CN116912586A - Substation intrusion event identification method, system, equipment and medium - Google Patents

Substation intrusion event identification method, system, equipment and medium Download PDF

Info

Publication number
CN116912586A
CN116912586A CN202310911993.7A CN202310911993A CN116912586A CN 116912586 A CN116912586 A CN 116912586A CN 202310911993 A CN202310911993 A CN 202310911993A CN 116912586 A CN116912586 A CN 116912586A
Authority
CN
China
Prior art keywords
convolution
target
intrusion
dimensional image
intrusion event
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
CN202310911993.7A
Other languages
Chinese (zh)
Inventor
江虹
姜永翔
李家成
赵一涵
邵向鑫
张琪
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.)
Changchun University of Technology
Original Assignee
Changchun University of Technology
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 Changchun University of Technology filed Critical Changchun University of Technology
Priority to CN202310911993.7A priority Critical patent/CN116912586A/en
Publication of CN116912586A publication Critical patent/CN116912586A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
    • 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
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • 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
    • G06N3/048—Activation functions
    • 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
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Business, Economics & Management (AREA)
  • Mathematical Physics (AREA)
  • Molecular Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Medical Informatics (AREA)
  • Multimedia (AREA)
  • Databases & Information Systems (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Economics (AREA)
  • Public Health (AREA)
  • Water Supply & Treatment (AREA)
  • Human Resources & Organizations (AREA)
  • Marketing (AREA)
  • Primary Health Care (AREA)
  • Strategic Management (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a method, a system, equipment and a medium for identifying an intrusion event of a transformer substation, which relate to the field of intrusion event identification, and comprise the following steps: acquiring a target vibration invasion signal; the target vibration signal is obtained by monitoring an intrusion event outside a fence of the target transformer substation by adopting an optical fiber sensor array; converting the target vibration intrusion signal into a two-dimensional image by adopting a Markov transfer field to obtain a target two-dimensional image; the event recognition model is adopted to recognize and classify the target two-dimensional image, so that the category of the invasion event outside the fence of the target transformer substation is obtained; the intrusion event recognition model is constructed based on a CNN-VGG19 model. The invention can improve the accuracy of classifying noise events and non-noise events and reduce false alarms of intrusion events.

