Cross-domain network fault positioning method for information physical system of power distribution network

Abstract

The invention relates to the technical field of safe operation of a power system, in particular to a cross-domain network fault positioning method for a power distribution network information physical system, which comprises the following steps of: constructing an information layer actual fault current array and a physical layer expected fault current array by collecting real-time current, a topological structure and power supply information of a power distribution network node; a particle swarm is generated by using an information-physical collaborative particle swarm optimization algorithm, particles represent the fault state of a feeder line section, the speed and position of the particles are iteratively updated by calculating the fitness value of the particles until convergence, an optimal solution is output to position the fault feeder line section, and a positioning result is output to a power distribution network data acquisition and monitoring control system. According to the method, single-point and multi-point faults in the power distribution network can be rapidly and accurately positioned, the fault positioning precision and speed are improved, long-time power failure caused by inaccurate fault positioning is reduced, and the power supply reliability is improved.

Description

Technical Field

[0001] This invention relates to the field of power system safe operation technology. Background Technology

[0002] In the field of power system technology, the development of smart grids has become a critical path. Through highly intelligent technologies, smart grids improve the operation and control performance of the power grid, ensuring a safe, reliable, clean, and high-quality power supply. The core characteristic of smart grids lies in the bidirectional interaction between power flow and information flow, constructing a highly automated and widely distributed energy exchange network. With the in-depth implementation of the smart grid strategy, a large number of intelligent electronic devices (IEDs) have been deployed in power systems, causing modern power systems to gradually evolve into power cyber-physical systems (CPPS) with deep coupling between physical and information systems. In CPPS, intelligent electronic devices are responsible for collecting real-time information on power system operation and controlling power equipment, while communication networks are responsible for information transmission, and the master station system performs data analysis and decision-making, forming a three-level interactive system. However, this deep integration also brings new challenges. The mutual coupling between information and physical systems allows fault risks to be transmitted across space. Failure of the information system may seriously affect the reliability of the power system, and even trigger cascading failures, causing huge economic losses.

[0003] Traditional power system fault location methods primarily target the physical system, failing to adequately consider the impact of information systems. This makes them unsuitable for the needs of power cyber-physical systems, especially in extreme scenarios such as cyberattacks. The vulnerability of information systems can lead to inaccurate fault location and low recovery efficiency, seriously threatening the safe operation of the power system. Therefore, researching fault location methods applicable to power cyber-physical systems has become a crucial issue in advancing the intelligentization of power grids. Summary of the Invention

[0004] To overcome the problem that existing power system fault location methods mainly target physical systems and do not fully consider the impact of information systems, this invention provides a cross-domain network fault location method for distribution network cyber-physical systems.

[0005] The technical solution adopted by the present invention to achieve the above objectives is: a method for locating cross-domain network faults in a distribution network cyber-physical system, comprising the following steps:

[0006] S1. Collect real-time current data of each node in the distribution network, and collect the topology data and power supply information of the distribution network; S2. Based on the real-time current data, obtain current fault information and construct the actual fault current array of the information layer.

[0007] Based on topology data and power information, construct a physical layer desired fault current array;

[0008] S3. Based on the information-physical collaborative particle swarm optimization algorithm, the actual fault current array of the information layer and the expected fault current array of the physical layer are compared to generate a particle swarm. Each particle represents the fault state of the feeder segment. The fitness value of each particle is calculated, and the velocity and position of the particles are iteratively updated until the convergence condition is met. The optimal solution is then output, which is the feeder segment where the distribution network fault occurred.

[0009] Preferably, in step S1, feeder terminal units are deployed at each key node of the distribution network to collect real-time current data of each node; the topology information of the distribution network is collected, including line impedance, line length, node connection relationship and distributed power source access location; and the power information of the system is collected, including the voltage and frequency of the system power source and the distributed power source.

