A data-driven automatic topology data management method for power grid

Abstract

The present invention proposes a data-driven automatic topology data management method for power grids, which includes: 1. Online acquisition of operating data of each node in the target area power grid within a time period, and repair of unreasonable parts of the operating data; 2. Setting the number of sampling times to obtain the power data at each node of each sampling; 3. Using the branch and bound optimization algorithm based on the power data at each node, calculate the node superior-subordinate relationship of each sampling; 4. Using the node current, verify the topology scheme of each sampling, and in each sampled topology scheme, take the topology scheme with the smallest error as the optimal topology scheme. The beneficial effects of the present invention: 1) It realizes automated equipment topology maintenance; 2) It solves the problem of inaccurate data and untimely update caused by manual maintenance of topology data of power companies; 3) The present invention does not involve the reconstruction and construction of distribution networks, and has low cost and high degree of practicality.

Description

Technical Field

[0001] The invention relates to a data-driven automatic topology data management method for a power grid, and belongs to the technical field of distribution network data management. Background Art

[0002] At present, when managing power grid data, most of them rely on manual maintenance to draw power grid wiring diagrams and maintain equipment parameters. The manual method is inefficient. When maintaining load power grid data of 10 kV and below, the manual error rate is high due to the complex structure of the distribution network and the large number of equipment. In addition, when new equipment is installed and put into operation, only manual maintenance is used, and the data update cycle is long, which cannot provide accurate data support for business applications.

[0003] Existing technologies cannot achieve automatic topological data management of the power grid by analyzing operating data such as current, power, and electricity meter readings at each node of the power grid. Summary of the invention

[0004] The present invention proposes a data-driven automatic topology data management method for a power grid, which aims to solve the problem that the prior art cannot realize automatic topology data management for a power grid.

[0005] The technical solution of the present invention is a data-driven automatic topology data management method for power grids, the method comprising:

[0006] 1. Online acquisition of operating data of each node in the target area power grid within a period of time, and repair of unreasonable operating data;

[0007] 2. Set the sampling times and obtain the power data at each node of each sampling;

[0008] 3. According to the power data at each node, use the branch and bound optimization algorithm to calculate the superior-subordinate relationship of each sampled node;

[0009] 4. Use the node current to verify each sampled topology solution. Among the sampled topology solutions, the topology solution with the smallest error is taken as the optimal topology solution.

[0010] Furthermore, the online acquisition of the operating data of each node in the target area power grid within a time period and repairing the unreasonable part of the operating data specifically includes:

[0011] 1-1. First, the quality of the operation data of each node of the power grid is analyzed to screen out unqualified data; the operation data of each node of the power grid includes the voltage, current, active power and reactive power of each node of the power grid;

[0012] 1-2. Use a method that combines business rules between data and data interpolation to repair unqualified data.

[0013] Furthermore, the unqualified data includes three types: empty, containing illegal characters, and abnormal numerical values ​​exceeding the limit.

[0014] Furthermore, the abnormal value exceeding the limit is obtained by setting the verification deviation according to the power design standard, and the upper and lower limits of the operating data of each branch are given according to the transformer capacity and the line current carrying capacity.

[0015] Furthermore, the data-to-data business rule is the relationship between active power, reactive power, voltage, and current, which is specifically expressed as formula (1):

[0016]

[0017] In the formula: S is the transformer capacity, U is the voltage, I is the line current carrying capacity, P is the active power, and Q is the reactive power.

[0018] Furthermore, the data interpolation method is specifically as follows: substitute data for unqualified data is derived from normal data; when the business rules contain unqualified data, the derived substitute data is replaced by the interpolation method to repair the unqualified data.

[0019] Furthermore, the setting of the sampling times and obtaining the power data at each node of each sampling specifically includes:

[0020] Set the sampling times to m, and calculate the power data of each node in the power grid in each sampling period according to the operating data within the sampling time; if the sampling includes frozen power data, the power in a certain sampling period is W:

[0021] W = W2-W1;

[0022] Among them, W2 is the frozen power at the end of sampling, and W1 is the frozen power at the start of sampling;

[0023] If the sampling does not contain frozen power data, the power in the sampling period is calculated by power integration:

[0024]

[0025] Among them, t1 is the starting time of a sampling period, t2 is the ending time of a sampling period, and P is the active power at the node.

