[go: up one dir, main page]

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 PDF

Info

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
arc fault
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
Application number
CN202310365153.5A
Other languages
Chinese (zh)
Inventor
刘松
李丹
迪心怡
马文骁
胡鹤龄
彭嘉怡
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
North China Electric Power University
Original Assignee
North China Electric Power University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by North China Electric Power University filed Critical North China Electric Power University
Priority to CN202310365153.5A priority Critical patent/CN116383630A/en
Publication of CN116383630A publication Critical patent/CN116383630A/en
Pending legal-status Critical Current

Links

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

基于改进灰狼算法的概率神经网络电弧故障检测方法Probabilistic Neural Network Arc Fault Detection Method Based on Improved Gray Wolf Algorithm

技术领域technical field

本发明涉及电弧故障诊断领域，特别是一种基于改进灰狼算法的概率神经网络电弧故障检测方法。The invention relates to the field of arc fault diagnosis, in particular to a probabilistic neural network arc fault detection method based on an improved gray wolf algorithm.

背景技术Background technique

随着人们用电需求的逐渐扩大，对家庭用电线路的平稳性也提出了更高标准。在当前人工智能迅速发展的大背景下，以智能化为导向发展的电力系统的正常安全运行是基本前提，电力系统的安全运行离不开高质量的故障诊断工作；且家用负载的种类越来越复杂，在日常使用过程中出现故障的频率也越来越高。绝缘层的老化和损坏能够导致电弧故障的产生，而电弧故障往往会引发设备燃烧甚至发生火灾事故。快速准确的诊断能够有效地保证供电系统的安全运行。由于在发生电弧故障的电路中，电流的大小较正常运行时的电流不会有明显改变，且波形特征也与接入非线性负载的电路中的电流波形特征存在一定程度上的相似，使得常规线路保护装置如断路器和剩余电流保护器等无法对其进行准确的诊断，因此其成为电弧检测技术的热点和难点。With the gradual expansion of people's demand for electricity, higher standards have been put forward for the stability of household electricity lines. Under the background of the current rapid development of artificial intelligence, the normal and safe operation of the intelligent power system is the basic premise. The safe operation of the power system is inseparable from high-quality fault diagnosis work; The more complex it is, the higher the frequency of failures during daily use. The aging and damage of the insulation layer can lead to arc faults, and arc faults often lead to equipment burning or even fire accidents. Fast and accurate diagnosis can effectively guarantee the safe operation of the power supply system. Because in the circuit where arc fault occurs, the magnitude of the current will not change significantly compared with the current in normal operation, and the waveform characteristics are also similar to the current waveform characteristics in the circuit connected to the nonlinear load to a certain extent, so that the conventional Line protection devices such as circuit breakers and residual current protectors cannot be diagnosed accurately, so it has become a hot and difficult point in arc detection technology.

目前，人工神经网络技术由于其能够产生记忆并将数据相应存储在数据库中，为后续工作提供参考，很大程度上减少人力物力，从而广泛应用于电力系统的故障诊断。人工神经网络技术包括BP神经网络、卷积神经网络、概率神经网络等。BP神经网络、卷积神经网络结构复杂，收敛速度不够理想，相比之下，概率神经网络因其原理简单、收敛速度快的优点更适宜解决故障诊断问题。概率神经网络的性能取决于模型内部平滑因子参数的取值。许多学者利用粒子群算法、遗传算法、海鸥优化算法等方法对概率神经网络的参数进行优化，但这些优化算法仍存在难以跳出局部最优的问题，导致优化后的概率神经网络的求解精度仍无法满足电弧故障检测的要求，无法获得较理想的概率神经网络模型参数。相较于传统优化算法，灰狼优化算法在收敛能力方面有明显的优越性。但是在求解复杂问题时，会出现求解效率低、优化效果不够理想的问题，并不适合直接用来优化概率神经网络的参数。因而，需要对灰狼优化算法进行适当的改进以克服上述问题。At present, artificial neural network technology is widely used in fault diagnosis of power systems because it can generate memory and store data in the database accordingly, providing reference for follow-up work and greatly reducing manpower and material resources. Artificial neural network technology includes BP neural network, convolutional neural network, probabilistic neural network, etc. The structure of BP neural network and convolutional neural network is complex, and the convergence speed is not ideal. In contrast, the probabilistic neural network is more suitable for solving fault diagnosis problems because of its simple principle and fast convergence speed. The performance of the probabilistic neural network depends on the value of the smoothing factor parameter inside the model. Many scholars use methods such as particle swarm optimization, genetic algorithm, and seagull optimization algorithm to optimize the parameters of the probabilistic neural network. To meet the requirements of arc fault detection, it is impossible to obtain ideal parameters of the probabilistic neural network model. Compared with the traditional optimization algorithm, the gray wolf optimization algorithm has obvious advantages in the convergence ability. However, when solving complex problems, there will be problems such as low solution efficiency and unsatisfactory optimization results, which are not suitable for directly optimizing the parameters of the probabilistic neural network. Therefore, it is necessary to make appropriate improvements to the gray wolf optimization algorithm to overcome the above problems.

发明内容Contents of the invention

本发明的目的是克服现有技术的上述不足而提供一种基于改进灰狼算法的概率神经网络电弧故障检测方法，能够稳定、可靠、高效地对电弧故障数据进行识别。The purpose of the present invention is to overcome the above shortcomings of the prior art and provide a probabilistic neural network arc fault detection method based on the improved gray wolf algorithm, which can identify arc fault data stably, reliably and efficiently.

本发明的技术方案是：基于改进灰狼算法的概率神经网络电弧故障检测方法，包括如下步骤：The technical solution of the present invention is: a probabilistic neural network arc fault detection method based on the improved gray wolf algorithm, comprising the following steps:

步骤S1、获取家庭用电线路中的正常运行和电弧故障两种情况下的不同负载组合的电流信号数据集。Step S1. Acquiring current signal data sets of different load combinations under two conditions of normal operation and arc fault in the household power line.

步骤S2、对获取的电流信号数据集进行预处理。Step S2, performing preprocessing on the acquired current signal data set.

步骤S3，基于改进的灰狼优化算法，对狼群中灰狼的位置参数进行参数优化，得到最优狼的最终位置参数。In step S3, based on the improved gray wolf optimization algorithm, parameter optimization is performed on the position parameters of the gray wolves in the wolf pack to obtain the final position parameters of the optimal wolf.

步骤S4，利用基于改进灰狼优化算法搭建的概率神经网络模型：将基于改进的灰狼优化算法得到的最优狼的最终位置参数作为概率神经网络的平滑因子，搭建概率神经网络模型。Step S4, using the probabilistic neural network model built based on the improved gray wolf optimization algorithm: the final position parameter of the optimal wolf obtained based on the improved gray wolf optimization algorithm is used as the smoothing factor of the probabilistic neural network to build a probabilistic neural network model.

