period is sufficient to reflect the differences in user electricity consumption behavior; typically, a period of one week or more is suitable. The sliding step S determines the degree of overlap between two adjacent sub-time periods; a smaller step size can generate more subsets of data, thus providing richer samples for subsequent probability aggregation.

[0040] Starting from day 1, data from day 1 to day 8 (including transformer phase data and all user data) is used as the dataset for the first sub-period. Then, the window slides forward one day, using data from day 2 to day 9 as the dataset for the second sub-period. This process continues until the end of the window reaches day 14. With 14 days of data, an 8-day window length, and a 1-day step, a total of 14 - 8 + 1 = 7 sub-period datasets can be generated. Each dataset contains 8 consecutive days of transformer three-phase electricity consumption data and electricity consumption data for 40 users. This process corresponds to... Figure 3 The process of generating multiple sub-time series t1, t2, and t3 from a complete time series T.

[0041] 103. For each of the aforementioned sub-time periods, based on the principle of energy conservation, independently solve a set of candidate user phase sequence attribution relationships; Subsequently, the independent solution step 103 is executed, and for each generated sub-time period dataset, a set of candidate user phase sequence attribution relationships is solved independently.

[0042] In one feasible implementation, independently solving for a set of candidate user phase sequence attribution relationships includes: employing a dynamic programming algorithm or a genetic algorithm, with the optimization objective of minimizing the difference between the total electricity consumption on the user side and the phase electricity consumption on the transformer side, to obtain the candidate user phase sequence attribution relationships. Specifically, independently solving for a set of candidate user phase sequence attribution relationships based on the principle of energy conservation includes: constructing the energy conservation principle into an optimization model with the user phase sequence attribution relationships as the variables to be solved, and using the minimum difference between the total electricity consumption on the user side and the sum of the phase electricity consumption on the transformer side and the estimated line loss as the optimization objective.

[0043] Step 103's core lies in constructing and solving an optimization problem based on the principle of energy conservation. In a distribution network, the principle of energy conservation is manifested as follows: at any given time, the total output power of a transformer phase should be approximately equal to the sum of the power consumption of all users connected to that phase, taking line losses into account. Based on this, the following optimization model can be constructed: for any dataset in a sub-time period (taking a subset containing L days of data as an example), the goal is to find an optimal user phase sequence assignment relationship. This relationship can be represented by an N x 3 matrix Y, where the matrix elements... It is either 0 or 1. If user u belongs to phase p, then =1, otherwise 0, and for each user u, satisfying .

[0044] The optimization objective is to minimize the difference between the total electricity consumption of each phase calculated by the user side based on the attribution relationship Y and the actual phase-specific electricity consumption measured by the transformer side. This difference can be quantified using a loss function L(Y). Specifically, a loss function can be defined as the sum of the squares of the phase errors at all sampling points: ; Among them, X d,u Y is the electricity consumption of user u on day d within the sub-time period. u,p W is the attribution variable to be solved. d,p This represents the total power consumption of the transformer on phase p on day d within the sub-time period. (W) lineloss,d,p This is the estimated line loss, which can be set as a fixed percentage (e.g., 2%-5%) of the total power consumption of the transformer based on experience, or calculated using a more complex line loss theoretical model. Incorporating line loss into the model makes the energy conservation equation more accurate, thereby improving the accuracy of the solution.

[0045] This is a combinatorial optimization problem. In this embodiment, step 103 uses a dynamic programming algorithm to solve this problem, which can effectively find the global optimal solution. Specifically, users can be assigned to phases A, B, and C one by one. Define the state dp(i, load...) A ,loadB After allocating the first i users, the cumulative load of phase A is load. A Phase B cumulative load is load B The minimum cumulative error that occurs at that time. Since the load of phase C can be obtained by subtracting the load from the total load. A and load B Since this is obtained, it does not need to be reflected in the state. The state transition equation describes the state changes and error increments corresponding to assigning the i-th user to phase A, phase B, or phase C. By filling the entire dynamic programming table, the allocation scheme that minimizes the total error can be found, and the corresponding ownership matrix Y can be obtained by backtracking.