[0026] Furthermore, the node superior-subordinate relationship of each sample is calculated using a branch and bound optimization algorithm according to the power data at each node, specifically including:

[0027] 3-1. According to Kirchhoff's current law, the sum of the current flowing into the lower level of the node is equal to the current collected by the node to the upper level. Suppose the number of branches in the power grid is n;

[0028] 3-2. According to the known sampling periods, the power of each branch of the power grid is [w1, w2, w3, w4, ..., w n ], denoted as W;

[0029] 3-3. The subordinate relationship between the branches of the power grid is:

[0030] It is denoted as X;

[0031] where x ij Indicates the subordinate relationship between branch i and branch j; if the superior branch of branch i is j, then x ij = -1, if the parent branch of branch i is not j, then x ij =0; when i=j, x ij =1;

[0032] 3-4. Establish the objective function. The objective function is to minimize the power imbalance at each node of the power grid, that is:

[0033] min Z = WX;

[0034] The constraints are:

[0035]

[0036] Furthermore, the node current is used to verify each sampled topology solution, and among each sampled topology solution, the topology solution with the smallest error is taken as the optimal topology solution, which specifically includes the following:

[0037] 4-1. During the analysis period of analyzing the quality of the operation data of each node of the power grid, the current data of each branch at a certain moment [I1, I2, I3...I n ], denoted as I t ;

[0038] 4-2. The branch topology solution obtained from m sampling cycles is [X1, X2, X3...X k ] Where k≤m, the current data within the analysis period is used to verify the accuracy of the branch topology scheme.

[0039] Further, the current data in the analysis period is used to verify the accuracy of the branch topology solution, specifically including:

[0040] Calculate the error of each topology solution during the analysis period, as follows:

[0041]

[0042] Among them, t1 is the starting time of the analysis period, t2 is the ending time of the analysis period, and the error set corresponding to the k alternative solutions is [δ1, δ2, δ3...δ k ];

[0043] When the error is the smallest, that is, when δ=min[δ1,δ2,δ3...δ k ] is taken as the final point network topological relationship.

[0044] Beneficial effects of the present invention:

[0045] 1) Based on the three-phase current, power, geographic coordinates and other data of each branch device of the distribution network, this paper considers the long-term operation of the power grid and uses machine learning algorithms to analyze the current, power, electricity meter reading and other operating data of each node of the power grid to realize automatic topological data management of the power grid, automatically analyze the power grid topological connection relationship between devices, and realize automatic equipment topology maintenance;

[0046] 2) Solved the problem of inaccurate data and untimely updates caused by manual maintenance of power company topology data;

[0047] 3) The present invention belongs to the distribution network operation data technology, does not involve the transformation and construction of the distribution network, has low cost and high degree of practicality. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Attached Figure 1 It is a schematic diagram of the process of the present invention.

[0049] Attached Figure 2 This is a schematic diagram of the topological structure of the power grid equipment formed by the method of the present invention in Example 1. DETAILED DESCRIPTION

[0050] A data-driven automatic topology data management method for a power grid, the method comprising:

[0051] 1. Online acquisition of operating data of each node in the target area power grid within a period of time, and repair of unreasonable operating data;

[0052] 2. Set the sampling times and obtain the power data at each node of each sampling;

[0053] 3. According to the power data at each node, use the branch and bound optimization algorithm to calculate the superior-subordinate relationship of each sampled node;

[0054] 4. Use the node current to verify each sampled topology solution. Among the sampled topology solutions, the topology solution with the smallest error is taken as the optimal topology solution.

[0055] The online acquisition of the operating data of each node in the target area power grid within a time period and repairing the unreasonable part of the operating data specifically includes:

[0056] 1-1. First, the quality of the operation data of each node of the power grid is analyzed to screen out unqualified data; the operation data of each node of the power grid includes the voltage, current, active power and reactive power of each node of the power grid;

[0057] 1-2. Use a method that combines business rules between data and data interpolation to repair unqualified data.