步骤S5，获取线路中运行时的实时电流信号数据，对该实时电流信号数据进行步骤S2的处理后，输入至步骤S4搭建的概率神经网络模型的输入层，并进行计算得到故障诊断的分类结果。Step S5, obtain the real-time current signal data during operation in the line, after processing the real-time current signal data in step S2, input it to the input layer of the probabilistic neural network model built in step S4, and perform calculation to obtain the classification result of fault diagnosis .

本发明进一步的技术方案是：所述电流信号数据集利用现有的数据库收集正常运行和电弧故障电流信号数据，或者采用专门的数据生成机生成正常运行和电弧故障时的电流信号数据。A further technical solution of the present invention is: the current signal data set uses the existing database to collect current signal data during normal operation and arc fault, or uses a special data generator to generate current signal data during normal operation and arc fault.

本发明再进一步的技术方案是：所述预处理包括进行归一化处理，以及提取时域特征值、频域特征值和能量特征值，并生成多维的特征向量。A further technical solution of the present invention is: the preprocessing includes performing normalization processing, extracting time-domain eigenvalues, frequency-domain eigenvalues and energy eigenvalues, and generating multidimensional eigenvectors.

本发明更进一步的技术方案是：所述时域特征值包括波形指标、峰值指标、脉冲指标、峭度指标、裕度指标；所述频域特征值包括重心频率、频率方差和均方频率。The further technical solution of the present invention is: the time domain characteristic value includes waveform index, peak index, pulse index, kurtosis index, margin index; the frequency domain characteristic value includes center of gravity frequency, frequency variance and mean square frequency.

本发明更进一步的技术方案是：所述基于改进的灰狼优化算法，对狼群中灰狼的位置参数进行参数优化具体为，The further technical solution of the present invention is: based on the improved gray wolf optimization algorithm, the parameter optimization of the position parameters of gray wolves in the wolf pack is specifically as follows:

S31，对狼群位置进行初始化，初始狼群位置根据下式产生：S31. Initialize the position of the wolves, and the initial position of the wolves is generated according to the following formula:

Figure SMS_1
（1）
Figure SMS_1
(1)

式中，

Figure SMS_2
为初始狼群位置，/>
Figure SMS_3
为[0,1]之间的随机数，/>
Figure SMS_4
为狼群位置的下界值，取值为0；/>
Figure SMS_5
为狼群位置的上界值，取值为1。In the formula,
Figure SMS_2
is the initial wolf pack position, />
Figure SMS_3
It is a random number between [0,1], />
Figure SMS_4
is the lower limit value of the position of wolves, and the value is 0; />
Figure SMS_5
is the upper limit value of the position of wolves, and the value is 1.

S32，计算每匹狼的适应度值，适应度函数的计算如下式所示：S32, calculate the fitness value of each wolf, the calculation of the fitness function is shown in the following formula:

Figure SMS_6
（2）
Figure SMS_6
(2)

其中，

Figure SMS_7
为发生电弧故障的数据中被正确检测出来的数量；/>
Figure SMS_8
为发生电弧故障的数据中被误判的数量；/>
Figure SMS_9
为正常运行的数据中被正确诊断的数量；/>
Figure SMS_10
为正常运行的数据中被误判为电弧故障的数量。in,
Figure SMS_7
The number of correctly detected arc faults in the data; />
Figure SMS_8
is the number of misjudgments in the data of arc fault; />
Figure SMS_9
The number of correctly diagnosed in the normal operation data; />
Figure SMS_10
The number of falsely judged arc faults in the normal operation data.

S33，计算在捕猎过程中狼群位置，每个个体的位置变化，公式如下：S33. Calculate the position of the wolf pack during the hunting process, and the position change of each individual, the formula is as follows:

Figure SMS_11
（3）
Figure SMS_11
(3)

式中，t为当前迭代次数；T为最大迭代次数；

Figure SMS_13
为灰狼i第t次迭代的位置；
Figure SMS_14
为猎物的当前位置；d为特征向量的维数；/>
Figure SMS_16
为步长权重，/>
Figure SMS_18
和/>
Figure SMS_20
分别为步长权重的最大值和最小值，；/>
Figure SMS_21
为步长动态因子，/>
Figure SMS_22
为第t-1次迭代时预测的猎物位置，
Figure SMS_12
为t次迭代时猎物的实际位置；A和C为系数，其中/>
Figure SMS_15
是值在/>
Figure SMS_17
之间的数；/>
Figure SMS_19
为控制因子。In the formula, t is the current iteration number; T is the maximum iteration number;
Figure SMS_13
is the position of gray wolf i in the t-th iteration;
Figure SMS_14
is the current position of the prey; d is the dimension of the feature vector; />
Figure SMS_16
is the step weight, />
Figure SMS_18
and />
Figure SMS_20
are the maximum and minimum values of the step weight, respectively; />
Figure SMS_21
is the step size dynamic factor, />
Figure SMS_22
is the predicted prey position at the t-1th iteration,
Figure SMS_12
is the actual position of the prey at t iterations; A and C are coefficients, where />
Figure SMS_15
is the value in />
Figure SMS_17
number between; />
Figure SMS_19
as the control factor.

S34，通过狼群个体

Figure SMS_23
与三类带领狼之间的距离来确定灰狼个体如何向猎物移动，计算公式如下：S34, through individual wolf packs
Figure SMS_23
The distance between the three types of leading wolves is used to determine how individual gray wolves move towards prey, and the calculation formula is as follows:

Figure SMS_24
（4）
Figure SMS_24
(4)

式中，

Figure SMS_38
、/>
Figure SMS_40
和/>
Figure SMS_41
分别为第i只/>
Figure SMS_42
狼第t次迭代朝向灰狼/>
Figure SMS_43
，灰狼/>
Figure SMS_44
，灰狼/>
Figure SMS_45
的更新步长，/>
Figure SMS_25
、/>
Figure SMS_28
和/>
Figure SMS_29
分别为本次迭代中灰狼/>
Figure SMS_32
，灰狼/>
Figure SMS_34
，灰狼/>
Figure SMS_36
的位置；/>
Figure SMS_37
，/>
Figure SMS_39
，/>
Figure SMS_26
和/>
Figure SMS_27
，/>
Figure SMS_30
，/>
Figure SMS_31
为本次迭代产生的系数；/>
Figure SMS_33
为第t+1次迭代第i只/>
Figure SMS_35
狼的更新位置。In the formula,
Figure SMS_38
, />
Figure SMS_40
and />
Figure SMS_41
Respectively for the i-th />
Figure SMS_42
wolf t iteration towards gray wolf />
Figure SMS_43
, gray wolf/>
Figure SMS_44
, gray wolf/>
Figure SMS_45
The update step size, />
Figure SMS_25
, />
Figure SMS_28
and />
Figure SMS_29
Respectively, gray wolf/> in this iteration
Figure SMS_32
, gray wolf/>
Figure SMS_34
, gray wolf/>
Figure SMS_36
location; />
Figure SMS_37
, />
Figure SMS_39
, />
Figure SMS_26
and />
Figure SMS_27
, />
Figure SMS_30
, />
Figure SMS_31
Coefficients generated for this iteration; />
Figure SMS_33
For the t+1th iteration i-th />
Figure SMS_35
The updated position of the wolf.