[0010] Preferably, in step S2, the Fourier algorithm is used to process the acquired real-time current data:

[0011] in, , Let represent the real and imaginary parts of the fundamental frequency of the node current. This represents the number of sampling points for the fundamental signal within one period. For the number of samples, For the first The sampled values ​​of the next time. This is the initial sample value. For the first Sub-sample value;

[0012] Calculate the current amplitude :

[0013] ;

[0014] Calculate the current phase angle :

[0015] ;

[0016] The continuous domain current data is converted into discrete domain data, with "0" representing no fault current, "1" representing fault current in the same direction as the positive current, and "-1" representing fault current in the opposite direction to the positive current, to construct the actual fault current array of the information layer.

[0017] Preferably, in step S2, when fault current information is lost, it is handled according to the following principles: when the switch that lost the information is located on the power supply side of the system, the information at the lost point is represented as "1"; when the switch that lost the information is located on the distributed power supply side, the information at the lost point is represented as "-1"; when the switch that lost the information is located in the middle of the distribution network, and the fault current information of the switches at two adjacent positions upstream and downstream of the switch is consistent, the information value of the adjacent positions is used to represent the missing position information value; when the switch that lost the information is located in the middle of the distribution network, and the fault current information of the switches at two adjacent positions upstream and downstream of the switch is inconsistent, the information at the lost point is represented as "0".

[0018] Preferably, in step S2, a switching function is defined. When the distribution network is radial, the switching function is:

[0019] ;

[0020] in, For the lower half of the switch The status of each feeder section; OR operation; For switch Desired switching function;

[0021] When distributed power sources are connected, the switching function is:

[0022] ;

[0023] in, , This indicates whether the power supply is connected to the upper and lower halves of the zone, respectively. A value of 1 indicates that the power supply is connected, and a value of 0 indicates that the power supply is not connected. , These represent the OR operations for all feeders in the upper and lower halves of the region, respectively. Indicates switch Calculations for the upper half of the power supply; Indicates switch Calculations for the lower half of the power supply;

[0024] The expected fault current of each node is calculated by switching functions to form a physical layer expected fault current array.

[0025] Preferably, in step S3, particles are randomly generated and their velocities are calculated. and location :

[0026] ;

[0027] ;

[0028] in, Number the particles; It is a particle dimension; and These represent the position and velocity of the previous generation particles, respectively. and These represent the updated particle position and velocity after each iteration; The inertial weight represents the degree to which the velocity of the previous generation of particles affects the velocity of the next generation of particles. , The learning factor of the particle represents its own learning ability, and all are positive real numbers. and These are random real numbers between 0 and 1 that appear in each iteration of the calculation; This is the optimal solution found by a particle after completing the iteration. This represents the optimal solution for all particles in each round of the iteration process.

[0029] Update particle velocity:

[0030] ;

[0031] in, Inertial weights;

[0032] ;

[0033] in, , These are the maximum and minimum values ​​of the inertia weight, respectively. To minimize fitness, This is the current fitness value. This represents the average fitness level.

[0034] Update particle positions:

[0035] ;

[0036] ;

[0037] in, For the mutation probability with a decreasing strategy, This represents the current iteration number. This represents the total number of iterations.

[0038] Calculate the particle's fitness:

[0039] ;

[0040] in, The number of nodes in the distribution network system. For actual fault conditions, when the first When an overcurrent signal is detected by a node The value is 1, otherwise it is not present. The value is 0; For switching functions, These are the weighting coefficients. This represents the status of each feeder segment; if there is a fault, use 1; otherwise, use 0.

[0041] Preferably, the range of particle motion speed is defined:

[0042] .

[0043] Preferably, the method further includes step S4, which outputs the feeder segment information of the distribution network fault accurately located by the algorithm to the data acquisition and monitoring control system of the distribution network.

