Power distribution network fault section positioning method considering distributed power supply dynamic switching

Abstract

The invention provides an improved subtraction optimizer algorithm-based power distribution network fault section positioning method considering distributed power supply dynamic switching, and belongs to the field of intelligent power distribution network fault diagnosis. The model can effectively deal with the influence of dynamic switching of the DG on the system, and makes up for the defects of a traditional single-power network model. Secondly, in order to solve the positioning model, a multi-strategy fusion improved method is adopted for an original algorithm, and a subtraction optimizer algorithm (ISABO) with excellent solving performance and de-trapping capability is obtained; and finally, the improved ISABO algorithm is applied to a 33-node power distribution network test system, so that the 33-node power distribution network test system operates in various designed complex fault working conditions. Analog simulation shows that the ISABO algorithm has feasibility in the fault positioning problem and can adapt to the on-off state of an interconnection switch and dynamic switching of a plurality of DGs; compared with an original SABO algorithm, the method has the advantages that the positioning precision can be always kept 100%, the positioning speed is also remarkably improved, and good search speed, positioning accuracy and fault tolerance are shown.

Description

Technical Field

[0001] The technical solution of the present invention belongs to the field of intelligent distribution network fault diagnosis, and specifically is a distribution network fault section locating method based on an improved subtraction optimizer algorithm taking into account the dynamic switching of distributed power sources. Background Art

[0002] In recent years, traditional energy has been increasingly exhausted due to over-exploitation, and the energy crisis has become increasingly serious. In addition, large-scale development and utilization have also brought environmental pollution, threatening human survival and safety. Therefore, it is very important to find and develop new energy as soon as possible to improve energy supply stability and energy security. At present, with the rapid development of renewable energy technology and the advancement of energy transformation, the integration and application of distributed generation (DG) in distribution networks are becoming more and more extensive. Distributed generation, with its flexible, efficient and environmentally friendly characteristics, provides new possibilities for the sustainable development of the power system and is of great help to the formation of an energy structure dominated by new energy.

[0003] However, with the large-scale access of distributed power sources, the operation of distribution networks has become more uncontrollable and random, making the distribution network from a traditional radial network to a complex network with multiple power sources, and the power flow direction has changed from unidirectional to bidirectional flow. The dynamic switching of DG and the randomness of output power make the traditional distribution network fault location method invalid; on the other hand, according to surveys, among various short-circuit faults, the probability of single-phase grounding fault is the highest; although the fault current of this fault is small and the system can continue to operate for a period of time, long-term operation with faults may cause insulation weak points to be broken down, further expanding the scope of fault impact; in addition, the harm of phase-to-phase short-circuit faults is the greatest, and the three-phase short-circuit fault is the most obvious. If the fault point cannot be quickly located and the fault cannot be removed, the power outage range will be expanded, causing greater losses to the power system; therefore, it is urgent to study the fast fault section location method suitable for DG access to the distribution network; on the one hand, accurate fault section location can shorten the fault power outage time, reduce power outage losses, and improve power supply reliability; on the other hand, through fault section location, the fault range can be quickly determined, providing decision support for subsequent fault isolation and recovery, and improving the intelligent level of fault handling;

[0004] In order to find the location of the fault accurately and timely, the present invention proposes a fault section location method based on an improved subtraction-average-based optimizer (ISABO) algorithm considering the dynamic switching of distributed power sources; firstly, a fault location model suitable for a distribution network containing DG is established, which takes into account the influence of the dynamic switching of DG on fault location, overcomes the limitations of the traditional single power model, and improves the accuracy of fault location; secondly, an ISABO algorithm with both optimization capability and fault tolerance is designed, which has the advantages of few set parameters, stable search performance, high optimization efficiency, etc., is suitable for multi-dimensional optimization problems, and its excellent performance is verified in performance test experiments; finally, through simulation experiments that comprehensively consider various influencing factors, it is verified that the fault location method is not affected by the number of faults and distorted signals; in addition, experiments are also carried out in a network containing tie switches, and the accuracy of its positioning is not affected, showing good adaptability and stability. Summary of the invention