S35，判断是否达到最大迭代次数，若达到则参数寻优结束，输出最优狼的最终位置参数；否则转向S33，利用灰狼更新后的位置参数继续进行迭代寻优。S35, judging whether the maximum number of iterations has been reached, if reached, the parameter optimization ends, and the final position parameter of the optimal wolf is output; otherwise, turn to S33, and use the updated position parameter of the gray wolf to continue iterative optimization.

本发明更进一步的技术方案是：所述控制因子在迭代早期的衰减速度小，在迭代后期的衰减速度大。A further technical solution of the present invention is: the decay speed of the control factor is small in the early stage of the iteration, and the decay speed in the late stage of the iteration is large.

本发明更进一步的技术方案是：所述概率神经网络模型分为四层，分别为输入层、隐含层、求和层和输出层；输入层接收数据并将其传递给隐含层，隐含层计算特征向量与训练样本类别的匹配程度，完成匹配程度计算之后将结果发送到求和层；求和层对结果进行加权平均得到该类别的估计概率密度函数；输出层输出估计概率最高的类别作为分类结果。The further technical scheme of the present invention is: described probabilistic neural network model is divided into four layers, is respectively input layer, hidden layer, summation layer and output layer; Input layer receives data and transmits it to hidden layer, hidden layer The containing layer calculates the matching degree between the feature vector and the training sample category, and sends the result to the summation layer after completing the matching degree calculation; the summation layer performs weighted average on the results to obtain the estimated probability density function of the category; the output layer outputs the highest estimated probability category as a classification result.

本发明更进一步的技术方案是：所述匹配程度计算公式如下式所示：A further technical solution of the present invention is: the formula for calculating the degree of matching is shown in the following formula:

Figure SMS_46
（5）
Figure SMS_46
(5)

其中，

Figure SMS_47
为输入到隐含层的向量x经隐含层中第i类的神经元j确定的匹配度；i=1，2，…，M，M表示训练样本中的类别数；/>
Figure SMS_48
为第i类样本的第j个中心，j的值与训练样本数相同；/>
Figure SMS_49
为平滑因子。in,
Figure SMS_47
is the matching degree determined by the neuron j of the i-th class in the hidden layer for the vector x input to the hidden layer; i=1, 2,..., M, M represents the number of categories in the training sample; />
Figure SMS_48
is the j-th center of the i-th sample, and the value of j is the same as the number of training samples; />
Figure SMS_49
is the smoothing factor.

本发明与现有技术相比具有如下特点：Compared with the prior art, the present invention has the following characteristics:

（1）本发明通过设置控制因子进行非线性递减，以跳出局部最优，提升了全局搜索能力。(1) In the present invention, the control factor is set to perform non-linear decrement to jump out of the local optimum and improve the global search ability.

（2）本发明对位置更新策略使用动态的自适应步长权重，不仅增强了算法的灵活性，还突出了最优狼的带领优势。(2) The present invention uses a dynamic adaptive step weight for the position update strategy, which not only enhances the flexibility of the algorithm, but also highlights the leading advantage of the optimal wolf.

（3）本发明采用改进后的灰狼算法优化平滑因子，并搭建概率神经网络模型，使得数据分类效果良好，准确率更高。(3) The present invention adopts the improved gray wolf algorithm to optimize the smoothing factor, and builds a probabilistic neural network model, so that the data classification effect is good and the accuracy rate is higher.

以下结合附图和具体实施方式对本发明的详细结构作进一步描述。The detailed structure of the present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

附图说明Description of drawings

附图1为本发明的概率神经网络电弧故障检测方法流程图；Accompanying drawing 1 is the flow chart of probabilistic neural network arc fault detection method of the present invention;

附图2为传统灰狼算法与改进灰狼算法控制因子

Figure SMS_50
的衰减曲线对比图；Attached Figure 2 shows the control factors of the traditional gray wolf algorithm and the improved gray wolf algorithm
Figure SMS_50
The attenuation curve comparison chart;

附图3为本发明方法对176组电流信号数据的验证集进行故障检测，其预测值和真实值的识别结果比较图；Accompanying drawing 3 is that the verification set of 176 groups of electric current signal data of the present invention carries out fault detection, its prediction value and the recognition result comparison figure of actual value;

附图4为本发明方法取值策略与线性衰减取值策略、余弦式衰减取值策略，在迭代过程中的迭代次数与步长权重变化对比曲线图；Accompanying drawing 4 is the value strategy of the present invention method and linear attenuation value strategy, cosine formula attenuation value strategy, the number of iterations and step size weight change comparison curve in the iterative process;

附图5为本发明方法与PNN、GWO-PNN的迭代次数与适应度值之间的曲线关系图；Accompanying drawing 5 is the curvilinear relationship figure between the number of iterations and the fitness value of the method of the present invention and PNN, GWO-PNN;

附图6为本发明方法与PNN、GWO-PNN的评价指标对比图。Accompanying drawing 6 is the comparison chart of the evaluation index of the method of the present invention and PNN, GWO-PNN.

具体实施方式Detailed ways

实施例一，如附图1-5所示，基于改进灰狼算法的概率神经网络电弧故障检测方法，具体包括如下步骤：Embodiment 1, as shown in accompanying drawing 1-5, based on the improved gray wolf algorithm probabilistic neural network arc fault detection method, specifically comprises the following steps:

步骤S1、获取家庭用电线路中的正常运行和电弧故障两种情况下的不同负载组合的电流信号数据集：利用现有的数据库收集正常运行和电弧故障电流信号数据，或者采用专门的数据生成机生成正常运行和电弧故障时的电流信号数据。Step S1. Obtain the current signal data sets of different load combinations in the normal operation and arc fault conditions of the household power line: use the existing database to collect the current signal data of normal operation and arc fault, or use special data generation The machine generates current signal data during normal operation and during arc faults.

步骤S2、对获取的电流信号数据集进行预处理：主要包括进行归一化处理，以及提取其时域特征值、频域特征值和能量特征值，并生成多维的特征向量。所述时域特征值包括波形指标、峰值指标、脉冲指标、峭度指标、裕度指标。所述频域特征值包括重心频率、频率方差和均方频率。Step S2. Preprocessing the acquired current signal data set: mainly including normalization processing, extracting its time-domain eigenvalues, frequency-domain eigenvalues and energy eigenvalues, and generating multi-dimensional eigenvectors. The time-domain characteristic value includes waveform index, peak index, pulse index, kurtosis index and margin index. The frequency-domain feature values include center of gravity frequency, frequency variance and mean square frequency.

