CN116383630A - Probabilistic neural network arc fault detection method based on improved wolf algorithm - Google Patents
Probabilistic neural network arc fault detection method based on improved wolf algorithm Download PDFInfo
- Publication number
- CN116383630A CN116383630A CN202310365153.5A CN202310365153A CN116383630A CN 116383630 A CN116383630 A CN 116383630A CN 202310365153 A CN202310365153 A CN 202310365153A CN 116383630 A CN116383630 A CN 116383630A
- Authority
- CN
- China
- Prior art keywords
- wolf
- neural network
- probabilistic neural
- algorithm
- wolves
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 241000282461 Canis lupus Species 0.000 title claims abstract description 88
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 42
- 238000001514 detection method Methods 0.000 title claims abstract description 22
- 238000000034 method Methods 0.000 claims abstract description 51
- 241000282421 Canidae Species 0.000 claims abstract description 37
- 238000005457 optimization Methods 0.000 claims abstract description 30
- 238000003062 neural network model Methods 0.000 claims abstract description 21
- 238000004364 calculation method Methods 0.000 claims abstract description 13
- 238000003745 diagnosis Methods 0.000 claims abstract description 10
- 238000009499 grossing Methods 0.000 claims abstract description 10
- 238000007781 pre-processing Methods 0.000 claims abstract description 7
- 230000008569 process Effects 0.000 claims description 16
- 239000013598 vector Substances 0.000 claims description 12
- 238000012549 training Methods 0.000 claims description 10
- 238000012545 processing Methods 0.000 claims description 6
- 230000008859 change Effects 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 4
- 230000005484 gravity Effects 0.000 claims description 3
- 210000002569 neuron Anatomy 0.000 claims description 3
- 230000000694 effects Effects 0.000 description 5
- 230000006870 function Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 238000012795 verification Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000005611 electricity Effects 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 230000032683 aging Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 238000009413 insulation Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 230000001012 protector Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
- Y04S10/52—Outage or fault management, e.g. fault detection or location
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Probability & Statistics with Applications (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The method for detecting the arc faults of the probabilistic neural network based on the improved wolf algorithm comprises the steps of obtaining current signal data sets of different load combinations under the two conditions of normal operation and arc faults in a household power line, preprocessing the data sets, and setting control factors of the wolf algorithm based on the improved wolf algorithm to optimize position parameters of the wolves in the wolf group, and constructing a probabilistic neural network model by utilizing parameter optimization results; and acquiring real-time current signal data, preprocessing the current signal data, and inputting the current signal data into a probabilistic neural network model for calculation to obtain a classification result of fault diagnosis. According to the invention, the control factor of the gray wolf algorithm is improved, and the dynamic self-adaptive step weight and the carrying weight are set, so that the convergence speed and the optimizing result of the algorithm are greatly improved, the optimizing result is used as the smoothing factor parameter of the AC arc fault detection model, the randomness of initial parameter selection is avoided, and the accuracy and the detection efficiency of the arc detection model are greatly improved.
Description
Technical Field
The invention relates to the field of arc fault diagnosis, in particular to a probability neural network arc fault detection method based on an improved gray wolf algorithm.
Background
With the gradual expansion of electricity demand of people, higher standards are also put forward for the stability of household electricity lines. Under the current large background of rapid development of artificial intelligence, the normal safe operation of an electric power system which is guided to develop by intelligence is a basic premise, and the safe operation of the electric power system is not separated from high-quality fault diagnosis work; and the types of household loads are more and more complex, and the frequency of faults in the daily use process is also higher and higher. Aging and damage to the insulation can lead to arc faults that often cause the device to burn or even fire. The rapid and accurate diagnosis can effectively ensure the safe operation of the power supply system. In the circuit with arc fault, the current is not obviously changed compared with the current in normal operation, and the waveform characteristics are similar to those of the current in the circuit connected with the nonlinear load to a certain extent, so that the conventional line protection device such as a breaker and a residual current protector cannot accurately diagnose the current, and therefore the current is a hot spot and a difficult point of an arc detection technology.
