Active power distribution network fault positioning method based on improved northern eagle search algorithm
By improving the Northern Goshawk search algorithm and adaptive Gauss-Caucy hybrid disturbance mutation strategy, combined with the multi-level partition positioning model, the problem of high computing resources and costs in the fault location of active distribution network is solved, and efficient and accurate faults are achieved position.
Patent Information
- Application Number
- CN202411061861.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-05
- Publication Date
- 2024-11-15
AI Technical Summary
In the fault location of active distribution networks, the existing technology has problems such as large demand for computing resources, high computing costs, insufficient convergence and stability, especially when the distribution network structure is complicated after the distributed power generation system is connected. It is difficult to effectively locate the fault.
The improved Northern Goshawk search algorithm is adopted, combined with the adaptive Gauss-Caucy hybrid disturbance mutation strategy and the cosine strategy, and the distribution network is divided into sub-regions through a multi-level partition positioning model, and refined layer by layer to finally determine the specific location of the fault. .
The algorithm's global search ability and local escape ability are improved, the positioning accuracy and convergence speed are enhanced, the population is avoided from falling into local optimal early, and the accuracy and efficiency of fault positioning are significantly improved.
Smart Images

Figure CN118962335A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of active distribution network fault location, and specifically relates to an active distribution network fault location method based on an improved northern goshawk search algorithm, using a northern goshawk optimization algorithm (INGO) and a northern goshawk algorithm (NGO), a gray wolf optimization algorithm (GWO) and a whale optimization algorithm (WOA) for positioning and conducting comparative experiments. Background Art
[0002] The distribution network is located at the end of the power system and is directly connected to the user. It is particularly important to ensure the operational stability of the distribution network. Under the influence of the dual carbon background, more and more distributed generation (DG) systems are connected to the distribution network. In this case, the direction of the system power flow is no longer unique. The traditional distribution network is transformed into a multi-directional complex active distribution network (ADN). The distribution network structure will be more complex, resulting in increased difficulty in fault location, which brings great challenges to the stable operation of the active distribution network. Therefore, it is of great research significance to study the fault location method suitable for ADN.
[0003] Due to the access of DG, when a fault occurs, the fault characteristics of ADN are very different from those of traditional distribution networks, which can be summarized as follows: (1) The location of the access point and the access capacity of the wind turbine will affect the direction of the system power flow and the amplitude of the fault current. (2) The output of each DG is uncertain, resulting in uncertainty in the transient process of the fault. (3) The low-voltage distribution network has many branches and the line parameters are unevenly distributed, which increases the complexity of fault analysis. In the context of the automation upgrade of distribution network equipment, the research on fault location methods based on the current information uploaded by the feeder terminal unit (FTU) has become a hot topic.
[0004] At present, a variety of distribution network fault location methods have been proposed, which can be divided into fault routing, fault distance measurement and fault segment location according to the location results: fault routing, fault distance measurement and fault segment location. The implementation of the research method is mainly based on matrix algorithm and intelligent algorithm. The matrix algorithm combines the topological structure of the distribution network with the current information uploaded by the FTU to generate a fault discrimination matrix and locate the fault segment through matrix operation; the intelligent algorithm is based on the minimum fault diagnosis set theory, which converts the fault segment location problem into a mathematical optimization problem and solves it using the intelligent algorithm.
[0005] In existing literature, a variety of fault location methods based on intelligent algorithms have been proposed and verified. For example, the author of reference [1] used the multiverse algorithm to locate faults in the distribution network and improved the algorithm by introducing an adaptive elite strategy and adaptive mutation operations. Although this method has strong positioning capabilities, it requires large computing resources and has certain limitations when dealing with some special faults. Reference [2] proposed a location method based on the vulture search algorithm, which improved the optimization capability of the algorithm by introducing a crossover operator, a non-uniform mutation operator, and a foraging strategy. However, the high computational cost of this method limits its popularity in practical applications.
[0006] Reference [3] uses a combination of matrix algorithm and chaotic binary particle swarm algorithm to solve the distribution network fault location problem. By establishing the causal association matrix and criteria of the region and node according to the actual structure of the distribution network, the fault location can be performed more effectively. However, the convergence and stability of the algorithm still need to be further considered, because these factors determine whether the method can achieve good results in practical applications. The authors of reference [4] applied an improved algorithm based on quantum ant colony algorithm to the distribution network fault location problem. This method improves the accuracy and efficiency of fault location to a certain extent.
[0007] In addition, reference [5] verified that the introduction of the improved sine-cosine algorithm into the local development stage of the fault location algorithm increased the population diversity in the later iteration, effectively prevented the algorithm from falling into the local optimum, and improved the solution accuracy and convergence speed. The authors of reference [6] proposed an improved adaptive differential evolution algorithm, which combines Gauss-Cauchy mutation and parameter adaptation strategy, significantly improving the performance of the algorithm. However, the above four algorithms have a great influence on the results in terms of parameter selection and adjustment, and need to be fully optimized and tested for specific problems to ensure their effectiveness in different application scenarios.
[0008] References are as follows:
[0009] [1].Li,LL,Ji,BX,Lim,MK,&Tseng,ML(2024).Activedistrib utionnetworkoperationaloptimizationproblem:Amulti-objectivetunas warmoptimizationmodel.AppliedSoftComputing,150,111087.
[0010] [2].Guohua,Y.,J,FENG.,&Xuan,L.(2022)Zonalfaultlocalizationin distributionnetworkswithdistributedpowersourcesbasedonimprovedvu lturesearchalgorithm.PowerSystemProtectionandControl,50(18):1-9.