其中归一化处理、时域特征值、频域特征值和能量特征值的提取均为现有技术，在此不再赘述。The normalization processing, the extraction of time-domain eigenvalues, frequency-domain eigenvalues, and energy eigenvalues are all prior art, and will not be repeated here.

步骤S3，基于改进的灰狼优化算法，对狼群中灰狼的位置参数进行参数优化。In step S3, based on the improved gray wolf optimization algorithm, parameter optimization is performed on the position parameters of gray wolves in the wolf pack.

S31，对狼群位置进行初始化，初始狼群位置根据下式产生：S31. Initialize the position of the wolves, and the initial position of the wolves is generated according to the following formula:

Figure SMS_51
（1）
Figure SMS_51
(1)

式中，

Figure SMS_52
为初始狼群位置，/>
Figure SMS_53
为[0,1]之间的随机数，/>
Figure SMS_54
为狼群位置的下界值，取值为0；/>
Figure SMS_55
为狼群位置的上界值，取值为1。In the formula,
Figure SMS_52
is the initial wolf pack position, />
Figure SMS_53
It is a random number between [0,1], />
Figure SMS_54
is the lower limit value of the position of wolves, and the value is 0; />
Figure SMS_55
is the upper limit value of the position of wolves, and the value is 1.

S32，计算每匹狼的适应度值，适应度函数的计算如下式所示：S32, calculate the fitness value of each wolf, the calculation of the fitness function is shown in the following formula:

Figure SMS_56
（2）
Figure SMS_56
(2)

其中，

Figure SMS_58
为发生电弧故障的数据中被正确检测出来的数量；/>
Figure SMS_59
为发生电弧故障的数据中被误判的数量；/>
Figure SMS_60
为正常运行的数据中被正确诊断的数量；/>
Figure SMS_61
为正常运行的数据中被误判为电弧故障的数量，本实施例中设置灰狼种群的数量为30。根据适应度值由大到小，将灰狼种群划分为/>
Figure SMS_62
，/>
Figure SMS_63
，/>
Figure SMS_64
和/>
Figure SMS_57
共４个等级。in,
Figure SMS_58
The number of correctly detected arc faults in the data; />
Figure SMS_59
is the number of misjudgments in the data of arc fault; />
Figure SMS_60
The number of correctly diagnosed in the normal operation data; />
Figure SMS_61
In this embodiment, the number of gray wolf populations is set to 30, which is the number of falsely judged as arc faults in the data of normal operation. According to the fitness value from large to small, the gray wolf population is divided into />
Figure SMS_62
, />
Figure SMS_63
, />
Figure SMS_64
and />
Figure SMS_57
There are 4 levels in total.

S33，计算在捕猎过程中狼群位置，每个个体的位置变化，公式如下：S33. Calculate the position of the wolf pack during the hunting process, and the position change of each individual, the formula is as follows:

Figure SMS_65
（3）
Figure SMS_65
(3)

式中，t为当前迭代次数；T为最大迭代次数；

Figure SMS_67
为灰狼i第t次迭代的位置；/>
Figure SMS_69
为猎物的当前位置；d为特征向量的维数；/>
Figure SMS_70
为步长权重，/>
Figure SMS_73
和/>
Figure SMS_74
分别为步长权重的最大值和最小值，；/>
Figure SMS_76
为步长动态因子，/>
Figure SMS_78
为t-1次迭代时，预测的猎物位置，/>
Figure SMS_66
为t次迭代时猎物的实际位置，这里把移动量最小的/>
Figure SMS_68
狼的位置近似看作本次迭代的猎物的实际位置，A和C为系数，其中/>
Figure SMS_71
是值在/>
Figure SMS_72
之间的数；/>
Figure SMS_75
为控制因子，设置控制因子在迭代早期的衰减速度较小，式中A的值波动较大，让灰狼群体有更广阔的搜索范围，有利于进行全局搜索，以跳出局部最优；在迭代后期的衰减速度较大，以提高算法的寻优效率和收敛速度，有利于快速得到最优解。本实施例中，最大迭代次数T设置为70，/>
Figure SMS_77
和/>
Figure SMS_79
的取值分别为1和0.4，传统灰狼算法与改进灰狼算法控制因子/>
Figure SMS_80
的衰减曲线对比图如附图2所示。In the formula, t is the current iteration number; T is the maximum iteration number;
Figure SMS_67
is the position of gray wolf i in the tth iteration; />
Figure SMS_69
is the current position of the prey; d is the dimension of the feature vector; />
Figure SMS_70
is the step weight, />
Figure SMS_73
and />
Figure SMS_74
are the maximum and minimum values of the step weight, respectively; />
Figure SMS_76
is the step size dynamic factor, />
Figure SMS_78
For t-1 iterations, the predicted prey position, />
Figure SMS_66
is the actual position of the prey during t iterations, where the smallest movement is taken
Figure SMS_68
The position of the wolf is approximately regarded as the actual position of the prey in this iteration, A and C are coefficients, where />
Figure SMS_71
is the value in />
Figure SMS_72
number between; />
Figure SMS_75
As the control factor, the decay rate of the control factor is set to be small in the early stage of the iteration, and the value of A in the formula fluctuates greatly, so that the gray wolf group has a wider search range, which is conducive to the global search to jump out of the local optimum; in the iteration The attenuation speed in the later period is larger to improve the optimization efficiency and convergence speed of the algorithm, which is conducive to quickly obtaining the optimal solution. In this embodiment, the maximum number of iterations T is set to 70, />
Figure SMS_77
and />
Figure SMS_79
The values of are 1 and 0.4 respectively, the traditional gray wolf algorithm and the improved gray wolf algorithm control factor/>
Figure SMS_80
The attenuation curve comparison chart is shown in Figure 2.

S34，通过狼群个体

Figure SMS_81
与三类带领狼之间的距离来确定灰狼个体如何向猎物移动，计算公式如下：S34, through individual wolf packs
Figure SMS_81
The distance between the three types of leading wolves is used to determine how individual gray wolves move towards prey, and the calculation formula is as follows:

Figure SMS_82
（4）
Figure SMS_82
(4)

