Low-voltage distribution network topology identification method based on constrained branch search

Abstract

The invention discloses a low-voltage power distribution network topology identification method based on constrained branch search. The method comprises the following steps: 1) constructing a low-voltage power distribution network node connection constraint; 2) identifying the topology of the low-voltage power distribution system through a limited branch search method; and 3) extracting node characteristics of the low-voltage power distribution system and writing the node characteristics into low-voltage power distribution system topology. According to the method, mutual features of voltage-load, voltage and access photovoltaic and self-features of node voltage are extracted by using a feature mining mathematical model, and accurate identification of the access position of the photovoltaic power station is realized in combination with a decision tree or a random forest classifier.

Description

Technical Field

[0001] The present invention relates to the field of distribution network topologies, and particularly to a method for identifying low-voltage distribution network topologies based on constrained branch search. Background Art

[0002] In modern power systems, the topology of the low-voltage distribution network (LVDN) is crucial for the stability and reliability of the power grid. Traditional impedance-based topology identification methods assume highly accurate impedance data, but this assumption is often difficult to achieve due to sensor quality, equipment aging, and complex load conditions. In addition, with the widespread access of distributed energy resources (DERs), especially photovoltaic (PV) systems, the unidirectional power flow characteristic of the distribution network has gradually changed, resulting in increased voltage fluctuations and frequent changes in the topology.

[0003] Currently, most mainstream topology identification technologies rely on centralized data processing, which requires large-scale data storage and computing capabilities, but faces the following defects and challenges:

[0004] 1. Low data quality: Data loss, bad data, and measurement noise may occur due to reasons such as sensor quality and communication network failures. These problems directly affect the accuracy of topology identification. Although existing research has proposed methods for filling in missing data and filtering noise, such as interpolation methods based on statistical analysis and machine learning and noise detection methods based on residual analysis, these methods are often computationally expensive when dealing with complex and large-scale networks and are difficult to process real-time data.

[0005] 2. High computational complexity: Traditional topology identification methods (such as mixed integer optimization methods) ensure global optimality through convex optimization models. However, since the problem of distribution network topology identification belongs to the NP (Non-deterministic Polynomial) hard problem, the computational cost increases rapidly with the increase in network scale. Although methods for decomposing the network to reduce the computational burden have been developed, these methods usually rely on accurate prior knowledge of the breaker states, and this information is not always available.

[0006] 3. Insufficient real-time performance: Centralized methods need to upload a large amount of data to the central data center for processing. This method not only increases the communication burden but also may cause high latency. Especially when facing the dynamic access of distributed PV systems, this latency may seriously affect the real-time response ability of the system.

[0007] 4. Poor dynamic adaptability: The intermittency and "plug and play" characteristics of photovoltaic systems require continuous adjustment of the topology of the distribution network. However, existing methods usually assume a static topology or a low change frequency, lacking the ability to adapt to dynamic topology changes.

[0008] The root causes of these problems lie in the dependence of distribution network topology identification technology on high-quality data and efficient calculation methods. When the data quality is low or the network scale is large, the calculation efficiency and identification accuracy of traditional methods decrease significantly. In addition, with the development of smart grid technology, the popularization of distributed intelligent terminals provides the possibility of stream computing for topology identification, but existing technologies have not fully utilized this architecture to solve the limitations of centralized computing. To solve the above problems, a more efficient, robust, and dynamically adaptable topology identification method is needed. Summary of the Invention

[0009] The object of the present invention is to provide a low-voltage distribution network topology identification method based on constrained branch search, including the following steps:

[0010] 1) Construct node connection constraints for the low-voltage distribution network;

[0011] 2) Identify the topology of the low-voltage distribution system through the restricted branch search method;

[0012] 3) Extract node features of the low-voltage distribution system and write them into the low-voltage distribution system topology.

[0013] Furthermore, the node connection constraints of the low-voltage distribution network include a voltage mutual information evaluation equation, a capacitor installation constraint, a one-to-one type topology constraint, and a one-to-many type topology constraint.

[0014] Furthermore, the voltage mutual information evaluation equation is as follows:

[0015]

[0016] Wherein, is the joint probability density function, and are the marginal probability density functions; is the voltage of node i; is the voltage of node j; m is the time period; MI(V i ,V j ) is the mutual information between voltage V i and V j ;

[0017] Furthermore, the capacitor installation constraint is as follows:

[0018] V up,i ≥V down,j ,i,j∈N (2)

[0019]

[0020] deg - (c i ) = 0, c i ∈N c (4)

[0021] V up,i ≤V down,j , i, j ∈ N uci (5)

[0022] Wherein, N represents the set of all nodes; deg - (*) represents the in-degree; N uci represents the set of affected upstream nodes; c i represents the node where the capacitor is installed; V up,i , V down,j respectively represent the voltages of the upstream node i and the downstream node j; I up,i , I down,j respectively represent the currents of the upstream node i and the downstream node j.

