CN114444395B - Quantum variation multi-universe optimized power supply line fault identification method - Google Patents

Abstract

A power supply line fault identification method based on quantum mutation multiverse optimization relates to the field of power supply line fault identification. A quantum particle swarm algorithm is introduced into a traditional multiverse optimization algorithm, and an average optimal position of particles is introduced into the algorithm, so that the algorithm has better convergence accuracy and speed. A Cauchy-Gauss mutation strategy is adopted to solve the problem of rapid assimilation of individuals and local optimal stagnation in the late iteration of the traditional multiverse optimization algorithm. Short-circuit faults are effectively identified, thereby providing maintenance personnel with good fault data information, timely removing the short-circuit fault of the line, avoiding the expansion of the accident, and ensuring the stable operation of the power system.

Description

A quantum variation multiverse optimized power supply line fault identification method

Technical Field

The present invention relates to the field of power supply line fault identification, and in particular to a power supply line fault identification method optimized by quantum variation multiverse.

Background Art

Coal is still the most important energy source in China. Coal is the main energy consumed by the people in my country. Among them, coal mine power supply lines are one of the parts with the highest fault incidence rate, which is mainly affected by external forces, line aging, etc. In order to speed up the speed of the protection device to resume normal operation after clearing the fault, it is very important to determine the fault location as soon as possible. In mine power supply lines, visual inspection is very time-consuming, delaying the maintenance of faulty lines. On the other hand, the accuracy of fault location methods depends on the quality of data acquisition and processing, including previous fault detection and classification results. Fault classification is the last step of fault identification and the most important step. Commonly used fault classification methods include random forests, support vector machines, extreme learning machines, kernel extreme learning machines, etc. Random forests have been proven to be over-fitting in some classification or regression problems with large noise. For data with attributes with different values, attributes with more value divisions will have a greater impact on random forests, so the attribute weights produced by random forests on such data are unreliable. Support vector machines are difficult to implement for large-scale training samples. The space consumption of support vector machines is mainly to store training samples and kernel matrices. Since support vector machines use quadratic programming to solve support vectors, and solving quadratic programming will involve the calculation of m-order matrices, when the number m is large, the storage and calculation of the matrix will consume a lot of machine memory and computing time. Support vector machines are difficult to solve multi-classification problems. The classic support vector machine algorithm only gives two-class classification algorithms, but in practical applications, multi-class classification problems are generally solved. Extreme learning machine is a simple, easy-to-use and effective single hidden layer feedforward neural network learning algorithm. Traditional neural network learning algorithms require a large number of network training parameters to be set manually, and it is easy to produce local optimal solutions. Extreme learning machines only need to set the number of hidden layer nodes of the network. During the execution of the algorithm, there is no need to adjust the input weights of the network and the bias of the hidden element, and it produces a unique optimal solution, so it has the advantages of fast learning speed and good generalization performance. The kernel extreme learning machine is an improved algorithm based on the extreme learning machine and combined with the kernel function. The kernel extreme learning machine can improve the prediction performance of the model while retaining the advantages of the extreme learning machine. Compared with the traditional training method, the kernel extreme learning machine has the advantages of fast learning rate and good generalization performance.

Summary of the invention

In order to solve the shortcomings of the existing technology, a power supply line fault identification method based on quantum variation multiverse optimization is provided;

The technical solution adopted by the present invention is:

A method for identifying power supply line faults based on quantum variation multiverse optimization comprises the following steps:

S1: Collecting transmission line fault voltage signals;

S2: performing variational modal decomposition on the fault voltage signal; selecting the mean of the instantaneous frequency of the modal component after variational modal decomposition, drawing a standardized instantaneous frequency mean curve, and obtaining the optimal K value in the variational modal decomposition algorithm;

S3: Decompose the fault voltage signal using the VMD method to obtain the IMF component; then use PE to calculate the multi-scale entropy value of each intrinsic mode component;

S4: The calculated permutation column entropy values of each eigenmode component are used to form a eigenvector;

S5: The quantum particle swarm algorithm is used in the standard multiverse optimization algorithm, and the average optimal position of particles Av(t) is introduced; the Cauchy-Gauss mutation strategy is introduced to construct the quantum mutation multiverse optimization algorithm;

The operation mode of the multiverse optimization algorithm is shown in formula (1):

Among them, _Xj is the jth variable of the current optimal universe, _ubj is the jth maximum value of the variable, and _lbj is the jth minimum value of the variable. is the jth variable of the i-th universe, r ₂ , r ₃ , r ₄ are all random numbers between 0 and 1;

WEP represents the probability coefficient of wormhole existence, as shown in formula (2):

Where min is the minimum WEP value; max is the maximum WEP value; l is the current iteration number; L is the maximum iteration number;

TDR represents the travel distance rate coefficient, as shown in formula (3)

Among them, p defines the detection speed that changes with the number of iterations. The higher the p value, the faster the local detection speed and the shorter the time.