[0058] The unqualified data includes three types: empty, containing illegal characters, and abnormal numerical values ​​exceeding the limit.

[0059] The abnormal value exceeding the limit can be obtained by setting the verification deviation according to the power design standard, and the upper and lower limits of the operation data of each branch line are given according to the transformer capacity and the line current carrying capacity, as shown in the following Table 1:

[0060] Table 1

[0061]

[0062] In Table 1: S is the transformer capacity, U is the voltage, I is the line current carrying capacity, W1 is the frozen power at the start time of sampling, W2 is the frozen power at the end time of sampling, and e is the calibration deviation.

[0063] The data service rule is the relationship between active power, reactive power, voltage and current, which is specifically expressed as formula (1):

[0064]

[0065] In the formula: In the formula: S is the transformer capacity, U is the voltage, I is the line current carrying capacity, P is the active power, and Q is the reactive power.

[0066] The data interpolation method is specifically: deriving substitute data for unqualified data from normal data. When the business rules contain unqualified data, the derived substitute data is replaced by the interpolation method to repair the unqualified data.

[0067] The setting of the sampling times and obtaining the power data at each node of each sampling specifically includes:

[0068] Set the sampling times to m, and calculate the power data of each node in the power grid in each sampling period according to the operating data within the sampling time; if the sampling includes frozen power data, the power in a certain sampling period is W:

[0069] W = W2-W1;

[0070] Among them, W2 is the frozen power at the end of sampling, and W1 is the frozen power at the start of sampling;

[0071] If the sampling does not contain frozen power data, the power in the sampling period is calculated by power integration:

[0072]

[0073] Among them, t1 is the starting time of a sampling period, t2 is the ending time of a sampling period, and P is the active power at the node.

[0074] The branch and bound optimization algorithm is used to calculate the node superior-subordinate relationship of each sample according to the power data at each node, specifically including:

[0075] 3 1. According to Kirchhoff's current law, the sum of the current flowing into the lower level of the node is equal to the current collected by the node to the upper level. Suppose the number of branches in the power grid is n;

[0076] 3-2. According to the known sampling periods, the power of each branch of the power grid is [w1, w2, w3, w4, ..., w n ], denoted as W;

[0077] 3-3. The subordinate relationship between the branches of the power grid is:

[0078] It is denoted as X;

[0079] where x ij Indicates the subordinate relationship between branch i and branch j; if the superior branch of branch i is j, then x ij = -1, if the parent branch of branch i is not j, then x ij =0; when i=j, x ij =1;

[0080] 3-4. Establish the objective function. The objective function is to minimize the power imbalance at each node of the power grid, that is:

[0081] min Z = WX;

[0082] The constraints are:

[0083]

[0084] Use the branch and bound algorithm to solve X corresponding to m samples, and set it as an alternative solution to the branch topology solution.

[0085] The node current is used to verify each sampled topology solution, and among each sampled topology solution, the topology solution with the smallest error is taken as the optimal topology solution, which specifically includes:

[0086] 4-1. During the analysis period of analyzing the quality of the operation data of each node of the power grid, the current data of each branch at a certain moment [I1, I2, I3...I n ], denoted as I t ;

[0087] 4-2. The branch topology solution obtained from m sampling cycles is [X1, X2, X3...X k ] Where k≤m, the current data within the analysis period is used to verify the accuracy of the branch topology scheme.

[0088] The use of current data within the analysis period to verify the accuracy of the branch topology scheme specifically includes:

[0089] Calculate the error of each topology solution during the analysis period, as follows:

[0090]

[0091] Among them, t1 is the starting time of the analysis period, t2 is the ending time of the analysis period, and the error set corresponding to the k alternative solutions is [δ1, δ2, δ3...δ k ];

[0092] When the error is the smallest, that is, when δ=min[δ1,δ2,δ3...δ k ] is taken as the final point network topological relationship.

[0093] Example 1

[0094] The present invention is described in detail below in conjunction with application examples.