式中，

Figure SMS_95
、/>
Figure SMS_97
和/>
Figure SMS_99
分别为第i只/>
Figure SMS_100
狼第t次迭代朝向灰狼/>
Figure SMS_101
，灰狼/>
Figure SMS_102
，灰狼/>
Figure SMS_103
的更新步长，/>
Figure SMS_83
、/>
Figure SMS_86
和/>
Figure SMS_88
分别为本次迭代中灰狼/>
Figure SMS_90
，灰狼/>
Figure SMS_92
，灰狼/>
Figure SMS_94
的位置；/>
Figure SMS_96
，/>
Figure SMS_98
，/>
Figure SMS_84
和/>
Figure SMS_85
，/>
Figure SMS_87
，/>
Figure SMS_89
为本次迭代产生的系数；/>
Figure SMS_91
为第t+1次迭代第i只/>
Figure SMS_93
狼的更新位置。In the formula,
Figure SMS_95
, />
Figure SMS_97
and />
Figure SMS_99
Respectively for the i-th />
Figure SMS_100
wolf t iteration towards gray wolf />
Figure SMS_101
, gray wolf/>
Figure SMS_102
, gray wolf/>
Figure SMS_103
The update step size, />
Figure SMS_83
, />
Figure SMS_86
and />
Figure SMS_88
Respectively, gray wolf/> in this iteration
Figure SMS_90
, gray wolf/>
Figure SMS_92
, gray wolf/>
Figure SMS_94
location; />
Figure SMS_96
, />
Figure SMS_98
, />
Figure SMS_84
and />
Figure SMS_85
, />
Figure SMS_87
, />
Figure SMS_89
Coefficients generated for this iteration; />
Figure SMS_91
For the t+1th iteration i-th />
Figure SMS_93
The updated position of the wolf.

S35，判断是否达到最大迭代次数，若达到则参数寻优结束，输出最优狼的最终位置参数；否则转向S33，利用灰狼更新后的位置参数继续进行迭代寻优。S35, judging whether the maximum number of iterations has been reached, if reached, the parameter optimization ends, and the final position parameter of the optimal wolf is output; otherwise, turn to S33, and use the updated position parameter of the gray wolf to continue iterative optimization.

步骤S4，利用基于改进灰狼优化算法搭建的概率神经网络模型：将基于改进的灰狼优化算法得到的最优狼的最终位置参数作为概率神经网络的平滑因子，搭建概率神经网络模型。Step S4, using the probabilistic neural network model built based on the improved gray wolf optimization algorithm: the final position parameter of the optimal wolf obtained based on the improved gray wolf optimization algorithm is used as the smoothing factor of the probabilistic neural network to build a probabilistic neural network model.

所述概率神经网络模型分为四层，分别为输入层、隐含层、求和层和输出层。输入层接收数据并将其传递给隐含层；隐含层计算特征向量与训练样本类别的匹配程度，其匹配程度计算公式如下式所示：The probabilistic neural network model is divided into four layers, namely an input layer, a hidden layer, a summation layer and an output layer. The input layer receives 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 as follows:

Figure SMS_104
（5）
Figure SMS_104
(5)

其中，

Figure SMS_105
为输入到隐含层的向量x经隐含层中第i类的神经元j确定的匹配度；i=1，2，…，M，M表示训练样本中的类别数，在这里M=2，M=1、2分别为正常运行、电弧故障两种类别；/>
Figure SMS_106
为第i类样本的第j个中心，j的值与训练样本数相同；/>
Figure SMS_107
为平滑因子，即步骤S3求得的最优狼的最终位置参数，它对概率神经网络模型的各项性能起着至关重要的作用，能够反映出概率神经网络模型的分类精度。in,
Figure SMS_105
is the matching degree determined by the vector x input to the hidden layer through the neuron j of the i-th class in the hidden layer; i=1, 2,..., M, 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; />
Figure SMS_106
is the j-th center of the i-th sample, and the value of j is the same as the number of training samples; />
Figure SMS_107
is the smoothing factor, that is, the final position parameter of the optimal wolf obtained in step S3, which plays a vital role in the performance of the probabilistic neural network model and can reflect the classification accuracy of the probabilistic neural network model.

完成匹配程度计算之后将结果发送到求和层；求和层对结果进行加权平均得到该类别的估计概率密度函数；输出层输出估计概率最高的类别作为分类结果。After completing the calculation of the matching degree, the result is sent to the summation layer; the summation layer performs weighted average on the results to obtain the estimated probability density function of the category; the output layer outputs the category with the highest estimated probability as the classification result.

步骤S5，获取线路中运行时的实时电流信号数据，对该实时电流信号数据进行步骤S2的处理后，输入至步骤S4搭建的概率神经网络模型的输入层，并进行计算得到故障诊断的分类结果。Step S5, obtain the real-time current signal data during operation in the line, after processing the real-time current signal data in step S2, input it to the input layer of the probabilistic neural network model built in step S4, and perform calculation to obtain the classification result of fault diagnosis .

如附图3所示，采用本实施例的基于改进灰狼算法的概率神经网络电弧故障检测方法，对176组电流信号数据的验证集进行故障检测，其预测值和真实值的识别结果比较图。其中0代表不存在电弧故障，1代表存在电弧故障。由附图3能够知道，本实施例的基于改进灰狼算法的概率神经网络电弧故障检测方法对验证集中电弧故障的识别率为97.7%，误判情况分别为2.3%和3.4%，误判概率非常少。As shown in Figure 3, using the probabilistic neural network arc fault detection method based on the improved gray wolf algorithm of this embodiment, the fault detection is performed on the verification set of 176 sets of current signal data, and the comparison chart of the recognition results between the predicted value and the real value . Where 0 means no arc fault exists and 1 means arc fault exists. It can be seen from accompanying drawing 3 that the probabilistic neural network arc fault detection method based on the improved gray wolf algorithm of this embodiment has a recognition rate of 97.7% for arc faults in the verification set, and the misjudgment cases are 2.3% and 3.4% respectively, and the misjudgment probability very few.

如附图4所示，为说明本实施例提出的步长权重取值策略相较其他的取值策略的优越性，绘制了本实施例提出的步长权重取值策略与线性衰减取值策略、余弦式衰减取值策略，在迭代过程中的迭代次数与步长权重变化对比曲线图。由附图4中能够看出相较之下，采用本实施例提出的步长权重取值策略在迭代初期维持在最大值状态，使狼群全局搜索的时间更长，搜索范围更大，在一定程度上避免了错漏搜索空间，过早陷入局部最优的情况。在迭代中期，步长权重的值设置为与预测的猎物位置和迭代猎物的实际位置相关的

Figure SMS_108
，采用本实施例提出的步长权重取值策略的曲线衰减速率较快，该/>
Figure SMS_109
的值反映当前预测准确度，其值越小准确度越高。当迭代到四十次时，采用本实施例提出的步长权重取值策略曲线就稳定在最小值状态。As shown in Figure 4, in order to illustrate the superiority of the step size weight value strategy proposed in this embodiment compared to other value strategies, the step size weight value strategy and the linear decay value strategy proposed in this embodiment are drawn , Cosine attenuation value strategy, the comparison curve of the number of iterations and the change of step weight in the iterative process. It can be seen from Figure 4 that by comparison, the step size weight value strategy proposed by this embodiment is maintained at the maximum value at the initial stage of iteration, so that the global search time of wolves is longer and the search range is larger. To a certain extent, it avoids the error and omission of the search space, and prematurely falls into the local optimal situation. In the middle of the iteration, the value of the step weight is set to be related to the predicted prey position and the actual position of the iterated prey.
Figure SMS_108
, the decay rate of the curve using the step size weight value strategy proposed in this embodiment is faster, the />
Figure SMS_109
The value of reflects the current prediction accuracy, and the smaller the value, the higher the accuracy. When the iteration reaches forty times, the curve of the step size weight value selection strategy proposed by this embodiment is stable at the minimum value state.