[0005] Aiming at the problem that a large number of distributed power sources are connected to the grid, resulting in a complex topology of the distribution network, variable fault currents, and difficulty in fault location, the present invention comprehensively considers the state of the DG switching switch, establishes a fault location model suitable for the distribution network containing DG, takes into account the impact of the distributed power source input and removal on the distribution network fault, and improves the accuracy and universality of fault location; in addition, in order to accurately solve the fault location model, an ISABO algorithm with excellent solution performance and good fault tolerance is proposed. On the basis of the original SABO algorithm, improved strategies such as chaos map initialization, adaptive dynamic inertia weight, out-of-bounds individual regression, Cauchy distribution variation and fusion of reverse learning strategy are integrated, which significantly improves the solution accuracy and iteration speed of the algorithm, and has excellent performance in solving the model;

[0006] The technical solution adopted by the present invention to solve the technical problem is: a method for locating a distribution network fault section based on an improved subtraction optimizer algorithm taking into account the dynamic switching of distributed power sources, comprising the following steps:

[0007] Step 1: Encode the fault status of each section in the distribution network and the fault current information flowing through each node, and use the encoded data as the input of the positioning model;

[0008] Step 2, taking into account the large-scale access of distributed generation and the dynamic switching of each DG switch, a switching function and evaluation function adapted to the large-scale access of DG are established;

[0009] Step 3: Set the initialization parameters of the ISABO algorithm according to the specific distribution network structure, including the population number N, the maximum number of iterations T maxAs well as the upper and lower limits of the search range, based on the newly proposed chaotic mapping strategy, the position of each search particle in the initial population is obtained and its corresponding fitness function value is calculated;

[0010] Step 4, start the iteration, and each particle in the search range continuously changes its position through the improved position update strategy (adaptive dynamic inertia weight), gradually approaching the global optimal value;

[0011] Step 5: Check whether the position of each search particle exceeds the solution space. If so, perform regression processing on the out-of-bounds particles.

[0012] Step 6: Calculate the fitness value of each search particle, select the current global optimal particle by comparison, use the mutation strategy that combines Cauchy distribution mutation and reverse learning to perturb the optimal particle, and check whether the fitness value corresponding to the perturbed particle is better;

[0013] Step 7, determine whether the loop termination condition is met, if the condition is met, output the current global optimal solution; otherwise, return to step 3 and continue to perform the corresponding operation;

[0014] Step 8: According to the output global optimal solution, the specific location of the final fault section is obtained;

[0015] Step 1: Encode the fault status of each section in the distribution network and the fault current information flowing through each node, and use the encoded data as the input of the positioning model;

[0016] (1.1) Information characterization

[0017] After a fault occurs in the active distribution network, the supervisory control and data acquisition system (SCADA) obtains the fault current information uploaded by the FTU installed at the switch node, and uses this information to start the algorithm to analyze, process, and locate the fault section; this study uses (i = 1, 2, ..., N) to represent the fault status information of the i-th section, where N is the total number of sections in the system, and the specific encoding method is:

[0018]

[0019] In the original single-power network, since there is no DG, the direction of the power flow is unidirectional, and the node status information only exists in two situations: when the node detects a fault current, the switch status information is 1; when the node does not flow through the fault current, the switch status information is 0. However, with the widespread access of distributed power sources, the network structure tends to be complex, and the direction of the power flow has also changed from the original unidirectional to bidirectional. In order to accurately determine the direction and source of the fault current when a fault occurs in the section, this study stipulates that the direction of the current flowing from the main power supply to the load side is the positive direction. Based on this stipulated positive direction, the switch status of each node is encoded, including the switch status S of the DG.j The code is:

[0020]

[0021] Step 2, taking into account the large-scale access of distributed generation and the dynamic switching of each DG switch, a switching function and evaluation function adapted to the large-scale access of DG are established;

[0022] (2.1) Switching function

[0023] In order to make full use of various known information, it is very important to establish the conversion between fault current information and section status. This conversion needs to be realized by a switch function. For a distribution network with only a main power supply, when a fault occurs at a certain point, the fault current always flows from the location of the power supply to the fault point, and the fault current will not flow through the switch from the fault point to the end of the feeder. Due to this characteristic, the fault location process is relatively easy. The specific switch function is:

[0024]

[0025] In the formula, is the expected value of the i-th switch state information; x j is the status information of a certain segment j in the area from node i to the end of this branch; ∏ is an OR operation;

[0026] Different from the above network, the switching function value of the network containing DG is not only related to the state of the downstream section, but also closely related to the state of the upstream section and the switching state of each DG. Based on this situation, this study proposes a new switching function that adapts to the large-scale access of DG. The specific formula is as follows:

[0027]

[0028] In the formula, is the expected function value of the i-th switch; N1 and N2 are the total number of upstream and downstream sections of switch i; N DG is the total number of connected DGs; K j is the switching state of the jth DG downstream of switch i; x ju and x jd are the state information of the j-th feeder section in the upstream and downstream areas of switch i, respectively, thereby generating the expected fault current array;

[0029] (2.2) Evaluation function

[0030] Since the switching function is constructed based on the causal relationship between the system sections and nodes, it is feasible and effective to use the node expected value calculated by the switching function to approximate the actual fault current information of the node collected by the FTU; therefore, in order to construct this approximation model, it is crucial to establish a suitable objective function; in addition, while constructing the approximation model to find the minimum value, reasonable measures must be taken to avoid the misjudgment problem caused by multiple solutions for one value; this study introduces a weight coefficient ω to deal with the misjudgment problem, and the specific fitness function formula is as follows:

[0031]

[0032] In the formula, W is a very large number, which is taken as 100 here; S i is the actual fault current information collected by the device at node i, which is the input data of the algorithm; is the calculated expected function value of node i; ω is the weight factor, which is taken as 0.5 here for the convenience of subsequent calculation; x j is the status information of the jth segment, which is the variable of this formula; D is the total number of nodes in the network; d is the total number of segments contained in the network; this formula is used to quantify the deviation between the actual fault current information and the expected fault current information. The smaller the deviation, the more accurate the positioning;

[0033] Step 3: Set the initialization parameters of the ISABO algorithm according to the specific distribution network structure, including the population number N, the maximum number of iterations T max As well as the upper and lower limits of the search range, based on the newly proposed chaotic mapping strategy, the position of each search particle in the initial population is obtained and its corresponding fitness function value is calculated;

[0034] (3.1) Chaotic map initialization

[0035] In the optimization process of the algorithm, the distribution of the initial population affects the balance between the algorithm's exploration space and development capabilities. A good population initialization can promote global search. The traditional SABO algorithm uses a random initialization strategy, which easily leads to uneven distribution of the initial population and is not conducive to covering more areas in the search space. Therefore, in order to introduce diversity in the initial stage of the population and help the algorithm explore a wider solution space, this study uses the Circle mapping strategy to initialize the population, and its expression is:

[0036]

[0037] In the formula, mod is the remainder function;

[0038] Step 4, start the iteration, and each particle in the search range continuously changes its position through the improved position update strategy (adaptive dynamic inertia weight), gradually approaching the global optimal value;

[0039] (4.1) Adaptive dynamic inertia weight

[0040] Adding a dynamic inertia factor to the position update formula of the traditional SABO algorithm can make the algorithm focus on global search in the early stage of the search and perform fine search in the later stage of the search, which greatly improves the algorithm's optimization ability and solution speed; the corresponding formula is as follows:

[0041]

[0042] Where t is the current iteration number; T max is the maximum number of iterations set; α1 and α0 are two coefficients, which are taken as 0.9 and 0.5 respectively;

[0043] Step 5: Check whether the position of each search particle exceeds the solution space. If so, perform regression processing on the out-of-bounds particles.

[0044] (5.1) Regression processing of out-of-bounds individuals

[0045] During the search process of the SABO algorithm, some particles may exceed the search boundary due to position updates or disturbances. In this case, if the position of the out-of-bounds particle is simply assigned to the boundary value of the search range, the out-of-bounds particles will gather at the boundary, resulting in uneven distribution of the particle population and affecting the algorithm's solution performance. In order to solve this problem, this study adopts a new boundary regression processing method, the specific formula is as follows:

[0046]

[0047] In the formula, u b and l b To solve the upper and lower boundary values ​​of the space; For particles that cross the boundary; is the particle after regression processing;

[0048] Step 6: Calculate the fitness value of each search particle, select the current global optimal particle by comparison, use the mutation strategy that combines Cauchy distribution mutation and reverse learning to perturb the optimal particle, and check whether the fitness value corresponding to the perturbed particle is better;

[0049] (6.1) Fusion of Cauchy distribution variation and reverse learning strategy

[0050] In view of the shortcoming of the traditional SABO algorithm that it is easy to fall into local extreme values, this study incorporates the Cauchy mutation operator and reverse learning strategy into the original algorithm to improve the position of the optimal solution, thereby avoiding the phenomenon of "premature maturity" of the algorithm; the reverse learning strategy is to reversely learn the current solution to generate a reverse solution, and then select a better solution by comparison. Its expression is as follows:

[0051]

[0052] Where u and d represent the upper and lower limits of the search boundary respectively; r is a random number matrix of (0, 1) standard uniform distribution 1*d (d is the spatial dimension); X best is the original optimal solution; is the optimal solution after mutation;

[0053] The random numbers generated by the Cauchy distribution can be used to perturb the optimal search particles to improve the global search capability of the algorithm and expand the search space. The formula is as follows:

[0054]

[0055] Where f(x) is the probability density function of the mutation strategy; is the multiplication symbol; cauchy(0,1) is the standard Cauchy mutation operator;

[0056] In order to improve the optimal solution of the original algorithm to the greatest extent, this study introduces the switching probability p, where p is a random number between 0.1 and 0.5. Now a number is randomly generated. If the number is less than p, the optimal solution position is updated using formula (9), otherwise, formula (10) is used.

[0057] Step 7, determine whether the loop termination condition is met, if the condition is met, output the current global optimal solution; otherwise, return to step 3 and continue to perform the corresponding operation;

[0058] Step 8: According to the output global optimal solution, the specific location of the final fault section is obtained;

[0059] Based on the output global optimal solution and the specific topological structure of the distribution network, the specific section where the fault occurred is located;

[0060] The above-mentioned method for locating faults in a distribution network containing distributed power sources based on an improved subtraction optimizer algorithm, the subtraction optimizer algorithm is an existing technology and is well known to those skilled in the art;

[0061] The advantages and positive effects of the present invention are:

[0062] 1. The present invention establishes a distribution network fault location model that takes into account a large number of DG connections, which makes up for the shortcomings of the traditional single-power distribution network fault location model. By introducing the impact of distributed generation on distribution network fault location and considering the dynamic switching of DG, the model's adaptability to the current new power system is significantly improved;

[0063] 2. The present invention designs an ISABO algorithm with excellent solving performance. By integrating a series of improvement measures into the original algorithm, various links such as initialization and update of the algorithm are improved, and the convergence and optimization capabilities of the algorithm are enhanced;

[0064] 3. In the example verification stage of the present invention, a processing method that considers the disconnection state of the tie switch in the fault location process is also proposed, and a fault location simulation experiment is carried out on the method, which proves that the proposed method can effectively cope with the change of the tie switch state and shows good adaptability in complex scenarios; BRIEF DESCRIPTION OF THE DRAWINGS

[0065] The present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0066] Figure 1 It is a flow chart of a fault location method for a distribution network containing distributed power sources based on an improved subtraction optimizer algorithm;

[0067] Figure 2 It is the topological structure diagram of IEEE 33-node distribution network including DG;

[0068] Figure 3 This is a comparison chart of the positioning accuracy of the four algorithms ISABO, SABO, ICS, and GA after running under fault cases Case 2, Case 3, Case 5, Case 6, and Case 7;

[0069] Figure 4 This is a comparison of the convergence curves of the four algorithms ISABO, SABO, ICS, and GA after running under fault cases Case 3, Case 6, and Case 8;

[0070] Figure 5 It is a schematic diagram of distribution network partition after considering the state of the tie switch; DETAILED DESCRIPTION

[0071] Figure 1The process of the fault location method of the distribution network containing distributed generation based on the improved subtraction optimizer algorithm shown is as follows: start → FTU equipment collects fault current information of each node → obtains the status information of each node through encoding → uploads to the control center → constructs the switch function and the evaluation function → sets the initial parameters of the algorithm according to the actual distribution network → obtains the initial population through the improved initialization strategy formula (6) → updates and iterates the position of the search particle using the improved update strategy formula (7) → uses formula (8) to perform regression processing on the particles beyond the boundary → uses formula (9) or (10) to disturb the current optimal solution → determines whether the convergence condition of the algorithm is met → if it is met, outputs the optimal solution → obtains the specific section with fault through the optimal solution → ends;

[0072] Figure 5 The figure shows a schematic diagram of distribution network partitioning after considering the state of the interconnecting switch. The specific process is: start → determine the position of the interconnecting switch → determine the open and closed state of the interconnecting switch → if the interconnecting switch is in the closed state, the interconnecting switch will not affect the fault location process, and the location process remains unchanged → if the interconnecting switch is in the open state, the entire system is divided into two independent subsystems with the interconnecting switch as the dividing point → construct respective switching functions and evaluation functions for the two subsystems → add the evaluation functions of the two subsystems to obtain the evaluation function of the entire network → use the algorithm to solve the aggregated function → obtain the final result of fault location → end.