[0011] [3].Hu,J.(2024).Tracematrixoptimizationforfaultlocalization.JournalofSystemsandSoftware,208,111900.
[0012] [4].Bi,Z.,Yu,X.,Wang,B.,Zhang,W.,Liu,Z.,&Zhang,D.(2023).Faul tLocationofDistributionNetworkBasedontheQuantumAntColonyAlgorith m.
[0013] [5].Abualigah,L.,&Diabat,A.(2021).Advancesinsinecosinealgori thm:Acomprehensivesurvey.ArtificialIntelligenceReview,54(4),2567-2608.
[0014] [6]. Chen,
[0015] In view of this, the present invention is proposed. Summary of the invention
[0016] The technical problem to be solved by the present invention is to overcome the deficiencies of the prior art and provide an active distribution network fault location method based on an improved northern goshawk search algorithm, thereby solving the problems raised in the above-mentioned background technology.
[0017] In order to solve the above technical problems, the basic concept of the technical solution adopted by the present invention is:
[0018] An active distribution network fault location method based on an improved northern goshawk search algorithm comprises the following steps:
[0019] Step S1: Read the fault current status information of the section switch, tie switch and circuit breaker elements detected by the feeder terminal unit, upload it to the SCADA system of the master station, and generate the actual fault current array of the switch node according to the number of nodes based on the acquired data;
[0020] Step S2: Initialize the population number, population dimension, variable range value and maximum iteration number of the northern goshawk optimization algorithm;
[0021] Step S3: Initializing a binary northern goshawk population, wherein each individual in the northern goshawk population represents a fault operation state of a group of feeder segments;
[0022] Step S4: Calculate the fitness value and update the position of the individual northern goshawk, and update the individual position through the prey recognition and attack stage and the pursuit and escape stage;
[0023] Step S5: Introduce the adaptive Gaussian-Cauchy mixed perturbation mutation strategy and the sine-cosine strategy, update the individual position through the mixed perturbation of Gaussian mutation and Cauchy mutation, determine whether the maximum number of iterations is reached or the convergence condition is met, and if not, return to recalculate the fitness value and update the individual position;
[0024] Step S6: Determine the fault area based on the multi-level partition positioning model, divide the distribution network into multiple sub-areas through the multi-level partition strategy, refine it layer by layer, and finally determine the specific location of the fault. Generate an initialization individual and calculate its fitness value;
[0025] Step S7: After determining the fault area, check whether the identified fault area and the actual fault area match. If they match, the fault area is determined and the process ends. Otherwise, return to generate the initialization individual and recalculate the fitness value.
[0026] Optionally, when the step S3 of binary optimization of the northern goshawk algorithm is to update the position of the northern goshawk, its expression is:
[0028]
[0029] Among them, α = 2.3; P i,d Represents the updated position of individual i in the d-dimensional space after 1 iteration.
[0030] Optionally, step S4 needs to complete the following steps during the process of updating the individual position in the prey identification stage and the pursuit and escape stage:
[0031] The northern goshawk population is evenly distributed in the search space through sinusoidal mapping, and the global detection ability of the algorithm is enhanced through crossover and non-uniform mutation.
[0032] Among them, the crossover operation swaps the positions of the northern goshawks and recalculates the fitness value. When the fitness value of the new northern goshawk position is better than the original northern goshawk, the original northern goshawk individual is replaced, and the non-uniform mutation operator disturbs the position of the northern goshawk.
[0033] When the non-uniform mutation strategy disturbs the position of the northern goshawk, k dimensions are randomly selected for each northern goshawk individual for disturbance in turn. When the new northern goshawk individual generated by each disturbance is better than the original northern goshawk individual, the original northern goshawk individual is replaced. The expression is:
[0034]
[0035] Among them, P t represents the position of the northern goshawk at the tth iteration; T represents the maximum number of iterations, which is 100; b is the system parameter that determines the non-uniformity, which is 2.
[0036] Optionally, in step S4, the process summary of the prey identification stage randomly selects a prey and quickly attacks it, and simulates the hunting behavior of the northern goshawk through a mathematical model to update the individual position. The mathematical expression of the behavior of the northern goshawk in this stage is as follows:
[0037] P i =X k ,i=1,2,...,N,k=1,2,...,i-1,i+1,...,N
[0038]
[0039]
[0040] Among them, P i is the location of the prey selected by the i-th northern goshawk; F pi is the objective function value, that is, the fitness value, K is a random natural number in [1, N], is the new state of the i-th northern goshawk, is the corresponding fitness value, R is a random number belonging to [0,1], and the value of i is 1 or 2. r and i are random numbers used to generate random northern goshawk algorithm behaviors in search and update.
[0041] Optionally, in step S4, during the chasing and escaping phase, the northern goshawk chases the prey, simulates the scenario of the prey escaping, and updates the individual position through a mathematical model, which is expressed as:
[0042]
[0043]
[0044]
[0045] Where t is the current number of iterations, T is the maximum number of iterations, is the new state of the i-th northern goshawk in the pursuit phase, is the new state of the i-th northern goshawk in the j-th dimension during the pursuit phase, and F i new,p2 is the fitness value in the new state.
[0046] Optionally, in step S5, the oscillation change characteristics of the sine and cosine models are used to act on the position, and the overall and local optimization are performed according to the oscillation change of the sine and cosine models to obtain the overall optimal value. The nonlinear decreasing search factor has a large weight in the early stage and a slow decreasing speed, which is conducive to improving the global optimization ability. When the weight factor is small, the advantage of the algorithm in local development is enhanced, and the speed of obtaining the optimal solution is accelerated. The expression is: Among them, η is the adjustment coefficient η≥1; a=1.