Description

Substation intrusion event identification method, system, equipment and medium
Technical Field
The present invention relates to the field of intrusion event recognition, and in particular, to a method, a system, an apparatus, and a medium for recognizing a substation intrusion event.
Background
Fiber optic sensor array perimeter or perimeter intrusion systems have evolved rapidly in recent years. The sensor has the advantages of high sensitivity, intrinsic safety required to be powered, lightning stroke resistance, various electromagnetic interferences and the like, and is widely applied to the fields of bridge vibration monitoring, underground pipeline transverse positioning, peripheral safety of important places such as airport oil reservoirs and the like.
The transformer substation is an important node in a power transmission system, and the safety of the transformer substation relates to safe production and normal power transmission, so that the stability of society and economy is affected. Traditional enclosure intrusion systems employ infrared fences, electronic fences, or video surveillance equipment, alone or in combination, for perimeter protection. The security measures are not enough when in use, and have prevention blind spots, such as high false alarm rate of an infrared peripheral fence, the electronic fence cannot warn invasion in advance and has poor positioning precision, the video monitoring density is low, the influence of weather and night illumination is large, most of substations are in open places, wind and sand are large, particularly the temperature difference between the summer and winter in northeast areas is large, the damage rate of monitoring equipment is high, and the security and theft prevention of power equipment and power lines are difficult to effectively execute.
In summary, some substations are located in suburban areas of the city, often subject to wind and rain, which may cause false positives in conventional enclosure intrusion systems. Thus, classification of how accurately noise events (e.g., beats, continuous shakes, continuous stomps, taps, etc.) and non-noise events (e.g., rain, wind, etc.) are achieved is particularly important for substation enclosure intrusion systems.
Disclosure of Invention
Based on the above, the embodiment of the invention provides a method, a system, equipment and a medium for identifying an intrusion event of a transformer substation, which can improve the accuracy of classifying noise events and non-noise events and reduce false alarms of the intrusion event.
In order to achieve the above object, the embodiment of the present invention provides the following solutions:
a substation intrusion event identification method, comprising:
acquiring a target vibration invasion signal; the target vibration signal is obtained by monitoring an intrusion event outside a fence of the target transformer substation by adopting an optical fiber sensor array;
converting the target vibration intrusion signal into a two-dimensional image by using a Markov transfer field (Markov Transition Field, MTF) to obtain a target two-dimensional image;
identifying and classifying the target two-dimensional image by adopting an event identification model to obtain the category of the intrusion event outside the fence of the target transformer substation; the intrusion event recognition model is constructed based on a CNN-VGG19 model.
Optionally, the method for determining the intrusion event recognition model specifically includes:
acquiring a training vibration invasion signal and a corresponding class label; the training vibration intrusion signal comprises: monitoring various different types of intrusion events outside the training transformer substation fence by adopting an optical fiber sensor array to obtain vibration signals;
converting the training vibration intrusion signal into a two-dimensional image by adopting a Markov transfer field to obtain a training two-dimensional image;
taking the training two-dimensional image and the corresponding class label as training data;
and training the CNN-VGG19 model by adopting the training data, and determining the trained CNN-VGG19 model as the intrusion event recognition model.
Optionally, the CNN-VGG19 model comprises: the input layer, the convolution module, the full connection module, the activation function layer and the output layer are sequentially connected;
the convolution module comprises: the five convolution units are sequentially connected and are respectively a first convolution unit, a second convolution unit, a third convolution unit, a fourth convolution unit and a fifth convolution unit;
the fully connected module comprises: the three full-connection layers are respectively a first full-connection layer, a second full-connection layer and a third full-connection layer;
the first convolution unit and the second convolution unit each comprise two convolution layers and a maximum pooling layer; the third convolution unit, the fourth convolution unit and the fifth convolution unit each comprise four convolution layers and a maximum pooling layer; the number of convolution kernels of the convolution layer in the latter convolution unit is not less than the number of convolution kernels of the convolution layer in the former convolution unit; the convolution kernels of the convolution layers in the five convolution units are all the same size.
Optionally, the number of convolution kernels of two convolution layers in the first convolution unit is 64; the number of convolution kernels of two convolution layers in the second convolution unit is 128; the number of convolution kernels of four convolution layers in the third convolution unit is 256; the number of convolution kernels of four convolution layers in the fourth convolution unit is 512; the number of convolution kernels of four convolution layers in the fifth convolution unit is 512;
the convolution kernel of the convolution layer in the five convolution units has a size of 3×3.
Optionally, the event recognition model is adopted to recognize and classify the target two-dimensional image, so as to obtain the category of the intrusion event outside the fence of the target substation, which specifically comprises the following steps:
modifying the size of the target two-dimensional image into a set size by adopting a cubic interpolation method;
and inputting the target two-dimensional image with the modified size into the event recognition model to obtain the category of the intrusion event outside the fence of the target transformer substation.