The quantum particle swarm algorithm is introduced into the multiverse optimization algorithm. Each particle is assumed to have quantum behavior. For particle i, its attractive potential is represented by p _i (p _i,1 ,p _i,2 ,…,p _i,n ). Considering that the particle position x _i,j changes with time and its state is described by the wave function, the particle evolution equation in the quantum particle swarm algorithm based on the δ potential well in the N-dimensional search space is shown in formula (4):

Among them, pi _,j represents the coordinates of the attractor _pi in the j dimension, Li _,j is the characteristic search length, and ui _,j ~U(0,1) are uniformly distributed random numbers; the average optimal position of particles Av(t) is introduced in the algorithm, which represents the average value of the optimal positions of all particles _Li,j = 2γ·| _Avj (t)-xi _,j (t)|, equation (4) is changed to equation (5):

Where γ represents the contraction-expansion factor;

The Cauchy-Gauss mutation strategy is adopted to select the individual with the best current fitness for mutation, and then compare its position before and after mutation, and select the optimal position to substitute for the next iteration, as shown in equations (6)-(8):

in, represents the position of the optimal individual after mutation; σ ² represents the standard deviation of the Cauchy-Gaussian mutation strategy; cauchy(0,σ ² ) is a random variable that satisfies the Cauchy distribution; Gaus(s0,σ ² ) is a random variable that satisfies the Gaussian distribution; and Where t represents the current iteration number, T _max represents the maximum iteration number; λ ₁ and λ ₂ are dynamic parameters that are adaptively adjusted with the iteration number. During the optimization process, λ ₁ gradually decreases and λ ₂ gradually increases;

S6: Optimize the regularization coefficient and kernel function parameters of the kernel extreme learning machine using the quantum mutation multiverse algorithm;

S6.1: Randomly generate initial positions to form an initial search population;

S6.2: In order to improve the performance of the native multiverse algorithm, the quantum particle swarm algorithm is introduced to accelerate the convergence speed;

S6.3: Use the Cauchy-Gauss mutation strategy to select the individual with the best current fitness for mutation, so that the algorithm can jump out of the local optimum;

S6.4: Update the optimal position of individual particles and the group, check whether the algorithm has reached the maximum number of iterations, and if so, output the optimal kernel extreme learning machine parameters; if not, return to S6.2 and continue the next iteration operation;

S7: Construct a prediction and classification model of line fault modal decomposition entropy and quantum mutation multiverse optimized nuclear extreme learning machine to identify the type of transmission line short-circuit fault.

Beneficial technical effects

(1) The quantum particle swarm algorithm is introduced, and the average optimal position of particles Av(t) is introduced into the algorithm, so that the algorithm has better convergence accuracy and speed.

(2) The Cauchy-Gauss mutation strategy is used to solve the problem of rapid assimilation of individuals and stagnation of local optimality in the late iteration of the traditional multiverse optimization algorithm.

(3) A quantum variation multiverse optimized power supply line fault identification method is used to effectively identify short-circuit faults, thereby providing maintenance personnel with good fault data information, timely removing the short-circuit faults in the line, avoiding the expansion of accidents, and ensuring the stable operation of the power system.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a flow chart of a method for identifying power line faults using quantum variation multiverse optimization provided by an embodiment of the present invention;

FIG2 is a flowchart of an execution of a quantum mutation multiverse optimization algorithm provided by an embodiment of the present invention.

DETAILED DESCRIPTION

The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments;

A method for identifying power line faults based on quantum variation multiverse optimization, as shown in FIG1 , comprises the following steps:

S1: Collecting transmission line fault voltage signals;

S2: performing variational modal decomposition on the fault voltage signal; selecting the mean of the instantaneous frequency of the modal component after variational modal decomposition, drawing a standardized instantaneous frequency mean curve, and obtaining the optimal K value in the variational modal decomposition algorithm;

S3: Decompose the fault voltage signal using the VMD method to obtain the IMF component; then use PE to calculate the multi-scale entropy value of each intrinsic mode component;

S4: The calculated permutation column entropy values of each eigenmode component are used to form a eigenvector;

S5: For example, a quantum particle swarm algorithm is used in a standard multiverse optimization algorithm, and the average optimal position of particles Av(t) is introduced; a Cauchy-Gauss mutation strategy is introduced to construct a quantum mutation multiverse optimization algorithm, as shown in FIG2 ;

The operation mode of the multiverse optimization algorithm is shown in formula (1):

Among them, _Xj is the jth variable of the current optimal universe, _ubj is the jth maximum value of the variable, and _lbj is the jth minimum value of the variable. is the jth variable of the i-th universe, r ₂ , r ₃ , r ₄ are all random numbers between 0 and 1;

WEP represents the probability coefficient of wormhole existence, as shown in formula (2):

Where min is the minimum WEP value; max is the maximum WEP value; l is the current iteration number; L is the maximum iteration number;

TDR represents the travel distance rate coefficient, as shown in formula (3)

Among them, p defines the detection speed that changes with the number of iterations. The higher the p value, the faster the local detection speed and the shorter the time.

The quantum particle swarm algorithm is introduced into the multiverse optimization algorithm. Each particle is assumed to have quantum behavior. For particle i, its attractive potential is represented by p _i (p _i,1 ,p _i,2 ,…,p _i,n ). Considering that the particle position x _i,j changes with time and its state is described by the wave function, the particle evolution equation in the quantum particle swarm algorithm based on the δ potential well in the N-dimensional search space is shown in formula (4):

Among them, pi _,j represents the coordinates of the attractor _pi in the j dimension, Li _,j is the characteristic search length, and ui _,j ~U(0,1) are uniformly distributed random numbers; the average optimal position of particles Av(t) is introduced in the algorithm, which represents the average value of the optimal positions of all particles _Li,j = 2γ·| _Avj (t)-xi _,j (t)|, equation (4) is changed to equation (5):

Where γ represents the contraction-expansion factor;

The Cauchy-Gauss mutation strategy is adopted to select the individual with the best current fitness for mutation, and then compare its position before and after mutation, and select the optimal position to substitute for the next iteration, as shown in equations (6)-(8):

in, represents the position of the optimal individual after mutation; σ ² represents the standard deviation of the Cauchy-Gaussian mutation strategy; cauchy(0,σ ² ) is a random variable that satisfies the Cauchy distribution; Gaus(s0,σ ² ) is a random variable that satisfies the Gaussian distribution; and Where t represents the current iteration number, T _max represents the maximum iteration number; λ ₁ and λ ₂ are dynamic parameters that are adaptively adjusted with the iteration number. During the optimization process, λ ₁ gradually decreases and λ ₂ gradually increases;

S6: Optimize the regularization coefficient and kernel function parameters of the kernel extreme learning machine using the quantum mutation multiverse algorithm;

S6.1: Randomly generate initial positions to form an initial search population;

S6.2: In order to improve the performance of the native multiverse algorithm, the quantum particle swarm algorithm is introduced to accelerate the convergence speed;

S6.3: Use the Cauchy-Gauss mutation strategy to select the individual with the best current fitness for mutation, so that the algorithm can jump out of the local optimum;

S6.4: Update the optimal position of individual particles and the group, check whether the algorithm has reached the maximum number of iterations, and if so, output the optimal kernel extreme learning machine parameters; if not, return to S6.2 and continue the next iteration operation;

S7: Construct a prediction and classification model of line fault modal decomposition entropy and quantum mutation multiverse optimized nuclear extreme learning machine to identify the type of transmission line short-circuit fault.

In this embodiment, in order to verify the effective identification of several typical short-circuit faults of the transmission line under different fault phase angles, the short-circuit fault voltage data of the transmission line when single-phase ground short circuit, two-phase short circuit, two-phase ground short circuit and three-phase short circuit occur are selected; 30 groups of sample data are selected for each fault type, a total of 120 groups of data, and 1000 data points of each of the four short-circuit voltage signals under the conditions of fault phase angles of 0°, 30°, 45° and 60° are collected, and the sampling frequency is 10KHz; the VMD algorithm is used to identify the fault The voltage signal is decomposed into modal components of different frequency bands under different fault phase angles, and then the permutation entropy algorithm is used to extract the fault features of each modal component to form a data sample. 120 feature samples can be obtained under each fault phase angle. 40 samples are selected as training samples from the 120 samples, and the remaining 80 samples are used as test samples to identify the four typical short-circuit faults of the transmission line under different fault phase angles. The final identification results are shown in Table 1. The prediction results of these four short-circuit fault types are given below.

Table 1 Prediction results of four short-circuit fault types

It can be seen from Table 1 that after the 80 fault feature quantities extracted are identified by the quantum variation multiverse optimization core extreme learning machine identification model, when the fault phase angle is 0°, there is no misjudgment in the single-phase ground short circuit, 1 misjudgment in the two-phase short circuit, 1 misjudgment in the two-phase ground short circuit, and 1 misjudgment in the three-phase short circuit. The recognition rates of the four short-circuit faults reach 100%, 95%, 100%, and 95% respectively, and the average recognition accuracy also reaches 96.25%; similarly, when the fault phase angles are 30°, 45°, and 60°, respectively, the average recognition accuracy also reaches 97.5%, 96.25%, and 97.5%, respectively, which shows that under the conditions of different fault phase angles, the fault recognition rate is basically not affected.

Claims (3)

1. A method for identifying power line faults based on quantum variation multiverse optimization, characterized in that it comprises the following steps: S1: Collecting transmission line fault voltage signals; S2: performing variational modal decomposition on the fault voltage signal; selecting the mean of the instantaneous frequency of the modal component after variational modal decomposition, drawing a standardized instantaneous frequency mean curve, and obtaining the optimal K value in the variational modal decomposition algorithm; S3: Decompose the fault voltage signal using the variational mode decomposition method to obtain K intrinsic mode components; then use the permutation entropy to calculate the multi-scale entropy value of each intrinsic mode component; S4: The calculated permutation entropy values of each eigenmode component are used to form a eigenvector; S5: The quantum particle swarm algorithm is used in the standard multiverse optimization algorithm, and the average optimal position of particles Av(t) is introduced; the Cauchy-Gauss mutation strategy is introduced to construct the quantum mutation multiverse optimization algorithm; S6: Optimize the regularization coefficient and kernel function parameters of the kernel extreme learning machine using the quantum mutation multiverse algorithm; S7: According to the feature vector and the regularization coefficient and kernel function parameters of the optimized kernel extreme learning machine, a line fault modal decomposition entropy and quantum mutation multiverse optimized kernel extreme learning machine prediction and classification model is constructed to identify the type of transmission line short circuit fault; The specific process of introducing the average optimal position of particles Av(t) is as follows: The quantum particle swarm algorithm is introduced into the multiverse optimization algorithm. Each particle is assumed to have quantum behavior. For particle i, its attractive potential is represented by p _i (p _i,1 ,p _i,2 ,…,p _i,n ). Considering that the particle position x _i,j changes with time and its state is described by the wave function, the particle evolution equation in the quantum particle swarm algorithm based on the δ potential well in the N-dimensional search space is shown in formula (4): Among them, pi _,j represents the coordinates of the attractor _pi in the j dimension, Li _,j is the characteristic search length, and ui _,j ~U(0,1) are uniformly distributed random numbers; the average optimal position of particles Av(t) is introduced in the algorithm, which represents the average value of the optimal positions of all particles _Li,j = 2γ·| _Avj (t)-xi _,j (t)|, equation (4) is changed to equation (5): Where γ represents the contraction-expansion factor; The specific process of introducing the Cauchy-Gauss mutation strategy to construct a quantum mutation multiverse optimization algorithm is as follows: The Cauchy-Gauss mutation strategy is adopted to select the individual with the best current fitness for mutation, and then compare its position before and after mutation, and select the optimal position to substitute for the next iteration, as shown in equations (6)-(8): λ ₂ Gauss(0,σ ² )] (7) in, represents the position of the optimal individual after mutation; σ ² represents the standard deviation of the Cauchy-Gaussian mutation strategy; cauchy(0,σ ² ) is a random variable that satisfies the Cauchy distribution; Gaus(s0,σ2) is a random variable that satisfies the Gaussian distribution; and Wherein, t represents the current number of iterations, T _max represents the maximum number of iterations, and λ ₁ and λ ₂ are dynamic parameters that are adaptively adjusted with the number of iterations; The specific process of S6 includes the following steps: S6.1: Randomly generate initial positions to form an initial search population; S6.2: In order to improve the performance of the native multiverse algorithm, the quantum particle swarm algorithm is introduced to accelerate the convergence speed; S6.3: Use the Cauchy-Gauss mutation strategy to select the individual with the best current fitness for mutation, so that the algorithm can jump out of the local optimum; S6.4: Update the optimal positions of individual particles and the group, check whether the algorithm has reached the maximum number of iterations, and if so, output the optimal kernel extreme learning machine parameters; if not, return to S6.2 and continue with the next iterative operation. 2. The method for identifying power line faults based on quantum variation multiverse optimization according to claim 1, wherein the multiverse optimization algorithm operates as shown in formula (1): Among them, _Xj is the jth variable of the current optimal universe, _ubj is the jth maximum value of the variable, and _lbj is the jth minimum value of the variable. is the jth variable of the i-th universe, r ₂ , r ₃ , r ₄ are all random numbers between 0 and 1; WEP represents the probability coefficient of wormhole existence, as shown in formula (2): Where min is the minimum WEP value; max is the maximum WEP value; l is the current iteration number; L is the maximum iteration number; TDR represents the travel distance rate coefficient, as shown in formula (3) Among them, p defines the detection speed that changes with the number of iterations. The higher the p value, the faster the local detection speed and the shorter the time. 3. The method for identifying power supply line faults using quantum mutation multiverse optimization according to claim 1, wherein during the optimization process of λ ₁ and λ ₂ , λ ₁ gradually decreases and λ ₂ gradually increases.