当采用本实施例提出的步长权重取值策略曲线，其迭代中期的步长权重与预测的猎物位置和迭代猎物的实际位置相关均相关，从而增强头狼的带领作用，令狼群在小范围内进行精细搜索。而在其他两种取值策略中，步长权重的取值只跟当前迭代次数有关，不能根据预测情况进行调整，使得早期会牺牲掉一定的搜索范围并且在后期无法充分发挥头狼的带领作用致使收敛速度缓慢。When adopting the step size weight value strategy curve proposed in this embodiment, the step size weight in the middle stage of the iteration is related to the predicted prey position and the actual position of the iterated prey, thereby enhancing the leading role of the alpha wolf and making the wolves in the small Perform a refined search within the range. In the other two value selection strategies, the value of the step weight is only related to the current number of iterations, and cannot be adjusted according to the forecast situation, so that a certain search range will be sacrificed in the early stage and the leading role of the wolf cannot be fully utilized in the later stage. resulting in slow convergence.

如附图5所示，为了进一步验证不同检测方法对电弧故障识别的有效性和优越性，分别构建了基于简单概率神经网络PNN、基于传统灰狼算法优化的概率神经网络模型GWO-PNN、以及基于改进灰狼算法的概率神经网络IGWO-PNN的迭代次数与适应度值之间的曲线关系图。其中参数条件分别为：在相同训练参数及历史数据下，对电弧故障检测模型进行训练，迭代次数为70次，学习率为0.001。由附图5中能够看出，相较于基于简单概率神经网络PNN、基于传统灰狼算法优化的概率神经网络模型GWO-PNN，基于改进灰狼算法的概率神经网络IGWO-PNN的适应度曲线达到最大的适应度值的时间较早且最大的适应度值的数值较高，由此可知其具有更快的收敛速度和更准确的分类效果。As shown in Figure 5, in order to further verify the effectiveness and superiority of different detection methods for arc fault identification, the simple probabilistic neural network PNN, the probabilistic neural network model GWO-PNN based on the optimization of the traditional gray wolf algorithm, and The curve relationship between the number of iterations and the fitness value of the probabilistic neural network IGWO-PNN based on the improved gray wolf algorithm. The parameter conditions are: under the same training parameters and historical data, the arc fault detection model is trained, the number of iterations is 70, and the learning rate is 0.001. As can be seen from Figure 5, compared with the probabilistic neural network model GWO-PNN based on the simple probabilistic neural network PNN and the traditional gray wolf algorithm optimization, the fitness curve of the probabilistic neural network IGWO-PNN based on the improved gray wolf algorithm The time to reach the maximum fitness value is earlier and the value of the maximum fitness value is higher, so it can be seen that it has a faster convergence speed and a more accurate classification effect.

如附图6所示，给出了分别采用基于简单概率神经网络PNN、基于传统灰狼算法优化的概率神经网络模型GWO-PNN、以及基于改进灰狼算法的概率神经网络IGWO-PNN方法的评价指标对比图。由附图6中能够看出，基于改进灰狼算法的概率神经网络IGWO-PNN方法的电弧故障检测准确率为97.159%，召回率为97.727%，精确率为96.629%，均高于另两种电弧故障检测的方法。由此可知，基于改进灰狼算法的概率神经网络IGWO-PNN方法识别电弧故障的能力较强。As shown in Figure 6, the evaluations of the probabilistic neural network model GWO-PNN based on the simple probabilistic neural network PNN, the optimized probabilistic neural network model GWO-PNN based on the traditional gray wolf algorithm, and the IGWO-PNN method based on the improved gray wolf algorithm are given. Index comparison chart. It can be seen from Figure 6 that the arc fault detection accuracy rate of the IGWO-PNN method based on the improved gray wolf algorithm is 97.159%, the recall rate is 97.727%, and the precision rate is 96.629%, which are higher than the other two methods. Methods of arc fault detection. It can be seen that the probabilistic neural network IGWO-PNN method based on the improved gray wolf algorithm has a strong ability to identify arc faults.

上述比较结果均给出了基于改进灰狼算法的概率神经网络IGWO-PNN方法的有效性和利用改进灰狼算法对平滑因子取值的必要性。该故障检测模型对电弧故障检测提供了有效诊断，有利于开展后续的电弧故障防治工作。The above comparison results show the effectiveness of the IGWO-PNN method based on the improved gray wolf algorithm and the necessity of using the improved gray wolf algorithm to select the smoothing factor. The fault detection model provides an effective diagnosis for arc fault detection, which is conducive to the subsequent prevention and control of arc faults.

Claims (8)