Optionally, training a CNN-VGG19 model by using the training data, and determining the trained CNN-VGG19 model as the intrusion event recognition model, which specifically includes:
modifying the size of the training two-dimensional image in the training data into a set size by adopting a cubic interpolation method;
and inputting the training two-dimensional image with the modified size and the corresponding category label into a CNN-VGG19 model to perform training with set iteration times, and determining the CNN-VGG19 model with the set iteration times as the intrusion event recognition model.
Optionally, acquiring the target vibration intrusion signal specifically includes:
acquiring a fiber Bragg grating vibration signal acquired by a fiber sensor array; the optical fiber sensor array is positioned on a fence of the target substation or buried in a peripheral area of the target substation;
demodulating the fiber Bragg grating vibration signal to obtain a target vibration invasion signal.
The invention also provides a transformer substation intrusion event recognition system, which comprises:
the intrusion signal acquisition module is used for acquiring a target vibration intrusion signal; the target vibration signal is obtained by monitoring an intrusion event outside a fence of the target transformer substation by adopting an optical fiber sensor array;
the image coding module is used for converting the target vibration intrusion signal into a two-dimensional image by adopting a Markov transfer field to obtain a target two-dimensional image;
the event recognition module is used for recognizing and classifying the target two-dimensional image by adopting an event recognition model to obtain the category of the intrusion event outside the fence of the target transformer substation; the intrusion event recognition model is constructed based on a CNN-VGG19 model.
The invention also provides electronic equipment, which comprises a memory and a processor, wherein the memory is used for storing a computer program, and the processor runs the computer program to enable the electronic equipment to execute the transformer substation intrusion event identification method.
The invention also provides a computer readable storage medium storing a computer program which when executed by a processor implements the substation intrusion event identification method described above.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
according to the embodiment of the invention, the Markov Transfer Field (MTF) is adopted to convert the target vibration intrusion signal into the two-dimensional image, the target two-dimensional image is obtained, then the event recognition model constructed based on the CNN-VGG19 model is adopted to recognize and classify the target two-dimensional image, and the category of the intrusion event outside the fence of the target transformer substation is obtained.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions of the prior art, the drawings that are needed in the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flowchart of a substation intrusion event identification method provided by an embodiment of the present invention;
FIG. 2 is a schematic diagram of a training vibration intrusion signal provided by an embodiment of the present invention;
FIG. 3 is a block diagram of a CNN-VGG19 model according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of a CNN-VGG19 model according to an embodiment of the present invention;
FIG. 5 is a graph of iteration number versus accuracy provided by an embodiment of the present invention;
FIG. 6 is a graph of iteration number versus loss rate provided by an embodiment of the present invention;
fig. 7 is a block diagram of a substation intrusion event identification system according to an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to the appended drawings and appended detailed description.
Example 1
Referring to fig. 1, the method for identifying a transformer substation intrusion event according to the present embodiment includes:
step 101: acquiring a target vibration invasion signal; the target vibration signal is obtained by monitoring an intrusion event outside the fence of the target transformer substation by adopting an optical fiber sensor array. Specific:
acquiring a fiber Bragg grating (Fiber Bragg Grating, FBG) vibration signal acquired by a fiber sensor array; the optical fiber sensor array is positioned on a fence of the target substation or buried in a peripheral area of the target substation. Demodulating the fiber Bragg grating vibration signal to obtain a target vibration invasion signal.
Step 102: and converting the target vibration intrusion signal into a two-dimensional image by adopting a Markov transfer field to obtain a target two-dimensional image.
Step 103: identifying and classifying the target two-dimensional image by adopting an event identification model to obtain the category of the intrusion event outside the fence of the target transformer substation; the intrusion event recognition model is constructed based on a CNN-VGG19 model. Specific:
modifying the size of the target two-dimensional image into a set size by adopting a cubic interpolation method; and inputting the target two-dimensional image with the modified size into the event recognition model to obtain the category of the intrusion event outside the fence of the target transformer substation.
In one example, the method for determining the intrusion event recognition model specifically includes:
(1) Acquiring a training vibration invasion signal and a corresponding class label; the training vibration intrusion signal comprises: and monitoring various different types of intrusion events outside the training transformer substation fence by adopting an optical fiber sensor array to obtain vibration signals.
For example, the fiber sensor array may be disposed on a fence of the training substation, for monitoring various intrusion events outside the fence of the training substation, including vibration of the fiber sensor array due to external interference: outside interference such as shaking the railing, knocking the railing, hand slapping the railing, demodulating the light signals reflected by different types of fiber gratings to obtain training vibration invasion signals, wherein fig. 2 shows several groups of training vibration invasion signals, the horizontal axis coordinates of fig. 2 show taking points, the vertical axis coordinates show amplitude values, the vibration signals corresponding to 6 types of invasion events are respectively six types of beating, continuous shaking, continuous stamping, beating, raining and wind blowing, in fig. 2, the amplitude values of the signals demodulated when knocking, hand slapping, shaking and stamping occur are seen to be changed drastically, the ascending attenuation rules of the waveforms of different types of signals with the increase of time are also different, the amplitude values of the vibration invasion signals of raining and heavy wind are reduced and are different as compared with the overall rule of knocking, but the vibration invasion signals are overlapped with the contrast of the signals of the like, and the signal identification classification can generate a certain false alarm rate by using the traditional mode. The next step requires preprocessing of the intrusion-like signals.