[0073] Example

[0074] The present invention uses a PC as a platform for model building, wherein the CPU is i5-12500H2.50GHz, the installed memory is 16GB, the operating system is Windows 11-64 bit, the MATLAB R2022a version is used, and the distribution network test system is an IEEE 33-node distribution network containing DG;

[0075] Step 1: Encode the fault current information of each node, build a distribution network fault location model containing DG, debug and set the initial parameters of the ISABO algorithm;

[0076] Step 2, taking the IEEE33 system with DG as the research object, 8 fault conditions were set up to verify the proposed method by integrating various influencing factors;

[0077] Step 3, set the initial parameters of each comparison algorithm: the traditional SABO algorithm, the improved cuckoo algorithm (ICS), and the genetic algorithm (GA); perform 40 simulation tests on ISABO and the above three comparison algorithms under the set fault cases, and record the average results of the 40 times;

[0078] Step 4, using the average number of iterations and positioning accuracy as indicators, the results of the four algorithms are compared to verify the superiority of the proposed method;

[0079] Step 5: Figure 2 The switch 21 in the figure is changed into a tie switch, and the tie switch is set to be in the closed and open states respectively for six possible fault conditions of the distribution network;

[0080] Step 6, using the proposed fault location method under the six fault conditions, solve the specific section where the fault exists, and verify the adaptability and effectiveness of the method in the distribution network considering the state of the tie switch;

[0081] Step 1: Encode the fault current information of each node, build a distribution network fault location model containing DG, debug and set the initial parameters of the ISABO algorithm;

[0082] The above formula is used to build the basic model of fault location, and the parameters of the ISABO algorithm are set to: N = 50, T max =30;

[0083] Step 2, taking the IEEE33 system with DG as the research object, 8 fault conditions were set up to verify the proposed method by integrating various influencing factors;

[0084] by Figure 2 The 33-node distribution network structure with DG shown in the figure is the research object, where 1 to 33 are switch nodes; x1 to x 33 is the feeder section; DG1, DG2 and DG3 are three distributed power sources; K is the state of each DG switching switch (K = [K1, K2, K3]), which is 1 when DG is connected to the network and 0 when DG is disconnected from the network; Taking into account the randomness of the fault section, information distortion nodes and the different access conditions of each DG, this study sets 8 fault conditions, covering single-section fault, multi-section fault, node information distortion and other situations;

[0085] Step 3, set the initial parameters of each comparison algorithm: the traditional SABO algorithm, the improved cuckoo algorithm (ICS), and the genetic algorithm (GA); perform 40 simulation tests on ISABO and the above three comparison algorithms under the set fault cases, and record the average results of the 40 times;

[0086] Under these 8 fault conditions, the ISABO algorithm, the traditional SABO algorithm, the improved cuckoo algorithm (ICS) and the genetic algorithm (GA) were used for comparative tests; the population size of each algorithm was set to 100, and other related parameters were kept consistent with the original literature; 40 simulation tests were carried out in the given 8 conditions when the global optimal solution of each algorithm remained unchanged for 10 consecutive times;

[0087] Step 4, using the average number of iterations and positioning accuracy as indicators, the results of the four algorithms are compared to verify the superiority of the proposed method;

[0088] The average number of iterations and accuracy of each algorithm are shown in Tables 1 and 2;

[0089] Table 1 Comparison of average iteration times of each algorithm

[0090]

[0091] Table 2 Comparison of positioning accuracy of each algorithm

[0092]

[0093]

[0094] It can be seen from Table 1 and Table 2 that, under various fault conditions, the ISABO algorithm requires the least number of iterations to find the fault location. When locating a single-segment fault condition, the number of iterations is within 10; when locating a multi-segment fault condition, the number of iterations increases slightly, but the change is not large. On the contrary, the number of iterations of other algorithms increases significantly with the increase in the complexity of the working conditions; in addition, in terms of positioning accuracy, the accuracy of the ISABO algorithm is 100% under various working conditions, while the positioning accuracy of other algorithms decreases significantly with the increase in the number of fault sections and the difference in distortion information; in order to more intuitively see the difference in positioning accuracy of each algorithm, this study selects the accuracy rates obtained by Case 2, Case 3, Case 5, Case 6, Case 7, and Case 8, and draws them into a bar graph, as shown Figure 3 As shown in the figure, it can be seen that the positioning performance of the ISABO algorithm has excellent stability and good adaptability to various complex working conditions; therefore, the ISABO algorithm has good rapidity, accuracy and fault tolerance when applied to the problem of locating the fault section of the distribution network;

