Fuzzy multi-target fault intelligent self-healing method for direct-current power grid

Abstract

The invention discloses a fuzzy multi-target fault intelligent self-healing method for a direct-current power grid, and relates to the technical field of power distribution network control. The method comprises the following steps: constructing a multi-objective optimization model comprising five objective functions according to a power grid state after a fault; determining a weight vector of each objective function in the multi-objective optimization model by adopting an analytic hierarchy process, and constructing a comprehensive objective function; and carrying out optimization solution on the comprehensive objective function by adopting a cloud theory adaptive genetic algorithm, and outputting a load state vector which meets a constraint condition and enables the comprehensive objective function to be maximum as an optimal strategy for fault self-healing. According to the method, comprehensive optimization under multiple objectives of important load power supply maximization, non-important load recovery, minimum switching operation times, line load balance and the like can be realized.

Description

Technical Field

[0001] This invention relates to the field of power distribution network control technology, and in particular to an intelligent self-healing method for fuzzy multi-objective faults in DC power grids. Background Technology

[0002] As the core energy supply network for ship operations, the safe and stable operation of the ship's electrical system directly affects its navigation safety and mission execution capabilities. Because ship electrical systems operate in the complex marine environment, they experience frequent load fluctuations and a high failure rate. If a failure occurs and cannot be restored promptly and effectively, it could lead to a complete power outage, failure of critical equipment, or even a safety accident. Therefore, researching rapid and intelligent fault self-healing technologies to achieve rapid power restoration after a failure is of great significance for improving the survivability and reliability of ship electrical systems.

[0003] Traditional fault recovery methods often rely on human experience or simple rule-based judgments, and can only optimize for a single objective, making it difficult to achieve multi-objective collaborative optimization under multiple constraints. With the increasing intelligence level of power systems, fault recovery strategies based on optimization algorithms have gradually become a research hotspot. Genetic algorithms, due to their powerful global search capabilities, are widely used in power system fault recovery; however, their fixed crossover and mutation rates can easily lead to premature convergence or slow convergence speeds. While adaptive genetic algorithms can dynamically adjust parameters, their adaptability to environmental changes remains limited. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent self-healing method for fuzzy multi-objective faults in DC power grids, which aims to solve or improve at least one of the above-mentioned technical problems.

[0005] To achieve the above objectives, the present invention provides the following solution: A method for intelligent self-healing of fuzzy multi-objective faults in DC power grids includes: Based on the power grid status after a fault, a multi-objective optimization model is constructed, comprising five objective functions: a function representing the total power supply to critical loads, a function representing the total power supply to non-critical loads, a function representing the number of manual transfer switch operations, a function representing the number of automatic transfer switch operations, and a function representing the line load distribution imbalance. The weight vectors of each objective function in the multi-objective optimization model are determined using the analytic hierarchy process (AHP), and a comprehensive objective function is constructed. An adaptive genetic algorithm based on cloud theory is used to optimize the comprehensive objective function and output a load state vector that satisfies the constraints and maximizes the comprehensive objective function, serving as the optimal strategy for fault self-healing. The adaptive genetic algorithm based on cloud theory includes: dynamically adjusting the crossover rate and mutation rate based on the cloud model, employing binary competitive selection, single-point crossover, and basic bit mutation operations, and performing constraint verification and correction on individuals in the population during the iteration process; the constraints include radial topology constraints, line current constraints, and generator capacity constraints.

[0006] Optionally, the process of determining the weight vector of each objective function in the multi-objective optimization model using the analytic hierarchy process includes: First, construct the objective pairwise comparison matrix, then calculate the maximum eigenvalue and the corresponding eigenvector, and finally perform a consistency check and normalize to obtain the weight vector of each objective function in the multi-objective optimization model.

[0007] Optionally, the comprehensive objective function is expressed as: in, The normalized function corresponding to the total power supply to critical loads; The normalized function corresponding to the function representing the total power supply to non-critical loads; The normalization function corresponding to the function representing the number of manual switching operations; This is the normalization function corresponding to the function representing the number of automatic switching operations; This is the normalized function corresponding to the line load distribution imbalance representation function.