(2) And converting the training vibration intrusion signal into a two-dimensional image by adopting a Markov transfer field to obtain a training two-dimensional image, and taking the training two-dimensional image and a corresponding class label as training data. The process of converting the one-dimensional training vibration intrusion signal into a two-dimensional image is the process of preprocessing the intrusion signal.
(3) And training the CNN-VGG19 model by adopting the training data, and determining the trained CNN-VGG19 model as the intrusion event recognition model. Specific:
modifying the size of the training two-dimensional image in the training data into a set size by adopting a cubic interpolation method; and inputting the training two-dimensional image with the modified size and the corresponding category label into a CNN-VGG19 model to perform training with set iteration times, and determining the CNN-VGG19 model with the set iteration times as the intrusion event recognition model.
In this embodiment, the intrusion event is discriminated based on the VGG-19 model, the VGG-19 model has a very compact structure, and the network with increased depth is comprehensively estimated by using a very small (3×3) convolution filter architecture, and compared with the early convolution neural network, the VGG-19 uses a smaller convolution kernel, but uses more convolution layers to construct the depth network.
Specifically, the CNN-VGG19 model comprises: the system comprises an input layer, a convolution module, a full connection module, an activation function layer and an output layer which are connected in sequence. The convolution module comprises: the five convolution units which are sequentially connected are a first convolution unit, a second convolution unit, a third convolution unit, a fourth convolution unit and a fifth convolution unit respectively. The fully connected module comprises: the three full-connection layers which are connected in sequence are a first full-connection layer, a second full-connection layer and a third full-connection layer respectively. The activation function layer may employ a soft-max function.
The first convolution unit and the second convolution unit each comprise two convolution layers (Conv) and one maximum pooling layer (maxpool); the third convolution unit, the fourth convolution unit, and the fifth convolution unit each include four convolution layers (Conv) and one maximum pooling layer (maxpool); the number of convolution kernels of the convolution layer in the latter convolution unit is not less than the number of convolution kernels of the convolution layer in the former convolution unit; the convolution kernels of the convolution layers in the five convolution units are all the same size.
For example, as shown in fig. 3, the convolution kernels of the convolution layers in the five convolution units are each 3×3 in size. The number of convolution kernels of two convolution layers (Conv 3-64) in the first convolution unit is 64; the number of convolution kernels of two convolution layers (Conv 3-128) in the second convolution unit is 128; the number of convolution kernels of four convolution layers (Conv 3-256) in the third convolution unit is 256; the number of convolution kernels of four convolution layers (Conv 3-512) in the fourth convolution unit is 512; the number of convolution kernels of four convolution layers (Conv 3-512) in the fifth convolution unit is 512. The first full connection layer (FC-4096), the second full connection layer (FC-4096) and the third full connection layer (FC-1000) are sequentially connected, and finally, an activation function layer (soft-max) is adopted for output. The number of convolution kernels in the convolution layer is 64, 128, 256 and 512 in sequence from left to right, and the design ensures that the feature extraction capacity of the VGG-19 model is stronger, and meanwhile, the risk of overfitting is reduced.
As shown in figure 4 of the drawings, the input layer, the first convolution unit (Conv 3-64, maxpool), the second convolution unit (Conv 3-128, maxpool), the third convolution unit (Conv 3-256, maxpool) and the fourth convolution unit (Conv 3-512 Conv3-512, maxpool), a fifth convolution unit (Conv 3-512, maxpool), a first full connection layer (FC-4096), a second full connection layer (FC-4096), a third full connection layer (FC-1000), an activation function layer, and an output layer. The full-connection layer converts all feature matrixes of the maximum pooling layer into one-dimensional feature large vectors, and finally enters the output layer, so that the information output by the full-connection layer is converted into corresponding category probabilities, and the input images are classified. The layer represented by the dashed line in fig. 4 is the largest pooling layer in the corresponding convolution unit.
The CNN-VGG19 has a 19-layer structure (16 convolution layers and 3 full connection layers), the whole CNN-VGG19 model uses the convolution kernel size (3 multiplied by 3) and the maximum pooling layer (2 multiplied by 2) with the same size, and for a given receptive field, the use of stacked small convolution kernels is superior to the use of large convolution kernels because the stacking of multiple nonlinear layers can increase the depth of the network, thereby ensuring that the model can learn more complex patterns, and the parameters of the small convolution kernels are fewer, so the calculation cost of the whole network is smaller.
In addition, the intrusion event recognition model can be constructed based on a CNN-VGG16 model.