1.基于改进灰狼算法的概率神经网络电弧故障检测方法，其特征在于，包括如下步骤：1. The probabilistic neural network arc fault detection method based on the improved gray wolf algorithm is characterized in that, comprising the steps: 步骤S1、获取家庭用电线路中的正常运行和电弧故障两种情况下的不同负载组合的电流信号数据集；Step S1. Obtain current signal data sets of different load combinations under normal operation and arc fault conditions in the household power line; 步骤S2、对获取的电流信号数据集进行预处理；Step S2, preprocessing the acquired current signal data set; 步骤S3，基于改进的灰狼优化算法，对狼群中灰狼的位置参数进行参数优化，得到最优狼的最终位置参数；Step S3, based on the improved gray wolf optimization algorithm, optimize the parameters of the position parameters of the gray wolves in the wolf pack to obtain the final position parameters of the optimal wolf; 步骤S4，利用基于改进灰狼优化算法搭建的概率神经网络模型：将基于改进的灰狼优化算法得到的最优狼的最终位置参数作为概率神经网络的平滑因子，搭建概率神经网络模型；Step S4, using the probabilistic neural network model built based on the improved gray wolf optimization algorithm: using the final position parameter of the optimal wolf obtained based on the improved gray wolf optimization algorithm as the smoothing factor of the probabilistic neural network to build a probabilistic neural network model; 步骤S5，获取线路中运行时的实时电流信号数据，对该实时电流信号数据进行步骤S2的处理后，输入至步骤S4搭建的概率神经网络模型的输入层，并进行计算得到故障诊断的分类结果。Step S5, obtain the real-time current signal data during operation in the line, after processing the real-time current signal data in step S2, input it to the input layer of the probabilistic neural network model built in step S4, and perform calculation to obtain the classification result of fault diagnosis . 2.如权利要求1所述的基于改进灰狼算法的概率神经网络电弧故障检测方法，其特征在于：所述电流信号数据集利用现有的数据库收集正常运行和电弧故障电流信号数据，或者采用专门的数据生成机生成正常运行和电弧故障时的电流信号数据。2. the probabilistic neural network arc fault detection method based on improved gray wolf algorithm as claimed in claim 1, is characterized in that: described current signal dataset utilizes existing database to collect normal operation and arc fault current signal data, or adopts A dedicated data generator generates current signal data during normal operation and arc fault conditions. 3.如权利要求1所述的基于改进灰狼算法的概率神经网络电弧故障检测方法，其特征在于：所述预处理包括进行归一化处理，以及提取时域特征值、频域特征值和能量特征值，并生成多维的特征向量。3. the probabilistic neural network arc fault detection method based on improved gray wolf algorithm as claimed in claim 1, is characterized in that: described preprocessing comprises carrying out normalization process, and extraction time domain characteristic value, frequency domain characteristic value and Energy eigenvalues, and generate multidimensional eigenvectors. 4.如权利要求3所述的基于改进灰狼算法的概率神经网络电弧故障检测方法，其特征在于：所述时域特征值包括波形指标、峰值指标、脉冲指标、峭度指标、裕度指标；所述频域特征值包括重心频率、频率方差和均方频率。4. the probabilistic neural network arc fault detection method based on the improved gray wolf algorithm as claimed in claim 3, wherein: said time domain characteristic value comprises waveform index, peak index, pulse index, kurtosis index, margin index ; The frequency domain eigenvalues include center of gravity frequency, frequency variance and mean square frequency. 5.如权利要求1所述的基于改进灰狼算法的概率神经网络电弧故障检测方法，其特征在于：所述基于改进的灰狼优化算法，对狼群中灰狼的位置参数进行参数优化具体为，5. the probabilistic neural network arc fault detection method based on improved gray wolf algorithm as claimed in claim 1, is characterized in that: described based on improved gray wolf optimization algorithm, carries out parameter optimization to the position parameter of gray wolf in wolf pack specifically for, S31，对狼群位置进行初始化，初始狼群位置根据下式产生：S31. Initialize the position of the wolves, and the initial position of the wolves is generated according to the following formula:
Figure QLYQS_1
（1）
Figure QLYQS_1
(1) 式中，
Figure QLYQS_2
为初始狼群位置，/>
Figure QLYQS_3
为[0,1]之间的随机数，/>
Figure QLYQS_4
为狼群位置的下界值，取值为0；/>
Figure QLYQS_5
为狼群位置的上界值，取值为1；In the formula,
Figure QLYQS_2
is the initial wolf pack position, />
Figure QLYQS_3
It is a random number between [0,1], />
Figure QLYQS_4
is the lower limit value of the position of wolves, and the value is 0; />
Figure QLYQS_5
is the upper limit value of the position of wolves, and the value is 1; S32，计算每匹狼的适应度值，适应度函数的计算如下式所示：S32, calculate the fitness value of each wolf, the calculation of the fitness function is shown in the following formula:
Figure QLYQS_6
（2）
Figure QLYQS_6
(2) 其中，
Figure QLYQS_7
为发生电弧故障的数据中被正确检测出来的数量；/>
Figure QLYQS_8
为发生电弧故障的数据中被误判的数量；/>
Figure QLYQS_9
为正常运行的数据中被正确诊断的数量；/>
Figure QLYQS_10
为正常运行的数据中被误判为电弧故障的数量；in,
Figure QLYQS_7
The number of correctly detected arc faults in the data; />
Figure QLYQS_8
is the number of misjudgments in the data of arc fault; />
Figure QLYQS_9
The number of correctly diagnosed in the normal operation data; />
Figure QLYQS_10
The number of misjudged arc faults in the normal operation data; S33，计算在捕猎过程中狼群位置，每个个体的位置变化，公式如下：S33. Calculate the position of the wolf pack during the hunting process, and the position change of each individual, the formula is as follows:
Figure QLYQS_11
（3）
Figure QLYQS_11
(3) 式中，t为当前迭代次数；T为最大迭代次数；
Figure QLYQS_13
为灰狼i第t次迭代的位置；/>
Figure QLYQS_14
为猎物的当前位置；d为特征向量的维数；/>
Figure QLYQS_17
为步长权重，/>
Figure QLYQS_18
和/>
Figure QLYQS_20
分别为步长权重的最大值和最小值，；/>
Figure QLYQS_21
为步长动态因子，/>
Figure QLYQS_22
为第t-1次迭代时预测的猎物位置，/>
Figure QLYQS_12
为t次迭代时猎物的实际位置；A和C为系数，其中/>
Figure QLYQS_15
是值在/>
Figure QLYQS_16
之间的数；/>
Figure QLYQS_19
为控制因子；In the formula, t is the current iteration number; T is the maximum iteration number;
Figure QLYQS_13
is the position of gray wolf i in the tth iteration; />
Figure QLYQS_14
is the current position of the prey; d is the dimension of the feature vector; />
Figure QLYQS_17
is the step weight, />
Figure QLYQS_18
and />
Figure QLYQS_20
are the maximum and minimum values of the step weight, respectively; />
Figure QLYQS_21
is the step size dynamic factor, />
Figure QLYQS_22
is the predicted prey position at the t-1th iteration, />
Figure QLYQS_12
is the actual position of the prey at t iterations; A and C are coefficients, where />
Figure QLYQS_15
is the value in />
Figure QLYQS_16
number between; />
Figure QLYQS_19
is the control factor; S34，通过狼群个体
Figure QLYQS_23
与三类带领狼之间的距离来确定灰狼个体如何向猎物移动，计算公式如下：S34, through individual wolf packs
Figure QLYQS_23
The distance between the three types of leading wolves is used to determine how individual gray wolves move towards prey, and the calculation formula is as follows:
Figure QLYQS_24
（4）
Figure QLYQS_24
(4) 式中，
Figure QLYQS_37
、/>
Figure QLYQS_40
和/>
Figure QLYQS_41
分别为第i只/>
Figure QLYQS_42
狼第t次迭代朝向灰狼/>
Figure QLYQS_43
，灰狼/>
Figure QLYQS_44
，灰狼
Figure QLYQS_45
的更新步长，/>
Figure QLYQS_26
、/>
Figure QLYQS_28
和/>
Figure QLYQS_30
分别为本次迭代中灰狼/>
Figure QLYQS_32
，灰狼/>
Figure QLYQS_34
，灰狼/>
Figure QLYQS_35
的位置；/>
Figure QLYQS_38
，/>
Figure QLYQS_39
，/>
Figure QLYQS_25
和/>
Figure QLYQS_27
，/>
Figure QLYQS_29
，/>
Figure QLYQS_31
为本次迭代产生的系数；/>
Figure QLYQS_33
为第t+1次迭代第i只/>
Figure QLYQS_36
狼的更新位置；In the formula,
Figure QLYQS_37
, />
Figure QLYQS_40
and />
Figure QLYQS_41
Respectively for the i-th />
Figure QLYQS_42
wolf t iteration towards gray wolf />
Figure QLYQS_43
, gray wolf/>
Figure QLYQS_44
, gray wolf
Figure QLYQS_45
The update step size, />
Figure QLYQS_26
, />
Figure QLYQS_28
and />
Figure QLYQS_30
Respectively, gray wolf/> in this iteration
Figure QLYQS_32
, gray wolf/>
Figure QLYQS_34
, gray wolf/>
Figure QLYQS_35
location; />
Figure QLYQS_38
, />
Figure QLYQS_39
, />
Figure QLYQS_25
and />
Figure QLYQS_27
, />
Figure QLYQS_29
, />
Figure QLYQS_31
Coefficients generated for this iteration; />
Figure QLYQS_33
For the t+1th iteration i-th />
Figure QLYQS_36
The updated position of the wolf; S35，判断是否达到最大迭代次数，若达到则参数寻优结束，输出最优狼的最终位置参数；否则转向S33，利用灰狼更新后的位置参数继续进行迭代寻优。S35, judging whether the maximum number of iterations has been reached, if reached, the parameter optimization ends, and the final position parameter of the optimal wolf is output; otherwise, turn to S33, and use the updated position parameter of the gray wolf to continue iterative optimization. 6.如权利要求5所述的基于改进灰狼算法的概率神经网络电弧故障检测方法，其特征在于：所述控制因子在迭代早期的衰减速度小，在迭代后期的衰减速度大。6. The probabilistic neural network arc fault detection method based on the improved gray wolf algorithm according to claim 5, characterized in that: the attenuation speed of the control factor is small in the early stage of iteration, and the attenuation speed in the later stage of iteration is large. 7.如权利要求1所述的基于改进灰狼算法的概率神经网络电弧故障检测方法，其特征在于：所述概率神经网络模型分为四层，分别为输入层、隐含层、求和层和输出层；输入层接收数据并将其传递给隐含层，隐含层计算特征向量与训练样本类别的匹配程度，完成匹配程度计算之后将结果发送到求和层；求和层对结果进行加权平均得到该类别的估计概率密度函数；输出层输出估计概率最高的类别作为分类结果。7. the probabilistic neural network arc fault detection method based on improved gray wolf algorithm as claimed in claim 1, is characterized in that: described probabilistic neural network model is divided into four layers, is respectively input layer, hidden layer, summation layer and the output layer; the input layer receives 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 sends the result to the summation layer after completing the matching degree calculation; the summation layer calculates the result The weighted average obtains the estimated probability density function of the category; the output layer outputs the category with the highest estimated probability as the classification result. 8.如权利要求7所述的基于改进灰狼算法的概率神经网络电弧故障检测方法，其特征在于：所述匹配程度计算公式如下式所示：8. the probabilistic neural network arc fault detection method based on the improved gray wolf algorithm as claimed in claim 7, characterized in that: the matching degree calculation formula is as follows:
Figure QLYQS_46
（5）
Figure QLYQS_46
(5) 其中，
Figure QLYQS_47
为输入到隐含层的向量x经隐含层中第i类的神经元j确定的匹配度；i=1，2，…，M，M表示训练样本中的类别数；/>
Figure QLYQS_48
为第i类样本的第j个中心，j的值与训练样本数相同；/>
Figure QLYQS_49
为平滑因子。in,
Figure QLYQS_47
is the matching degree determined by the neuron j of the i-th class in the hidden layer for the vector x input to the hidden layer; i=1, 2,..., M, M represents the number of categories in the training sample; />
Figure QLYQS_48
is the j-th center of the i-th sample, and the value of j is the same as the number of training samples; />
Figure QLYQS_49
is the smoothing factor.
CN202310365153.5A 2023-04-07 2023-04-07 Probabilistic neural network arc fault detection method based on improved wolf algorithm Pending CN116383630A (en)