[0044] The beneficial effects of this invention are as follows:

[0045] This invention proposes an information-physical collaborative fault location method by constructing a fault current information array and a physical system expected current array. This method can quickly and accurately locate single-point and multi-point faults in the distribution network, improve the accuracy and speed of fault location, reduce long-term power outages caused by inaccurate fault location, reduce the impact of faults on the power system, and improve power supply reliability. Attached Figure Description

[0046] Figure 1 This is a schematic diagram of the overall process of an embodiment of the present invention;

[0047] Figure 2 This is a schematic diagram of a simplified physical layer power distribution network according to an embodiment of the present invention;

[0048] Figure 3 This is a schematic diagram of a multi-source power distribution network according to an embodiment of the present invention;

[0049] Figure 4 This is a schematic diagram of a radial distribution network model according to an embodiment of the present invention;

[0050] Figure 5 This is a schematic diagram of a distributed power generation distribution network model according to an embodiment of the present invention;

[0051] Figure 6 This is a schematic diagram of the 21-node distribution network CPS system structure according to an embodiment of the present invention;

[0052] Figure 7 This is a schematic diagram of a 21-node distribution network CPS simulation model according to an embodiment of the present invention. Detailed Implementation

[0053] Embodiments of the present invention provide a method for locating cross-domain network faults in a distribution network cyber-physical system, such as... Figure 1As shown, it includes the following steps:

[0054] S1. Collect basic data from the information and physical layers of the power distribution network:

[0055] S1-1. Data Acquisition: Feeder Terminal Units (FTUs) are deployed at each key node of the distribution network to collect real-time current data from each node. The feeder terminal units will then upload the collected real-time current data to the Supervisory Control and Data Acquisition (SCADA) system of the distribution network.

[0056] S1-2. Collect basic physical layer data: Collect the topology information of the distribution network, including parameters such as line impedance, line length, node connection relationship and distributed power source access location, and collect the power information of the system, including parameters such as voltage and frequency of the system power source and distributed power source.

[0057] S2. Based on real-time current data, obtain current fault information and construct the actual fault current array at the information layer; based on topology data and power supply information, construct the desired fault current array at the physical layer.

[0058] S2-1. Preprocess the data uploaded from the FTU to SCADA, and use Fourier's algorithm to calculate the real and imaginary parts of the fundamental frequency of the node current:

[0059] ;

[0060] ;

[0061] in, , Let represent the real and imaginary parts of the fundamental frequency of the node current. This represents the number of sampling points for the fundamental signal within one period. For the number of samples, For the first The sampled values ​​of the next time. This is the initial sample value. For the first Sub-sample value;

[0062] Calculate the current amplitude :

[0063] ;

[0064] Calculate the current phase angle :

[0065] ;

[0066] The preprocessed values ​​are shown in the table below:

[0067]

[0068] Table 1. Preprocessing conditions for FTU sampling data

[0069] In the table, This is the maximum rated current of the line. The FTU overcurrent setting factor is set between 1.2 and 1.4. In this embodiment, the direction of the overcurrent flowing from the system power supply to the load is defined as the positive direction to avoid the influence of distributed power supply access at different locations. "0" represents no fault current, "1" represents the fault current in the same direction as the positive current, and "-1" represents the fault current in the opposite direction to the positive current. In this way, the current signal collected by the FTU is converted into discrete data describing whether there is overcurrent and the direction of overcurrent.

[0070] S2-2, as shown Figure 2 The diagram shows a simple physical layer distribution network with only a single power source. G represents the system power source, S1 to S6 are switching nodes, and L1 to L6 are the corresponding feeder sections. When only a single power source is connected to the distribution network, the fault current flows from the system power source to the load. When a fault occurs in a section, the FTU can detect that a fault overcurrent is flowing through the section, and the detected state value is "1". Nodes without fault overcurrent flowing through them have a state value of "0". Therefore, the actual state value detected by the FTU is:

[0071] ;

[0072] When there are multiple power sources in the distribution network, the upstream and downstream relationships between feeders are affected by the access of distributed generation. Therefore, the distribution network fault model needs to be adjusted. This fault model includes three parameters: "0", "1", and "-1". When the fault current transmitted by the FTU is in the same direction as the positive direction, ... If the detected fault current direction is opposite to the positive direction, If no fault current is detected, then .like Figure 3 In the distributed power distribution network with multiple power sources shown, when a fault occurs at section L7, the fault current flowing through node S7 flows from the system power source G to the load, which is the same as the positive direction, so the status value monitored by the FTU at S7 is "1"; when a fault occurs at section L6, the fault current flowing through node S7 flows from DG1 to the load, which is the opposite of the positive direction, so the status value monitored by the FTU at S7 is "-1".