[0046] Step 103 will execute the above dynamic programming solution process on the datasets of the 7 sub-time periods respectively, with each process corresponding to... Figure 3 One of the solvers (such as solvers 301, 302, and 303) is used. Therefore, 7 independent candidate user phase sequence attribution relationships will be obtained in the end, that is, 7 attribution matrices Y, which constitute 7 solution sets (such as solution sets 304, 305, and 306).

[0047] 104. Aggregate all candidate user phase sequence attribution relationships generated for the sub-time period to determine the connection probability of each user belonging to each phase; Then, the probability aggregation step 104 is performed, which receives all 7 sets of candidate affiliations and aggregates them to determine the connection probability of each user belonging to each phase. This process corresponds to... Figure 3 The determination of the aggregation analysis results. In one feasible implementation, determining the connection probability of each user belonging to each phase includes: counting the number of times each user is assigned to each phase in all generated candidate user phase sequence assignment relationships; and determining the ratio of the number of times to the total number of candidate user phase sequence assignment relationships as the connection probability of the user belonging to the corresponding phase.

[0048] The specific calculation method is as follows: For each user u (from 1 to 40), traverse 7 sets of candidate solutions. Count the number of times user u is assigned to phase A in these 7 sets of solutions, denoted as . The number of times user u is assigned to phase B in these 7 solutions is recorded as follows: The number of times user u is assigned to phase C in these 7 solutions is recorded as follows: Obviously, there are + + =7.

[0049] Then, the connection probability of user u belonging to each phase is calculated: the connection probability of user u belonging to phase A. = / 7. Connection probability of user u belonging to phase B = / 7. Connection probability of user u belonging to phase C = / 7.

[0050] For example, for user U1, if they are identified as belonging to phase C in all 7 solutions, then their connection probability of belonging to phase C is 7 / 7 = 100%. For user U2, if they are identified as phase B 5 times and phase C 2 times in the 7 solutions, then their connection probability of belonging to phase B is 5 / 7 ≈ 71.4%, and their connection probability of belonging to phase C is 2 / 7 ≈ 28.6%.

[0051] 105. Based on the connection probability and at least one preset confidence threshold, determine the final phase sequence assignment result and the corresponding confidence level for each user.

[0052] Finally, confidence decision step 105 is executed. This step, based on the calculated connection probability and combined with a preset confidence threshold, determines the final phase sequence assignment result and corresponding confidence level for each user. In this embodiment, a two-layer threshold decision logic is adopted. In one feasible implementation, the confidence threshold includes a first confidence threshold and a second confidence threshold, where the second confidence threshold is less than the first confidence threshold. Then, determining the final phase sequence assignment result and corresponding confidence level for each user based on the connection probability and at least one preset confidence threshold includes: if the user's highest connection probability is 100%, then the phase corresponding to the highest connection probability is determined to be at the first confidence level, where the first confidence level indicates that the user is absolutely and accurately assigned to the phase corresponding to the highest connection probability, and the first confidence threshold includes 100%; if the user's highest connection probability is less than 100% but higher than the second confidence threshold, then the phase corresponding to the highest connection probability is determined to be at the second confidence level, where the second confidence level indicates that the user is assigned to the phase corresponding to the highest connection probability with high confidence; if the user's highest connection probability is not higher than the second confidence threshold, then the phase corresponding to the highest connection probability is determined to be at the third confidence level, where the third confidence level indicates that the user is assigned to the phase corresponding to the highest connection probability with low confidence.

[0053] Specifically, first, a first and a second confidence threshold are set. The first confidence threshold is 100%, and the second confidence threshold is a preset value less than 100%, such as 50%. Then, the following decision is made for each user: 1. For user u, find the highest probability of connection: and the corresponding phase p max . 2. If Equal to 100% (i.e.) If the user's final phase sequence assignment is determined to be... p max It is assigned a first confidence level, representing absolute accuracy. This means that the user's affiliation remains highly consistent across all different subsets of data, making the identification results extremely reliable and requiring no manual verification.