At present, the artificial neural network technology can generate memory and correspondingly store data in a database, so that references are provided for subsequent work, manpower and material resources are reduced to a great extent, and the artificial neural network technology is widely applied to fault diagnosis of an electric power system. Artificial neural network techniques include BP neural networks, convolutional neural networks, probabilistic neural networks, and the like. BP neural network, convolution neural network structure are complicated, and convergence rate is not ideal enough, and probability neural network is more suitable for solving the fault diagnosis problem because of its principle is simple, advantage that convergence rate is fast. The performance of the probabilistic neural network depends on the value of the smoothing factor parameter inside the model. Many scholars optimize parameters of the probabilistic neural network by using particle swarm algorithm, genetic algorithm, seagull optimization algorithm and other methods, but the optimization algorithms still have the problem that local optimization is difficult to jump out, so that the solving precision of the optimized probabilistic neural network still cannot meet the requirement of arc fault detection, and ideal probabilistic neural network model parameters cannot be obtained. Compared with the traditional optimization algorithm, the gray wolf optimization algorithm has obvious superiority in convergence capacity. However, when solving complex problems, problems of low solving efficiency and unsatisfactory optimizing effect occur, and the method is not suitable for directly optimizing parameters of the probabilistic neural network. Thus, there is a need for suitable improvements to the gray wolf optimization algorithm to overcome the above problems.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide the probability neural network arc fault detection method based on the improved gray wolf algorithm, which can stably, reliably and efficiently identify arc fault data.
The technical scheme of the invention is as follows: the probabilistic neural network arc fault detection method based on the improved wolf algorithm comprises the following steps:
and S1, acquiring current signal data sets of different load combinations under the two conditions of normal operation and arc faults in a household power line.
And step S2, preprocessing the acquired current signal data set.
And step S3, carrying out parameter optimization on the position parameters of the wolves in the wolves based on an improved wolf optimization algorithm to obtain the final position parameters of the optimal wolves.
Step S4, utilizing a probabilistic neural network model built based on an improved gray wolf optimization algorithm: and taking the final position parameter of the optimal wolves obtained based on the improved wolf optimization algorithm as a smoothing factor of the probabilistic neural network, and constructing a probabilistic neural network model.
And S5, acquiring real-time current signal data in the running process in the line, processing the real-time current signal data in the step S2, inputting the real-time current signal data into an input layer of the probabilistic neural network model built in the step S4, and calculating to obtain a classification result of fault diagnosis.
The invention further adopts the technical scheme that: the current signal data set utilizes the existing database to collect the current signal data of normal operation and arc faults, or adopts a special data generator to generate the current signal data of the normal operation and arc faults.
The invention further adopts the technical scheme that: the preprocessing comprises normalization processing, extracting time domain characteristic values, frequency domain characteristic values and energy characteristic values, and generating multidimensional characteristic vectors.
The invention further adopts the technical scheme that: the time domain characteristic value comprises a waveform index, a peak value index, a pulse index, a kurtosis index and a margin index; the frequency domain eigenvalues include center of gravity frequency, frequency variance, and mean square frequency.
The invention further adopts the technical scheme that: the improved wolf optimization algorithm is based on the parameter optimization of the position parameters of the wolves in the wolf group, specifically,
s31, initializing the wolf group position, wherein the initial wolf group position is generated according to the following formula:
in the method, in the process of the invention,for the initial wolf position->Is [0,1]Random number between->The lower limit value of the wolf group position is 0; />Is wolfThe upper bound of the group position is 1.
S32, calculating the fitness value of each wolf, wherein the calculation of the fitness function is shown as follows:
wherein,,the number of correctly detected arc faults in the data; />The number of misinterpretations in the data of the arc faults; />Is the number of correctly diagnosed data in the normal operation data; />The data for normal operation is misjudged as the number of arc faults.
S33, calculating the position of the wolf group in the hunting process, wherein the position change of each individual is as follows:
wherein t is the current iteration number; t is the maximum iteration number;the position of the t-th iteration of the gray wolf i;is the current location of the prey; d is the dimension of the feature vector; />For step weight, ++>And->Respectively a maximum value and a minimum value of step weights; />For step dynamic factor, +.>For the predicted prey location at iteration t-1,the actual position of the prey at t iterations; a and C are coefficients, wherein->Is of the value +.>A number therebetween; />Is a control factor.