[0047] Optionally, in step S5, an adaptive Gaussian-Cauchy mixed perturbation mutation perturbation strategy is used to perform mutation perturbation on the optimal individual, and then the positions before and after the mutation are compared to select a better position to enter the next iteration, and the expression is:
[0048] H b (t) = X b (t) * (1+μ1*Gauss(σ)+μ2*cauchy(σ))
[0049] Among them, X b (t) is the optimal position of individual X in the tth iteration, H b (t) is the optimal position X in the tth iteration b (t) is the position after the Gaussian-Cauchy mixed perturbation, Gauss(σ) is the Gaussian mutation operator, cauchy(σ) is the Cauchy mutation operator, μ2 = t / T max , μ2=1-t / T max , the weight coefficients μ1, μ2 of the mutation operator change gradually in a one-dimensional linear manner to ensure that each iterative disturbance is balanced and smooth.
[0050] Optionally, the fault area is determined based on a multi-level partitioning location model, and the distribution network is divided into multiple sub-areas through a multi-level partitioning strategy, which is refined layer by layer to finally determine the specific location of the fault. The steps of generating the initialization individual and calculating its fitness value are:
[0051] Step S6.1, dividing the entire distribution network into multiple sub-areas according to the topological structure of the distribution network.
[0052] Step S6.2, perform preliminary fault detection on the nodes in each sub-area to determine the sub-area where the fault may occur. The expression is: F i ={Node j ||I upload,j -I actual,j |>∈}, where F i represents the set of nodes that may fail, I upload,j is the upload current value of the jth node, I actual,j is the actual current value of the jth node, ∈ is the fault judgment threshold;
[0053] Step S6.3, for the initially determined fault sub-area, further refine the fault location range, divide the sub-area into smaller sub-areas, repeat the fault detection process, and narrow the fault location range layer by layer;
[0054] Step S6.4, by refining layer by layer, finally determine the specific location of the fault, generate the initialization individual, and calculate its fitness value, which is expressed as Among them, P i represents the i-th initialized individual, and N is the total number of nodes.
[0055] Optionally, after determining the fault area in step S7, the step of checking whether the identified fault area matches the actual fault area is:
[0056] Minimize the sum of the differences between the uploaded current value of each node and the actual value. The smaller the sum of the differences, the higher the similarity between the resolved fault condition and the actual fault condition. The expression is:
[0057]
[0058] Among them, θ is a factor to prevent misjudgment, θ|X j |] is to prevent node X j The misjudgment factor, θ|X j |] has a value of 0.5.
[0059] After adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art. Of course, any product implementing the present invention does not necessarily need to achieve all the advantages described below at the same time:
[0060] By using the chaos initialization strategy, the population quality of the sample is improved. The chaos initialization strategy uses the ergodicity and randomness of the chaotic sequence to make the initial population more evenly distributed in the search space, thereby avoiding the situation where the population falls into the local optimum in the early stage. At the same time, the sine-cosine strategy and the adaptive Gauss-Cauchy mixed variational perturbation strategy are introduced into the northern goshawk search algorithm. These two strategies use the periodicity of the sine-cosine function and the randomness of the Gauss-Cauchy distribution to adaptively perturb individuals. Specifically, the sine-cosine strategy searches for individual positions over a large range through periodic fluctuations, enhancing the global search capability of the algorithm; while the adaptive Gauss-Cauchy mixed variational perturbation strategy improves the diversity of the population by locally perturbing the optimal individual, which helps to jump out of the local optimum and enhance the local escape capability. This not only maintains the diversity of the population, but also effectively improves the convergence speed and positioning accuracy of the algorithm.
[0061] The specific implementation modes of the present invention are further described in detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0062] The drawings described below are only some embodiments. For those skilled in the art, other drawings can be obtained based on these drawings without creative work. In the drawings:
[0063] Figure 1 It is a schematic diagram of the dual-source power distribution network structure;
[0064] Figure 2 is a flowchart of a fault location method;
[0065] Figure 3 This is a 33-node distribution network partition model diagram;
[0066] Figure 4 This is one of the schematic diagrams of the fault location convergence curve without information distortion;
[0067] Figure 5 This is the second schematic diagram of the fault location convergence curve without information distortion;
[0068] Figure 6 It is a schematic diagram of the convergence curve of fault location when the node information is distorted;
[0069] Figure 7 A schematic diagram showing the comparison of single-point fault location algorithms when the number of DGs connected to the distribution network is different;
[0070] Figure 8 Schematic diagram of the comparison of algorithms for multi-point fault location with different numbers of DGs connected to the distribution network;
[0071] Fig. 9Schematic diagram of the algorithm comparison for multi-point fault location with distorted information when different numbers of DGs are connected to the distribution network.
[0072] It should be noted that these drawings and textual descriptions are not intended to limit the conceptual scope of the present invention in any way, but are intended to illustrate the concept of the present invention for those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION
[0073] The present invention will now be described in further detail with reference to the accompanying drawings.