[0054] 3. If If the confidence level is less than 100% but higher than the set second confidence threshold of 50%, then the user's final phase sequence assignment result is determined to be... p max It is then assigned a second confidence level, indicating high confidence. This means that the user's affiliation is consistent across most sub-periods, and the result has high reliability, but there may be a few instances of inconsistency.

[0055] 4. If If the result is not higher than 50% of the second confidence threshold, the user's result is classified as a third confidence level, indicating low confidence. This typically means that the user's electricity usage behavior is not easily distinguishable from other users, or that there are data quality issues, resulting in highly dispersed solution results across different sub-time periods. Such users are flagged to recommend that operations and maintenance personnel conduct focused manual checks.

[0056] Ultimately, a complete transformer substation topology identification report can be output, which may include each user's ID, recommended assigned phase, and corresponding confidence level (e.g., "Absolutely Accurate," "High Confidence," "Low Confidence"). This report provides clear and quantifiable decision support for power grid operation and maintenance, enabling maintenance personnel to concentrate limited human resources on a few low-confidence users, thereby greatly improving work efficiency.

[0057] Further refer to Table 1, which shows the phase sequence identification results for a transformer substation. An example is shown with phase sequence identification results for a transformer substation containing 40 users: Table 1: Phase Sequence Identification Results of Transit Areas

[0058] This embodiment uses a transformer substation with 40 users as the subject. During the experiment, user electricity consumption data and corresponding transformer phase-by-phase power data were collected for 14 consecutive days. A sliding window method was used for time series division, with a window step size of 1 and a sequence length of 8, ultimately generating 8 sub-time series for phase sequence identification and topology inference analysis. The final results are shown in Table 1. As can be seen from Table 1, for the 40 users in this substation, except for 3 users whose electricity consumption information was unavailable and therefore could not be determined, the other 37 users all made a determination. Among the 37 users who made a determination, 18 users were considered to be 100% located in a certain phase, and all determinations were correct. After comparing the probabilities of the other users, two users made incorrect determinations, resulting in an overall accuracy rate of 95%.

[0059] In a test of a real-world distribution area with 40 users, 3 users could not be analyzed due to a lack of long-term electricity usage records. For the remaining 37 users, using the method described in this embodiment, 18 users achieved a 100% highest connection probability, and their attribution results were deemed "absolutely accurate." On-site verification confirmed that the identification results for these 18 users were all correct. Among the remaining 19 users, most also had a highest connection probability far exceeding 50%, and were deemed "high confidence." Overall, the identification accuracy reached over 95%, far surpassing the traditional energy conservation method that uses only a single long-term data period for a one-time solution (its accuracy typically hovers around 70%-80%). The number of users requiring manual verification was drastically reduced from all 37 to single digits, demonstrating significant technical advantages and practical value.

[0060] This invention provides a method for topology identification of a transformer substation. The method includes: acquiring total power consumption data of each phase of the transformer substation and power consumption data of multiple users within the substation within a preset time period; dividing the total power consumption data of each phase of the transformer substation and the power consumption data of the multiple users into multiple sub-time period datasets; for each sub-time period dataset, independently solving a set of candidate user phase sequence attribution relationships based on the principle of energy conservation; aggregating all candidate user phase sequence attribution relationships generated for the sub-time period to determine the connection probability of each user belonging to each phase; and determining the final phase sequence attribution result and corresponding confidence level for each user based on the connection probability and at least one preset confidence threshold. By dividing the dataset into multiple sub-time period datasets according to time; independently solving a set of candidate user phase sequence attribution relationships for each sub-time period dataset based on the principle of energy conservation; aggregating the candidate user phase sequence attribution relationships to determine the connection probability of each user belonging to each phase; and significantly improving the accuracy of the identification results through stability evaluation of results at multiple time scales.

[0061] Furthermore, Embodiment 2 is provided, which is a variation of Embodiment 1 and is intended to illustrate the flexibility of the data segmentation steps. The main difference from Embodiment 1 is that the data segmentation in this embodiment uses a non-overlapping window approach.