S34, passing the wolf group individualsThe distances between the wolves and the three types of the lead wolves are used for determining how the individual wolves move towards the prey, and the calculation formula is as follows:
in the method, in the process of the invention,、/>and->I < th->The t-th iteration of wolf is directed to gray wolf->Grey wolf->Grey wolf->Update step size of ∈10->、/>And->Respectively the gray wolves in the iteration>Grey wolf->Grey wolf->Is a position of (2); />,/>,/>And->,/>,/>Coefficients generated for the current iteration; />I is only +.1 for the t+1th iteration>The updated position of wolves.
S35, judging whether the maximum iteration times are reached, if so, finishing parameter optimization, and outputting final position parameters of the optimal wolves; otherwise turning to S33, continuing to perform iterative optimization by using the updated position parameters of the wolves.
The invention further adopts the technical scheme that: the attenuation speed of the control factor in the early stage of iteration is small, and the attenuation speed in the later stage of iteration is large.
The invention further adopts the technical scheme that: the probabilistic neural network model is divided into four layers, namely an input layer, an implicit layer, a summation layer and an output layer; the input layer receives data and transmits the data to the hidden layer, the hidden layer calculates the matching degree of the feature vector and the training sample class, and the result is sent to the summation layer after the matching degree calculation is completed; the summation layer carries out weighted average on the results to obtain an estimated probability density function of the category; the output layer outputs the category with the highest estimated probability as a classification result.
The invention further adopts the technical scheme that: the matching degree calculation formula is shown as follows:
wherein,,a degree of matching determined for the vector x input to the hidden layer via the neuron j of the ith class in the hidden layer; i=1, 2, …, M represents the number of categories in the training sample; />The j center is the j center of the i-th sample, and the j value is the same as the training sample number; />Is a smoothing factor.
Compared with the prior art, the invention has the following characteristics:
(1) According to the invention, the control factor is set for nonlinear decrement so as to jump out local optimum, and the global searching capability is improved.
(2) The invention uses dynamic self-adaptive step weight for the position updating strategy, which not only enhances the flexibility of the algorithm, but also highlights the leading advantage of the optimal wolf.
(3) The improved gray wolf algorithm is adopted to optimize the smoothing factors, and a probabilistic neural network model is built, so that the data classification effect is good, and the accuracy is higher.
The detailed structure of the present invention is further described below with reference to the accompanying drawings and detailed description.
Drawings
FIG. 1 is a flow chart of a probabilistic neural network arc fault detection method of the present invention;
FIG. 2 shows a conventional wolf algorithm and an improved control factor for the wolf algorithmIs a graph comparing attenuation curves of the first and second pairs of the first and;
FIG. 3 is a graph showing the comparison of the identification results of the predicted value and the actual value of the fault detection performed on the verification set of 176 groups of current signal data by the method of the invention;
FIG. 4 is a graph showing the comparison of iteration times and step weight changes in the iteration process of the method of the invention with a linear attenuation value strategy and a cosine attenuation value strategy;
FIG. 5 is a graph of the relationship between the number of iterations of the method of the present invention and PNN, GWO-PNN and fitness values;
FIG. 6 is a graph comparing the evaluation index of the method of the present invention with that of PNN and GWO-PNN.
Detailed Description
1-5, the method for detecting the arc faults of the probabilistic neural network based on the improved gray wolf algorithm specifically comprises the following steps:
step S1, acquiring current signal data sets of different load combinations under two conditions of normal operation and arc faults in a household power line: and collecting normal operation and arc fault current signal data by using an existing database, or generating current signal data during normal operation and arc fault by using a special data generator.
Step S2, preprocessing the acquired current signal data set: the method mainly comprises the steps of carrying out normalization processing, extracting a time domain characteristic value, a frequency domain characteristic value and an energy characteristic value of the time domain characteristic value, and generating a multidimensional characteristic vector. The time domain characteristic value comprises a waveform index, a peak value index, a pulse index, a kurtosis index and a margin index. The frequency domain eigenvalues include center of gravity frequency, frequency variance, and mean square frequency.
The normalization process, the time domain feature value, the frequency domain feature value and the energy feature value are all in the prior art, and are not described herein.
And step S3, carrying out parameter optimization on the position parameters of the wolves in the wolves based on an improved wolf optimization algorithm.
S31, initializing the wolf group position, wherein the initial wolf group position is generated according to the following formula:
in the method, in the process of the invention,for the initial wolf position->Is [0,1]Random number between->The lower limit value of the wolf group position is 0; />The upper limit value of the wolf group position is 1.