[0095] The number of iterations of each algorithm is guaranteed to be 100. Experiments are conducted on the three working conditions of Case 3, Case 6, and Case 8 in Table 1. The experimental comparison results of the four algorithms are as follows: Figure 4 As shown;

[0096] Depend on Figure 4It can be seen that in these three fault conditions, the ISABO algorithm tends to converge first. Thanks to a series of improvement measures introduced in the original algorithm, the convergence curve of the ISABO algorithm almost achieves a straight rise in the early stage of positioning, reflecting a strong traversal ability. As the fault conditions become worse and worse, in the later stage of the search, although ISABO may fall into the local optimum, the duration is very short. It can jump out of the local optimum at the fastest speed and move closer to the global optimum, while other algorithms fall into the local optimum for a relatively long time and cannot meet the requirements of fault location for rapidity. Therefore, the above comparison results show that the ISABO algorithm has a higher search efficiency to obtain high-quality solutions.

[0097] Step 5: Figure 2 The switch 21 in the figure is changed into a tie switch, and the tie switch is set to be in the closed and open states respectively for six possible fault conditions of the distribution network;

[0098] In order to meet the needs of operation and reduce fault losses, the distribution network sometimes involves the operation of tie switches. In order to study the impact of tie switch operation on fault location, this study Figure 2 A main power supply S2 is added to the 33 section of the network shown in the figure, and the switch 21 is changed to a tie switch. The modified network diagram is as follows Figure 5 As shown;

[0099] When the tie switch 21 is closed or open, simulation experiments are carried out using the ISABO algorithm under the designed six fault conditions.

[0100] Step 6, using the proposed fault location method under the six fault conditions, solve the specific section where the fault exists, and verify the adaptability and effectiveness of the method in the distribution network considering the state of the tie switch;

[0101] Simulation experiments were carried out under 6 fault conditions, and the experimental results are shown in Table 3;

[0102] Table 3 Experimental results considering tie switch operation

[0103]

[0104]

[0105] It can be seen from Table 3 that when the tie switch 21 is closed, no matter it is a single fault or a double fault, and no matter whether the uploaded information is distorted, the ISABO algorithm can always correctly locate the section where the fault is located, showing the accuracy of positioning; when the tie switch 21 is disconnected, when a short circuit fault occurs in any subnet or both subnets at the same time, the ISABO algorithm can still accurately find the fault section and is not affected by the distortion of individual information, showing good fault tolerance performance; therefore, the switch state of the tie switch does not affect the solution performance of the ISABO algorithm, and correct positioning results can be obtained in both cases; compared with the traditional positioning method, the present invention has obvious advantages in shortening the calculation time and improving the positioning accuracy, and provides a new idea and method for the development of distribution network fault positioning technology;

[0106] The above is an implementation example of the present invention, but the protection scope of the present invention is not limited to this example. Every technician familiar with the technical field is within the technical scope disclosed by the present invention. Any equivalent replacement or change according to the technical solution and inventive concept of the present invention should fall within the protection scope of the present invention.