[0008] Optionally, the radial topological constraints specifically include: To avoid circulating currents and facilitate fault location and isolation, the ship's electrical network maintains a radial topology before and after fault recovery. The radial topology is determined as follows: if each load has at most one connected path, the ship's power distribution system is considered to be radial.

[0009] Optionally, the line current constraint specifically includes: Limit the line current to Within the operating range, among which, I i Let I be the current of the i-th line. imax This is the maximum value during normal operation.

[0010] Optionally, the generator capacity constraint specifically includes: Limit generator capacity to Within the operating range, among which, L ij For generatorG j The i-th load that supplies power, C j For generator G j Rated capacity.

[0011] Optionally, the process of using a cloud-theoretic adaptive genetic algorithm to optimize the comprehensive objective function includes: a. Given initial data, including load data, switch type, network structure before the fault, faulty line, population size, chromosome length, and maximum number of iterations; b. Establish an initial population P0, implant chromosomes corresponding to the load states before the fault into P0 which is mainly generated randomly, and determine whether the fault load is isolated or the fault path is broken for each individual in P0. If not, correct the individual. c. Check whether the ship distribution network structure corresponding to each body in P0 is radial; if not, correct it. d. Evaluate each individual in P0 individually: First, calculate the constraint violation value. Then, calculate the sum of the constraint violation values ​​of each line current and the sum of the constraint violation values ​​of each generator capacity. After normalizing the two, calculate the weighted sum to obtain the comprehensive constraint violation value. Next, determine whether the constraint violation value is greater than 0. If it is, set all target values ​​to 0. Otherwise, calculate the target values ​​one by one. Next, calculate the five objective function values, their corresponding normalized values, the combined objective function value, and the combined constraint violation value. Then, link the obtained objective function values, constraint violation values, combined objective function values, and combined constraint violation values ​​to the corresponding individuals in sequence. At this point, the individual length increases to N+O+C+2, where N is the number of variables to be calculated, O is the number of objectives, and C is the number of computable constraints. The resulting population after evaluation is denoted as P. t t represents the number of iterations; e. For population P t Perform selection, crossover, and mutation, and establish a progeny population P. t+1 The system employs binary competitive selection, single-point crossover, and basic bit mutation operations to constrain and correct individuals in the population during the iteration process. f. Detecting population P t+1 Check whether the distribution network structure of each ship corresponds to a radial pattern; if not, make corrections. g. Evaluate P using the same method as in step d. t+1 ; h. Detecting whether the algorithm has converged: If P t+1 =P t If convergence is achieved, proceed to step k; otherwise, proceed to step i. i. Update population Pt ; j. Update the iteration count t, and determine whether the maximum iteration count has been reached. If not, jump to step e; if so, jump to step k. k. Output results: From Pt+1 The individual corresponding to the maximum value of the comprehensive objective function is identified as the global optimal solution.

[0012] Optionally, during the iteration process, the offspring population P obtained after selection, crossover, and mutation is... t+1 All individuals in the system determine whether the faulty load has been isolated and whether the faulty path has been disconnected. If any of the judgment results are negative, the individual is corrected.