[0074] In order to accurately determine the direction and source of the fault current when a fault occurs in a certain area, this article defines the current direction flowing from the main power supply to the load end as the forward direction. When the FTU detects a forward fault current, Ij = 1; when the FTU detects a reverse fault current, Ij = -1; when the FTU does not detect a fault current, Ij = 0, as shown in the following formula
[0075]
[0076] See also Figure 1-9 As shown, in this embodiment, a method for actively locating faults in a power distribution network based on an improved northern goshawk search algorithm is provided, comprising the following steps:
[0077] Step S1: Read the fault current status information of the section switch, tie switch, and circuit breaker elements detected by the feeder terminal unit, upload it to the SCADA system of the master station, and generate the actual fault current array of the switch node according to the number of nodes based on the acquired data; distribution system problems include short circuit, overload, ground fault and other fault problems. Partition fault location in the distribution network refers to dividing the distribution network into a series of areas and monitoring and analyzing the power flow in each area to help quickly locate the location of the fault. This partition fault location method helps to improve the accuracy and efficiency of fault diagnosis, shorten the fault handling time, and ensure the stable operation of the power system.
[0078] by Figure 1 Taking the dual-source power distribution network shown as an example, it is stipulated that the state of the terminal node in each area represents the state of the entire area. The first layer of the hierarchical positioning model uses an algorithm to locate the fault area according to the regional state, and the second layer locates the specific fault area in the area obtained by positioning. For example, a fault occurs in area 9 in the distribution network. The first level uses an algorithm to start regional positioning based on the state of each area, and provides the positioning result of the area 3 fault; the second level starts specific network segment positioning based on the state of nodes 8 and 9 in area 3 to determine the fault area.
[0079] The switching function can realize the conversion between line fault status and switch fault information. Considering the switching of distributed power sources, the switching function can be defined as follows
[19] :
[0080]
[0081] in, is the switch function value of node j, also called the expected state value; K u and K d are the power switching coefficients of the upstream and downstream areas of node j, respectively. When there is power input, the coefficient is set to 1, otherwise it is set to 0; are all regions or operating values between node j and each upstream power source; is the value of all regions or operations between node j and each downstream power source; and are the values of the upstream and downstream regions or operations of node j, respectively; M1 and M2 are the numbers of power supplies upstream and downstream of node j, respectively; N1 and N2 are the numbers of regions upstream and downstream of the node, respectively.
[0082] Step S2: Initialize the population number, population dimension, variable range value and maximum iteration number of the northern goshawk optimization algorithm;
[0083] Step S3: Initializing a binary northern goshawk population, wherein each individual in the northern goshawk population represents a fault operation state of a group of feeder segments;
[0084] Step S4: Calculate the fitness value and update the position of the individual northern goshawk, and update the individual position through the prey recognition and attack stage and the pursuit and escape stage;
[0085] Step S5: Introduce the adaptive Gaussian-Cauchy mixed perturbation mutation strategy and the sine-cosine strategy, update the individual position through the mixed perturbation of Gaussian mutation and Cauchy mutation, determine whether the maximum number of iterations is reached or the convergence condition is met, and if not, return to recalculate the fitness value and update the individual position;
[0086] Step S6: Determine the fault area based on the multi-level partition positioning model, divide the distribution network into multiple sub-areas through the multi-level partition strategy, refine it layer by layer, and finally determine the specific location of the fault. Generate an initialization individual and calculate its fitness value;
[0087] Step S7: After determining the fault area, check whether the identified fault area and the actual fault area match. If they match, the fault area is determined and the process ends. Otherwise, return to generate the initialization individual and recalculate the fitness value.
[0088] In this embodiment, when the step S3 of the binary optimization of the northern goshawk algorithm updates the position of the northern goshawk, its expression is:
[0090]
[0091] Among them, α = 2.3; P i,d Represents the updated position of individual i in the d-dimensional space after 1 iteration.
[0092] In the present embodiment, step S4 is in the process of updating individual position in the prey identification stage and the pursuit and escape stage because the northern goshawk individual can determine the next search area according to the available information of the previous stage when selecting another search area, if the northern goshawk colony is trapped in local search, then the prey cannot be accurately captured in the global search optimization process. To make up for this deficiency, firstly, the northern goshawk colony is evenly distributed in the search space by Sinusoidal mapping, which can solve the problem of "premature convergence" to a certain extent, and then the global detection ability of the algorithm is enhanced by crossover and non-uniform mutation. The crossover operation interchanges the position of the northern goshawk and recalculates the fitness value, and when the fitness value of the new northern goshawk position is better than the original northern goshawk, the original northern goshawk individual is replaced, thereby increasing the diversity of the population individual after iteration. The non-uniform mutation operator disturbs the position of the northern goshawk, further improves the diversity of the population individual, and increases the search range and search accuracy of the northern goshawk algorithm. When the non-uniform mutation strategy disturbs the position of the northern goshawk, k dimensions are randomly selected for each northern goshawk individual for disturbance in turn. When the new northern goshawk individual generated by each disturbance is better than the original northern goshawk individual, the original northern goshawk individual is replaced. The expression is:
[0093]
[0094] Among them, P t represents the position of the northern goshawk at the tth iteration; T represents the maximum number of iterations, which is 100; b is the system parameter that determines the non-uniformity, which is 2.
[0095] In this embodiment, the process summary of step S4 in the prey identification stage randomly selects a prey and quickly attacks it, and simulates the hunting behavior of the northern goshawk through a mathematical model to update the individual position. The mathematical expression of the behavior of the northern goshawk in this stage is as follows:
[0096] P i =X k ,i=1,2,...,N,k=1,2,...,i-1,i+1,...,N
[0097]
[0098]
[0099] Among them, P i is the location of the prey selected by the i-th northern goshawk; Fpi is the objective function value, that is, the fitness value, K is a random natural number in [1, N], is the new state of the i-th northern goshawk, is the corresponding fitness value, R is a random number belonging to [0,1], and the value of i is 1 or 2. r and i are random numbers used to generate random northern goshawk algorithm behaviors in search and update.