S32, calculating the fitness value of each wolf, wherein the calculation of the fitness function is shown as follows:
wherein,,the number of correctly detected arc faults in the data; />The number of misinterpretations in the data of the arc faults; />Is the number of correctly diagnosed data in the normal operation data; />In order to misjudge the number of arc faults in the data of normal operation, the number of the wolf population is set to be 30 in the embodiment. Dividing the sirius population into ++s according to the fitness value from large to small>,/>,/>And->A total of 4 grades.
S33, calculating the position of the wolf group in the hunting process, wherein the position change of each individual is as follows:
wherein t is the current iteration number; t is the maximum iteration number;the position of the t-th iteration of the gray wolf i; />Is the current location of the prey; d is the dimension of the feature vector; />For step weight, ++>And->Respectively a maximum value and a minimum value of step weights; />For step dynamic factor, +.>Predicted prey location for t-1 iterations,>for the actual position of the prey at t iterations, here the shift is minimized +.>The position of wolf is approximately taken as the actual position of the prey of this iteration, A and C are coefficients, where +.>Is of the value +.>A number therebetween; />Setting a control factor for the control factor, wherein the attenuation speed of the control factor in the early iteration stage is smaller, the value fluctuation of A in the formula is larger, so that the wolf group has a wider search range, and global search is facilitated to jump out of local optimum; the attenuation speed in the later iteration stage is high, so that the optimizing efficiency and the convergence speed of the algorithm are improved, and the method is beneficial to quickly obtaining the optimal solution. In this embodiment, the maximum number of iterations T is set to 70 +.>And->The values of (2) are 1 and 0.4 respectively, and the control factors of the traditional wolf algorithm and the improved wolf algorithm are +.>A comparison of the attenuation curves of (a) is shown in figure 2.
S34, passing the wolf group individualsThe distances between the wolves and the three types of the lead wolves are used for determining how the individual wolves move towards the prey, and the calculation formula is as follows:
in the method, in the process of the invention,、/>and->I < th->The t-th iteration of wolf is directed to gray wolf->Grey wolf->Grey wolf->Update step size of ∈10->、/>And->Respectively the gray wolves in the iteration>Grey wolf->Grey wolf->Is a position of (2); />,/>,/>And->,/>,/>Coefficients generated for the current iteration; />I is only +.1 for the t+1th iteration>The updated position of wolves.
S35, judging whether the maximum iteration times are reached, if so, finishing parameter optimization, and outputting final position parameters of the optimal wolves; otherwise turning to S33, continuing to perform iterative optimization by using the updated position parameters of the wolves.
Step S4, utilizing a probabilistic neural network model built based on an improved gray wolf optimization algorithm: and taking the final position parameter of the optimal wolves obtained based on the improved wolf optimization algorithm as a smoothing factor of the probabilistic neural network, and constructing a probabilistic neural network model.
The probabilistic neural network model is divided into four layers, namely an input layer, an implicit layer, a summation layer and an output layer. The input layer receives the data and passes it to the hidden layer; the hidden layer calculates the matching degree of the feature vector and the training sample category, and the matching degree calculation formula is shown as follows:
wherein,,a degree of matching determined for the vector x input to the hidden layer via the neuron j of the ith class in the hidden layer; i=1, 2, …, M represents the number of categories in the training sample, where m=2, m=1, 2 are two categories of normal operation and arc fault, respectively; />The j center is the j center of the i-th sample, and the j value is the same as the training sample number; />The final position parameter of the optimal wolf obtained in the step S3 is a smoothing factor, plays a vital role in each performance of the probabilistic neural network model, and can reflect the classification accuracy of the probabilistic neural network model.
After the matching degree calculation is completed, the result is sent to a summation layer; the summation layer carries out weighted average on the results to obtain an estimated probability density function of the category; the output layer outputs the category with the highest estimated probability as a classification result.
And S5, acquiring real-time current signal data in the running process in the line, processing the real-time current signal data in the step S2, inputting the real-time current signal data into an input layer of the probabilistic neural network model built in the step S4, and calculating to obtain a classification result of fault diagnosis.