[0013] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects: This invention discloses a fuzzy multi-objective intelligent self-healing method for DC power grid faults. The method includes constructing a multi-objective optimization model comprising five objective functions based on the post-fault grid state. These objective functions include a function representing the total power supply to critical loads, a function representing the total power supply to non-critical loads, a function representing the number of manual transfer switch operations, a function representing the number of automatic transfer switch operations, and a function representing the line load distribution imbalance. The method employs the analytic hierarchy process (AHP) to determine the weight vectors of each objective function in the multi-objective optimization model and constructs a comprehensive objective function. A cloud-theory adaptive genetic algorithm is then used to optimize and solve the comprehensive objective function, outputting a load state vector that satisfies constraints and maximizes the comprehensive objective function, serving as the optimal strategy for fault self-healing. The cloud-theory adaptive genetic algorithm includes: dynamically adjusting the crossover and mutation rates based on the cloud model; employing binary competitive selection, single-point crossover, and basic bit mutation operations; and performing constraint verification and correction on individuals in the population during the iteration process. These constraints include radial topology constraints, line current constraints, and generator capacity constraints. This invention can achieve comprehensive optimization under multiple objectives, including maximizing power supply to critical loads, restoring non-critical loads, minimizing the number of switch operations, and balancing line loads. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a flowchart illustrating the intelligent self-healing method for fuzzy multi-objective faults in DC power grids according to the present invention. Figure 2 This is a schematic diagram of the AHP model structure for fault recovery in this embodiment; Figure 3 This is a schematic diagram of the ship's integrated electric power system in this embodiment; Figure 4 This is a schematic diagram of the ship's electrical system symbols in this embodiment; Figure 5(a) is a schematic diagram of the results of Case 1 when SGA is used in this embodiment; Figure 5(b) is a schematic diagram of the results of Case 1 when AGA is used in this embodiment; Figure 5(c) is a schematic diagram of the results of Case 1 when CAGA is used in this embodiment; Figure 6(a) is a schematic diagram of the results of Case 3 when SGA is used in this embodiment; Figure 6(b) is a schematic diagram of the results of Case 3 when AGA is used in this embodiment; Figure 6(c) is a schematic diagram of the results of Case 3 when CAGA is used in this embodiment; Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] The purpose of this invention is to provide an intelligent self-healing method for fuzzy multi-objective faults in DC power grids, which aims to solve or improve at least one of the above-mentioned technical problems.

[0018] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] like Figure 1 As shown, this invention provides an intelligent self-healing method for fuzzy multi-objective faults in DC power grids, comprising: Step 100: Based on the power grid status after the fault, construct a multi-objective optimization model including five objective functions; the objective functions include the total power supply of important loads, the total power supply of non-important loads, the number of manual transfer switch operations, the number of automatic transfer switch operations, and the line load distribution imbalance function.

[0020] a. Total power supply to critical loads f1(X).

[0021] Where f1(X) is the function representing the total power supply to critical loads after fault recovery; μ1(X) is the normalization function corresponding to f1(X); S Li Power supply for the i-th load; N 1R N1 is the set of important load labels for power supply after fault recovery; N2 is the set of all important load labels; the decision variable X is the load sequence state vector, X=[x1, x2, ..., x...]. N ], the i-th component x i Indicates the i-th load state, x i The values ​​0, 1, and 2 represent the i-th load being in a fault state, under normal power supply, and under standby power supply, respectively; i = 1, 2, ..., N (N is the total number of loads).

[0022] b. Total power supply for non-critical loads f2(X).

[0023] Where f2(X) is the function representing the total power supply to non-critical loads after fault recovery, μ2(X) is the normalization function corresponding to f2(X), and N2 R N1 represents the set of labels for non-critical loads that will be powered after the fault is recovered, and N2 represents the set of labels for all non-critical loads.

[0024] c. Number of manual switch operations f3(X).

[0025] Where f3(X) is the function representing the number of operations of the manual bus transfer (MBT) switch, μ3(X) is the normalization function corresponding to f3(X), and n m n represents the number of MBT operations. M This represents the total number of MBTs.

[0026] d. Number of automatic transfer switch operations f4(X).

[0027] Where f4(X) is the function representing the number of operations of the Automatic Bus Transfer (ABT) switch, μ4(X) is the normalization function corresponding to f4(X), and n a n represents the number of ABT operations. A This represents the total number of ABTs.

[0028] e. Line load distribution imbalance f5(X).

[0029] Where f5(X) is the function representing the line load distribution imbalance, N b S represents the total number of lines. Lb Let f5(X) be the total load supplied to the line, and let μ5(X) be the normalized function corresponding to f5(X).

[0030] Fault recovery operations will alter the power grid topology, causing fluctuations in the total load supplied to the lines, while S Lb The objective function f5(X) describes the degree of dynamic imbalance in the line load distribution, thus objectively reflecting the actual balance of load distribution after the fault is restored.

[0031] Step 200: The weight vectors of each objective function in the multi-objective optimization model are determined using the analytic hierarchy process (AHP), and a comprehensive objective function is constructed. Specifically: First, a pairwise comparison matrix of objectives is constructed, then the largest eigenvalue and its corresponding eigenvector are calculated, and finally, after a consistency check, the weight vectors of each objective function in the multi-objective optimization model are obtained through normalization.