[0023] Furthermore, the one-to-one type topological constraint is as follows:

[0024]

[0025]

[0026] A = [R 12 X 12 R 23 X 23 … R (n-1)n X (n-1)n (9)

[0027]

[0028] Wherein, RSS is the total residual; RSS min1 is the preset residual threshold; P j , Q j are the active power and reactive power; Y is the voltage difference matrix, A is the circuit parameter matrix, X is the Jacobian matrix, R ij and X ij are the resistance and reactance of the line between nodes i and j respectively. is the fitted value; Y is the true value.

[0029] Furthermore, the one-to-many type topological constraint is as follows:

[0030]

[0031] A = [R i1 X i1 R i2 X i2 … R in X in (14)

[0032]

[0033] where RSS is the total sum of residuals; RSS min2 is the preset residual threshold;

[0034] Further, in step 2), the steps of identifying the topology of the low-voltage distribution system by the restricted branch search method include:

[0035] 2.1) Identifying the one-to-one type branch connection topology in the low-voltage distribution system, the steps include:

[0036] 2.1.1) Traverse each node in the node set N. If there is no capacitor and on-load tap-changer transformer at this node, then select formulas (1), (2), (3), (6)-(10) as the constraint conditions; if there are capacitor and on-load tap-changer transformer at this node, then select formulas (1), (4)-(10) as the constraint conditions;

[0037] 2.1.2) Add the lines that meet the constraint conditions described in step 2.1.1) to the set L;

[0038] 2.2) Identifying the one-to-many type branch connection topology in the low-voltage distribution system, the steps include:

[0039] 2.2.1) In the set L, identify and merge the rows of shared nodes;

[0040] 2.2.2) Sort the rows in the set L in descending order according to the voltage value of the first node in each row;

[0041] 2.2.3) Put the h-th row of the set L into the set T, traverse the remaining rows of the set L, and use the node connection constraints of the low-voltage distribution network to determine the most likely connection node of the row in T, and update T; the initial value of h is 1;

[0042] 2.2.4) Let h = h + 1, and return to step 2.2.3), traverse all rows of the set L, and obtain the set T containing the topology of the low-voltage distribution system.

[0043] Further, the node characteristics of the low-voltage distribution system include the cross-characteristics of voltage-load, voltage, access to photovoltaic, and the self-characteristics of node voltage.

[0044] Further, the self-characteristics of the node voltage are as follows:

[0045] V p-p = max(V i ) - min(V i ) (16)

[0046]

[0047] Wherein, MAD is the average absolute deviation of voltage; V p-p is the node voltage deviation; is the voltage mean value;

[0048] The cross - characteristics of voltage - load, voltage, and grid - connected PV are as follows:

[0049] P out - P in = P PV - P load (18)

[0050]

[0051] U 2 = C V-PV P PV + C V-P P load (21)

[0052] Wherein, U 2 is the square of the node voltage; Z equ = Z in Z out / (Z in + Z out ); γ is the angle of Z equ after parallel connection; P load is the load; P out is the outgoing power; P in is the input power; U i ∠θ is the node voltage; Z i ∠β is the impedance of the line connected to node i; C V-PV , C V-P respectively represent the cross - characteristics of voltage - plug PV power generation and voltage - node power, P PV is the PV power generation power, Z equ is the equivalent impedance value. S is the voltage cross - characteristic.

[0053] Furthermore, the steps of extracting the node characteristics of the low - voltage distribution system include:

[0054] 3.1) Determine the access points of PV grid connection in the low - voltage distribution system;

[0055] 3.2) Based on the topology of the low-voltage distribution system, use simulated data to train a classifier for the location selection of photovoltaic power generation;

[0056] Input the daytime data of the low-voltage distribution system into the classifier for the location selection of photovoltaic power generation, so as to determine the access location of the photovoltaic power station;

[0057] 3.3) Combine plug-in photovoltaic access to extract the node characteristics of the low-voltage distribution system.

[0058] The technical effects of the present invention are beyond doubt, and the beneficial effects of the present invention are as follows:

[0059] 1. Improve the accuracy of topology recognition and reduce computational complexity: By proposing four node connection constraint rules, the present invention can more accurately evaluate the voltage similarity and connection relationship between nodes based on voltage mutual information, capacitor installation effect, one-to-one and one-to-many types of topology constraint equations. This not only effectively corrects the errors in topology recognition, but also improves the overall recognition accuracy.