[0095] The known operation data of each node on site are shown in Table 2-Table 5:

[0096] Table 2

[0097] Branch Road time P Q I U 1# 1 point Null 19.20442 3.52 10.605 1# 2 o'clock 102.4599 34.15331 6.26 10.71 1# 3 o'clock 195.9178 65.30593 11.97 10.605 1# 4 o'clock 194.3629 64.78763 11.875 11.025 1# 5 o'clock 159.5822 53.19405 9.75 11.025 1# 6 o'clock 212.6125 70.87084 12.99 10.815 1# 7 o'clock 202.7921 67.59736 12.39 10.92 1# 8 o'clock 165.9654 Null 10.14 11.025 1# 9 o'clock 147.7976 49.26587 9.03 11.025 1# 10 o'clock 156.145 52.04833 9.54 10.92 2# 1 point 30.93439 10.31146 1.89 10.605 2# 2 o'clock 59.24999 19.75 3.62 10.71 2# 3 o'clock 104.0967 34.69889 6.36 10.395 2# 4 o'clock 100.6595 33.55317 6.15 10.5 2# 5 o'clock 44.19198 14.73066 2.7 11.025 2# 6 o'clock 92.80316 30.93439 5.67 10.815 2# 7 o'clock 122.2645 40.75483 7.47 10.605 2# 8 o'clock 94.27622 31.42541 5.76 10.92 2# 9 o'clock 66.77899 22.25966 4.08 11.025 2# 10 o'clock Null 14.07596 2.58 10.815 ;

[0098] Table 3

[0099] Branch Road time P Q I U 3# 1 point 26.67886 8.892954 1.63 10.605 3# 2 o'clock 43.20994 14.40331 2.64 10.395 3# 3 o'clock 91.82111 30.60704 5.61 10.605 3# 4 o'clock 93.70337 31.23446 5.725 11.025 3# 5 o'clock 115.3902 38.46339 7.05 10.71 3# 6 o'clock 119.8094 39.93646 7.32 10.605 3# 7 o'clock 80.52761 26.84254 4.92 10.92 3# 8 o'clock 71.68921 23.8964 4.38 Null 3# 9 o'clock 81.01863 27.00621 4.95 10.5 3# 10 o'clock 113.9171 37.97237 6.96 10.92 4# 1 point 24.5511 8.1837 1.5 10.605 4# 2 o'clock 34.04419 11.34806 2.08 10.71 4# 3 o'clock 49.59322 16.53107 3.03 10.395 4# 4 o'clock Null 15.68543 2.875 10.29 4# 5 o'clock 21.11395 7.037982 1.29 11.025 4# 6 o'clock 46.15607 15.38536 2.82 10.815 4# 7 o'clock 70.21615 23.40538 4.29 10.605 4# 8 o'clock 54.01242 18.00414 3.3 10.92 4# 9 o'clock 39.28176 13.09392 2.4 10.92 4# 10 o'clock 20.1319 6.710634 1.23 10.815 ;

[0100] Table 4

[0101] Branch Road time P Q I U 5# 1 point 6.383286 2.127762 0.39 10.605 5# 2 o'clock 25.2058 8.401932 1.54 10.29 5# 3 o'clock 54.50344 18.16781 3.33 10.29 5# 4 o'clock 53.60324 17.86775 3.275 10.5 5# 5 o'clock 23.07803 7.692678 1.41 10.29 5# 6 o'clock 46.64709 15.54903 2.85 10.605 5# 7 o'clock 52.04833 17.34944 null 10.5 5# 8 o'clock 40.2638 13.42127 2.46 10.5 5# 9 o'clock 27.49723 9.165744 1.68 11.025 5# 10 o'clock 22.09599 7.36533 1.35 10.71 6# 1 point 11.45718 3.81906 0.7 10.605 6# 2 o'clock 23.24171 7.747236 1.42 10.395 6# 3 o'clock 67.76104 22.58701 4.14 10.605 6# 4 o'clock 61.78694 20.59565 3.775 11.025 6# 5 o'clock 46.64709 15.54903 2.85 10.71 6# 6 o'clock 82.98273 27.66091 5.07 10.605 6# 7 o'clock 34.37154 11.45718 2.1 10.815 6# 8 o'clock 33.3895 11.12983 2.04 10.815 6# 9 o'clock 42.71891 14.23964 2.61 10.5 6# 10 o'clock 50.08424 16.69475 3.06 10.92 ;