[0032] First, construct based on each objective function: Where f(X) is the comprehensive objective function, w i For μ i The weight factor corresponding to (X) is represented by the weight vector W = [w1, w2, w3, w4, w5]. T Using AHP to achieve multi-objective synthesis, the AHP model for ship electrical system fault recovery is as follows: Figure 2 As shown.

[0033] The target pairwise comparison matrix A is constructed using a scale of 1 to 9 as shown in the following formula: The largest eigenvalue of A is λ max =5.0681, corresponding feature vector v=[0.7868, 0.4935, 0.3007, 0.1828, 0.1161]T, number of targets in A n=5, consistency index δ CI =(λ max -n) / (n-1)=0.0170, average consistency index δ RI =1.12, Consistency ratio δ CR =δ CI / δ RI=0.0152 < 0.1, therefore it meets the consistency requirement. Normalizing v yields the weight vector W = [0.4185, 0.2625, 0.1599, 0.0973, 0.0618]. T .

[0034] Therefore, the comprehensive objective function can be further expressed as follows: in, The normalized function corresponding to the total power supply to critical loads; The normalized function corresponding to the function representing the total power supply to non-critical loads; The normalization function corresponding to the function representing the number of manual switching operations; This is the normalization function corresponding to the function representing the number of automatic switching operations; This is the normalized function corresponding to the function representing the load distribution imbalance of the line. Furthermore, the X corresponding to the maximum value of f(X) is the globally optimal solution.

[0035] Step 300: The cloud theory adaptive genetic algorithm is used to optimize the comprehensive objective function and output the load state vector that satisfies the constraints and maximizes the comprehensive objective function, which serves as the optimal strategy for fault self-healing. The cloud theory adaptive genetic algorithm includes: dynamically adjusting the crossover rate and mutation rate based on the cloud model, using binary competitive selection, single-point crossover, and basic bit mutation operations, and performing constraint verification and correction on individuals in the population during the iteration process. The constraints include radial topology constraints, line current constraints, and generator capacity constraints.

[0036] As a specific implementation method for this step, the constraints are first constructed: Radial topological constraints, specifically including: To avoid circulating currents and facilitate fault location and isolation, the ship's electrical network maintains a radial topology before and after fault recovery. The radial topology is determined as follows: if each load has at most one connected path, the ship's power distribution system is considered to be radial.

[0037] Line current constraints specifically include: Limit the line current to Within the operating range, among which, I i Let I be the current of the i-th line. imax This is the maximum value during normal operation.

[0038] The generator capacity constraints are as follows: Limit generator capacity to Within the operating range, among which, Lij For generator G j The i-th load that supplies power, C j For generator G j Rated capacity.

[0039] Then, the cloud theory adaptive genetic algorithm CAGA is used to optimize and solve the comprehensive objective function: a. Given initial data, including load data (load size and priority), switch type (MBT or ABT), network structure before the fault, faulty line, population size, chromosome length, and maximum number of iterations; b. Establish an initial population P0. Implant chromosomes corresponding to the load states before the fault into P0, which is mainly generated randomly. For each individual in P0, determine whether the fault load has been isolated and whether the fault path (including normal and backup power supply paths) has been disconnected. If not, make corrections to the individual.