[0060] The restricted branch search method developed by the present invention significantly reduces the computational complexity of topology recognition by reducing unnecessary calculation steps and simplifying the model structure. This method focuses on overall topology recovery, forms connection constraints by using the node voltage to calculate the mutual information matrix, and traverses the nodes to judge the correctness of the lines, thereby further improving the recognition efficiency and accuracy.

[0061] 2. Enhance dynamic adaptability: The present invention fully considers the wide access and intermittent characteristics of distributed energy (especially photovoltaic systems), and realizes effective adaptation to the frequent changes in the topology structure of the distribution network by dynamically adjusting the topology structure recognition algorithm. The algorithm can process the radial LVDN with plug-in photovoltaic integration, accurately identify the overall topology structure, and locate the access location of the photovoltaic power station.

[0062] In particular, the present invention uses a feature mining mathematical model to extract the mutual features of voltage-load, voltage, access photovoltaic, and the self-features of node voltage, and combines a decision tree or a random forest classifier to realize the accurate identification of the access location of the photovoltaic power station. This dynamic adaptability provides strong support for the stable operation of the power system. Description of the Drawings

[0063] Figure 1 It is the influence of the capacitor installation position on the voltage amplitude;

[0064] Figure 2 They are two connections that make up the topology in the present invention Figure 2 (a) is a one-to-one connection structure Figure 2 (b) is a one-to-many connection structure;

[0065] Figure 3It is a schematic diagram of the power flow of node i;

[0066] Figure 4 It is the topology recognition process of the present invention. Specific embodiments

[0067] The present invention will be further described below in conjunction with embodiments, but it should not be understood that the above-mentioned subject matter scope of the present invention is limited to the following embodiments. Without departing from the above-mentioned technical idea of the present invention, various substitutions and changes made according to ordinary technical knowledge and customary means in the art should be included within the protection scope of the present invention.

[0068] Embodiment 1:

[0069] See Figures 1 to 4 , a low-voltage distribution network topology recognition method based on constrained branch search, including the following steps:

[0070] 1) Construct the node connection constraints of the low-voltage distribution network;

[0071] 2) Identify the topology of the low-voltage distribution system by the restricted branch search method;

[0072] 3) Extract the node characteristics of the low-voltage distribution system and write them into the topology of the low-voltage distribution system.

[0073] The node connection constraints of the low-voltage distribution network include a voltage mutual information evaluation equation, a capacitor installation constraint, a one-to-one type topology constraint, and a one-to-many type topology constraint.

[0074] The voltage mutual information evaluation equation is as follows:

[0075]

[0076] Among them, is the joint probability density function, and are the marginal probability density functions; is the voltage of node i; is the voltage of node j; m is the time period; MI(V i ,V j ) is the mutual information between voltages V i and V j ;

[0077] The capacitor installation constraint is as follows:

[0078] V up,i ≥V down,j ,i,j∈N (2)

[0079]

[0080] deg- (c i ) = 0, c i ∈N c (4)

[0081] V up,i ≤V down,j , i, j ∈ N uci (5)

[0082] In the formula, N represents the set of all nodes; deg - (*) represents the in-degree; N uci represents the set of affected upstream nodes; c i represents the node where the capacitor is installed; V up,i 、V down,j represent the voltages of upstream node i and downstream node j respectively; I up,i 、I down,j represent the currents of upstream node i and downstream node j respectively.

[0083] The one-to-one type topological constraint is as follows:

[0084]

[0085] A = [R 12 X 12 R 23 X 23 … R (n-1)n X (n-1)n (9)

[0086]

[0087] In the formula, RSS is the total residual; RSS min1 is the preset residual threshold; P j 、Q j are the active power and reactive power; Y is the voltage difference matrix, A is the circuit parameter matrix, X is the Jacobian matrix, R ij and X ij are the resistance and reactance of the line between nodes i and j respectively. is the fitting value; Y is the true value.

[0088] The one-to-many type topological constraint is as follows:

[0089]

[0090]

[0091] A = [R i1 X i1 R i2 X i2… R in X in (14)

[0092]

[0093] Wherein, RSS is the total residual sum; RSS min2 is the preset residual threshold;

[0094] In step 2), the steps of identifying the topology of the low-voltage distribution system by the restricted branch search method include:

[0095] 2.1) Identifying the one-to-one type branch connection topology in the low-voltage distribution system, the steps include:

[0096] 2.1.1) Traverse each node in the node set N. If there is no capacitor and on-load tap-changer transformer at this node, select formulas (1), (2), (3), (6)-(11) as the constraint conditions; if there is a capacitor and on-load tap-changer transformer at this node, select formulas (1), (4)-(11) as the constraint conditions;

[0097] 2.1.2) Add the lines that meet the constraint conditions described in step 2.1.1) to the set L;

[0098] 2.2) Identifying the one-to-many type branch connection topology in the low-voltage distribution system, the steps include:

[0099] 2.2.1) In the set L, identify and merge the rows of shared nodes;

[0100] 2.2.2) Sort the rows in the set L in descending order according to the voltage value of the first node in each row;

[0101] 2.2.3) Put the h-th row of the set L into the set T, traverse the remaining rows of the set L, and use the node connection constraints of the low-voltage distribution network to determine the most likely connection node of the row in T, and update T; the initial value of h is 1;

[0102] 2.2.4) Let h = h + 1, and return to step 2.2.3), traverse all rows of the set L, and obtain the set T containing the topology of the low-voltage distribution system.