[0100] In this embodiment, step S4 is that the northern goshawk chases the prey during the chasing and escaping stage, simulates the scene of the prey escaping, and updates the individual position through the mathematical model, which is expressed as:
[0101]
[0102]
[0103]
[0104] Where t is the current number of iterations, T is the maximum number of iterations, is the new state of the i-th northern goshawk in the pursuit phase, is the new state of the i-th northern goshawk in the j-th dimension during the pursuit phase, and F i new,p2 is the fitness value in the new state.
[0105] In this embodiment, the oscillation change characteristics of the sine-cosine model are used to act on the position in step S5. In order to solve the problem of premature convergence of the algorithm and falling into the local optimum, this article quotes the sine-cosine algorithm (SCA) to act on the position by using the oscillation change characteristics of the sine-cosine model to maintain individual diversity and thus improve the global search capability of INGO. The central idea of SCA is to optimize the whole and the local according to the oscillation change of the sine-cosine model to obtain the overall optimal value. The step search factor for the basic sine-cosine algorithm is r1=a-at / Iter max (a is a constant, t is the number of iterations, and this paper sets a=1) and shows a linear decreasing trend, which is not conducive to further balancing the global search and local development capabilities of NGOs. The new nonlinear decreasing search factor is as follows:
[0106]
[0107] In the early stage, the weight is large and the decreasing speed is slow, which is conducive to improving the global optimization ability. When the weight factor is small, the algorithm's advantage in local development is enhanced, and the speed of obtaining the optimal solution is accelerated. Among them, η is the adjustment coefficient η≥1; a=1.
[0108] In this embodiment, the unique pursuit and escape phase of the NGO algorithm has good traversal properties, but the NGO algorithm is prone to fall into local optimality in the later stages of iteration. Therefore, the adaptive Gaussian-Cauchy mixed perturbation mutation perturbation strategy
[22] is introduced to enhance the algorithm's local development and global search capabilities, thereby increasing the probability of finding the best prey position. Since the results of the mutation perturbation operation are random, if all individuals are subjected to mutation perturbation operations, the complexity of the algorithm will inevitably increase. Therefore, this paper only performs mutation perturbation on the optimal individual, and then compares its position before and after the mutation, selects a better position to enter the next iteration, fully increases the diversity of individuals, and expands the population search range. Its expression is:
[0109] H b (t) = X b (t) * (1+μ1*Gauss(σ)+μ2*cauchy(σ))
[0110] Among them, X b (t) is the optimal position of individual X in the tth iteration, H b (t) is the optimal position X in the tth iteration b (t) is the position after the Gaussian-Cauchy mixed perturbation, Gauss(σ) is the Gaussian mutation operator, cauchy(σ) is the Cauchy mutation operator, μ2 = t / T max , μ2=1-t / T max , the weight coefficients μ1, μ2 of the mutation operator change gradually in a one-dimensional linear manner to ensure that each iterative disturbance is balanced and smooth.
[0111] In this embodiment, the fault area is determined based on the multi-level partition positioning model, and the distribution network is divided into multiple sub-areas through a multi-level partition strategy, which is refined layer by layer to finally determine the specific location of the fault.
[0112] Step S6.1, dividing the entire distribution network into multiple sub-areas according to the topological structure of the distribution network.
[0113] Step S6.2, perform preliminary fault detection on the nodes in each sub-area to determine the sub-area where the fault may occur. The expression is: F i ={Node j ||I upload,j -I actual,j |>∈}, where F i represents the set of nodes that may fail, I upload,j is the upload current value of the jth node, I actual,j is the actual current value of the jth node, ∈ is the fault judgment threshold;
[0114] Step S6.3, for the initially determined fault sub-area, further refine the fault location range, divide the sub-area into smaller sub-areas, repeat the fault detection process, and narrow the fault location range layer by layer;
[0115] Step S6.4, by refining layer by layer, finally determine the specific location of the fault, generate the initialization individual, and calculate its fitness value, which is expressed as Among them, P i represents the i-th initialized individual, and N is the total number of nodes.
[0116] It should be noted that the multi-level partition positioning model can adapt to distribution network structures of different sizes and complexities. Whether it is a large distribution network or a small distribution network, it can be adapted by adjusting the size of the sub-area and the level of refinement. Through the layer-by-layer refinement of the fault detection method, the non-fault area can be gradually excluded, reducing the probability of false alarms. At the same time, through the refinement of the detection, it can be ensured that the real fault area is accurately located and the possibility of missed alarms is reduced. This method can minimize false alarms and missed alarms while ensuring the positioning accuracy, and improve the accuracy and reliability of fault detection and positioning. This adaptability makes this method have a wide range of application prospects and can be applied to various different actual distribution network environments.
[0117] In this embodiment, after determining the fault area in step S7, the step of checking whether the identified fault area matches the actual fault area is as follows:
[0118] Minimize the sum of the differences between the uploaded current value of each node and the actual value. The smaller the sum of the differences, the higher the similarity between the resolved fault condition and the actual fault condition. The expression is:
[0119]
[0120] Among them, θ is a factor to prevent misjudgment, θ|X j |] is anti-node X j The misjudgment factor, θ|X j |] is 0.5. Figure 1 For example, when the fault occurs in area X5 in area 2, X = [X1, X2, L, X16] = [0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], the switching function values of nodes 1 and 2 in area 1 can be obtained by formula (2) as follows:
[0121]
[0122]
[0123] Similarly, the switching function values of the nodes in regions 3, 4, and 5 can be calculated as follows:
[0124] It can be found that the switch function values of the nodes in the fault area are not equal, while the switch function values of the nodes in the non-fault area are equal to the switch function values of the end nodes of their respective areas. To further verify the partition basis, assuming that different regional faults and multiple faults occur in area 3, and there is no information distortion, the switch function values of all nodes in the non-fault area can be calculated, as shown in Table 1.