[0040] c. Check whether the ship distribution network structure corresponding to each body in P0 is radial; if not, correct it. d. Evaluate each individual in P0 individually: First, calculate the constraint violation value, then calculate the sum of the constraint violation values ​​of each line current and the sum of the constraint violation values ​​of each generator capacity. After normalizing the two, calculate the weighted sum to obtain the comprehensive constraint violation value. Next, determine whether the constraint violation value is greater than 0. If it is, set all target values ​​to 0; otherwise, calculate the target values ​​one by one. Next, calculate the five objective function values, their corresponding normalized values, the combined objective function value, and the combined constraint violation value. Then, link the obtained objective function values, constraint violation values, combined objective function values, and combined constraint violation values ​​to the corresponding individuals in sequence. At this point, the individual length increases to N+O+C+2, where N is the number of variables to be calculated, O is the number of objectives, and C is the number of computable constraints. The resulting population after evaluation is denoted as P. t (i.e., population of generation t), where t represents the number of iterations; e. For population P t Perform selection, crossover, and mutation, and establish a progeny population P. t+1 (i.e., generation t+1), employing binary tournament selection (BTS), one-point crossover (OC), and simple mutation (SM) operations, the population individuals are constrained, verified, and corrected during the iteration process; during the iteration process, the offspring population P obtained after selection, crossover, and mutation is... t+1All individuals in the system determine whether the faulty load has been isolated and whether the faulty path has been disconnected. If any of the judgment results are negative, the individual is corrected. f. Detecting population P t+1 Check whether the distribution network structure of each ship corresponds to a radial pattern; if not, make corrections. g. Evaluate P using the same method as in step d. t+1 ; h. Detecting whether the algorithm has converged: If P t+1 =P t If convergence is achieved, proceed to step k; otherwise, proceed to step i. i. Update population P t ; j. Update the iteration count t and determine whether the maximum iteration count has been reached. If not, jump to step e; otherwise, jump to step k.

[0041] k. Output results: From Pt+1 The individual corresponding to the maximum value of the comprehensive objective function is identified as the global optimal solution.

[0042] The descriptions of each operator are as follows: The BTS operator is described below.

[0043] Definition 1: Let X i With X j Compete and choose X i Then one of the following conditions is true: Condition 1: v(X) i ) <v(X j ); Condition 2: v(X) i )=v(X j )=0 and f(X) i )>f(X j ).

[0044] In Definition 1, v(X) i ), v(X) j ) are decision variables X i X j The comprehensive constraint violation value; f(X) i f(X) j ) are decision variables X i X j The comprehensive objective function value.

[0045] OC operator crossover rate P c The following is confirmed: Among them, f c The parent with the larger overall objective function value among the participating cross-parent pairs; f c avg and f c max E represents the average and maximum values ​​of the integrated objective function for all individuals in the mating pool used for crossover; x E represents the mathematical expectation. n Entropy; H e This is hyperentropy, i.e., entropy E. n Entropy; E n ′~N(E n H e randn(E) n H e The expected value of the generated mathematical expression is E. x The standard deviation is H e Normal random numbers; k1 and k2 are constants; c1 and c2 are control parameters.

[0046] SM operator mutation rate P m The method for determining the crossover rate P is similar. c , but f c f c avg f c max k1, k2, c1, and c2 need to be replaced with f accordingly. m f m avg f m max k3, k4, c3, c4. It is worth noting that since the population changes after crossover, the resulting population needs to be re-evaluated before performing mutation.

[0047] Based on the above technical solution, the following example is provided for verification.

[0048] The fault recovery process was implemented using Matlab and tested on four different examples of a ship's integrated power system. As shown in the table, all loads are constant power loads. It is assumed that all loads were powered via normal paths before the fault. In the following three genetic algorithms, the population size is set to 100 and the maximum number of generations is set to 200. For simplicity, it is assumed that the current amplitude of branches other than the generator branch is unrestricted. Therefore, only three constraints need to be considered: generator capacity, connectivity, and radial structure.

[0049] The current amplitude and priority level of each load in the system are set as shown in Table 1.

[0050] In Case 1, branches 19 and 57 are faulty branches. The capacities of the four generators in the system are: cg1=855A; cg2=959A; cg3=799A; cg4=770A.

[0051] After the fault, the loads L5 and L 11 Disconnected from their respective normal power supply paths, the corresponding individuals are [11110 1111101111 11111].

[0052] Case 2 is the same as Case 1 except for the capacity of the four generators. cg1=700A, cg2=734A, cg3=699A, cg4=700A.

[0053] Case 3 is the same as Case 1 except that the faulty branches are branches 10 and 63 of the given system. After the fault, L3 and L... 13 The normal power supply path was disconnected, L 12 Power outage, L 10 The backup power supply path has been disconnected. The corresponding individual is [1101111111 10011 11111].

[0054] Case 4 is the same as Case 3 except for the four generator capacities. cg1=700A, cg2=734A, cg3=699A, cg4=700A.