[0103] The node characteristics of the low-voltage distribution system include the cross-characteristics of voltage-load, voltage, access to photovoltaic, and the self-characteristics of node voltage.

[0104] The self-characteristics of node voltage are as follows:

[0105] V p-p = max(V i ) - min(V i ) (16)

[0106]

[0107] Wherein, MAD is the mean absolute deviation of voltage; V p-p is the node voltage deviation; is the voltage mean value;

[0108] The cross - characteristics of voltage - load, voltage, and grid - connected PV are as follows:

[0109] P out -P in =P PV -P load (18)

[0110]

[0111] U 2 =C V-PV P PV +C V-P P load (21)

[0112] Wherein, U 2 is the square of the node voltage; Z equ =Z in Z out / (Z in +Z out ); γ is the angle of Z equ after parallel connection; P load is the load; P out is the out - flowing power; P in is the input power; U i ∠θ is the node voltage; Z i ∠β is the impedance of the line connected to node i; C V-PV , C V-P respectively represent the cross - characteristics of voltage - plug PV power generation and voltage - node power; P PV is the PV power generation power, Z equ is the equivalent impedance value.

[0113] The steps of extracting the node characteristics of the low - voltage distribution system include:

[0114] 3.1) Determine the access points of PV grid connection in the low - voltage distribution system;

[0115] 3.2) Based on the topology of the low - voltage distribution system, use simulation data to train a classifier for PV power generation site selection;

[0116] Input the day - time data of the low - voltage distribution system into the classifier for PV power generation site selection, so as to determine the access location of the PV power station;

[0117] 3.3) Combine the plug-in PV access to extract the node characteristics of the low-voltage distribution system.

[0118] Embodiment 2:

[0119] A method for identifying the topology of a low-voltage distribution network based on constrained branch search, comprising the following steps:

[0120] 1) Construct the node connection constraints of the low-voltage distribution network;

[0121] 2) Identify the topology of the low-voltage distribution system through the restricted branch search method;

[0122] 3) Extract the node characteristics of the low-voltage distribution system and write them into the topology of the low-voltage distribution system.

[0123] Embodiment 3:

[0124] A method for identifying the topology of a low-voltage distribution network based on constrained branch search, the technical content of which is the same as that of Embodiment 2. Further, the node connection constraints of the low-voltage distribution network include a voltage mutual information evaluation equation, a capacitor installation constraint, a one-to-one type topology constraint, and a one-to-many type topology constraint.

[0125] Embodiment 4:

[0126] A method for identifying the topology of a low-voltage distribution network based on constrained branch search, the technical content of which is the same as any one of Embodiments 2-3. Further, the voltage mutual information evaluation equation is as follows:

[0127]

[0128] Wherein, is the joint probability density function, and are the marginal probability density functions; is the voltage of node i; is the voltage of node j; m is the time period; MI(V i ,V j ) is the mutual information between voltage V i and V j ;

[0129] Embodiment 5:

[0130] A method for identifying the topology of a low-voltage distribution network based on constrained branch search, the technical content of which is the same as any one of Embodiments 2-4. Further, the capacitor installation constraint is as follows:

[0131] V up,i ≥V down,j ,i,j∈N (2)

[0132]

[0133] deg - (c i )=0, c i ∈N c (4)

[0134] V up,i ≤V down,j , i, j ∈ N uci (5)

[0135] Wherein, N represents the set of all nodes; deg - (*) represents the in-degree; N uci represents the set of affected upstream nodes; c i represents the node where the capacitor is installed; V up,i 、V down,j respectively represent the voltages of upstream node i and downstream node j; I up,i 、I down,j respectively represent respectively.

[0136] Example 6:

[0137] A method for identifying the topology of a low-voltage distribution network based on constrained branch search, the technical content is the same as any one of Examples 2-5. Further, the one-to-one type topology constraints are as follows:

[0138]

[0139] A = [R 12 X 12 R 23 X 23 … R (n-1)n X (n-1)n (9)

[0140]

[0141] Wherein, RSS is the total residual; RSS min1 is the preset residual threshold; P j 、Q j are the active power and reactive power; Y is the voltage difference matrix, A is the circuit parameter matrix, X is the Jacobian matrix, R ij and X ij are the resistance and reactance of the line between nodes i and j respectively.. is the fitting value; Y is the true value.