As shown in fig. 3, the probability neural network arc fault detection method based on the improved wolf algorithm of the embodiment is adopted to perform fault detection on the verification set of 176 groups of current signal data, and the identification result comparison graph of the predicted value and the true value is obtained. Where 0 represents the absence of an arc fault and 1 represents the presence of an arc fault. As can be seen from fig. 3, the probability neural network arc fault detection method based on the improved gray wolf algorithm in this embodiment has an identification rate of 97.7% for the arc faults in the verification set, and the misjudgment conditions are respectively 2.3% and 3.4%, so that the misjudgment probability is very low.
As shown in fig. 4, to illustrate the superiority of the step size weight value policy proposed in this embodiment compared with other value policies, a graph of iteration times and step size weight change in the iteration process is drawn for the step size weight value policy, the linear attenuation value policy and the cosine attenuation value policy proposed in this embodiment. Compared with the figure 4, the step weight value strategy provided by the embodiment is maintained in the maximum state at the initial stage of iteration, so that the time of global searching of the wolf cluster is longer, the searching range is larger, and the situation that the searching space is missed and falls into local optimum prematurely is avoided to a certain extent. In the middle of the iteration, the value of the step weight is set to be related to the predicted position of the prey and the actual position of the iterated preyThe curve decay rate is faster by adopting the step weight value strategy provided by the embodiment, the +.>The value of (2) reflects the current prediction accuracy, with smaller values being more accurate. When iteration is performed for forty times, the step weight value taking strategy curve provided by the embodiment is stabilized in the minimum state.
When the step weight value strategy curve provided by the embodiment is adopted, the step weight in the iteration middle period is related to the predicted position of the prey and the actual position of the iterative prey, so that the leading effect of the head wolves is enhanced, and the wolves are subjected to fine search in a small range. In other two values strategies, the value of the step weight is only related to the current iteration number and cannot be adjusted according to the prediction situation, so that a certain search range is sacrificed in the early stage, and the leading effect of the wolf cannot be fully exerted in the later stage, so that the convergence speed is slow.
As shown in fig. 5, in order to further verify the effectiveness and superiority of different detection methods for arc fault identification, a graph of the iteration number and fitness value of the probabilistic neural network model GWO-PNN based on simple probabilistic neural network PNN, based on traditional wolf algorithm optimization, and based on improved wolf algorithm, probabilistic neural network IGWO-PNN, is respectively constructed. Wherein the parameter conditions are respectively as follows: under the same training parameters and historical data, the arc fault detection model is trained, the iteration number is 70, and the learning rate is 0.001. As can be seen from fig. 5, compared with the probabilistic neural network model GWO-PNN optimized based on the simple probabilistic neural network PNN and based on the traditional wolf algorithm, the probability neural network IGWO-PNN based on the improved wolf algorithm has earlier time for reaching the maximum fitness value and higher value of the maximum fitness value, thereby being known to have faster convergence speed and more accurate classification effect.
As shown in fig. 6, a comparison graph of evaluation indexes of probabilistic neural network IGWO-PNN methods based on a simple probabilistic neural network PNN, a probabilistic neural network model GWO-PNN optimized based on a traditional wolf algorithm and an improved wolf algorithm is provided. As can be seen from fig. 6, the arc fault detection accuracy of the probabilistic neural network IGWO-PNN method based on the improved wolf algorithm is 97.159%, the recall rate is 97.727%, and the accuracy rate is 96.629% higher than that of the other two arc fault detection methods. From the above, the probability neural network IGWO-PNN method based on the improved gray wolf algorithm has strong arc fault identification capability.
The above comparison results all give the effectiveness of the probabilistic neural network IGWO-PNN method based on the modified wolf algorithm and the necessity of taking the value of the smoothing factor by using the modified wolf algorithm. The fault detection model provides effective diagnosis for arc fault detection, and is beneficial to developing subsequent arc fault prevention and control work.