[0125] Table 1 Switch function values of nodes in each area under different fault conditions
[0126]
[0127] Experimental Example 1:
[0128] Case Analysis
[0129] In order to verify the effectiveness of the method proposed in this paper, a mathematical model of the IEEE 33-node ADN structure was established on the Matlab platform. Figure 3 As shown in the figure. X1-X33 represents 33 feeder segments, 1-33 represents 33 switch nodes, and K1-K3 represents the access switch of each distributed power source. The access of distributed power sources increases the complexity of fault location. In the simulation experiment, distributed power sources are connected to the ADN from random nodes, and the number of distributed power sources is different. It should also be noted that the data of the fault diagnosis algorithm in this paper comes from a data center and control center in Jilin Province.
[0130] Simulation test analysis
[0131] Assumptions Figure 3 A fault occurs in area 7 of area 6. All three distributed power sources are in operation and the fault current information is not distorted. First, according to formula (3), the fault current information uploaded by the FTU can be calculated as
[0132] [1,1,1,1,1,0,0,0,0,-1,-1,-1,0,0,0,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-0,0,0,0], then use the improved Northern Eagle algorithm to extract the current information of the end nodes in each area as [1,1,-1,0,-1,0,-1,-1,-1,0], search the fault area, and the corresponding result is [0,0,0,0,0,0,1,0,0,0,0,0,0,0]. It can be inferred that the fault occurred in an area in area 6. Then, apply the INGO algorithm, Figure 4-5The comparison results of the fit and the number of iterations are shown. The positioning result is [1, 0, 0, 0], which means that the fault in area 7 is consistent with the hypothesis.
[0133] In order to better illustrate the fault tolerance of the proposed method, information distortion is further added to the faulty nodes. Assuming that faults occur simultaneously in zones 9, 12, and 22, the state of node 3 changes from 1 to 0, the state of node 18 changes from -1 to 0, and the state of node 32 changes from 0 to -1. Then, the convergence curve of fault location when node information is distorted can be calculated and obtained, as shown in Figure 6 As shown. At the same time, the state matrix [0,0,1,0,0,1,1,0,0,0,0,0,0] can be obtained, that is, the faults occurred in area 3, area 6 and area 7 respectively. On this basis, the state values of each area in area 3, area 6 and area 7 can be further calculated, and the results are [0,1,0], [0,0,1,0] and [0,1] respectively. It can be reasonably inferred that the faults occurred in area 9, area 12 and area 22, which is consistent with the set fault assumption location.
[0134] from Figure 4-5 It can be seen that the convergence curve presents a straight line after the first iteration, which means that the proposed method finds the optimal solution at the beginning of the iteration, and then the system converges to a stable state. Figure 5 It can be seen that the convergence curve presents a straight line after the third iteration, indicating that the method proposed in this paper can find the optimal solution after two iterations. Therefore, a positive conclusion can be drawn: the INGO algorithm partition location model proposed in this paper can accurately locate the fault in the ADN within the maximum number of iterations under the preset no information distortion double fault and no information distortion triple fault conditions. In addition, this method also has great advantages in convergence speed.
[0135] Experimental Example 2
[0136] This paper selects the Northern Eagle Optimization Algorithm (NGO), the Gray Wolf Optimization Algorithm (GWO) and the Whale Optimization Algorithm (WOA) for comparative experiments. Single-point and multi-point fault comparison simulation experiments were carried out when distributed power sources of different numbers and locations were connected to the distribution network. At the same time, the positioning accuracy and average convergence algebra were used as performance evaluation indicators of the algorithm. Since the above algorithms are all random optimization algorithms, each algorithm is repeated 20 times in the experiment, and then the average value of the performance evaluation indicators of each algorithm is calculated.
[0137] Single Point of Failure
[0138] by Figure 3Taking the IEEE33-node ADN model in the example, in order to simulate the situation where distributed generation with different numbers and access locations is connected to the distribution network, four experiments are set up as shown in Table 2. For example, [K1, K2, K3] = [0, 0, 0] means that no distributed generation is connected to the distribution network. When a fault occurs in the X11 area, the information recorded by the FTU device is [1,
[0139] 1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0], the output positioning information is [0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0]. The comparison curve of the algorithm iteration is as follows: Figure 7 a. Similarly, [K1, K2, K3] = [1, 0, 0], [K1, K2, K3] = [0, 1, 1],
[0140] The simulation results of the experiment [K1, K2, K3] = [1, 1, 1] are as follows Figure 7 b to Figure 7 As shown in d.
[0141] Table 2. Single point failure simulation example
[0142]
[0143] Multiple points of failure
[0144] by Figure 3 Taking the IEEE33-node ADN model in as an example, in order to simulate the situation where distributed generation sources with different numbers and access locations are connected to the distribution network, four experiments are set up as shown in Table 3. For example, [K1, K2, K3] = [1,1,1] means that there are three distributed generation sources connected to the distribution network. When the fault occurs in area X10, area X28 and area X32, the information recorded by the FTU device is [1, ...