[0102] Table 5

[0103] Branch Road time P Q I U 7# 1 point 15.22168 5.073894 0.93 10.395 7# 2 o'clock 19.96823 6.656076 1.22 10.395 7# 3 o'clock 24.06008 8.020026 null 10.29 7# 4 o'clock 31.91643 10.63881 1.95 11.025 7# 5 o'clock 68.74308 22.91436 4.2 10.5 7# 6 o'clock 36.82665 12.27555 2.25 10.29 7# 7 o'clock 46.15607 15.38536 2.82 10.92 7# 8 o'clock 38.29972 12.76657 2.34 11.025 7# 9 o'clock 38.29972 12.76657 2.34 10.29 7# 10 o'clock 63.83286 21.27762 3.9 10.815 .

[0104] Step 1:

[0105] First, the quality of the operating data such as voltage, current, active power, reactive power, etc. of each node in the power grid is analyzed to filter out unqualified data; as shown in Table 6 below, there is an abnormality of missing null operating data:

[0106] Table 6

[0107] Branch Road time P Q I U 1# 1 point null 19.20442 3.52 10.605 1# 8 o'clock 165.9654 null 10.14 11.025 2# 10 o'clock null 14.07596 2.58 10.815 4# 4 o'clock null 15.68543 2.875 10.29 5# 7 o'clock 52.04833 17.34944 null 10.5 7# 3 o'clock 24.06008 8.020026 null 10.29 ;

[0108] According to the relationship between power, voltage and current The missing data is repaired by the relationship, as shown in Table 7:

[0109] Table 7

[0110] Branch Road time P Q I U 1# 1 point 57.61325 19.20442 3.52 10.605 1# 8 o'clock 165.9654 55.32181 10.14 11.025 2# 10 o'clock 42.22789 14.07596 2.58 10.815 4# 4 o'clock 47.05628 15.68543 2.875 10.29 5# 7 o'clock 52.04833 17.34944 3.18 10.5 7# 3 o'clock 24.06008 8.020026 1.47 10.29 .

[0111] Step 2:

[0112] Set the sampling times to 4, randomly sample 4 times, and each sample is adjacent to the time. If the sampling does not include frozen power data, the power data is derived based on the active power integral. As shown in Table 8 below:

[0113] Table 8

[0114]

[0115] Step 3:

[0116] Assume that the number of branches in the power grid is n=7, as in step 2, taking the first sampling period as an example:

[0117] (1) Electricity of each branch of the power grid

[0118] W=[w1,w2,w3,w4,…,w n ] = [550.35, 294.94, 255.41, 155.24, 139.69, 164.24, 91.16];

[0119] (2) The subordinate relationship between the branches of the power grid is:

[0120] Denoted as X, it is the variable to be solved;

[0121] where x ij Indicates the subordinate relationship between branch i and branch j; if the superior branch of branch i is j, then x ij = -1, if the parent branch of branch i is not j, then x ij =0; when i=j, x ij =1;

[0122] (3) The objective function is to minimize the power imbalance at each node of the power grid, that is:

[0123]

[0124] (4) The constraint condition is: denoted as:

[0125]

[0126] Wherein, i and j are 1 to 7;

[0127] (5) Use the branch and bound algorithm to solve X corresponding to sample 1, as shown below:

[0128]

[0129] (6) Similarly, the corresponding X is obtained for samples 2, 3, and 4, as shown below:

[0130]

[0131]

[0132]

[0133] Step 4:

[0134] During the analysis period, the current data of each branch at a certain time is known.

[0135]

[0136] As in step 3, the branch topology solution obtained from 4 sampling cycles is:

[0137] [X1, X2, X3, X4],

[0138] in

[0139]

[0140] Calculate the error of each solution during the analysis period, δ = I t X;

[0141] δ X1\2 =0.001,δ X3\4 =1.345;

[0142] Therefore, X1 and X2 are selected as the final topology solutions. The final topology is shown in the attached figure. Figure 2 shown.