[0142] Example 7:

[0143] A method for identifying the topology of a low-voltage distribution network based on constrained branch search, the technical content is the same as any one of Examples 2-6. Further, the one-to-many type topology constraints are as follows:

[0144]

[0145] A = [R i1 X i1 R i2 X i2 … R in X in (14)

[0146]

[0147] Wherein, RSS is the total residual; RSS min2 is a preset residual threshold;

[0148] Example 8:

[0149] A method for identifying the topology of a low-voltage distribution network based on constrained branch search, the technical content is the same as any one of Examples 2-7. Further, in step 2), the steps of identifying the topology of the low-voltage distribution system by the restricted branch search method include:

[0150] 2.1) Identifying the one-to-one type branch connection topology in the low-voltage distribution system, the steps include:

[0151] 2.1.1) Traverse each node in the node set N. If there is no capacitor and on-load tap-changer at this node, select formulas (1), (2), (3), (6)-(10) as the constraint conditions; if there is a capacitor and on-load tap-changer at this node, select formulas (1), (4)-(10) as the constraint conditions;

[0152] 2.1.2) Add the lines that meet the constraint conditions described in step 2.1.1) to the set L;

[0153] 2.2) Identifying the one-to-many type branch connection topology in the low-voltage distribution system, the steps include:

[0154] 2.2.1) In the set L, identify and merge the rows of shared nodes;

[0155] 2.2.2) Sort the rows in the set L in descending order according to the voltage value of the first node in each row;

[0156] 2.2.3) Put the h-th row of the set L into the set T, traverse the remaining rows of the set L, and use the node connection constraints of the low-voltage distribution network to determine the most likely connection node of the row in T, and update T; the initial value of h is 1;

[0157] 2.2.4) Let h = h + 1, and return to step 2.2.3), traverse all rows of the set L, and obtain the set T containing the topology of the low-voltage distribution system.

[0158] Example 9:

[0159] A low-voltage distribution network topology identification method based on constrained branch search, the technical content is the same as any one of Examples 2-8. Further, the node characteristics of the low-voltage distribution system include the mutual characteristics of voltage-load, voltage, and photovoltaic access, as well as the self-characteristics of the node voltage.

[0160] Example 10:

[0161] A low-voltage distribution network topology identification method based on constrained branch search, the technical content is the same as any one of Examples 2-9. Further, the self-characteristics of the node voltage are as follows:

[0162] V p-p = max(V i ) - min(V i ) (16)

[0163]

[0164] In the formula, MAD is the mean absolute deviation of voltage; V p-p is the node voltage deviation;

[0165] The mutual characteristics of voltage-load, voltage, and photovoltaic access are as follows:

[0166] P out -P in = P PV -P load (18)

[0167]

[0168]

[0169] U 2 = C V-PV P PV + C V-P P load (21)

[0170] In the formula, U 2 is the square of the node voltage; Z equ = Z in Z out / (Z in + Z out ); γ is the angle of Z equ after parallel connection; P load is the load; P out is the out-flowing power; P in is the input power; U i ∠θ is the node voltage; Z i∠β is the impedance of the line connected to node i; C V-PV and C V-P respectively represent the mutual characteristics of voltage plug-in photovoltaic power generation and voltage node power. P PV is the photovoltaic power generation power, and Z equ is the equivalent impedance value. S is the voltage mutual characteristic.

[0171] Example 11:

[0172] A method for identifying the topology of a low-voltage distribution network based on constrained branch search, the technical content is the same as any one of Examples 2-10. Further, the steps of extracting the node characteristics of the low-voltage distribution system include:

[0173] 3.1) Determine the access points of photovoltaic grid connection in the low-voltage distribution system;

[0174] 3.2) Based on the topology of the low-voltage distribution system, use simulation data to train a classifier for photovoltaic power generation site selection;

[0175] Input the daytime data of the low-voltage distribution system into the classifier for photovoltaic power generation site selection, so as to determine the access location of the photovoltaic power station;

[0176] 3.3) Combine the plug-in photovoltaic access to extract the node characteristics of the low-voltage distribution system.

[0177] Example 12:

[0178] A method for identifying the topology of a low-voltage distribution network based on constrained branch search, the steps include:

[0179] S1. Propose 4 node connection constraint rules.

[0180] S2. Propose a constrained branch search method to identify the radial low-voltage distribution system.

[0181] S3. Develop a feature mining mathematical model for extracting the mutual characteristics of voltage-load, voltage, access photovoltaic, and the self-characteristics of node voltage.

[0182] Among them, S1 includes the following steps:

[0183] S101: Propose node connection constraint 1: Use non-constraint to evaluate the voltage mutual information to evaluate the voltage similarity;

[0184] S102: Propose node connection constraint 2: In the low-voltage grid with long feeders, add capacitors to maintain an appropriate voltage distribution;

[0185] S103: Propose node connection constraint 3: Establish a constraint equation for the one-to-one type of topology;

[0186] S104: Propose node connection constraint 4: Establish a constraint equation for a one-to-many type of topology;

[0187] S2 includes the following steps:

[0188] S201: Correct the one-to-one type of branch connection topology;

[0189] S202: Correct the one-to-many type of branch connection topology.

[0190] S3 includes the following steps:

[0191] S301: Determine the access point for photovoltaic grid connection;

[0192] S302: Process the actual daylight data to determine the access location of the photovoltaic power station;

[0193] S303: Combine the plug-in photovoltaic access to achieve topology recognition.

[0194] Specific implementation example:

[0195] In specific implementation, S1 includes the following steps:

[0196] S101: Use the Node Connection (NC) constraint to evaluate the voltage mutual information.

[0197] The voltages of nodes i and j are expressed as: and

[0198] The discrete random variables V i and V j The calculation formula for the mutual information (MI) between them is:

[0199]

[0200] Among them, is the joint probability density function, and are the marginal probability density functions.

[0201] The present invention uses kernel density approximation to integrate the probability density function:

[0202]

[0203] where the Gaussian kernel is h represents the bandwidth.

[0204] S102: The installation of capacitors will cause changes in the voltage amplitude, as Figure 1 shown.

[0205] Assume that the capacitor is installed at node c i (c i ∈N c ). The four cases are as follows:

[0206] Case 0: Natural characteristics of voltage and current. In this case, no capacitor is installed, and the downstream voltage and current can be expressed by equations (3) and (4), where N represents the set of all nodes,

[0207] V up,i ≥V down,j , i, j ∈ N

[0208]

[0209] Case 1: Slight distortion of the voltage waveform. This case does not change the characteristics of the voltage drop waveform.

[0210] Case 2: Partial distortion of the voltage waveform. There is a voltage rise from the first upstream node to c i .

[0211] Case 3: Severe distortion of the voltage waveform. Adding a capacitor bank at c i may cause voltage rises at multiple upstream nodes.

[0212] Temporarily disconnect node c i ∈ N from the topology using equation (5) because they may violate (3) and (4).

[0213] deg - (c i ) = 0, c i ∈ N c

[0214] where deg - (*) represents the in-degree, i.e., the number of lines pointing to a node. For Case 3, due to the voltage rises at multiple upstream nodes, we define these affected upstream nodes as N uci , which satisfies (6):

[0215] V up,i ≤V down,j , i, j ∈ N uci

[0216] When there is a on-load tap-changing transformer, the goal is to promote voltage rise. This case is similar to Case 2, so we stipulate that when there is an on-load tap-changing transformer, it conforms to (5). We define (3)-(6) as NC Constraint 2.

[0217] S103: The one-to-one type is equivalent to a straight line without branches, such as Figure 2(a), where n≥2. According to the linear voltage drop estimation formula, the relationship between the two nodes is obtained:

[0218]

[0219] Adding each term in Equation (7) gives Equation (8):

[0220]

[0221] Written in matrix form for m time periods, the matrices of Y, X, and A are given:

[0222]

[0223] A = [R 12 X 12 R 23 X 23 … R (n-1)n X (n-1)n

[0224]

[0225] The Residual Sum of Squares (RSS) is introduced to calculate the linearity of Y = AX. The smaller the RSS, the higher the goodness of fit. The calculation formula for RSS is where the true value is Y and the fitted value is In the following content, the one-to-one type of constraints (9), (10), and (11) are collectively referred to as NC constraint 3.

[0226] S104: The one-to-many type is as shown in Figure 2 (b). Taking the voltage of node 1 as the reference and subtracting the voltage of node i, we list the formulas for each branch:

[0227]

[0228] Eliminating V i The constraint equation for the one-to-many type is obtained:

[0229]

[0230] Expressed in matrix form as:

[0231]

[0232] A = [R i1 X i1 R i2 X i2 … R in X in ​​

[0233]

[0234] Still use RSS to evaluate the linearity, so as to verify the correctness of the connection. The one-to-many type constraints (14), (15), (16) are used as NC constraint 4.

[0235] In specific implementation, S2 includes the following steps:

[0236] S201: Based on previous theoretical derivations, develop an algorithm to implement the topological recognition of one-to-one type branch connections. This step is performed by calculating the MI matrix using the node voltages, forming NC constraint 1. Subsequently, traverse each node in N. When there are no capacitors and on-load tap-changers, this situation belongs to case 0. We use (3) and (4) in NC constraint 2. When there are capacitors and on-load tap-changers, this refers to Figure 1 cases 1 - 3 in. We use (5) in NC constraint 2 to process nodes c i and t i . Then, apply NC constraint 3 to calculate RSS to judge the correctness of the line, and add the correct lines to L. In addition, in case 3, we need to determine N uci . After traversing all nodes to find one-to-one type connections, these nodes will be automatically isolated and called N iso . Therefore, in order to determine whether each node merged into c i has a corresponding N uci , we use:

[0237]

[0238] where N s can be obtained through the mutual information (MI) of the node voltages. The purpose is to find connection points similar to c i .

[0239] Ns = {j∣MI(V(c i ),V(j)) > α·MI(V(c i ),V(c i ))}

[0240] where α is the selection coefficient, 0 < α < 1. For the nodes in N uci , use (6) in NC constraint 2. Finally, merge all the correct rows into L.

[0241] S202. Implement the topology recognition for one-to-many type branch connections. In L, we identify the rows of shared nodes, enabling us to merge the shorter rows into an extended row. Then, we sort the rows in L according to the voltage values of the first nodes in each row, in descending order. Put the first row in L into T. Then traverse the remaining rows in L, and use NC constraints 1 - 4 to determine the most likely connection nodes in T for each row and update T. Correspondingly, if the first node of a row is any one of N c , N t , N uci , we only use NC constraints 1, 3, and 4 for judgment. Specifically, for t i , when determining its upstream node, the known transformation ratio k needs to be used to adjust its voltage and current. After completing the traversal of L, the resulting overall topology is included in T. After completing the traversal of L, the resulting overall topology is included in T.

[0242] Specifically in implementation, S3 includes the following steps:

[0243] S301: For its own characteristics, the present invention only extracts V p-p and the Mean Absolute Deviation (MAD) value for voltage data. When these two values are large, it indicates large voltage fluctuations.

[0244] V p-p = max(V i ) - min(V i )

[0245]

[0246] A mathematical model that correlates voltage, total active power, and plug-in photovoltaic power generation is established to derive mutual characteristics, with the correlation coefficient as the solution. As Figure 3 shown, with node i as the reference, the node voltage of this node is known as U i ∠θ, the input power is P in , and the load is P load . The impedance of the line connected to node i is Z i ∠β, and the outgoing power is P out .

[0247] According to the node power conservation, their relationship is expressed as:

[0248] P out - P in = P PV - P load

[0249] Apply Ohm's law to derive the relationship between node voltage, power, and plug-in photovoltaic power generation:

[0250]

[0251] where Z equ = Z in Z out / (Z in + Z out ), and γ is the angle of Z equ after parallel connection. If only active power is considered, it is expressed as:

[0252]

[0253] Based on the simplified network model mentioned above, cos(2θ - γ) is almost constant, making the square of the node power and the node voltage can be regarded as linearly related. Using this linear relationship, a general formula for cross-feature description is derived, especially for distinguishing plug-in photovoltaic power generation:

[0254] U 2 = C V-PV P PV + C V-P P load

[0255] S302: Using the topology structure identified in the previous stage, train the classifier for photovoltaic power generation site selection with simulation data. These data include accurate load curves, photovoltaic power generation output, and the positions of photovoltaic power stations randomly assigned in the simulation software. Subsequently, select decision tree or random forest as the classifier to process the actual daylight data to determine the access positions of photovoltaic power stations.

[0256] S303: In the radial LVDN with plug-in photovoltaic integration, topology identification is divided into two stages: overall topology recovery and plug-in photovoltaic positioning. The overall recovery uses night data, and in the case where the topology structure has been identified, the positioning uses daylight data. As Figure 4 shown, this process includes three calculation steps in two stages. This algorithm is conducive to the deployment of the stream computing framework, which is an edge computing method that prioritizes real-time data processing and can alleviate the defects of centralized data processing and computing.

Claims

1. A low voltage distribution network topology identification method based on constrained branch search, characterized in that: The following steps are involved: 1) Construct low-voltage distribution network node connection constraints. 2) Identify the low-voltage distribution system topology through a restricted branch search method; 3) Extract the node characteristics of the low-voltage distribution system and write them into the low-voltage distribution system topology.

2. A low voltage distribution network topology identification method based on constrained branch search according to claim 1, characterized in that: The low-voltage distribution network node connection constraints include voltage mutual information evaluation equations, capacitor installation constraints, one-to-one type topology constraints, and one-to-many type topology constraints.

3. A low voltage distribution network topology identification method based on constrained branch search according to claim 2, characterized in that: The voltage mutual information evaluation equation is as follows: in, is the joint probability density function, and is the marginal probability density function; is the voltage at node i; is the voltage at node j; m is the time period; MI (V i ,V j ) is the voltage V i and V j The mutual information between them.