[0145] The comparison curve of the algorithm iteration is as follows: [0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0]. Figure 8d. Similarly, the simulation results of the experiments [K1, K2, K3] = [0, 0, 0], [K1, K2, K3] = [1, 0, 0], and [K1, K2, K3] = [0, 1, 1] are shown in Figure 8 a and Figure 8 As shown in c.
[0146] Table 3. Multi-point fault simulation example
[0147]
[0148] Multi-point distortion fault
[0149] Since the actual working environment of the distribution network is complex and uncontrollable, the FTU device nodes are often exposed to harsh environments, which may cause data loss and data distortion when the detection device nodes transmit fault current information. When a fault occurs in the actual active distribution network, the FTU device at the node may not be able to upload the corresponding fault information due to the fault, and there may be false alarms, missed alarms, and wrong alarms. In the simulation experiment, under the premise that single-point and multi-point faults occur in the distribution network, the FTU reports the fault current information distortion point, and the distribution network containing distributed power sources is analyzed for fault tolerance. For example, [K1, K2, K3] = [0, 0, 0], section X11 fails, the single-point distortion position is 8, and the algorithm iteration comparison curve is shown in Figure 2. Fig. 9 (a); [K1, K2, K3] = [1, 0, 0], if the segment X9 fails, the multi-point distortion positions are 6 and 12, and the algorithm iteration comparison curve is shown in Fig. 9 (b); [K1, K2, K3] = [0, 1, 1], if faults occur in sections X15 and X27, the multi-point distortion positions are 3 and 33, and the algorithm iteration comparison curve is shown in Fig. 9 (c); [K1, K2, K3] = [1, 1, 1], if faults occur in sections X13 and X25, the multi-point distortion positions are 5 and 23, and the algorithm iteration comparison curve is shown in Fig. 9 (d)
[0150] from Figure 7 , 8 , 9, it can be found that the INGO algorithm has obvious advantages over the NGO, GWO, and WOA algorithms in terms of convergence speed. INGO can find the optimal solution in about 5 iterations. WOA, SCA, and FFOA can search for the optimal solution, but because of their weak global optimization ability, they need 6, 24, and 75 iterations to find the optimal solution, respectively, indicating that the INGO algorithm has a good solution speed. In order to prevent the occurrence of accidents, the INGO, NGO, GWO, and WOA algorithms are respectively operated 100 times, and the accuracy and average number of iterations of each algorithm are counted when faults occur in different locations and information distortion exists. The performance comparison results of the four algorithms are shown in Table 4.
[0151] Table 4 Positioning results of different algorithms
[0152]
[0153] From the analysis of Table 4, we can see that as the complexity of the fault type increases, the average number of iterations of the fault location process of the Northern Goshawk algorithm, the Gray Wolf search algorithm and the Whale Optimization Algorithm increases and the accuracy decreases. After the introduction of the Northern Goshawk algorithm partition model, the dimension of the algorithm solution is reduced while the speed of solution is accelerated and the accuracy is improved, so that the partitioned improved Northern Goshawk algorithm has an average accuracy improvement of 4.5% in the solution process compared with the unpartitioned Northern Goshawk algorithm. Compared with other optimization search algorithms, the improved Northern Goshawk algorithm performs well in terms of solution speed and accuracy, indicating that the INGO partition strategy model achieves higher accuracy and faster fault section location, which has good engineering practice significance.
[0154] The present invention is not limited to the above-mentioned embodiments. Anyone should be aware that any structural changes made under the enlightenment of the present invention, and any technical solutions that are the same or similar to the present invention, fall within the protection scope of the present invention. The technology, shape, and structural parts not described in detail in the present invention are all well-known technologies.
Claims
1. An active distribution network fault location method based on an improved northern goshawk search algorithm, characterized in that: The following steps are involved: Step S1: Read the fault current status information of the section switch, tie switch and circuit breaker elements detected by the feeder terminal unit, upload it to the SCADA system of the master station, and generate the actual fault current array of the switch node according to the number of nodes based on the acquired data; Step S2: Initialize the population number, population dimension, variable range value and maximum iteration number of the northern goshawk optimization algorithm; Step S3: Initializing a binary northern goshawk population, wherein each individual in the northern goshawk population represents a fault operation state of a group of feeder segments; Step S4: Calculate the fitness value and update the position of the individual northern goshawk, and update the individual position through the prey recognition and attack stage and the pursuit and escape stage; Step S5: Introduce the adaptive Gaussian-Cauchy mixed perturbation mutation strategy and the sine-cosine strategy, update the individual position through the mixed perturbation of Gaussian mutation and Cauchy mutation, determine whether the maximum number of iterations is reached or the convergence condition is met, and if not, return to recalculate the fitness value and update the individual position; Step S6: Determine the fault area based on the multi-level partition positioning model, divide the distribution network into multiple sub-areas through the multi-level partition strategy, refine it layer by layer, and finally determine the specific location of the fault. Generate an initialization individual and calculate its fitness value; Step S7: After determining the fault area, check whether the identified fault area and the actual fault area match. If they match, the fault area is determined and the process ends. Otherwise, return to generate the initialization individual and recalculate the fitness value.
2. The active distribution network fault location method based on the improved northern goshawk search algorithm according to claim 1 is characterized in that: When the step S3 of binary optimization of the northern goshawk algorithm updates the position of the northern goshawk, its expression is: Among them, α = 2.3; P i,d Represents the updated position of individual i in the d-dimensional space after 1 iteration.
3. The method for locating faults in an active distribution network based on an improved northern goshawk search algorithm according to claim 1, characterized in that: Step S4: During the updating of individual positions in the prey identification phase and the pursuit and escape phase, the following steps need to be completed: The northern goshawk population is evenly distributed in the search space through sinusoidal mapping, and the global detection ability of the algorithm is enhanced through crossover and non-uniform mutation. Among them, the crossover operation swaps the positions of the northern goshawks and recalculates the fitness value. When the fitness value of the new northern goshawk position is better than the original northern goshawk, the original northern goshawk individual is replaced, and the non-uniform mutation operator disturbs the position of the northern goshawk. When the non-uniform mutation strategy disturbs the position of the northern goshawk, k dimensions are randomly selected for each northern goshawk individual for disturbance in turn. When the new northern goshawk individual generated by each disturbance is better than the original northern goshawk individual, the original northern goshawk individual is replaced. The expression is: Among them, P t represents the position of the northern goshawk at the tth iteration; T represents the maximum number of iterations, which is 100; b is the system parameter that determines the non-uniformity, which is 2.
4. The method for locating faults in an active distribution network based on an improved northern goshawk search algorithm according to claim 1, characterized in that: In step S4, the process summary of the prey identification stage randomly selects a prey and quickly attacks it, and simulates the hunting behavior of the northern goshawk through a mathematical model to update the individual position. The mathematical expression of the behavior of the northern goshawk in this stage is as follows: P i =X k ,i=1,2,...,N,k=1,2,...,i-1,i+1,...,N Among them, P i is the location of the prey selected by the i-th northern goshawk; F pi is the objective function value, that is, the fitness value, K is a random natural number in [1, N], is the new state of the i-th northern goshawk, is the corresponding fitness value, R is a random number belonging to [0,1], and the value of i is 1 or 2. r and i are random numbers used to generate random northern goshawk algorithm behaviors in search and update.
5. The method for locating faults in an active distribution network based on an improved northern goshawk search algorithm according to claim 1, characterized in that: Step S4: During the chasing and escaping phase, the northern goshawk chases the prey, simulates the escape scenario of the prey, and updates the individual position through a mathematical model, which is expressed as: Where t is the current number of iterations, T is the maximum number of iterations, is the new state of the i-th northern goshawk in the pursuit phase, is the new state of the i-th northern goshawk in the j-dimension during the pursuit phase, is the fitness value in the new state.
6. The method for locating faults in an active distribution network based on an improved northern goshawk search algorithm according to claim 1, characterized in that: In step S5, the oscillation change characteristics of the sine and cosine model are used to act on the position, and the overall and local optimization are performed according to the oscillation change of the sine and cosine model to obtain the overall optimal value. The nonlinear decreasing search factor has a large weight in the early stage and a slow decreasing speed, which is conducive to improving the global optimization ability. When the weight factor is small, the advantage of the algorithm in local development is enhanced, and the speed of obtaining the optimal solution is accelerated. The expression is: Among them, η is the adjustment coefficient η≥1; a=1.
7. The method for locating faults in an active distribution network based on an improved northern goshawk search algorithm according to claim 1, characterized in that: In step S5, the adaptive Gaussian-Cauchy mixed perturbation mutation perturbation strategy is used to mutate the optimal individual, and then the positions before and after the mutation are compared to select a better position to enter the next iteration. The expression is: H b (t)=X b (t) * (1+μ1*Gauss(σ)+μ2*cauchy(σ)) Among them, X b (t) is the optimal position of individual X in the tth iteration, H b (t) is the optimal position X in the tth iteration b (t) is the position after the Gaussian-Cauchy mixed perturbation, Gauss(σ) is the Gaussian mutation operator, cauchy(σ) is the Cauchy mutation operator, μ2 = t / T max , μ2=1-t / T max , the weight coefficients μ1, μ2 of the mutation operator change gradually in a one-dimensional linear manner to ensure that each iterative disturbance is balanced and smooth.
8. The method for locating faults in an active distribution network based on an improved northern goshawk search algorithm according to claim 1, characterized in that: The fault area is determined based on the multi-level partition positioning model. The distribution network is divided into multiple sub-areas through a multi-level partition strategy, which is refined layer by layer to finally determine the specific location of the fault. The steps to generate the initialization individual and calculate its fitness value are as follows: Step S6.1, dividing the entire distribution network into multiple sub-areas according to the topological structure of the distribution network. Step S6.2, perform preliminary fault detection on the nodes in each sub-area to determine the sub-area where the fault may occur. The expression is: F i ={Node j ||I upload,j -I actual,j |>∈}, where F i represents the set of nodes that may fail, I upload,j is the upload current value of the jth node, I actual,j is the actual current value of the jth node, ∈ is the fault judgment threshold; Step S6.3, for the initially determined fault sub-area, further refine the fault location range, divide the sub-area into smaller sub-areas, repeat the fault detection process, and narrow the fault location range layer by layer; Step S6.4, by refining layer by layer, finally determine the specific location of the fault, generate the initialization individual, and calculate its fitness value, which is expressed as Among them, P i represents the i-th initialized individual, and N is the total number of nodes.
9. The method for locating faults in an active distribution network based on an improved northern goshawk search algorithm according to claim 1, characterized in that: After the fault area is determined in step S7, the steps of checking whether the identified fault area matches the actual fault area are as follows: Minimize the sum of the differences between the uploaded current value of each node and the actual value. The smaller the sum of the differences, the higher the similarity between the resolved fault condition and the actual fault condition. The expression is: Among them, θ is a factor to prevent misjudgment, θ|X j |] is to prevent node X j The misjudgment factor, θ|X j |] has a value of 0.5.
Citation Information
Cited By
Personal perception network structure optimization method based on improved snake vulture algorithm
CN119729556A