Claims

1. A method for locating fault sections in a distribution network based on an improved subtraction optimizer algorithm considering dynamic switching of distributed generation, characterized in that The following steps are involved: Step 1: Encode the fault status of each section in the distribution network and the fault current information flowing through each node, and use the encoded data as the input of the positioning model; Step 2, taking into account the large-scale access of distributed generation and the dynamic switching of each DG switch, a switching function and evaluation function adapted to the large-scale access of DG are established; Step 3: Set the initialization parameters of the ISABO algorithm according to the specific distribution network structure, including the population number N, the maximum number of iterations T max As well as the upper and lower limits of the search range, based on the newly proposed chaotic mapping strategy, the position of each search particle in the initial population is obtained and its corresponding fitness function value is calculated; Step 4, start the iteration, and each particle in the search range continuously changes its position through the improved position update strategy (adaptive dynamic inertia weight), gradually approaching the global optimal value; Step 5: Check whether the position of each search particle exceeds the solution space. If so, perform regression processing on the out-of-bounds particles. Step 6: Calculate the fitness value of each search particle, select the current global optimal particle by comparison, use the mutation strategy that combines Cauchy distribution mutation and reverse learning to perturb the optimal particle, and check whether the fitness value corresponding to the perturbed particle is better; Step 7, determine whether the loop termination condition is met, and if so, output the current global optimal solution; Otherwise, return to step 3 and continue to perform corresponding operations; Step 8: According to the output global optimal solution, the specific location of the final fault section is obtained; Step 1: Encode the fault status of each section in the distribution network and the fault current information flowing through each node, and use the encoded data as the input of the positioning model; (1.1) Information characterization After a fault occurs in the active distribution network, the supervisory control and data acquisition system (SCADA) obtains the fault current information uploaded by the FTU installed at the switch node, and uses this information to start the algorithm to analyze, process, and locate the fault section; this study uses (i = 1, 2, ..., N) to represent the fault status information of the i-th section, where N is the total number of sections in the system, and the specific encoding method is: In the original single-power network, since there is no DG, the direction of the power flow is unidirectional, and the node status information only exists in two situations: when the node detects a fault current, the switch status information is 1; when the node does not flow through the fault current, the switch status information is 0. However, with the widespread access of distributed power sources, the network structure tends to be complex, and the direction of the power flow has also changed from the original unidirectional to bidirectional. In order to accurately determine the direction and source of the fault current when a fault occurs in the section, this study stipulates that the direction of the current flowing from the main power supply to the load side is the positive direction. Based on this stipulated positive direction, the switch status of each node is encoded, including the switch status S of the DG. j The code is: Step 2, taking into account the large-scale access of distributed generation and the dynamic switching of each DG switch, a switching function and evaluation function adapted to the large-scale access of DG are established; (2.1) Switching function In order to make full use of various known information, it is very important to establish the conversion between fault current information and section status. This conversion needs to be realized by a switch function. For a distribution network with only a main power supply, when a fault occurs at a certain point, the fault current always flows from the location of the power supply to the fault point, and the fault current will not flow through the switch from the fault point to the end of the feeder. Due to this characteristic, the fault location process is relatively easy. The specific switch function is: In the formula, is the expected value of the i-th switch state information; x j is the status information of a certain segment j in the area from node i to the end of this branch; ∏ is an OR operation; Different from the above network, the switching function value of the network containing DG is not only related to the state of the downstream section, but also closely related to the state of the upstream section and the switching state of each DG. Based on this situation, this study proposes a new switching function that adapts to the large-scale access of DG. The specific formula is as follows: In the formula, is the expected function value of the i-th switch; N1 and N2 are the total number of upstream and downstream sections of switch i; N DG is the total number of connected DGs; K j is the switching state of the jth DG downstream of switch i; x ju and x jd are the state information of the j-th feeder section in the upstream and downstream areas of switch i, respectively, thereby generating the expected fault current array; (2.2) Evaluation function Since the switching function is constructed based on the causal relationship between the system sections and nodes, it is feasible and effective to use the node expected value calculated by the switching function to approximate the actual fault current information of the node collected by the FTU; therefore, in order to construct this approximation model, it is crucial to establish a suitable objective function; in addition, while constructing the approximation model to find the minimum value, reasonable measures must be taken to avoid the misjudgment problem caused by multiple solutions for one value; this study introduces a weight coefficient ω to deal with the misjudgment problem, and the specific fitness function formula is as follows: In the formula, W is a very large number, which is taken as 100 here; S i is the actual fault current information collected by the device at node i, which is the input data of the algorithm; is the calculated expected function value of node i; ω is the weight factor, which is taken as 0.5 here for