Claims

1. A data-driven automatic topology data management method for power grid, characterized by: include: (1) Online acquisition of operating data of each node in the target area power grid within a period of time, and repair of unreasonable operating data; (2) Set the sampling times and obtain the power data at each node of each sampling; (3) Based on the power data at each node, the branch and bound optimization algorithm is used to calculate the superior-subordinate relationship of each sampled node; (4) Using the node current, verify each sampled topology solution. Among the sampled topology solutions, the topology solution with the smallest error is selected as the optimal topology solution; The online acquisition of the operating data of each node in the target area power grid within a time period and repairing the unreasonable part of the operating data specifically includes: (1-1) First, the quality of the operation data of each node of the power grid is analyzed to screen out unqualified data; the operation data of each node of the power grid includes the voltage, current, active power and reactive power of each node of the power grid; (1-2) Use a method that combines business rules between data and data interpolation to repair unqualified data; The unqualified data includes three types: empty, containing illegal characters, and abnormal value exceeding the limit; The abnormal value exceeding the limit is obtained according to the verification deviation set in the power design standard, and the upper and lower limits of the operation data of each branch are given according to the transformer capacity and the line current carrying capacity; The data service rule is the relationship between active power, reactive power, voltage and current, which is specifically expressed as formula (1): In the formula: S is the transformer capacity, U is the voltage, I is the line current carrying capacity, P is the active power, and Q is the reactive power; The data interpolation method specifically includes: deriving substitute data for unqualified data from normal data, and when the business rules contain unqualified data, using the interpolation method to replace the unqualified data with the derived substitute data to repair the unqualified data; The branch and bound optimization algorithm is used to calculate the node superior-subordinate relationship of each sample according to the power data at each node, specifically including: (3-1) According to Kirchhoff's current law, the sum of the current flowing into the node's lower level is equal to the current collected by the node to the upper level. Suppose the number of branches in the power grid is n; (3-2) According to the known sampling periods, the power of each branch of the power grid is [w1, w2, w3, w4, ..., w n ], denoted as W; (3-3) The subordinate relationship between the branches of the power grid is: It is denoted as X; where x ij Indicates the subordinate relationship between branch i and branch j; if the superior branch of branch i is j, then x ij = -1, if the parent branch of branch i is not j, then x ij =0; when i=j, x ij =1; (3-4) Establish the objective function, which is to minimize the power imbalance at each node of the power grid, that is: min Z = WX; The constraints are: The node current is used to verify each sampled topology solution. Among the sampled topology solutions, the topology solution with the smallest error is taken as the optimal topology solution, which specifically includes the following: (4-1) During the analysis period of analyzing the quality of the operation data of each node of the power grid, the current data of each branch at a certain moment [I1, I2, I3...I n ], denoted as I t ; (4-2) Based on m sampling cycles, the branch topology solution is [X1, X2, X3...X k ] Where k≤m, the accuracy of the branch topology scheme is verified using the current data within the analysis period; The use of current data within the analysis period to verify the accuracy of the branch topology scheme specifically includes: Calculate the error of each topology solution during the analysis period, as follows: Among them, t1 is the starting time of the analysis period, t2 is the ending time of the analysis period, and the error set corresponding to the k alternative solutions is [δ1, δ2, δ3…δ k ]; When the error is the smallest, that is, when δ=min[δ1,δ2,δ3...δ k ] is taken as the final point network topological relationship.

2. A data-driven automatic topology data management method for power grid according to claim 1, characterized in that The setting of the sampling times and obtaining the power data at each node of each sampling specifically includes: Set the sampling times to m, and calculate the power data of each node in the power grid in each sampling period according to the operating data within the sampling time; if the sampling includes frozen power data, the power in a certain sampling period is W: W = W2-W1; Among them, W2 is the frozen power at the end of sampling, and W1 is the frozen power at the start of sampling; If the sampling does not contain frozen power data, the power in the sampling period is calculated by power integration: Among them, t1 is the starting time of a sampling period, t2 is the ending time of a sampling period, and P is the active power at the node.

Images