[0062] As an optional implementation, assume that step 101 obtains electricity consumption data for 16 consecutive days. In step 102, the window length is set to 8 days, and a non-overlapping method is used for segmentation. That is, the data from day 1 to day 8 is used as one sub-period dataset, and the data from day 9 to day 16 is used as a second sub-period dataset. In this way, the complete 16-day data is divided into two completely independent sub-period datasets with no data overlap.

[0063] The subsequent steps are basically the same as those in Example 1 above. Step 103 will run a dynamic programming algorithm on the two sub-time period datasets respectively to obtain two sets of candidate user phase sequence attribution relationships.

[0064] Next, step 104 aggregates these two sets of candidate solutions. For each user u, the number of times they are assigned to phases A, B, and C in these two sets of solutions is counted. For example, if user U3 is assigned to phase A in both sets of solutions, then the probability of their connection to phase A is 2 / 2 = 100%. If user U4 is in phase B in the first set of solutions and phase C in the second set of solutions, then the probability of their connection to phase B and phase C is 1 / 2 = 50% each.

[0065] Finally, step 105 also makes a decision based on these probabilities and a preset threshold (e.g., 50%).

[0066] Compared to Example 1, this example generates fewer subsets (2 vs. 7), thus significantly reducing the total computation in step 103 and shortening the overall recognition process. Although the reduced sample size may slightly affect the smoothing effect of probability statistics, the time span of each subset (8 days) is still long enough to ensure the relative accuracy of a single solution. Therefore, through cross-time period consistency checks, this method can still effectively improve the reliability of recognition. This example is suitable for application scenarios with limited computing resources or higher requirements for recognition efficiency. This example also demonstrates that the data segmentation method in this application has broad applicability; both overlapping and non-overlapping segmentation methods fall within the protection scope of this application.

[0067] Example 3 is provided, which is another variation of Example 1, and is intended to illustrate the flexibility of step 103. The main difference from Example 1 is that step 103 in this example uses a genetic algorithm instead of a dynamic programming algorithm to solve the energy conservation optimization problem.

[0068] Steps 101 and 102 are exactly the same as in Example 1, and similarly generate 7 overlapping sub-time period datasets, each containing 8 days of electricity consumption data.

[0069] In step 103, for each sub-time period of the dataset, step 103 may execute the following solution process based on a genetic algorithm: 1. Chromosome Encoding: Encoding a potential solution, i.e., a complete user phase assignment scheme, into a chromosome. For N users, a chromosome can be an integer array of length N, where the value of the i-th element of the array is taken from {0, 1, 2}, representing whether user i is assigned to phase A, phase B, or phase C, respectively.

[0070] 2. Fitness Function: The fitness function is used to evaluate the quality of each chromosome (i.e., each solution). Its core is the loss function L(Y) defined in Example 1. To transform the minimization problem into a maximization problem, the fitness function can be defined as the reciprocal of the loss function, for example: Fitness(Y) = 1 / (1 + L(Y)); The calculation method for L(Y) is exactly the same as in Example 1, and the estimated line loss is also taken into account. The addition of 1 here is to prevent the denominator from being zero. A chromosome with higher fitness indicates that its corresponding assignment scheme is more in line with energy conservation.

[0071] 3. Genetic Operations: * Initialization: Randomly generate an initial population containing several chromosomes. * Selection: Based on the fitness value of each chromosome, use strategies such as roulette wheel selection or tournament selection to select superior chromosomes for the next generation. Individuals with higher fitness have a greater probability of being selected. * Crossover: Pair the selected chromosomes and perform a crossover operation with a certain probability (e.g., single-point crossover or uniform crossover) to generate new offspring chromosomes. This operation simulates gene recombination, aiming to combine the superior parts of different excellent solutions. * Mutation: Randomly change certain gene positions (i.e., the sequence allocation of a user) in the offspring chromosomes with a small probability. This operation simulates gene mutation, which helps to escape local optima and increase population diversity.

[0072] 4. Iteration: The selection, crossover, and mutation operations are repeatedly performed to allow the population to evolve continuously. The algorithm terminates after a preset number of iterations or when the optimal fitness of the population stabilizes. At this point, the chromosome with the highest fitness in the population is considered the optimal candidate solution for the current sub-time period dataset.