[0073] S2-3. The fault location method in this embodiment relies on information uploaded from the CPS system information layer for location. In the CPS system, attacks on the information side can lead to missing or tampered uploaded information. When a fault occurs in an actual active distribution network, the FTU device at the node may be unable to transmit fault information to the master station SCADA system due to disturbances in the information layer. In this case, the master station SCADA system has the following four processing principles:

[0074] ① When the switch that loses information is located on the system power supply side, regardless of which feeder segment the fault occurs in, the information for that point of loss will be reported as 1;

[0075] ② When the switch that loses information is located on the distributed power supply side, regardless of which feeder segment the fault occurs in, the information of the lost point will be reported as -1;

[0076] ③ When the switch that lost information is located in the middle of the distribution network, and the FTU devices at two adjacent upstream and downstream locations of the switch report the same information, the master station SCADA system will fill in the missing location information with the information from the adjacent locations.

[0077] ④ When the switch that lost information is located in the middle of the distribution network, and the information reported by the switch FTU devices at two adjacent upstream and downstream locations of the switch is inconsistent, the system will report the information of the lost point as 0 and handle it as a false alarm (data distortion).

[0078] By filling in the missing fault information of the CPS system using the above criteria, and judging the overall state of the system based on the proposed fault location algorithm, fault location is performed.

[0079] Thus, an actual fault current array for the information layer is constructed.

[0080] S2-4. Define the switch function

[0081] For radial distribution networks, the switching function is generated by performing an OR operation on the feeder section states:

[0082] ;

[0083] in, For the lower half of the switch The status of each feeder section; OR operation; For switch Desired switching function;

[0084] When distributed power sources are connected, the connection status of the upper and lower half-zone power sources (kS1, kS2) and feeder OR operation need to be considered. The switching function is defined as follows:

[0085] ;

[0086] in, , This indicates whether the power supply is connected to the upper and lower halves of the zone, respectively. A value of 1 indicates that the power supply is connected, and a value of 0 indicates that the power supply is not connected. , These represent the OR operations for all feeders in the upper and lower halves of the region, respectively. Indicates switch Calculations for the upper half of the power supply; Indicates switch Calculations for the lower half of the power supply;

[0087] Radial distribution networks calculate the expected fault current at each node based on the state of the faulted section using switching functions, forming a physical layer expected fault current array. For example... Figure 4 As shown, S1-S9 are switch nodes, and L1-L9 are feeder sections. Assuming that a fault has occurred in section L4, by substituting each section, the expected fault current array is obtained as [1 1 1 1 0 0 0 0 0]. After processing by the algorithm, the fault will be located in section L4, and the expected fault current array [0 0 0 1 0 0 0 00] will be output.

[0088] In distribution networks incorporating distributed generation, the switching function is recalculated based on the power supply connection status and feeder status, generating a more complex physical layer expected fault current array. For example... Figure 5 As shown, S1-S10 are switching nodes and L1-L10 are feeder sections. In a distribution network system where both the main power source and distributed power sources are connected, when a fault occurs in section L8, the expected fault current array is [1 1 0 0 0 1 0 1 -1 -1 0].

[0089] Thus, the desired fault current array of the physical layer is constructed.

[0090] S3. Based on the information-physical collaborative particle swarm optimization algorithm, the actual fault current array of the information layer and the expected fault current array of the physical layer are compared to generate a particle swarm. Each particle represents the fault state of the feeder segment. The fitness value of each particle is calculated, and the particle velocity and position are updated. The particle velocity and position are iteratively updated until the convergence condition is met, and the optimal solution, i.e. the feeder segment where the distribution network fault occurred, is output.