[0055] The comparison of the fault recovery results obtained by SGA, AGA and CAGA for Case 1 and Case 3 is shown in Tables 2-5.

[0056] Table 2 Parameter settings for the three solutions

[0057] Table 3 Comparison of Fault Recovery Results in Case 1

[0058] Table 4 Comparison of Fault Recovery Results in Case 3

[0059] Table 5. Fault recovery results obtained from CAGA

[0060] The optimal individual obtained by CAGA satisfies the two objectives mentioned above, namely, the maximum power supply load and the minimum number of switching operations, and takes into account the priority of load and switches.

[0061] 2) Intelligent self-healing of faults based on cloud theory adaptive genetic algorithm This embodiment implements the above-mentioned fault recovery strategy in the MATLAB environment and verifies it using four case studies (Case 1-4). (Ship electrical system diagram) Figure 3 In this configuration, the MSB (Main Switch Board) is the main switchboard, and the LCP (Load Center Panel) is the load center. The maximum capacity of generator G1 is 770.0A, generator G2 is 807.4A, generator G3 is 768.9A, and generator G4 is 770.0A. In the load current amplitude and priority settings, "1" indicates high priority and "0" indicates low priority. In the transfer switch types, "1" indicates ABT and "0" indicates MBT. i For the i-th load, the symbolic representation of the ship's electrical system is provided as follows: Figure 4 As shown in Table 6, the current amplitude, priority, and switch type of each load are as follows.

[0062] Table 6. Load Current Amplitude, Priority, and Switch Type

[0063] The algorithm parameters are set as follows: P in SGA c =0.85, P m =0.15; in AGA, k1=0.6, k2=0.9, k3=0.2, k4=0.3; in CAGA, c1=2.9, c2=10, c3=3.0, c4=10, k1=0.9, k2=0.9, k3=0.2, k4=0.2. In the above three algorithms, the population size is 50, the chromosome length is 20, and the maximum number of iterations is 200. It is assumed that all loads were powered via normal paths before the fault, and that the line current capacity was unrestricted.

[0064] Let the faulty lines in Case 1 be 10 and 63; the faulty lines in Case 2 be 19 and 57; the faulty lines in Case 3 be 36 and 40; and the faulty lines in Case 4 be 35 and 60. It can be seen that the fault condition in Case 1 is L. 12 Power outage, L3 and L 13 Normal power supply path damaged, L 10 The backup power supply path is damaged; Case 2 fault conditions are L5 and L 11The normal power supply path is damaged; Case 3 fault condition is L7 power failure and L8 normal power supply path damage; Case 4 fault condition is L7 power failure and L8 and L 11 The normal power supply path is damaged, and the L5 backup power supply path is damaged.

[0065] In the fault recovery results of Cases 1-4, the constraint violation values ​​are all 0. Looking at the comprehensive objective function value f(X), except for Case 4 where the AGA is smaller than the SGA, the CAGA is the largest and the SGA is the smallest in the other cases. The results for Case 1 are shown in Figures 5(a)-5(c), and the results for Case 3 are shown in Figures 6(a)-6(c).

[0066] From the perspective of the optimal solution X, for Case 1, the decision schemes given by SGA, AGA, and CAGA are L, ... 12 Uninstall, L3 and L 13 Recovered from backup path, L5, L8, L 16 L 20 Power was switched to the backup path, L 12 Uninstall, L3 and L 13 Recovered from backup path, L1, L6, L 17 Power was switched to the backup path, L 12 Uninstall, L3 and L 13 Recovered from backup path, L6, L 17 Power is switched to the backup path; for Case 2, the decision schemes given by SGA, AGA, and CAGA are L5, L... 11 Uninstall, L5 and L 11 Recovered from backup path, L3, L6, L8, L 17 Power was switched to the backup path, L5, L 11 Recovered from backup path, L3, L8, L 10 L 17 Power was switched to an alternate path; Cases 3 and 4 will not be discussed further. Of these schemes, except for Case 3 where AGA is inferior to SGA, the other examples show CAGA as the optimal choice and SGA as the worst. The objective function values ​​f1(X) to f5(X) essentially confirm the judgment regarding the superiority or inferiority of the above decision schemes.