the convenience of subsequent calculation; x j is the status information of the jth segment, which is the variable of this formula; D is the total number of nodes in the network; d is the total number of segments contained in the network; this formula is used to quantify the deviation between the actual fault current information and the expected fault current information. The smaller the deviation, the more accurate the positioning; Step 3: Set the initialization parameters of the ISABO algorithm according to the specific distribution network structure, including the population number N, the maximum number of iterations T max As well as the upper and lower limits of the search range, based on the newly proposed chaotic mapping strategy, the position of each search particle in the initial population is obtained and its corresponding fitness function value is calculated; (3.1) Chaotic map initialization In the optimization process of the algorithm, the distribution of the initial population affects the balance between the algorithm's exploration space and development capabilities. A good population initialization can promote global search. The traditional SABO algorithm uses a random initialization strategy, which easily leads to uneven distribution of the initial population and is not conducive to covering more areas in the search space. Therefore, in order to introduce diversity in the initial stage of the population and help the algorithm explore a wider solution space, this study uses the Circle mapping strategy to initialize the population, and its expression is: In the formula, mod is the remainder function; Step 4, start the iteration, and each particle in the search range continuously changes its position through the improved position update strategy (adaptive dynamic inertia weight), gradually approaching the global optimal value; (4.1) Adaptive dynamic inertia weight Adding a dynamic inertia factor to the position update formula of the traditional SABO algorithm can make the algorithm focus on global search in the early stage of the search and perform fine search in the later stage of the search, which greatly improves the algorithm's optimization ability and solution speed. The corresponding formula is as follows: Where t is the current iteration number; T max is the maximum number of iterations set; α1 and α0 are two coefficients, which are taken as 0.9 and 0.5 respectively; Step 5: Check whether the position of each search particle exceeds the solution space. If so, perform regression processing on the out-of-bounds particles. (5.1) Regression processing of out-of-bounds individuals During the search process of the SABO algorithm, some particles may exceed the search boundary due to position updates or disturbances. In this case, if the position of the out-of-bounds particle is simply assigned to the boundary value of the search range, the out-of-bounds particles will gather at the boundary, resulting in uneven distribution of the particle population and affecting the algorithm's solution performance. In order to solve this problem, this study adopts a new boundary regression processing method, the specific formula is as follows: In the formula, ub and lb are the upper and lower boundary values ​​of the solution space; For particles that cross the boundary; is the particle after regression processing; Step 6: Calculate the fitness value of each search particle, select the current global optimal particle by comparison, use the mutation strategy that combines Cauchy distribution mutation and reverse learning to perturb the optimal particle, and check whether the fitness value corresponding to the perturbed particle is better; (6.1) Fusion of Cauchy distribution variation and reverse learning strategy In view of the disadvantage that the traditional SABO algorithm is prone to fall into local extreme values, this study incorporates the Cauchy mutation operator and reverse learning strategy into the original algorithm to improve the position of the optimal solution, thereby avoiding the "premature" phenomenon of the algorithm; the reverse learning strategy is to reversely learn the current solution to generate a reverse solution, and then select a better solution by comparison. Its expression is as follows: Where u and d represent the upper and lower limits of the search boundary respectively; r is a random number matrix of (0, 1) standard uniform distribution 1*d (d is the spatial dimension); X best is the original optimal solution; is the optimal solution after mutation; The random numbers generated by the Cauchy distribution can be used to perturb the optimal search particles to improve the global search capability of the algorithm and expand the search space. The formula is as follows: Where f(x) is the probability density function of the mutation strategy; is the multiplication symbol; cauchy(0,1) is the standard Cauchy variation operator; In order to improve the optimal solution of the original algorithm to the greatest extent, this study introduces the switching probability p, where p is a random number between 0.1 and 0.

5. Now a number is randomly generated. If the number is less than p, the optimal solution position is updated using formula (9), otherwise, formula (10) is used. Step 7, determine whether the loop termination condition is met, and if so, output the current global optimal solution; Otherwise, return to step 3 and continue to perform corresponding operations; Step 8: According to the output global optimal solution, the specific location of the final fault section is obtained; Based on the output global optimal solution and the specific topological structure of the distribution network, the specific section where the fault occurred is located; The above-mentioned method for locating faults in a distribution network containing distributed power sources based on an improved subtraction optimizer algorithm, the subtraction optimizer algorithm is an existing technology and is well known to technicians in this technical field.

2. A method for locating faults in a distribution network containing distributed power sources based on an improved subtraction optimizer algorithm according to claim 1, characterized in that: An improved subtraction optimizer algorithm is proposed, which introduces the improvement measures such as chaos map initialization, adaptive dynamic inertia weight, out-of-bounds individual regression processing, Cauchy distribution mutation and the integration of reverse learning strategy, and improves the initialization and update of the algorithm. Compared with the original algorithm, the improved algorithm has stronger convergence and better optimization ability. In addition, the algorithm can also be extended to other fields.

Images