[0091] S3-1 is the particle encoding, where each particle is a binary string representing the fault status of each feeder segment, with 1 indicating a fault and 0 indicating normal operation. The particle velocity is then calculated. and location :

[0092] ;

[0093] ;

[0094] in, Number the particles; It is a particle dimension; and These represent the position and velocity of the previous generation particles, respectively. and These represent the updated particle position and velocity after each iteration; The inertial weight represents the degree to which the velocity of the previous generation of particles affects the velocity of the next generation of particles. , The learning factor of the particle represents its own learning ability, and all are positive real numbers. and These are random real numbers between 0 and 1 that appear in each iteration of the calculation; This is the optimal solution found by a particle after completing the iteration. This represents the optimal solution that appears in each round of the iteration process for all particles;

[0095] Limit the range of particle velocity to prevent it from becoming too large:

[0096] ;

[0097] The Sigmoid function is:

[0098] ;

[0099] The continuous and discrete versions of the particle swarm optimization (PSO) algorithm are distinguished by two distinct parts: the transfer function and the different position update processes. The transfer function, in the continuous search space, is designed to switch particle positions in a binary search space, selecting between "0" and "1". For the discrete PSO algorithm, the particle positions are encoded in binary, thus restricting the particle position information to "0" or "1". Furthermore, the Sigmoid function constrains the particle velocities to the interval [0, 1].

[0100] ;

[0101] in, This represents the random numbers generated in each iteration, uniformly distributed in the interval [0,1].

[0102] The particle's velocity is understood as the probability of position change; that is, each component of the particle's velocity represents the probability that the corresponding position component chooses "0" or "1". The value of the sigmoid function tends to 1 as the particle's velocity increases, and tends to 0 when the particle's velocity is very small. To prevent saturation of the sigmoid function, the particle's velocity is set within the range [𝑉𝑚𝑎𝑥, 𝑉𝑚𝑖𝑛]. Therefore, its corresponding sigmoid function is:

[0103] .

[0104] S3-2. In the particle swarm optimization algorithm, a fitness function is used to determine whether a particle is moving towards the optimal value during the particle optimization process. In this embodiment, the fitness function used to determine whether the fault location is reasonable and accurate is:

[0105] ;

[0106] in, This refers to the number of nodes in the distribution network system, i.e., the number of FTUs. The actual fault condition reported by the FTU, i.e., the input item, when the... When an overcurrent signal is detected by a node The value is 1, otherwise it is not present. The value is 0; The switching function constructed in the preceding steps is a node The expected value, that is, the reflection of the operating state of each feeder at the node at the current position of the particle. This is the weighting coefficient, and its value is taken as 0.5 in the calculation. The state of each feeder segment is set to 1 if there is a fault and 0 if there is no fault. The fitness function includes a part for the state of each feeder, which acts as a penalty function. This can effectively avoid errors and minimize the difference between the fault information uploaded by the FTU and the expected value derived from the switching function, thus avoiding misjudgment.

[0107] S3-3. The particle swarm is initialized using a uniform distribution. This represents the particle's ability to learn. This represents the learning ability of a particle society. , All of these are related to the particle's optimization ability. When the particle size is large, its ability to seek optimization decreases. When the value is large, particles accelerating towards the optimal particle are prone to getting trapped in local optima. A contraction factor is established to avoid this situation.

[0108] ;

[0109] The updated formula for particle velocity iteration is:

[0110] ;

[0111] The inertia weight is related to the convergence of the algorithm. When the particle size is larger, the particle speed increases and the step size increases, allowing the particle to search a larger spatial region and discover new areas; when... When the fitness level is low, the particle speed decreases and the step size becomes shorter, allowing the particle to perform a more refined search within a local region and find a better solution. Therefore, the particle's inertia weight needs to change with the fitness level, which requires setting a dynamic inertia weight, called the adaptive weight. The adaptive weight is:

[0112] ;

[0113] in, , These are the maximum and minimum values ​​of the inertia weight, respectively. To minimize fitness, This is the current fitness value. This represents the average fitness score.

[0114] When the fitness value of each particle is about to converge to a local optimum, the particle's inertia weight automatically increases; when the fitness values ​​of each particle are relatively dispersed, the particle's inertia weight automatically decreases. Furthermore, when the fitness value is better than the average, the corresponding inertia weight is relatively small, and the particle has better local convergence ability; when the fitness value is lower than the average, its inertia weight is relatively large, and the particle has better overall convergence ability. In summary, the adaptive weight setting is significantly better than the traditional linear weight setting.

[0115] To expand the search range in the early iterations of the algorithm and enhance its position search capability in later iterations, the following mutation method is proposed to update the particle position:

[0116] ;

[0117] ;

[0118] in, For the mutation probability with a decreasing strategy, This represents the current iteration number. The total number of iterations is used. After a particle updates its position, it undergoes mutation. The system checks if the number of iterations exceeds 5% of the total number of iterations. Particles within this 5% range have a certain probability of mutating; particles within this 5% range have a zero probability of mutation and can no longer mutate.

[0119] S3-4, The iteration process is as follows:

[0120] S3-41. Set various parameters for the particles, including inertial weights. Current and preset iteration counts , Learning factor , ;

[0121] S3-42. Randomly generate particles, update the particle velocity and position, and calculate the particle fitness.

[0122] S3-43. Based on the calculated fitness values ​​of the particles, update the optimal position of the individual particles and the optimal position of the population.

[0123] S3-44. Update the particle's velocity and position.

[0124] S3-45. Calculate the fitness value of each particle.

[0125] S3-46. Update the individual best position and individual best fitness value of each particle, and update the global best position and global best fitness value of each particle.

[0126] S3-47. Determine whether the maximum number of iterations or the convergence condition has been reached. If the condition is met, the optimal solution is obtained and the algorithm ends. If the condition is not met, return to step S3-43 and continue running until the maximum number of iterations or the convergence condition is met, and output the optimal solution, which is the feeder segment where the distribution network fault occurred.

[0127] S4. Output the feeder segment information of the distribution network fault located by the algorithm to the data acquisition and monitoring control system of the distribution network.

[0128] The following calculation example is provided in this embodiment:

[0129] This embodiment is based on the MATLAB platform and uses data from the 21-node CPS system of the Datang Substation in Fengxian County, Chengde, Hebei Province for a case study analysis. The system structure is as follows: Figure 6 As shown, L1-L21 are 21 feeder segments, S1-S21 are 21 nodes, K1-K3 are the access switches for each distributed power source, and nodes 11, 12, 15, 16, 17, and 21 are dynamically adjustable loads. A Simulink simulation model is built, as shown in the figure. Figure 7As shown, a three-phase fault module is added to the model to simulate phase-to-phase short-circuit faults; the green module is an adjustable load module built based on actual system operating data to simulate the constantly changing characteristics of the distribution network; the red module is an FTU model built according to the aforementioned method, simulating its acquisition of current information and uploading the current information to the SCADA system module in the upper left corner of the model; the fault location algorithm proposed in this embodiment is embedded in the SCADA system module, which receives the information uploaded by the FTU and outputs the fault location.

[0130] For single-point-of-failure scenarios in a multi-distributed-source power distribution network model, to simulate the connection of distributed sources to the distribution network at different numbers and locations, eight categories are used to classify single-point-of-failure localization simulations. K1, K2, and K3 are categorized as follows: a state of 1 indicates the connection of distributed sources to the system, and a state of 0 indicates the opposite. Simulation results are shown in the table below.

[0131]

[0132] In addition, fault location algorithms based on immune algorithm (IA), genetic algorithm (GA), quantum genetic algorithm (QGA), binary dragonfly algorithm (BDA), and slime mold algorithm (SM) were compared with the algorithm in this embodiment. Performance evaluation indicators of the location algorithms were statistically analyzed under different numbers and locations of distributed power sources connected to the distribution network. The results are shown in the table below:

[0133]

[0134] The algorithm proposed in this embodiment can accurately and quickly locate single-point faults even when distributed power sources are connected to the distribution network in different numbers and locations. Comparative analysis with other optimization algorithms shows that: compared to traditional intelligent algorithms such as IA and GA, the improved PSO algorithm with adaptive weights has higher population quality and better location accuracy; compared with the QGA algorithm, their optimization capabilities are similar, but the improved PSO algorithm has stronger local exploration capabilities and shorter iteration time, ensuring rapid location; furthermore, the study found that when the fault occurs in the S1 feeder segment, more advanced artificial intelligence algorithms such as BDA and SM algorithms may experience inaccurate location. Analysis reveals that because the fault point is close to the power source, the fault causes a significant change in the overall power flow of the distribution network, resulting in a large difference in the current characteristics of the distribution network before and after the fault. BDA and SM algorithms struggle to consider such extreme cases when generating candidate populations, leading to location deviations. The improved PSO algorithm, while having a similar iteration speed, offers better location accuracy and is more suitable for fault location algorithms.

[0135] In the case of multiple faults, the number of faulty lines was increased sequentially to 6. The performance of the fault location algorithm was evaluated, and the algorithm results are shown in the table below:

[0136]

[0137] The performance of multi-point fault location was evaluated, and the results are shown in the table below:

[0138]

[0139] The simulation results under multi-point faults are similar to those under single-point faults. The algorithm proposed in this embodiment can also quickly and accurately locate multi-point faults.

[0140] The simulation model was set up to simulate a scenario where the FTU fails to transmit fault information normally. The algorithm's ability to withstand disturbances was analyzed, and the simulation results are shown in the table below:

[0141]

[0142] As shown in the table, the algorithm still demonstrates good performance even under information disturbances, with a positioning accuracy rate remaining above 97%.

[0143] Three different scales of computational tests (IEEE-14, IEEE-33, and IEEE-118) were set up to verify the impact of parameter settings on the algorithm performance. The results are shown in the table below:

[0144]

[0145] As the system scale increases, the population dimension of the particle swarm optimization algorithm increases accordingly, and the computation time of the fault location algorithm increases. At this time, increasing the initial inertia weight can speed up the convergence speed, but excessively increasing it will cause the model to converge prematurely, resulting in location errors. Therefore, when using the fault location algorithm proposed in this embodiment, it is necessary to adjust the initial weight according to the system's own scale and various parameters to achieve the optimal combination of accuracy and speed of the location algorithm.

[0146] This invention has been described through embodiments. Those skilled in the art will understand that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of the invention. Furthermore, under the teachings of this invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the invention. Therefore, this invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of this invention.

Claims

1. A method for locating cross-domain network faults in a distribution network cyber-physical system, characterized in that, Includes the following steps: S1. Collect real-time current data of each node in the distribution network, and collect the topology data and power information of the distribution network. S2. Based on real-time current data, obtain current fault information and construct an information layer actual fault current array; Based on topology data and power information, construct a physical layer desired fault current array; S3. Based on the information-physical collaborative particle swarm optimization algorithm, the actual fault current array of the information layer and the expected fault current array of the physical layer are compared to generate a particle swarm. Each particle represents the fault state of the feeder segment. The fitness value of each particle is calculated, and the velocity and position of the particles are iteratively updated until the convergence condition is met. The optimal solution is then output, which is the feeder segment where the distribution network fault occurred.

2. The cross-domain network fault location method for a distribution network cyber-physical system according to claim 1, characterized in that, In step S1, feeder terminal units are deployed at each key node of the distribution network to collect real-time current data of each node; the topology information of the distribution network is collected, including line impedance, line length, node connection relationship and distributed power source access location; the power information of the system is collected, including the voltage and frequency of the system power source and distributed power source.

3. The cross-domain network fault location method for a distribution network cyber-physical system according to claim 1, characterized in that, In step S2, the Fourier algorithm is used to process the acquired real-time current data: ； ； in, , Let represent the real and imaginary parts of the fundamental frequency of the node current. This represents the number of sampling points for the fundamental signal within one period. For the number of samples, For the first The sampled values ​​of the next time. This is the initial sample value. For the first Sub-sample value; Calculate the current amplitude : ； Calculate the current phase angle : ； The continuous domain current data is converted into discrete domain data, with "0" representing no fault current, "1" representing fault current in the same direction as the positive current, and "-1" representing fault current in the opposite direction to the positive current, to construct the actual fault current array of the information layer.

4. The cross-domain network fault location method for a distribution network cyber-physical system according to claim 3, characterized in that, In step S2, when fault current information is lost, it is handled according to the following principle: when the switch that lost the information is located on the system power supply side, the information lost at that point is represented as "1"; When the switch that loses information is located on the distributed power supply side, the information at the point of loss is represented as "-1"; When the switch that lost information is located in the middle of the distribution network, and the fault current information of the switches at two adjacent positions upstream and downstream of the switch is consistent, the information value of the adjacent position is used to represent the missing position information value. When the switch that lost information is located in the middle of the distribution network, and the fault current information of the switches at two adjacent locations upstream and downstream of the switch is inconsistent, the information of the lost point is represented as "0".

5. The cross-domain network fault location method for a distribution network cyber-physical system according to claim 1, characterized in that, In step S2, the switching function is defined. When the distribution network is radial, the switching function is: ； in, For the lower half of the switch The status of each feeder section; OR operation; For switch Desired switching function; When distributed power sources are connected, the switching function is: ； in, , This indicates whether the power supply is connected to the upper and lower halves of the zone, respectively. If connected, the value is 1; otherwise, the value is 0. , These represent the OR operations for all feeders in the upper and lower halves of the region, respectively. Indicates switch Calculations for the upper half of the power supply; Indicates switch Calculations for the lower half of the power supply; The expected fault current of each node is calculated by switching functions to form a physical layer expected fault current array.

6. The cross-domain network fault location method for a distribution network cyber-physical system according to claim 1, characterized in that, In step S3, particles are randomly generated, and their velocities are calculated. and location : ； ； in, Number the particles; It is a particle dimension; and These represent the position and velocity of the previous generation particles, respectively. and These represent the updated particle position and velocity after each iteration; The inertial weight represents the degree to which the velocity of the previous generation of particles affects the velocity of the next generation of particles. , The learning factor of the particle represents its own learning ability, and all are positive real numbers. and These are random real numbers between 0 and 1 that appear in each iteration of the calculation; This is the optimal solution found by a particle after completing the iteration. This represents the optimal solution for all particles in each round of the iteration process. Update particle velocity: ； in, Inertial weights; ； in, , These are the maximum and minimum values ​​of the inertia weight, respectively. To minimize fitness, This is the current fitness value. This represents the average fitness level. Update particle positions: ； ； in, For the mutation probability with a decreasing strategy, This represents the current iteration number. This represents the total number of iterations. Calculate the particle's fitness: ； in, The number of nodes in the distribution network system. For actual fault conditions, when the first When an overcurrent signal is detected by a node The value is 1, otherwise it is not present. The value is 0; For switching functions, These are the weighting coefficients. This represents the status of each feeder segment; if there is a fault, use 1; otherwise, use 0.

7. The cross-domain network fault location method for a distribution network cyber-physical system according to claim 6, characterized in that, Limiting the range of particle motion speed: 。 8. The cross-domain network fault location method for a distribution network cyber-physical system according to claim 1, characterized in that, It also includes step S4, which outputs the feeder segment information of the distribution network fault that the algorithm accurately locates to the data acquisition and monitoring control system of the distribution network.

Images