[0073] Step 103 runs the above genetic algorithm on the datasets of the 7 sub-time periods respectively, thereby obtaining 7 groups of candidate user phase sequence attribution relationships.

[0074] The subsequent steps 104 and 105 are exactly the same as those in Example 1 above, which involve statistical analysis and decision-making on the seven candidate solutions generated by the genetic algorithm.

[0075] Heuristic optimization algorithms such as genetic algorithms are particularly suitable for solving large-scale combinatorial optimization problems. When the number of users N in a distribution area is extremely large, dynamic programming algorithms may face problems such as state space explosion and excessive computational cost. Genetic algorithms, on the other hand, have good scalability; although they do not guarantee finding the global optimum, they can usually find a high-quality approximate optimum within a reasonable time. Therefore, this embodiment provides a more efficient and practical solution for handling ultra-large-scale distribution areas, thus proving that the "independent solution" step in this application is compatible with various different optimization algorithms.

[0076] Example 4 is provided, which is another variation of Example 1, designed to illustrate the flexibility of step 105 and demonstrate a more refined multi-level confidence partitioning strategy. The first four steps of this example, namely data acquisition step 101, data segmentation step 102, independent solution step 103, and probability aggregation step 104, can be the same as any of the schemes in Examples 1, 2, or 3. It is assumed that after these steps, step 104 has calculated the connection probability belonging to phases A, B, and C for each user.

[0077] In confidence decision step 105, multiple confidence thresholds can be set to achieve more refined confidence level classification. For example, three thresholds can be set: 100%, 80%, and 50%. The decision logic is as follows: For each user u, find its highest connection probability. P max and corresponding phase p max .

[0078] 1. If P max If the value equals 100%, then the user's affiliation is determined as follows: p max The confidence level is "confirmed" or "first confidence level". These results can be directly written to the production database as the final topology profile without any manual intervention.

[0079] 2. If P max If the user's affiliation is within the range [80%, 100%), then the user's affiliation is determined as follows: p max The confidence level is either "high confidence" or "second confidence level". These results are considered to be highly likely to be correct and can be adopted by the system by default, but they can also be marked as low-priority review tasks and checked randomly when resources permit.

[0080] 3. If P max If the user's affiliation falls within the range [50%, 80%), then the user's affiliation is determined as follows: p max The confidence level is "medium confidence level" or "third confidence level". These results have some reference value, but there is uncertainty, and it is recommended that operations and maintenance personnel verify them.

[0081] 4. If P max If the confidence level is below 50%, the user's affiliation is considered uncertain, with a confidence level of "low confidence" or "fourth confidence level." These users must be manually verified on-site to determine their affiliation.

[0082] By employing this multi-level confidence classification, this embodiment provides richer and more refined decision-making support for operation and maintenance (O&M) work. The O&M management system can automatically trigger different work order processes based on different confidence levels. For example, it can create high-priority on-site verification work orders for users with "low confidence" and "medium confidence," while archiving or creating low-priority spot check work orders for users with "high confidence." This strategy facilitates the optimized allocation of O&M resources, enabling smarter and more efficient power grid management.

[0083] In summary, the method and apparatus for transformer topology identification based on energy conservation and cross-time period consistency analysis provided in this application successfully solves the problems of low accuracy, unstable results, and inability to quantify reliability in existing technologies that rely solely on low-frequency data for topology identification. This is achieved by segmenting long-term data, solving for each sub-time period independently, and then performing probability aggregation and confidence level judgment on multiple sets of results. This solution requires no additional hardware, is low-cost, easy to deploy, and can output highly accurate identification results with clear confidence levels, thereby greatly reducing the workload of manual verification and providing solid technical support for the automated and intelligent maintenance of distribution network topology archives.