In order to improve the recognition rate of the intrusion event of the surrounding area of the transformer substation, the method for recognizing the intrusion event of the transformer substation is realized based on the combination of MTF and Convolutional Neural Network (CNN). The intrusion signals are complex and various, in order to better reflect the characteristics of the intrusion signals, the false alarm rate is reduced, the MTF is used for preprocessing the intrusion signals, one-dimensional signals are converted into two-dimensional images, deeper signal details can be presented compared with the traditional characteristic extraction mode, then the CNN-VGG19 model is used for identifying the two-dimensional characteristic images, and the intrusion events are classified.
The above-described process of acquiring and demodulating the vibration signal and the MTF employed for preprocessing the demodulated signal are described below.
1. Vibration signals are acquired and demodulated.
The optical fiber grating sensing array is used for collecting signals, the optical fiber grating sensing array sends external light into the modulation area through the incident optical fiber, and the modulation area is influenced by physical quantities such as external strong light and the like, so that certain characteristics of the light are changed, and the optical detector is used for detecting the signal light, so that measured parameters in the signals can be obtained.
When external vibration is sensed on the sensing optical fiber, the structure inside the optical fiber is changed, so that the phase of transmitted light is changed. The length L of the fiber is varied by the longitudinal strain effect, the refractive index n of the core is varied by the elasto-optical effect, and the core a is varied by the transverse poisson effect.
If no external vibration is sensed on the sensing fiber, the phase delay is calculated as follows:
where β represents the propagation constant of light in the fiber and λ represents the wavelength of light as it propagates. If the sensing optical fiber senses external vibration, the expression formula of the phase delay is as follows:
wherein, the liquid crystal display device comprises a liquid crystal display device,phase retardation is caused by longitudinal strain, elasto-optical effects and poisson effects.
Only the phase delay caused by the longitudinal strain is typically considered, and the other two are not considered. Therefore, the phase delay caused by the sensing fiber under the action of the vibration pressure P can be approximately obtained by the formula (2-3):
wherein the method comprises the steps ofIndicating the phase difference caused by the change of the optical fiber length, and +.>Indicating the phase difference resulting from the change in the refractive index of the core. Equation (2-4) represents a positive strain vector, ε, which is obtained by strain theory:
wherein the physical quantities represented by μ and E are poisson's constant and young's modulus of elasticity of the optical fiber, respectively. If there is no other change, only the change in the length of the optical fiber due to the longitudinal strain of the optical fiber is examined, and the Z-direction vector is taken to obtain:
where β=nk 0 And k is 0 Representative is the propagation constant of light in vacuum, then:
according to the elasto-optical effect, there are:
wherein p is ij Representing the elasto-tensor component. Since the sensing optical fiber is subjected to uniform vibration pressure, no shearing stress exists, thus epsilon j =0, j=4, 5,6, there are:
and then can obtain:
from formulas (2-6) and (2-9):
the change of the phase of the sensing optical fiber when the sensing optical fiber is subjected to vibration or pressure can be obtained by the formulas (2-3), (2-5) and (2-10), namely:
by adopting the detection mode of the photodetector, only the intensity of light can be obtained, and the phase difference of light cannot be obtained, so that the change of the optical phase can be obtained by adopting an interference method. Because the optical field is simple harmonic vibration, the field strengths of the signal light and the reference light can be expressed as the following formula:
E 1 ＝E 0 cos(ωt+s(t)+θ 1 ) (2-12)
E 2 ＝E 0 cos(ωt+θ 2 ) (2-13)
wherein E is 1 、E 2 Representing the field strengths of the signal light and the reference light, and E 0 The light field amplitude of the signal light and the reference light is represented, omega represents the angular frequency of the two light beams, theta 1 、θ 2 The initial phases of the two beams are shown, and s (t) is the phase modulation amount of the signal light generated by the external action.
Let I 1 Representing signal light, I 2 The light intensity of the reference light is represented, and the light intensity distribution of interference fringes formed by interference of two coherent light beams is represented as follows:
wherein Δθ=θ 1 -θ 2 Is the initial phase difference of the two beams of coherent light, I 0 Representing the total light intensity of the two beams of coherent light, and alpha represents the visibility coefficient of the interference fringes.
The final output electrical signal can be obtained by conversion by a photoelectric converter. It is expressed as:
i(t)＝KI 0 αcos(s(t)+Δθ) (2-15)
wherein K is a photoelectric conversion coefficient, the actually measured optical fiber vibration signals are collected by a data acquisition card in a mode of sampling for 2 seconds with a sampling rate of 750Hz, and a group of the optical fiber vibration signals are composed of 1500 sampling points.
2. The demodulated signal is preprocessed using MTF.
The demodulated intrusion signal is an irregular time domain signal and the demodulated intrusion-like signal is not stable, where the intrusion signal is converted into image codes using MTF, which converted image codes can manifest the characteristics of different kinds of intrusion signals more obviously.
MTF is a time-series image coding method based on markov transfer matrices. The method regards the time lapse of the time series as a markov process, i.e.: under the condition of knowing the current state, the future evolution of the method is independent of the previous evolution of the method, so that a Markov transfer matrix is constructed, the Markov transfer matrix is further expanded into a Markov transfer field, and image coding is realized.
First, a set of data sequences x= { X is given 1 ，x 2 ，x i ，...x n Discretizing X into Q quantile units, consisting of quantile Q j (j E {1,2, …, Q }) quantizing each value in a one-dimensional data sequence, the value x in the sequence by identifying quantiles i (i.epsilon. {1,2, …, n }) corresponds to a unique q j Wherein w is ij From q i Is q i The probability of following P determines:
w ij ＝P(x t ∈q i |x t-1 ∈q j )
if Q quantiles are used, then this matrix is q×q, resulting in a transfer matrix W:
it can be seen that the markov state transition matrix converted from the original time series data is less sensitive to the distribution of the original data and loses the time information, so the MTF transition matrix is constructed as follows. MTF, the markov transition field, denoted M, is an n×n matrix, N being the time length:
wherein q k Is x k Is divided into a plurality of position barrels q l Is x l And x is time sequence data.
Note that here M and W are not identical, the subscript in W is a state, and the subscript in M is a time in the time series data. In the sense of relative to W, M kl Is x k The located bitbucket is transferred to x l At which is locatedProbability of a split bucket. Such a representation may be easier to understand as follows:
M i,j||i-j|＝k
representing transition probabilities for k points in time interval, e.g. M i,j||i-j|＝1 The transition probabilities differing by one step on the time axis are represented. M on diagonal i,i It is a special case that means (k=0) the probability that each of the quantile buckets transitions to itself at point in time i. MTF represents a relationship between data at any two time points in time series data, and relatively gives whether or not they are often adjacent from a state point of view.
The effectiveness of the substation intrusion event identification method of the above embodiment is verified as follows.
The training process, the acquired intrusion signal is converted into 2310 two-dimensional images with 2000 x 2000 size by using MTF, according to 3: the ratio of 1 is divided into a training set and a test set, in order to avoid excessive parameters introduced in the training process, the image size is changed to 224 x 224 by using a cubic interpolation method, the initial learning rate is set to 0.00001, and the values at epoch {10, 20, 40, 80, 150} are attenuated to 90% of the previous values. The accuracy of the test was about 97% using a specific training procedure after 200 iterations as shown in fig. 5 and 6.
In the test process, an intrusion vibration signal is collected by an optical fiber sensing array set fixed on a fence, the collected vibration signal is converted into a two-dimensional image by using MTF, and the two-dimensional image is identified and classified by using a trained CNN-VGG19 model. The color picture with the size of 224 multiplied by 224 after three interpolation modification is input into a trained CNN-VGG19 model, firstly, the color picture enters into the first two layers of a structure, each layer uses 64 convolution kernels with the size of 3 multiplied by 3, the color picture enters into a maximum pooling layer after convolution, then the first part is processed to obtain the picture with the size of 112 multiplied by 112, the picture is input into the second layer with 128 convolution kernels with the size of 3 multiplied by 3, the color picture enters into the maximum pooling layer after convolution, the convolution is analogically performed according to the structure shown in fig. 4, each layer uses convolution kernels with the size of 3 multiplied by 3, all feature matrixes of the pooling layer are converted into one-dimensional feature large vectors by entering into a full-connection layer after final maximum pooling, and finally, the color picture enters into a prediction layer, so that the information output by the full-connection layer is converted into corresponding category probabilities, and the input images are classified.
The invention adopts the MTF method to preprocess the demodulated intrusion signals, converts the intrusion-like signals into characteristic images, and identifies and classifies the characteristic images in a deep learning (2 DCNN-VGG-19) mode, so as to distinguish disturbance signals, and experimental results show that the average identification rate of the scheme to the intrusion-like signals is 96.7%, the identification rate to noise events is 99%, the feasibility of the proposed scheme is verified, and the scheme can play an important role in improving the safety protection of transformer substations.
Example two
In order to execute a corresponding method of the above embodiment to achieve corresponding functions and technical effects, a substation intrusion event recognition system is provided below.
Referring to fig. 7, the system includes:
an intrusion signal acquisition module 701, configured to acquire a target vibration intrusion signal; the target vibration signal is obtained by monitoring an intrusion event outside the fence of the target transformer substation by adopting an optical fiber sensor array.
The image encoding module 702 is configured to convert the target vibration intrusion signal into a two-dimensional image by using a markov transfer field, so as to obtain a target two-dimensional image.
The event recognition module is used for recognizing and classifying the target two-dimensional image by adopting an event recognition model to obtain the category of the intrusion event outside the fence of the target transformer substation; the intrusion event recognition model is constructed based on a CNN-VGG19 model.
Example III
The embodiment provides an electronic device, including a memory and a processor, where the memory is configured to store a computer program, and the processor runs the computer program to enable the electronic device to execute the substation intrusion event identification method of the first embodiment.
Alternatively, the electronic device may be a server.
In addition, the embodiment of the invention also provides a computer readable storage medium, which stores a computer program, and the computer program realizes the substation intrusion event identification method of the first embodiment when being executed by a processor.
In the present specification, each embodiment is described in a progressive manner, and each embodiment is mainly described in a different point from other embodiments, and identical and similar parts between the embodiments are all enough to refer to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant points refer to the description of the method section.
The principles and embodiments of the present invention have been described herein with reference to specific examples, the description of which is intended only to assist in understanding the methods of the present invention and the core ideas thereof; also, it is within the scope of the present invention to be modified by those of ordinary skill in the art in light of the present teachings. In view of the foregoing, this description should not be construed as limiting the invention.

Claims (10)

1. A method for identifying an intrusion event of a substation, comprising:
acquiring a target vibration invasion signal; the target vibration signal is obtained by monitoring an intrusion event outside a fence of the target transformer substation by adopting an optical fiber sensor array;
converting the target vibration intrusion signal into a two-dimensional image by adopting a Markov transfer field to obtain a target two-dimensional image;
identifying and classifying the target two-dimensional image by adopting an event identification model to obtain the category of the intrusion event outside the fence of the target transformer substation; the intrusion event recognition model is constructed based on a CNN-VGG19 model.
2. The method for identifying the intrusion event of the transformer substation according to claim 1, wherein the method for determining the intrusion event identification model specifically comprises the following steps:
acquiring a training vibration invasion signal and a corresponding class label; the training vibration intrusion signal comprises: monitoring various different types of intrusion events outside the training transformer substation fence by adopting an optical fiber sensor array to obtain vibration signals;
converting the training vibration intrusion signal into a two-dimensional image by adopting a Markov transfer field to obtain a training two-dimensional image;
taking the training two-dimensional image and the corresponding class label as training data;
and training the CNN-VGG19 model by adopting the training data, and determining the trained CNN-VGG19 model as the intrusion event recognition model.
3. The substation intrusion event identification method according to claim 1, wherein the CNN-VGG19 model comprises: the input layer, the convolution module, the full connection module, the activation function layer and the output layer are sequentially connected;
the convolution module comprises: the five convolution units are sequentially connected and are respectively a first convolution unit, a second convolution unit, a third convolution unit, a fourth convolution unit and a fifth convolution unit;
the fully connected module comprises: the three full-connection layers are respectively a first full-connection layer, a second full-connection layer and a third full-connection layer;
the first convolution unit and the second convolution unit each comprise two convolution layers and a maximum pooling layer; the third convolution unit, the fourth convolution unit and the fifth convolution unit each comprise four convolution layers and a maximum pooling layer; the number of convolution kernels of the convolution layer in the latter convolution unit is not less than the number of convolution kernels of the convolution layer in the former convolution unit; the convolution kernels of the convolution layers in the five convolution units are all the same size.
4. The substation intrusion event identification method according to claim 3, wherein the number of convolution kernels of two convolution layers in the first convolution unit is 64; the number of convolution kernels of two convolution layers in the second convolution unit is 128; the number of convolution kernels of four convolution layers in the third convolution unit is 256; the number of convolution kernels of four convolution layers in the fourth convolution unit is 512; the number of convolution kernels of four convolution layers in the fifth convolution unit is 512;
the convolution kernels of the convolution layers in the five convolution units are all 3×3 in size.
5. The method for identifying the intrusion event of the transformer substation according to claim 3, wherein the event identification model is adopted to identify and classify the target two-dimensional image, so as to obtain the category of the intrusion event outside the fence of the target transformer substation, and the method specifically comprises the following steps:
modifying the size of the target two-dimensional image into a set size by adopting a cubic interpolation method;
and inputting the target two-dimensional image with the modified size into the event recognition model to obtain the category of the intrusion event outside the fence of the target transformer substation.
6. The substation intrusion event identification method according to claim 2, wherein the training data is used to train a CNN-VGG19 model, and the trained CNN-VGG19 model is determined as the intrusion event identification model, specifically comprising:
modifying the size of the training two-dimensional image in the training data into a set size by adopting a cubic interpolation method;
and inputting the training two-dimensional image with the modified size and the corresponding category label into a CNN-VGG19 model to perform training with set iteration times, and determining the CNN-VGG19 model with the set iteration times as the intrusion event recognition model.
7. The method for identifying a transformer substation intrusion event according to claim 1, wherein the step of acquiring the target vibration intrusion signal specifically comprises:
acquiring a fiber Bragg grating vibration signal acquired by a fiber sensor array; the optical fiber sensor array is positioned on a fence of the target substation or buried in a peripheral area of the target substation;
demodulating the fiber Bragg grating vibration signal to obtain a target vibration invasion signal.
8. A substation intrusion event identification system, comprising:
the intrusion signal acquisition module is used for acquiring a target vibration intrusion signal; the target vibration signal is obtained by monitoring an intrusion event outside a fence of the target transformer substation by adopting an optical fiber sensor array;
the image coding module is used for converting the target vibration intrusion signal into a two-dimensional image by adopting a Markov transfer field to obtain a target two-dimensional image;
the event recognition module is used for recognizing and classifying the target two-dimensional image by adopting an event recognition model to obtain the category of the intrusion event outside the fence of the target transformer substation; the intrusion event recognition model is constructed based on a CNN-VGG19 model.
9. An electronic device comprising a memory for storing a computer program and a processor that runs the computer program to cause the electronic device to perform the substation intrusion event identification method according to any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that it stores a computer program, which when executed by a processor implements the substation intrusion event identification method according to any one of claims 1 to 7.
CN202310911993.7A 2023-07-25 2023-07-25 Substation intrusion event identification method, system, equipment and medium Pending CN116912586A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310911993.7A CN116912586A (en) 2023-07-25 2023-07-25 Substation intrusion event identification method, system, equipment and medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310911993.7A CN116912586A (en) 2023-07-25 2023-07-25 Substation intrusion event identification method, system, equipment and medium