[0067] It should be noted that in Case 3, although the overall objective f(X) obtained by AGA is better than that of SGA, for the critical objective f1(X), AGA is inferior to SGA. Therefore, the decision proposed by AGA is considered inferior to that of SGA. Conversely, in Case 4, although the overall objective f(X) obtained by AGA is inferior to that of SGA, for the critical objective f1(X), AGA is superior to SGA. Therefore, the decision proposed by AGA is considered superior to that of SGA. This also reflects the limitations of using a multi-objective compromise strategy that integrates objectives.

[0068] The results of the four examples are shown in Table 7.

[0069] Table 7 Results of the four examples

[0070] To test the convergence performance of the algorithms, Case 1 was used as an example. The population size was 100, the maximum number of iterations was 200, and other parameters remained unchanged. The solution with f(X) = 0.9010 was considered the global optimum, and each algorithm was run 10 times consecutively. It can be seen that the probabilities of SGA, AGA, and CAGA converging to the global optimum are 30%, 20%, and 90%, respectively; the average minimum number of iterations to converge to the global optimum are 57, 74.5, and 28.6, respectively.

[0071] The minimum number of iterations required for Case 1 to converge to the global optimum is shown in Table 8.

[0072] Table 8 shows the minimum number of iterations required for Case 1 to converge to the global optimum.

[0073] This embodiment provides examples Case 1 and 3, which yield the population P after 100 iterations. 100 The results show f1(X)~f5(X), f(X) and P 100The relationship between the individual indices n. For f1(X), SGA and CAGA are slightly better than AGA; for f2(X), the three are basically the same; for f3(X), AGA is the best and SGA is the worst; for f4(X), AGA is the best and SGA is the worst; for f5(X), the three are basically the same; for f(X), CAGA is better than SGA and AGA; for constraint violation, the three are the same, with only 1 individual violating the constraint. For f1(X), SGA and CAGA are significantly better than AGA; for f2(X), CAGA is the best and SGA is the worst; for f3(X), CAGA is better than SGA and AGA; for f4(X), AGA is better than SGA and CAGA; for f5(X), SGA and CAGA are better than AGA; for f(X), CAGA is slightly better than SGA and AGA; for constraint violation, CAGA is the best, with only 1 individual violating the constraint, AGA is next, with 2 individuals violating the constraint, and SGA is the worst, with 3 individuals violating the constraint.

[0074] During the solution process of the example, it was found that, in addition to the crossover rate P c and the rate of variation P m In addition, population size is a major factor affecting the quality of fault recovery. A population that is too small will make it difficult to select individuals in the mating pool, while a population that is too large will lead to slow algorithm convergence.

[0075] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0076] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A DC grid fuzzy multi-objective fault intelligent self-healing method, characterized in that, The application relates to a method for realizing fault self-healing of a ship power grid. According to a post-fault power grid state, a multi-objective optimization model comprising five objective functions is constructed; the objective functions comprise an important load power supply total amount representation function, a non-important load power supply total amount representation function, a manual transfer switch operation number representation function, an automatic transfer switch operation number representation function and a line load distribution imbalance degree representation function; An analytic hierarchy process is used to determine the weight vector of each objective function in the multi-objective optimization model, and a comprehensive objective function is constructed; A cloud theory adaptive genetic algorithm is used to optimize and solve the comprehensive objective function, and a load state vector satisfying a constraint condition and making the comprehensive objective function maximum is output as an optimal strategy for fault self-healing; the cloud theory adaptive genetic algorithm comprises the following steps: based on a cloud model, a cross rate and a mutation rate are dynamically adjusted, binary competitive selection, single-point cross and basic bit mutation operations are adopted, and population individuals are subjected to constraint checking and correction in an iteration process; the constraint condition comprises a radial topological structure constraint, a line current constraint and a generator capacity constraint.

2. The DC grid fuzzy multi-objective fault intelligent self-healing method according to claim 1, characterized in that, The weight vector of each objective function in the multi-objective optimization model is determined by using the analytic hierarchy process, and the specific process comprises the following steps: First, a target pair comparison matrix is constructed, then the maximum eigenvalue and the corresponding eigenvector are calculated, and finally the weight vector of each objective function in the multi-objective optimization model is obtained through normalization after consistency checking.