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 (2)

* Cited by examiner, † Cited by third party
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
CN117809300A (en) * 2023-12-29 2024-04-02 中国人民解放军陆军军医大学第二附属医院 Machine vision-based immunoelectrophoresis typing detection method and system

Cited By (3)

* Cited by examiner, † Cited by third party
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 Distribution network single-phase grounding fault detection method, system, device and storage medium
CN117809300A (en) * 2023-12-29 2024-04-02 中国人民解放军陆军军医大学第二附属医院 Machine vision-based immunoelectrophoresis typing detection method and system

Similar Documents

Publication Publication Date Title
CN113469253B (en) Electric larceny detection method based on triple twinning network
CN113255848B (en) Identification method of hydraulic turbine cavitation acoustic signal based on big data learning
CN116383630A (en) Probabilistic neural network arc fault detection method based on improved wolf algorithm
CN102903007B (en) Method for optimizing disaggregated model by adopting genetic algorithm
CN114239725B (en) Electric larceny detection method for data poisoning attack
CN114781573A (en) A Transformer Fault Diagnosis Method Based on Improved Grey Wolf Algorithm Optimization
CN110609200B (en) A Ground Fault Protection Method for Distribution Network Based on Fuzzy Metric Fusion Criterion
CN110879377B (en) Metering device fault tracing method based on deep belief network
CN111191835A (en) IES incomplete data load prediction method and system based on C-GAN transfer learning
CN102938562B (en) Prediction method of total wind electricity power in area
CN106203723A (en) Wind power short-term interval prediction method based on RT reconstruct EEMD RVM built-up pattern
CN115115090A (en) A Short-Term Prediction Method of Wind Power Based on Improved LSTM-CNN
CN105203869A (en) Microgrid island detection method based on extreme learning machine
CN112241605A (en) Method for identifying state of circuit breaker energy storage process by constructing CNN characteristic matrix through acoustic vibration signals
CN115712064B (en) An excitation system fault diagnosis method based on LSTM-CNN hybrid neural network
CN116224081A (en) Battery SOH estimation method based on LOF-Pearson detection multidimensional feature vector
CN106529741A (en) Space relevant characteristic-based ultra-short-period wind power prediction method
CN114049014A (en) Method, device and system for evaluating operating status of offshore wind turbines
CN116881714A (en) Grid false data injection detection method and device based on dual-branch CNN-LSTM
CN113420912B (en) A method for identifying users with abnormal low voltage in distribution network
CN114818817B (en) Weak fault identification system and method for capacitive voltage transformer
CN113762591B (en) Short-term electric quantity prediction method and system based on GRU and multi-core SVM countermeasure learning
Wang et al. Power quality disturbance recognition method in park distribution network based on one-dimensional VGGNet and multi-label classification
CN112949720B (en) Unknown radiation source identification method based on triple loss
Momani et al. Short-term load forecasting based on NARX and radial basis neural networks approaches for the Jordanian power grid

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