Publications (1)

Publication Number Publication Date
CN116912586A true CN116912586A (en) 2023-10-20

Family

ID=88362702

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310911993.7A Pending CN116912586A (en) 2023-07-25 2023-07-25 Substation intrusion event identification method, system, equipment and medium

Country Status (1)

Country Link
CN (1) CN116912586A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117830053A (en) * 2023-12-22 2024-04-05 中国电子信息产业集团有限公司第六研究所 Perimeter security alarm system and method
CN118051830A (en) * 2024-04-16 2024-05-17 齐鲁工业大学(山东省科学院) Perimeter security intrusion event identification method

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117830053A (en) * 2023-12-22 2024-04-05 中国电子信息产业集团有限公司第六研究所 Perimeter security alarm system and method
CN118051830A (en) * 2024-04-16 2024-05-17 齐鲁工业大学(山东省科学院) Perimeter security intrusion event identification method

Similar Documents

Publication Publication Date Title
US11562224B2 (en) 1D-CNN-based distributed optical fiber sensing signal feature learning and classification method
CN105931402B (en) Optical fiber perimeter intrusion detection method based on image identification
CN101556724B (en) Safety management system of optical fiber perimeter and pattern recognition method thereof
CN111157099B (en) Distributed optical fiber sensor vibration signal classification method and identification classification system
CN110570613A (en) Fence vibration intrusion positioning and mode identification method based on distributed optical fiber system
CN102045120B (en) Vibration signal identification method for optical fiber perimeter system
CN103617684A (en) Interference type optical fiber perimeter vibration intrusion recognition algorithm
CN114692681B (en) SCNN-based distributed optical fiber vibration and acoustic wave sensing signal identification method
CN116912586A (en) Substation intrusion event identification method, system, equipment and medium
Lyu et al. Abnormal events detection based on RP and inception network using distributed optical fiber perimeter system
CN114510960A (en) Method for recognizing distributed optical fiber sensor system mode
CN111104891A (en) Composite characteristic optical fiber sensing disturbing signal mode identification method based on BiLSTM
Tian et al. Intrusion signal recognition in OFPS under multi-level wavelet decomposition based on RVFL neural network
Wang et al. Multi-branch long short-time memory convolution neural network for event identification in fiber-optic distributed disturbance sensor based on φ-OTDR
CN111951505B (en) Fence vibration intrusion positioning and mode identification method based on distributed optical fiber system
Marie et al. A hybrid model integrating MPSE and IGNN for events recognition along submarine cables
CN116186642B (en) Distributed optical fiber sensing event early warning method based on multidimensional feature fusion
CN116091897A (en) Distributed optical fiber sensing event identification method and system based on light weight
Liang Study on the fiber-optic perimeter sensor signal processor based on neural network classifier
CN114354744B (en) Power cable external damage prevention monitoring method, system and device
CN106404016A (en) Fiber sampling signal filtering method and device and fiber sensing system
CN115060184A (en) Optical fiber perimeter intrusion detection method and system based on recursive graph
Wu et al. Intelligent target recognition for distributed acoustic sensors by using both manual and deep features
Bublin Machine learning for distributed acoustic sensors, classic versus image and deep neural networks approach
CN103903270B (en) Regularized valid characteristic section selecting method of optical fiber link monitoring signals

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