[0084] Compared with existing technologies, this application has the following beneficial effects: 1. By dividing long-term data into multiple sub-periods and solving them independently, and then aggregating and analyzing the results, it utilizes an ensemble learning-like approach to effectively smooth and suppress misjudgments caused by data randomness or similar user electricity consumption behavior within a single period, significantly improving the overall accuracy and robustness of the identification results. 2. This application can output identification results with confidence level classifications. For users with a connection probability of 100%, it can be judged as "absolutely accurate," requiring no manual verification; for users with a probability higher than the threshold, it can be judged as "high confidence," greatly reducing the scope requiring manual verification and significantly reducing the manpower cost and workload of power grid topology verification, achieving a high degree of automation. 3. This application is entirely based on existing low-frequency metering system data, requiring no additional high-frequency acquisition hardware or power grid modifications, resulting in extremely low deployment costs and economic viability for large-scale application in distribution networks. 4. By providing accurate, reliable, and automatically updated transformer topology data, it provides solid data support for advanced applications such as intelligent operation and maintenance, precise fault location, load forecasting, and analysis of distribution networks.

[0085] Please see Figure 4 , Figure 4 This is a structural block diagram of a transformer area topology identification device according to an embodiment of the present invention, as shown below. Figure 4 The apparatus shown includes: Data acquisition module 401 is used to acquire the total power consumption data of each phase of the transformer in the distribution area and the power consumption data of multiple users in the distribution area within a preset time period; The data segmentation module 402 is used to divide the total power consumption data of each phase of the transformer in the distribution area and the power consumption data of the multiple users into multiple sub-time period datasets according to time. The independent solution module 403 is used to independently solve a set of candidate user phase sequence attribution relationships for each of the sub-time periods based on the principle of energy conservation. The probability aggregation module 404 is used to aggregate all candidate user phase sequence attribution relationships generated for the sub-time period to determine the connection probability of each user belonging to each phase. The confidence decision module 405 is used to determine the final phase sequence assignment result and the corresponding confidence level for each user based on the connection probability and at least one preset confidence threshold.

[0086] It should be noted that, Figure 4 The function of each module in the device shown is as follows: Figure 1 The steps in the method shown are similar, and will not be repeated here to avoid repetition. For details, please refer to [reference needed]. Figure 1 The content of each step in the method shown.

[0087] This invention provides a transformer substation topology identification device, comprising: a data acquisition module for acquiring total power consumption data of each phase of the transformer in the substation and power consumption data of multiple users within the substation within a preset time period; a data segmentation module for segmenting the total power consumption data of each phase of the transformer in the substation and the power consumption data of the multiple users into multiple sub-time period datasets according to time; an independent solution module for independently solving a set of candidate user phase sequence attribution relationships for each sub-time period dataset based on the principle of energy conservation; a probability aggregation module for aggregating all candidate user phase sequence attribution relationships generated for the sub-time periods to determine the connection probability of each user belonging to each phase; and a confidence level decision module for determining the final phase sequence attribution result and corresponding confidence level for each user based on the connection probability and at least one preset confidence level threshold. The dataset is divided into multiple sub-time periods using the above method. For each sub-time period dataset, a set of candidate user phase sequence attribution relationships is independently solved based on the principle of energy conservation. The candidate user phase sequence attribution relationships are aggregated to determine the connection probability of each user belonging to each phase. By evaluating the stability of the results under multiple time scales, the accuracy of the recognition results is significantly improved.

[0088] Figure 5 An internal structural diagram of a computer device in one embodiment is shown. This computer device can specifically be a terminal or a server. Figure 5 As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and may also store a computer program, which, when executed by the processor, causes the processor to perform the aforementioned methods. The internal memory may also store a computer program, which, when executed by the processor, causes the processor to perform the aforementioned methods. Those skilled in the art will understand that… Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0089] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform actions such as... Figure 1 The steps of the method shown in any embodiment.

[0090] In one embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, causes the processor to perform the following actions: Figure 1 The steps of the method shown in any embodiment.