4. A low voltage distribution network topology identification method based on constrained branch search according to claim 2, characterized in that: The capacitor installation constraints are as follows: V up,i ≥V down,j ,i,j∈N (2) deg - (c i )=0,c i ∈N c (4) V up,i ≤V down,j ,i,j∈N uci (5) Where N represents the set of all nodes; deg - (*) indicates in-degree; N uci Indicates the set of affected upstream nodes; c i Indicates the node where the capacitor is installed; V up,i 、V down,j I represents the voltage of the upstream node i and the downstream node j respectively; up,i ,I down,j They represent the currents at the upstream node i and the downstream node j respectively.

5. A low voltage distribution network topology identification method based on constrained branch search according to claim 2, characterized in that: The one-to-one type topology constraints are as follows: A=[R 12 X 12 R 23 X 23 …R (n-1)n X (n-1)n ](9) Where, RSS is the residual sum; min1 is the preset residual threshold; P j , Q j is the active power and reactive power; Y is the voltage difference matrix, A is the circuit parameter matrix, X is the Jacobian matrix, R ij and X ij are the resistance and reactance of the line between nodes i and j respectively. is the fitted value; Y is the true value.

6. A low voltage distribution network topology identification method based on constrained branch search according to claim 2, characterized in that: The one-to-many type topology constraints are as follows: A=[R i1 X i1 R i2 X i2 …R in X in ](14) Where, RSS is the residual sum; min2 is the preset residual threshold.

7. A method for identifying low-voltage distribution network topology based on constrained branch search according to claim 2, characterized in that: In step 2), the step of identifying the low voltage power distribution system topology by a restricted branch search method includes: 2.1) Identify the one-to-one branch connection topology in the low voltage distribution system, the steps include: 2.1.1) Traverse each node in the node set N. If there is no capacitor and on-load tap-changing transformer at the node, select formula (1), formula (2), formula (3), formula (6)-(10) as constraint conditions; if there is a capacitor and on-load tap-changing transformer at the node, select formula (1), formula (4)-(10) as constraint conditions; 2.1.2) adding the lines satisfying the constraints in step 2.1.1) to the set L; 2.2) Identifying a one-to-many branch connection topology in a low voltage power distribution system, the steps include: 2.2.1) In the set L, identify and merge rows that share nodes; 2.2.2) Sort the rows in set L in descending order according to the voltage value of the first node in each row; 2.2.3) Put the hth row of set L into set T, traverse the remaining rows of set L, and use the low-voltage distribution network node connection constraints to determine the most likely connection node of the row in T, and update T; the initial value of h is 1; 2.2.4) Let h=h+1, and return to step 2.2.3), traverse all rows of the set L, and obtain the set T containing the low-voltage distribution system topology.

8. A method for identifying low-voltage distribution network topology based on constrained branch search according to claim 1, characterized in that: The node characteristics of the low-voltage distribution system include the mutual characteristics of voltage-load, voltage, connected photovoltaics, and the self-characteristics of node voltage.

9. A method for identifying low-voltage distribution network topology based on constrained branch search according to claim 8, characterized in that: The node voltage self-characteristics are as follows: V p-p =max(V i )-min(V i )(16) Where MAD is the mean absolute deviation of voltage; V p-p is the node voltage deviation; is the voltage mean; The mutual characteristics of voltage-load, voltage, and access to photovoltaic are as follows: P out -P in =P PV -P load (18) U 2 =C V-PV P PV +C V-P P load (21) Where U 2 is the square of the node voltage; Z equ =Z in Z out / (Z in +Z out );γ is the Z after parallel connection equ Angle; P load is the load; P out is the outflow power; P in is the input power; U i ∠θ is the node voltage; Z i ∠β is the impedance of the line connected to node i; C V-PV , C V-P They represent the mutual characteristics of voltage plug-in photovoltaic power generation and voltage node power, respectively, P PV is the photovoltaic power generation power, Z equ is the equivalent impedance value; S is the voltage mutual characteristic.

10. A low voltage distribution network topology identification method based on constrained branch search according to claim 1, characterized in that: The steps of extracting the node features of the low-voltage distribution system include: 3.1) Determine the access point for photovoltaic grid connection in the low-voltage distribution system; 3.2) Based on the low-voltage distribution system topology, a classifier for photovoltaic power generation site selection is trained using simulated data; Input the daytime data of the low-voltage distribution system into the classifier for photovoltaic power generation site selection to determine the access location of the photovoltaic power station; 3.3) Combined with plug-in photovoltaic access, the node characteristics of the low-voltage distribution system are extracted.

Images