3. The DC grid fuzzy multi-objective fault intelligent self-healing method according to claim 1, characterized in that, The comprehensive objective function is represented as: wherein, is a normalized function corresponding to the total amount of power supply for important loads representation function; is a normalized function corresponding to the total amount of power supply for non-important loads representation function; is a normalized function corresponding to the manual transfer switch operation number representation function; is a normalized function corresponding to the automatic transfer switch operation number representation function; is a normalized function corresponding to the line load distribution imbalance degree representation function.

4. The DC grid fuzzy multi-objective fault intelligent self-healing method according to claim 1, characterized in that, The radial topological structure constraint specifically comprises the following steps: In order to avoid circulating current and facilitate fault positioning and isolation, the ship power grid maintains a radial topological structure before and after fault recovery; the judgment of the radial topological structure is as follows: if at most only one path of each load is connected, the ship power distribution system is determined to be radial.

5. The DC grid fuzzy multi-objective fault intelligent self-healing method according to claim 1, characterized in that, The line current constraint specifically comprises the following steps: to limit the line current to an operating range, wherein I i is the i-th line current, I imax is the maximum value at normal operation.

6. The DC grid fuzzy multi-objective fault intelligent self-healing method of claim 1, wherein, The generator capacity constraint is specifically as follows: The generator capacity is limited to an operating range, wherein L ij is the rated capacity of the generator G j the i-th load supplied by the generator, C j is the rated capacity of the generator G j .

7. The DC grid fuzzy multi-objective fault intelligent self-healing method of claim 1, wherein, The cloud theory adaptive genetic algorithm is used to optimize and solve the comprehensive objective function, and the specific process comprises the following steps: a. initial data is given, including load data, switch type, network structure before fault, fault line, population size, chromosome length and maximum iteration number; b. an initial population P0 is established, chromosomes corresponding to load states before fault are implanted in P0 which is mainly generated randomly, and whether the fault load is isolated and the fault path is disconnected is judged for each individual in P0, and the individual is modified if not; c. whether the ship power distribution structure corresponding to each individual in P0 is radial is detected, and the individual is modified if not; d. each individual in P0 is evaluated one by one: First, constraint violation values are calculated, the sum of line current constraint violation values and the sum of generator capacity constraint violation values are calculated, the two are normalized, and a weighted sum is calculated to obtain a comprehensive constraint violation value; then, whether the constraint violation value is greater than 0 is judged, all objective values are set to 0 if yes, and the objective values are calculated one by one if no; The five objective function values, the corresponding normalized values, the comprehensive objective function values and the comprehensive constraint violation values are calculated respectively, and then the obtained objective function values, constraint violation values, comprehensive objective function values and comprehensive constraint violation values are sequentially linked to the corresponding individuals, at this time the length of the individual is increased to N+O+C+2, wherein N is the number of variables to be calculated, O is the number of objectives, C is the number of computable constraints, and the population obtained after evaluation is recorded as P t , t represents the number of iterations; e. Selecting, crossing and mutating, and establishing the offspring population P t Selecting, crossing and mutating, and establishing the offspring population P t+1 , using binary competitive selection, single-point crossing and basic bit mutation operations, and performing constraint checking and correction on the population individuals in the iteration process; f. detecting the population P t+1 whether the network structure of each ship corresponds to a radial structure, and if not, correcting it; g. The P is evaluated using the same method as in step d t+1 ; h. Check if the algorithm converges: if P t+1 = P t , then consider it has converged, jump to step k, if it is not converged, jump to step i; i. updating the population P t ; j. the iteration number t is updated, and whether the maximum iteration number is reached is judged, step e is jumped to if not, and step k is jumped to if yes; k. Output result: find the individual corresponding to the maximum value of the comprehensive objective function from Pt+1 global optimal solution.

8. The DC grid fuzzy multi-objective fault intelligent self-healing method according to claim 7, characterized in that, In the iteration process, all individuals in the offspring population P t+1 obtained after selection, crossover and mutation are judged whether the fault load is isolated and whether the fault path is disconnected, and if there is any negative judgment result, the individual is modified.

Images