[0091] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0092] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0093] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for identifying transformer substation topology, characterized in that, The method includes: Obtain the total power consumption data of each phase of the transformer in the distribution area and the power consumption data of multiple users in the distribution area within a preset time period; The total power consumption data of each phase of the transformer in the distribution area and the power consumption data of the multiple users are divided into multiple sub-time periods of data. For each of the aforementioned sub-time periods, based on the principle of energy conservation, a set of candidate user phase sequence attribution relationships are independently solved; Aggregate all candidate user phase sequence attribution relationships generated for the sub-time period to determine the connection probability of each user belonging to each phase; Based on the connection probability and at least one preset confidence threshold, the final phase sequence assignment result and corresponding confidence level are determined for each user.

2. The method according to claim 1, characterized in that, The dataset, which divides the total power consumption data of each phase of the transformer substation and the power consumption data of the multiple users into multiple sub-time periods according to time, includes: The total power consumption data of each phase of the transformer in the distribution area and the power consumption data of the multiple users are segmented using overlapping sliding windows or non-overlapping windows to generate datasets for the multiple sub-time periods.

3. The method according to claim 1, characterized in that, The independent solution of a set of candidate user phase sequence attribution relationships includes: Using dynamic programming or genetic algorithms, with the optimization objective of minimizing the difference between the total electricity consumption on the user side and the phase electricity consumption on the transformer side, the candidate user phase sequence assignment relationship is obtained.

4. The method according to claim 1 or 3, characterized in that, The method of independently solving a set of candidate user phase sequence attribution relationships based on the principle of energy conservation includes: The energy conservation principle is constructed into an optimization model with the user phase sequence attribution relationship as the variable to be determined, and the optimization objective is to minimize the difference between the total electricity consumption on the user side and the sum of the phase electricity consumption and the estimated line loss on the transformer side.

5. The method according to claim 1, characterized in that, Determining the connection probability of each user belonging to each phase includes: Count the number of times each user is assigned to each phase in all the generated candidate user phase sequence attribution relationships; The ratio of the number of times to the total number of candidate user phase sequence associations is determined as the connection probability of a user belonging to the corresponding phase.

6. The method according to claim 1, characterized in that, The confidence threshold includes a first confidence threshold and a second confidence threshold. If the second confidence threshold is less than the first confidence threshold, then determining the final phase sequence assignment result and corresponding confidence level for each user based on the connection probability and at least one preset confidence threshold includes: If a user's highest connection probability is equal to 100%, then the phase corresponding to the highest connection probability is determined to be the first confidence level. The first confidence level is used to indicate that the user absolutely and accurately belongs to the phase corresponding to the highest connection probability. The first confidence threshold includes 100%. If a user's highest connection probability is less than 100% but higher than the second confidence threshold, then the phase corresponding to the highest connection probability is determined to be the second confidence level. The second confidence level is used to indicate that the user's high confidence belongs to the phase corresponding to the highest connection probability. If the user's highest connection probability is not higher than the second confidence threshold, then the phase corresponding to the highest connection probability is determined to be the third confidence level. The third confidence level is used to indicate that the user's low confidence belongs to the phase corresponding to the highest connection probability.

7. The method according to claim 1, characterized in that, The electricity consumption data is low-frequency metering data, and the data collection period is 15 minutes, 1 hour, or 1 day.

8. A transformer area topology identification device, characterized in that, The device includes: The data acquisition module is used to acquire the total power consumption data of each phase of the transformer in the distribution area and the power consumption data of multiple users in the distribution area within a preset time period. The data segmentation module is used to divide the total power consumption data of each phase of the transformer in the distribution area and the power consumption data of the multiple users into multiple sub-time period datasets according to time. An independent solution module is used to independently solve a set of candidate user phase sequence attribution relationships for each of the sub-time periods, based on the principle of energy conservation. The probability aggregation module is used to aggregate all candidate user phase sequence attribution relationships generated for the sub-time period to determine the connection probability of each user belonging to each phase. The confidence level decision module is used to determine the final phase sequence assignment result and the corresponding confidence level for each user based on the connection probability and at least one preset confidence level threshold.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it causes the processor to perform the steps of the method as described in any one of claims 1 to 7.

10. A computer device, comprising a memory and a processor, characterized in that, The memory stores a computer program that, when executed by the processor, causes the processor to perform the steps of the method as described in any one of claims 1 to 7.

Images