Claims (8)
1. The probabilistic neural network arc fault detection method based on the improved wolf algorithm is characterized by comprising the following steps of:
step S1, acquiring current signal data sets of different load combinations under the two conditions of normal operation and arc faults in a household power line;
s2, preprocessing an acquired current signal data set;
step S3, carrying out parameter optimization on position parameters of the wolves in the wolves based on an improved wolf optimization algorithm to obtain final position parameters of the optimal wolves;
step S4, utilizing a probabilistic neural network model built based on an improved gray wolf optimization algorithm: setting up a probabilistic neural network model by taking the final position parameter of the optimal wolves obtained based on the improved wolf optimization algorithm as a smoothing factor of the probabilistic neural network;
and S5, acquiring real-time current signal data in the running process in the line, processing the real-time current signal data in the step S2, inputting the real-time current signal data into an input layer of the probabilistic neural network model built in the step S4, and calculating to obtain a classification result of fault diagnosis.
2. The method for detecting arc faults of the probabilistic neural network based on the improved wolf algorithm as claimed in claim 1, wherein the method comprises the following steps: the current signal data set utilizes the existing database to collect the current signal data of normal operation and arc faults, or adopts a special data generator to generate the current signal data of the normal operation and arc faults.
3. The method for detecting arc faults of the probabilistic neural network based on the improved wolf algorithm as claimed in claim 1, wherein the method comprises the following steps: the preprocessing comprises normalization processing, extracting time domain characteristic values, frequency domain characteristic values and energy characteristic values, and generating multidimensional characteristic vectors.
4. The method for detecting arc faults of the probabilistic neural network based on the improved wolf algorithm as claimed in claim 3, wherein the method comprises the following steps: the time domain characteristic value comprises a waveform index, a peak value index, a pulse index, a kurtosis index and a margin index; the frequency domain eigenvalues include center of gravity frequency, frequency variance, and mean square frequency.
5. The method for detecting arc faults of the probabilistic neural network based on the improved wolf algorithm as claimed in claim 1, wherein the method comprises the following steps: the improved wolf optimization algorithm is based on the parameter optimization of the position parameters of the wolves in the wolf group, specifically,
s31, initializing the wolf group position, wherein the initial wolf group position is generated according to the following formula:
in the method, in the process of the invention,for the initial wolf position->Is [0,1]Random number between->The lower limit value of the wolf group position is 0; />The upper limit value of the position of the wolf group is 1;
s32, calculating the fitness value of each wolf, wherein the calculation of the fitness function is shown as follows:
wherein,,the number of correctly detected arc faults in the data; />The number of misinterpretations in the data of the arc faults; />Is the number of correctly diagnosed data in the normal operation data; />The number of arc faults misjudged in the data of normal operation;
s33, calculating the position of the wolf group in the hunting process, wherein the position change of each individual is as follows:
wherein t is the current iteration number; t is the maximum iteration number;the position of the t-th iteration of the gray wolf i; />Is the current location of the prey; d is the dimension of the feature vector; />For step weight, ++>And->Respectively a maximum value and a minimum value of step weights; />For step dynamic factor, +.>For predicted prey position at iteration t-1,/for the target position>The actual position of the prey at t iterations; a and C are coefficients, wherein->Is of the value +.>A number therebetween; />Is a control factor;
s34, passing the wolf group individualsThe distances between the wolves and the three types of the lead wolves are used for determining how the individual wolves move towards the prey, and the calculation formula is as follows:
in the method, in the process of the invention,、/>and->I < th->The t-th iteration of wolf is directed to gray wolf->Grey wolf->Wolf of grayUpdate step size of ∈10->、/>And->Respectively the gray wolves in the iteration>Grey wolf->Grey wolf->Is a position of (2); />,/>,/>And->,/>,/>Coefficients generated for the current iteration; />I is only +.1 for the t+1th iteration>The updated position of wolves;
s35, judging whether the maximum iteration times are reached, if so, finishing parameter optimization, and outputting final position parameters of the optimal wolves; otherwise turning to S33, continuing to perform iterative optimization by using the updated position parameters of the wolves.
6. The method for detecting arc faults of the probabilistic neural network based on the improved wolf algorithm as claimed in claim 5, wherein the method comprises the following steps: the attenuation speed of the control factor in the early stage of iteration is small, and the attenuation speed in the later stage of iteration is large.
7. The method for detecting arc faults of the probabilistic neural network based on the improved wolf algorithm as claimed in claim 1, wherein the method comprises the following steps: the probabilistic neural network model is divided into four layers, namely an input layer, an implicit layer, a summation layer and an output layer; the input layer receives data and transmits the data to the hidden layer, the hidden layer calculates the matching degree of the feature vector and the training sample class, and the result is sent to the summation layer after the matching degree calculation is completed; the summation layer carries out weighted average on the results to obtain an estimated probability density function of the category; the output layer outputs the category with the highest estimated probability as a classification result.
8. The method for detecting arc faults of the probabilistic neural network based on the improved wolf algorithm as claimed in claim 7, wherein the method comprises the following steps: the matching degree calculation formula is shown as follows:
wherein,,determination of the i-th class of neurons j in the hidden layer for vector x input to the hidden layerMatching degree of (3); i=1, 2, …, M represents the number of categories in the training sample; />The j center is the j center of the i-th sample, and the j value is the same as the training sample number; />Is a smoothing factor.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310365153.5A CN116383630A (en) | 2023-04-07 | 2023-04-07 | Probabilistic neural network arc fault detection method based on improved wolf algorithm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310365153.5A CN116383630A (en) | 2023-04-07 | 2023-04-07 | Probabilistic neural network arc fault detection method based on improved wolf algorithm |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116383630A true CN116383630A (en) | 2023-07-04 |
Family
ID=86980324
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310365153.5A Pending CN116383630A (en) | 2023-04-07 | 2023-04-07 | Probabilistic neural network arc fault detection method based on improved wolf algorithm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116383630A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117368648A (en) * | 2023-11-08 | 2024-01-09 | 国网四川省电力公司电力科学研究院 | Power distribution network single-phase earth fault detection method, system, equipment and storage medium |
-
2023
- 2023-04-07 CN CN202310365153.5A patent/CN116383630A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117368648A (en) * | 2023-11-08 | 2024-01-09 | 国网四川省电力公司电力科学研究院 | Power distribution network single-phase earth fault detection method, system, equipment and storage medium |
CN117368648B (en) * | 2023-11-08 | 2024-06-04 | 国网四川省电力公司电力科学研究院 | Power distribution network single-phase earth fault detection method, system, equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2021068454A1 (en) | Method for identifying energy of micro-energy device on basis of bp neural network | |
CN109389058B (en) | Sea clutter and noise signal classification method and system | |
CN110082640B (en) | Distribution network single-phase earth fault identification method based on long-time memory network | |
CN108062572A (en) | A kind of Fault Diagnosis Method of Hydro-generating Unit and system based on DdAE deep learning models | |
CN113255848B (en) | Water turbine cavitation sound signal identification method based on big data learning | |
CN103926526A (en) | Analog circuit fault diagnosis method based on improved RBF neural network | |
CN104809722A (en) | Electrical device fault diagnosis method based on infrared thermography | |
CN112580471A (en) | Non-invasive load identification method based on AdaBoost feature extraction and RNN model | |
CN111783845B (en) | Hidden false data injection attack detection method based on local linear embedding and extreme learning machine | |
CN110909977A (en) | Power grid fault diagnosis method based on ADASYN-DHSD-ET | |
CN111191835A (en) | IES incomplete data load prediction method and system based on C-GAN transfer learning | |
CN110702966B (en) | Fault arc detection method, device and system based on probabilistic neural network | |
CN116383630A (en) | Probabilistic neural network arc fault detection method based on improved wolf algorithm | |
CN113203914A (en) | Underground cable early fault detection and identification method based on DAE-CNN | |
CN116248392A (en) | Network malicious traffic detection system and method based on multi-head attention mechanism | |
CN109901064B (en) | ICA-LVQ-based high-voltage circuit breaker fault diagnosis method | |
CN116756575B (en) | Non-invasive load decomposition method based on BGAIN-DD network | |
CN117874615A (en) | Photovoltaic fault diagnosis method and system based on depth digital twin | |
CN115659258B (en) | Power distribution network fault detection method based on multi-scale graph roll-up twin network | |
CN116663414A (en) | Fault diagnosis method and system for power transformer | |
Wang et al. | Power quality disturbance recognition method in park distribution network based on one-dimensional vggnet and multi-label classification | |
Jiajia et al. | Intrusion detection model based on sae and balstm | |
CN111079509A (en) | Abnormal behavior detection method based on self-attention mechanism | |
CN114881176A (en) | Non-invasive load identification method based on self-adaptive optimization random forest | |
CN114841266A (en) | Voltage sag identification method based on triple prototype network under small sample |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |