CN117496650B - Distributed optical fiber intrusion early warning method and system based on environment embedding - Google Patents
Distributed optical fiber intrusion early warning method and system based on environment embedding Download PDFInfo
- Publication number
- CN117496650B CN117496650B CN202410000986.6A CN202410000986A CN117496650B CN 117496650 B CN117496650 B CN 117496650B CN 202410000986 A CN202410000986 A CN 202410000986A CN 117496650 B CN117496650 B CN 117496650B
- Authority
- CN
- China
- Prior art keywords
- intrusion
- feature vector
- vector
- optical fiber
- environment
- 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.)
- Active
Links
- 239000013307 optical fiber Substances 0.000 title claims abstract description 60
- 238000000034 method Methods 0.000 title claims abstract description 25
- 239000013598 vector Substances 0.000 claims abstract description 137
- 230000007613 environmental effect Effects 0.000 claims abstract description 42
- 238000004364 calculation method Methods 0.000 claims abstract description 31
- 230000003287 optical effect Effects 0.000 claims abstract description 30
- 230000007123 defense Effects 0.000 claims abstract description 19
- 238000007781 pre-processing Methods 0.000 claims abstract description 17
- 238000013528 artificial neural network Methods 0.000 claims abstract description 15
- 238000001514 detection method Methods 0.000 claims abstract description 15
- 238000012545 processing Methods 0.000 claims description 12
- 230000004927 fusion Effects 0.000 claims description 7
- 239000000835 fiber Substances 0.000 claims description 6
- 230000009466 transformation Effects 0.000 claims description 4
- 238000012549 training Methods 0.000 description 10
- 230000006870 function Effects 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 239000011159 matrix material Substances 0.000 description 5
- 125000004122 cyclic group Chemical group 0.000 description 4
- 238000000354 decomposition reaction Methods 0.000 description 4
- 238000007689 inspection Methods 0.000 description 4
- 238000007726 management method Methods 0.000 description 4
- 238000012544 monitoring process Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 230000004913 activation Effects 0.000 description 3
- 230000009467 reduction Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000005856 abnormality Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 239000012530 fluid Substances 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000000253 optical time-domain reflectometry Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B13/00—Burglar, theft or intruder alarms
- G08B13/18—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength
- G08B13/181—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using active radiation detection systems
- G08B13/183—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using active radiation detection systems by interruption of a radiation beam or barrier
- G08B13/186—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using active radiation detection systems by interruption of a radiation beam or barrier using light guides, e.g. optical fibres
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
-
- 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
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
- G06F18/2148—Generating training patterns; Bootstrap methods, e.g. bagging or boosting characterised by the process organisation or structure, e.g. boosting cascade
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2431—Multiple classes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/20—Ensemble learning
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2123/00—Data types
- G06F2123/02—Data types in the time domain, e.g. time-series data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
- G06F2218/06—Denoising by applying a scale-space analysis, e.g. using wavelet analysis
-
- 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
- G06F2218/10—Feature extraction by analysing the shape of a waveform, e.g. extracting parameters relating to peaks
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Medical Informatics (AREA)
- Probability & Statistics with Applications (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Burglar Alarm Systems (AREA)
Abstract
The invention discloses a distributed optical fiber intrusion early warning method and system based on environment embedding, comprising the following steps: acquiring reverse Rayleigh scattering light signal data in a detection optical fiber; preprocessing the optical signal data to obtain a daily defense area alarm sequence; dividing an alarm sequence, introducing an embedded vector, and embedding the embedded vector into the alarm sequence; inputting the alarm sequence added with the embedded vector into a first neural network to obtain an environment characteristic vector; performing first intrusion alarm probability calculation based on the environmental feature vector and the time feature vector; performing second intrusion alarm probability calculation based on the environmental feature vector, the time feature vector and the first intrusion alarm probability, and performing intrusion early warning based on the calculation result; according to the invention, the full-space distribution of the intrusion event is modeled by fusing the environmental feature vector and the time feature vector, so that the accuracy of the distributed optical fiber early warning is improved.
Description
Technical Field
The invention relates to the technical field of pipeline monitoring, in particular to a distributed optical fiber intrusion early warning method and system based on environment embedding.
Background
In recent years, distributed optical fiber intrusion early warning systems have attracted extensive attention in the fields of perimeter security, pipeline monitoring and the like. The pipeline is one of the most important transportation modes of oil gas, but different from single scene environments in the fields of security protection, high-speed rails and the like, the environments along the long-distance pipeline are complex and changeable, including plain, mountain areas, lakes and the like. Different geological conditions have different effects on the propagation of vibrations, which presents a great challenge for the external perception of the distributed optical fiber sensing system.
However, the prior art mostly does not consider the influence of the spatial environment information on intrusion recognition. The natural and artificial environment surrounding the defense area has an important influence on the modal distribution of the vibration waveform. For example, in a guard area near a farmland, the average amplitude of the waveform is about 50, whereas a guard area near a high speed may have an average amplitude of 2000. The same model can not realize effective intrusion detection in farmland and nearby high speed, and the usability and practicability of the model are severely limited. Therefore, the application of the distributed optical fiber intrusion early warning technology to long-distance pipelines needs to overcome the influence of environmental factors.
For example, chinese patent CN201810475691.9 discloses an on-line monitoring system for fluid delivery pipes based on optical fiber sensing, which links the on-line detection of pipes, the management of pipe information and the inspection of pipes together, and after the optical fiber sensing pipe detection subsystem detects that the pipe has an abnormality, sends alarm information to the pipe information management subsystem, the pipe information management subsystem generates an inspection task and sends the task to the pipe inspection subsystem, the pipe inspection subsystem takes measures in time to process the abnormal pipe, and returns the execution result of the task to the pipe information management subsystem; however, the system still does not overcome the influence of the application of the optical fiber intrusion early warning technology on the long-distance pipeline on environmental factors.
Disclosure of Invention
The invention mainly solves the problem of low early warning accuracy caused by the influence of environmental factors in the optical fiber intrusion early warning technology in the prior art; the distributed optical fiber intrusion early warning method and system based on environment embedding are provided, the pipeline intrusion detection precision is improved, and the early warning accuracy is high.
The technical problems of the invention are mainly solved by the following technical proposal: a distributed optical fiber intrusion early warning method based on environment embedding comprises the following steps:
acquiring reverse Rayleigh scattering light signal data in a detection optical fiber;
preprocessing the optical signal data to obtain a daily defense area alarm sequence;
dividing an alarm sequence, introducing an embedded vector, and embedding the embedded vector into the alarm sequence;
inputting the alarm sequence added with the embedded vector into a first neural network to obtain an environment characteristic vector;
performing first intrusion alarm probability calculation based on the environmental feature vector and the time feature vector;
and performing second intrusion alarm probability calculation based on the environmental feature vector, the time feature vector and the first intrusion alarm probability, and performing intrusion early warning based on the calculation result.
Preferably, the method for preprocessing the optical signal data comprises the following steps:
performing discrete wavelet transformation on the initial optical signal data to obtain wavelet coefficients;
modifying wavelet coefficients by threshold rules;
wavelet reconstruction is performed using an inverse discrete wavelet transform based on the modified wavelet coefficients.
Preferably, the alarm sequence is divided by using a window K.
Preferably, the embedded vector is determined according to the environmental weight coefficient and the alarm sequence of the current day.
In another aspect, the present invention further provides an environment-embedded distributed optical fiber intrusion early warning system, including: the optical signal transmitting unit modulates the light source and transmits pulse light to the circulator; the circulator is used for transmitting the pulse light to the detection optical fiber and obtaining the reverse Rayleigh scattering light signal data of the detection optical fiber; the data acquisition module acquires reverse Rayleigh scattering light signal data of the circulator and transmits the data to the processing module; the processing module is used for preprocessing the optical signal data to obtain a daily defense area alarm sequence; dividing an alarm sequence, introducing an embedded vector, and embedding the embedded vector into the alarm sequence; inputting the alarm sequence added with the embedded vector into a first neural network to obtain an environment characteristic vector; performing first intrusion alarm probability calculation based on the environmental feature vector and the time feature vector; and performing second intrusion alarm probability calculation based on the environmental feature vector, the time feature vector and the first intrusion alarm probability, and performing intrusion early warning based on the calculation result.
Preferably, the optical signal transmitting unit includes: a narrow linewidth laser producing a continuous light source; an acousto-optic modulator modulating continuous light of the narrow linewidth laser into pulse light; and the optical fiber amplifier is used for amplifying the pulse light transmitted by the acousto-optic modulator.
Preferably, the processing module includes: a preprocessing unit for preprocessing the optical signal data; an environmental feature vector generation unit generating a plurality of environmental feature vectors using the preprocessed optical signal data; the optical fiber intrusion early warning unit is used for calculating the probability of the first intrusion alarm based on the environmental feature vector and the time feature vector; and performing second intrusion alarm probability calculation based on the environmental feature vector, the time feature vector and the first intrusion alarm probability, and performing intrusion early warning based on the calculation result.
Preferably, the optical fiber intrusion early warning unit includes: the time feature module stores time feature vectors; the environment feature module stores environment feature vectors; the space-time feature judging module is used for carrying out first intrusion alarm probability calculation based on the XGBoost model according to the input time feature vector and the environment feature vector; and the fusion module takes the time feature vector, the environment feature vector and the first intrusion alarm probability as input, adopts an MLP network model to calculate the second intrusion alarm probability, and obtains an intrusion early warning result.
The beneficial effects of the invention are as follows: by introducing an embedded vector into an alarm sequence signal, obtaining an environment vector similar to an actual environment after model training, fully excavating environment information around a defending area, effectively characterizing the environment vector, modeling the whole space distribution of an intrusion event by fusing an environment feature vector and a time feature vector, designing an integrated learning model based on environment embedding and time sequence features aiming at random noise and environment interference, greatly improving the generalization capability and practicability of the model, and improving the accuracy of distributed optical fiber early warning.
Drawings
Fig. 1 is a flow chart of a distributed optical fiber intrusion early warning method according to an embodiment of the invention.
FIG. 2 is a schematic diagram of a fiber optic data intercept segment according to an embodiment of the present invention.
Fig. 3 is a waveform diagram of a wavelet high frequency coefficient according to an embodiment of the present invention.
Fig. 4 is a schematic diagram of signals before denoising according to an embodiment of the present invention.
Fig. 5 is a schematic diagram of a denoised signal according to an embodiment of the present invention.
Fig. 6 is a schematic diagram of training loss variation of an MLP network model according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, further detailed description of the technical solutions in the embodiments of the present invention will be given by the following examples with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
Examples: as shown in FIG. 1, the distributed optical fiber intrusion early warning method specifically comprises the following steps:
s1: acquiring reverse Rayleigh scattering light signal data in a detection optical fiber; the acquisition frequency of the data of the reverse Rayleigh scattered light signal is 2kHz.
S2: preprocessing the optical signal data to obtain a daily defense area alarm sequence; firstly, all signals are subjected to noise reduction, and because random noise such as system noise, environmental interference and the like exists in the distributed optical fiber sensing system, the waveform data is required to be subjected to noise reduction so as to improve the signal to noise ratio of the system.
The method for preprocessing the optical signal data comprises the following steps:
s21: performing discrete wavelet transformation on the initial optical signal data to obtain wavelet coefficients;
s22: modifying wavelet coefficients by threshold rules; the invention is realized by adopting a soft threshold value;
s23: wavelet reconstruction is performed using an inverse discrete wavelet transform based on the modified wavelet coefficients.
Specifically, as shown in fig. 2-5, 6000 continuous values are intercepted from real-time data of the optical fiber for explanation, the real-time data of the optical fiber contains a large amount of noise, and it can be seen that the waveform is not smooth and has a lot of clutters, and the wavelet transformation denoising is performed first.
The data is subjected to wavelet decomposition, five layers of wavelet decomposition are performed by using 8-order Symlet wavelet, and the decomposition high-frequency coefficient result is shown in figure 3. After five levels of wavelet are obtained, the high frequency coefficients need to be thresholded. The soft threshold is obtained by subtracting the threshold value when the wavelet threshold value is larger than the threshold value, and setting the wavelet to 0 when the wavelet threshold value is smaller than the threshold value, as shown in the following formula:
wherein,is wavelet coefficient +.>Is a threshold value. The signal obtained through the soft threshold function has better continuity and is not easy to generate large fluctuation. />Selecting a VisuShrink threshold, and calculating by the following formula
Wherein N is the number of wavelet layers, ,/>the value of (2) is set to an intermediate value of the absolute value of the first layer wavelet decomposition coefficient. Finally, the reconstruction is performed through inverse discrete wavelet transform to obtain a noise-reduced signal, as shown in fig. 5. By usingThe noise reduction effect was evaluated by signal-to-noise ratio (SNR) and calculated as follows:
for maximum amplitude of the effective vibration signal, +.>Is the maximum amplitude of the background noise. The calculated SNR was 7.27db.
The natural and artificial environment surrounding the defense area has a significant impact on the waveform distribution of the event. For example, the a defence area is adjacent to a highway and traffic flow is large. The vibration amplitude of the optical fiber is generally maintained at a high level of 2000. In contrast, the B guard zone is immediately adjacent to the farmland and the surrounding environment is relatively quiet. The vibration amplitude of the optical fiber is generally about 50. For a vibration segment with an a-defense area amplitude of about 2000, a dangerous intrusion event cannot be accurately identified. While for a vibration segment with an amplitude of about 2000 for the B defense area, an intrusion event is considered to have occurred. It can be seen that environmental characteristics are important factors in determining intrusion detection accuracy.
S3: dividing an alarm sequence, introducing an embedded vector, and embedding the embedded vector into the alarm sequence; embedding is a distributed representation method that converts raw input data into a linear combination of a series of features. The method solves the problem of oversized representation dimensions and provides very efficient representation capabilities. Embedding has a wide range of applications in natural language processing, for representing basic characters and words. Embedding techniques based on single thermal coding train entity embedding vectors with meaningful sequence prediction tasks. Compared with Shan Re coding, the embedded vector has the characteristics of flexible dimension setting and high feature abstraction. In a distributed fiber optic system, it is theoretically possible to set a class-based Shan Re code for each defense area. For example, categories such as expressways, highways, farmlands, factories, and the like may be set. The single thermal code is then determined from the surrounding of the particular defensive area. However, this method of extracting environmental features is inefficient and has limited accuracy for the following reasons:
(1) The surrounding environment of a plurality of defense areas is investigated, and a large amount of manpower and material resources are consumed;
(2) The optimal category setting cannot be determined;
(3) A prejudice to humans is introduced.
Therefore, the invention provides an environment embedding technology, and reasonable and efficient environment representation vectors are set for specific defense areas. The specific implementation steps are as follows:
an alarm sequence for a distributed fiber optic monitoring system having n defense areas may be expressed as a determination of an alarm based on the day of the defense area. Assuming that a t-day alarm sequence result is provided, a k window segmentation alarm sequence is set, and an embedded vector is determined according to an environment weight coefficient and the alarm sequence of the same day, a weight coefficient matrix of an environment embedded layer can be expressed as follows:
wherein the t-th time step embeds the vectorCan be expressed as +.>Embedding the whole alarm sequence to obtain an embedding matrix +.>The context embedding vector, in which each row corresponds to a time step, is expressed as:
s4: inputting the alarm sequence added with the embedded vector into a first neural network to obtain an environment characteristic vector; will add the embedded vectorRNN neural network training is carried out by utilizing the defense areas at two sides of the sequence to predict the middle defense area, and training a three-layer neural network to obtain the environmental feature vector of the defense area, wherein the environmental embedding vector and the environmental feature vector are respectively mentioned, the environmental embedding vector is the environmental feature vector before the neural network training, a three-layer RNN cyclic neural network model is introduced, and the hidden state of the model is->The update rule may be expressed as:
wherein,is a weight matrix from hidden state to hidden state, < +.>Is a weight matrix of environment embedded vectors to hidden states, < >>Is a sigmoid activation function.
The cyclic neural network (Recurrent Neural Network, RNN) is a type of cyclic neural network which takes sequence data as input, performs recursion (recovery) in the evolution direction of the sequence, and all nodes (cyclic units) are connected in a chained mode, wherein the output of the cyclic neural network can be one-to-one single output or one-to-one single output, and can also be output according to a sequence corresponding sequence.
Further, the output of the RNN may be used to measure the similarity of the alarm sequences, expressed as:
wherein,、/>output values of hidden layers, which are RNN, ">The similarity measurement value can be used for comparing the similarity between alarm sequences on different days to extract the similarity characteristics of the alarm sequences.
The aim of the whole process of training the environmental feature vector by utilizing the RNN network model is to enable the learned environmental feature vector to well express the similarity between alarm sequences by adjusting the parameters of an embedded layer and the model. In the training process, the loss function is minimized through a back propagation algorithm, model parameters are optimized, and the model parameters mainly comprise the input of a model, a weight matrix and the like.
The method comprises the steps of selecting an alarm sequence according to actual observation, wherein a defending area which is frequently alarmed at about 8 a.m. is positioned near a highway or a highway and other areas with large traffic flow. The environment feature vector can span the ground surface positioning features such as roadsides, high-speed edges and the like, so that the high-level abstract features of large traffic flow are deeply excavated, and the high-level abstract features have stronger characterization and generalization capabilities.
S5: performing first intrusion alarm probability calculation based on the environmental feature vector and the time feature vector; the time feature vector acquisition method comprises the following steps: and (3) calculating and combining characteristic values of the windowed waveforms to obtain a time characteristic vector, and extracting high-level time characteristics by adopting a time characteristic module, wherein the extracted high-level time characteristics comprise peak values, minimum values, energy, average values, variances, peak-to-peak values, root mean square, standard deviations, peak factors, skewness factors, waveform factors, pulse factors and margin factors.
And storing a plurality of time feature vectors in the time feature module, selecting the time feature vectors through the environment feature vectors, and similarly, summarizing and storing the environment feature vectors in the environment feature module, selecting the environment feature vectors through the time feature vectors, and obtaining a probability value through an n-layer XGBoost model by combining the time sequence feature vectors and the environment embedded vectors as input by the space-time feature judging module. The introduction of the module can significantly improve the final recognition accuracy. In order to avoid random noise interference, the invention takes n=10, and the overall output can be expressed as:
wherein,indicating the output of the i-th layer. The objective function is expressed as:
where n is the number of samples,alarm prediction result representing sample i at time t, < >>Is the alarm prediction value given by the model of step t-1, < >>Is the predicted value of the new model that needs to be added in step t,/->Representing regularization terms. The first intrusion alert probability is derived based on the output of this module.
S6: and performing second intrusion alarm probability calculation based on the environmental feature vector, the time feature vector and the first intrusion alarm probability, and performing intrusion early warning based on the calculation result. If the value of the second intrusion alarm probability exceeds the early warning threshold, performing intrusion early warning prompt, otherwise, returning to the step S5 to reselect the environment feature vector and the time feature vector to perform new intrusion alarm probability calculation until all intrusion alarm probabilities of all environment feature vectors and all time feature vectors are calculated.
The fusion module receives the outputs of the first three modules as input variablesFeature fusion is performed through a multi-layer neural network. The MLP network model can be expressed as:
wherein H represents a hidden layer, O represents an output layer,representing the input variable +.>;/>,/>,/>,Respectively representing network layer parameters,/-, and>representing a classifier; />Representing the activation function, since the alarm is judged to be a two-class problem, using the sigmoid function as the activation function, the result can be mapped to the (0, 1) interval, expressed as:
finally, model training is carried out through the optical fiber alarm data set, and the model training process is shown in fig. 6.
The invention also provides a distributed optical fiber intrusion early warning system based on environment embedding, which comprises an optical signal transmitting unit, a circulator, a data acquisition module and a processing module, wherein the optical signal transmitting unit is connected with the circulator, the circulator is connected with an optical fiber, the data acquisition module is connected with the circulator, the processing module is connected with the data acquisition module, and the optical signal transmitting unit modulates a light source and then transmits pulse light to the circulator; the circulator transmits the pulse light to the detection optical fiber and obtains reverse Rayleigh scattering light signal data of the detection optical fiber; the data acquisition module acquires reverse Rayleigh scattering light signal data of the circulator and transmits the data to the processing module; the processing module is used for preprocessing the optical signal data to obtain a daily defense area alarm sequence; the alarm sequence is segmented, an embedded vector is introduced, the embedded vector is embedded into the alarm sequence, the alarm sequence added with the embedded vector is input into a first neural network to obtain an environment feature vector, first intrusion alarm probability calculation is conducted based on the environment feature vector and the time feature vector, second intrusion alarm probability calculation is conducted based on the environment feature vector, the time feature vector and the first intrusion alarm probability, and intrusion early warning is conducted based on a calculation result.
The optical signal transmitting unit includes: the device comprises a narrow linewidth laser, an acousto-optic modulator and an optical fiber amplifier, wherein the narrow linewidth laser is connected with the acousto-optic modulator, and the acousto-optic modulator is connected with the optical fiber amplifier, and the narrow linewidth laser generates a continuous light source; the acousto-optic modulator modulates continuous light of the narrow linewidth laser into pulse light; the optical fiber amplifier amplifies the pulse light transmitted by the acousto-optic modulator.
The distributed optical fiber sensing technology adopted by the invention is based on a phase sensitive optical time domain reflectometer (ϕ -OTDR) technology. The method is to detect the intensity of the reverse Rayleigh scattered light in the optical fiber and classify the vibration sources according to the difference of interference waveforms. The sensing optical fibers are distributed along the oil-gas pipe network in a large area, and each sensing unit is arranged as a defense area. Other components are integrated into the chassis. The laser source with narrow linewidth is switched into pulse light by the modulator and then sent into the sensing optical fiber through the circulator. The Rayleigh back scattering light output from the other end of the circulator is received by the photoelectric detector and is transmitted to the processing module through the data acquisition card.
The distributed optical fiber vibration sensing system has the advantages of high precision, high reliability, high efficiency and the like. The vibration source can effectively detect the pipeline vibration signal and position the vibration source. When the sensing optical fiber is disturbed, the optical phase of the disturbed position changes due to the elastic optical effect. Therefore, the phase of the back-scattered light at the corresponding position changes, and the internal pulse width of the interference light intensity of the scattered light also changes accordingly. The main parameters of the system are that the central wavelength of the narrow linewidth laser is 1550nm, the linewidth is 3kHz, the total length of the distributed optical fiber is 48km, the signal loss is 0.27dB/km, and the data acquisition frequency is 2kHz.
The processing module comprises: the device comprises a preprocessing unit, an environment feature vector generating unit and an optical fiber intrusion early warning unit; the preprocessing unit is used for preprocessing the optical signal data; the environment feature vector generating unit generates a plurality of environment feature vectors by using the preprocessed optical signal data; the optical fiber intrusion early warning unit calculates the probability of the first intrusion alarm based on the environmental feature vector and the time feature vector; and performing second intrusion alarm probability calculation based on the environmental feature vector, the time feature vector and the first intrusion alarm probability, and performing intrusion early warning based on the calculation result.
The optical fiber intrusion early warning unit comprises: the system comprises a time feature module, an environment feature module, a space-time feature judging module and a fusion module, wherein the time feature module and the environment feature module are connected with the space-time feature judging module, and the time feature module, the environment feature module and the space-time feature judging module are connected with the fusion module; the time feature module stores a time feature vector; the environment feature module stores environment feature vectors; the space-time feature judging module carries out first intrusion alarm probability calculation based on the XGBoost model according to the input time feature vector and the environment feature vector; the fusion module takes the time feature vector, the environment feature vector and the first intrusion alarm probability as input, adopts an MLP network model to calculate the second intrusion alarm probability, and obtains an intrusion early warning result.
According to the invention, the embedded vector is introduced into the alarm sequence signal, the environment vector similar to the realization environment is obtained after model training, the environment information around the defending area is fully mined and effectively characterized, the full space distribution of the intrusion event is modeled by fusing the environment feature vector and the time feature vector, and the integrated learning model based on the environment embedding and the time sequence feature is designed aiming at random noise and environment interference, so that the generalization capability and the practicability of the model are greatly improved, and the accuracy of the distributed optical fiber early warning is improved.
The above-described embodiment is only a preferred embodiment of the present invention, and is not limited in any way, and other variations and modifications may be made without departing from the technical aspects set forth in the claims.
Claims (7)
1. The distributed optical fiber intrusion early warning method based on environment embedding is characterized by comprising the following steps of:
acquiring reverse Rayleigh scattering light signal data in a detection optical fiber;
preprocessing the optical signal data to obtain a daily defense area alarm sequence;
dividing an alarm sequence, introducing an embedded vector, determining the embedded vector according to an environment weight coefficient and an alarm sequence of the current day, and embedding the embedded vector into the alarm sequence;
inputting the alarm sequence added with the embedded vector into a first neural network to obtain an environment characteristic vector;
performing first intrusion alarm probability calculation based on the environmental feature vector and the time feature vector;
and performing second intrusion alarm probability calculation based on the environmental feature vector, the time feature vector and the first intrusion alarm probability, and performing intrusion early warning based on the calculation result.
2. The method for environmental embedding-based distributed fiber intrusion alert of claim 1, wherein,
the method for preprocessing the optical signal data comprises the following steps:
performing discrete wavelet transformation on the initial optical signal data to obtain wavelet coefficients;
modifying wavelet coefficients by threshold rules;
wavelet reconstruction is performed using an inverse discrete wavelet transform based on the modified wavelet coefficients.
3. The method for environmental embedding-based distributed fiber intrusion alert of claim 1, wherein,
the alarm sequence is divided by a window K.
4. The utility model provides a distributed optical fiber intrusion early warning system based on environment embedding which characterized in that includes:
the optical signal transmitting unit modulates the light source and transmits pulse light to the circulator;
the circulator is used for transmitting the pulse light to the detection optical fiber and obtaining the reverse Rayleigh scattering light signal data of the detection optical fiber;
the data acquisition module acquires reverse Rayleigh scattering light signal data of the circulator and transmits the data to the processing module;
the processing module is used for preprocessing the optical signal data to obtain a daily defense area alarm sequence; dividing an alarm sequence, introducing an embedded vector, determining the embedded vector according to an environment weight coefficient and an alarm sequence of the current day, and embedding the embedded vector into the alarm sequence; inputting the alarm sequence added with the embedded vector into a first neural network to obtain an environment characteristic vector; performing first intrusion alarm probability calculation based on the environmental feature vector and the time feature vector; and performing second intrusion alarm probability calculation based on the environmental feature vector, the time feature vector and the first intrusion alarm probability, and performing intrusion early warning based on the calculation result.
5. The distributed optical fiber intrusion alert system based on environmental embedding of claim 4, wherein,
the optical signal transmitting unit includes:
a narrow linewidth laser producing a continuous light source;
an acousto-optic modulator modulating continuous light of the narrow linewidth laser into pulse light;
and the optical fiber amplifier is used for amplifying the pulse light transmitted by the acousto-optic modulator.
6. The distributed optical fiber intrusion alert system based on environmental embedding of claim 4 or 5, wherein,
the processing module comprises:
a preprocessing unit for preprocessing the optical signal data;
an environmental feature vector generation unit generating a plurality of environmental feature vectors using the preprocessed optical signal data;
the optical fiber intrusion early warning unit is used for calculating the probability of the first intrusion alarm based on the environmental feature vector and the time feature vector; and performing second intrusion alarm probability calculation based on the environmental feature vector, the time feature vector and the first intrusion alarm probability, and performing intrusion early warning based on the calculation result.
7. The distributed fiber intrusion alert system based on environmental embedding of claim 6, wherein,
the optical fiber intrusion early warning unit comprises:
the time feature module stores time feature vectors;
the environment feature module stores environment feature vectors;
the space-time feature judging module is used for carrying out first intrusion alarm probability calculation based on the XGBoost model according to the input time feature vector and the environment feature vector;
and the fusion module takes the time feature vector, the environment feature vector and the first intrusion alarm probability as input, adopts an MLP network model to calculate the second intrusion alarm probability, and obtains an intrusion early warning result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410000986.6A CN117496650B (en) | 2024-01-02 | 2024-01-02 | Distributed optical fiber intrusion early warning method and system based on environment embedding |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410000986.6A CN117496650B (en) | 2024-01-02 | 2024-01-02 | Distributed optical fiber intrusion early warning method and system based on environment embedding |
Publications (2)
Publication Number | Publication Date |
---|---|
CN117496650A CN117496650A (en) | 2024-02-02 |
CN117496650B true CN117496650B (en) | 2024-03-26 |
Family
ID=89674725
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202410000986.6A Active CN117496650B (en) | 2024-01-02 | 2024-01-02 | Distributed optical fiber intrusion early warning method and system based on environment embedding |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117496650B (en) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110852187A (en) * | 2019-10-22 | 2020-02-28 | 华侨大学 | Method and system for identifying perimeter intrusion event |
CN115983497A (en) * | 2023-02-16 | 2023-04-18 | 华润数字科技有限公司 | Time sequence data prediction method and device, computer equipment and storage medium |
CN116186642A (en) * | 2023-04-27 | 2023-05-30 | 山东汇英信息科技有限公司 | Distributed optical fiber sensing event early warning method based on multidimensional feature fusion |
CN116798198A (en) * | 2022-10-12 | 2023-09-22 | 云南电网有限责任公司昆明供电局 | Sensor abnormality detection and early warning method based on multivariate time sequence prediction model |
CN116957585A (en) * | 2023-02-10 | 2023-10-27 | 腾讯科技(深圳)有限公司 | Data processing method, device, equipment and storage medium |
CN117275202A (en) * | 2023-09-20 | 2023-12-22 | 天津大学 | Omnibearing real-time intelligent early warning method and system for dangerous sources in important areas of cultural relics |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109974835B (en) * | 2018-12-29 | 2021-06-04 | 无锡联河光子技术有限公司 | Vibration detection identification and space-time positioning method and system based on optical fiber signal characteristics |
-
2024
- 2024-01-02 CN CN202410000986.6A patent/CN117496650B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110852187A (en) * | 2019-10-22 | 2020-02-28 | 华侨大学 | Method and system for identifying perimeter intrusion event |
CN116798198A (en) * | 2022-10-12 | 2023-09-22 | 云南电网有限责任公司昆明供电局 | Sensor abnormality detection and early warning method based on multivariate time sequence prediction model |
CN116957585A (en) * | 2023-02-10 | 2023-10-27 | 腾讯科技(深圳)有限公司 | Data processing method, device, equipment and storage medium |
CN115983497A (en) * | 2023-02-16 | 2023-04-18 | 华润数字科技有限公司 | Time sequence data prediction method and device, computer equipment and storage medium |
CN116186642A (en) * | 2023-04-27 | 2023-05-30 | 山东汇英信息科技有限公司 | Distributed optical fiber sensing event early warning method based on multidimensional feature fusion |
CN117275202A (en) * | 2023-09-20 | 2023-12-22 | 天津大学 | Omnibearing real-time intelligent early warning method and system for dangerous sources in important areas of cultural relics |
Non-Patent Citations (2)
Title |
---|
Φ-OTDR分布式光纤传感系统的关键技术研究;杨斌;皋魏;席刚;;光通信研究;20120810(04);第23-26页 * |
基于实体嵌入和长短时记忆网络的入侵检测方法;赖训飞;梁旭文;谢卓辰;李宗旺;;中国科学院大学学报;20200708(04);第124-132页 * |
Also Published As
Publication number | Publication date |
---|---|
CN117496650A (en) | 2024-02-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Liu et al. | Vehicle detection and classification using distributed fiber optic acoustic sensing | |
Wu et al. | A dynamic time sequence recognition and knowledge mining method based on the hidden Markov models (HMMs) for pipeline safety monitoring with Φ-OTDR | |
Li et al. | Pattern recognition for distributed optical fiber vibration sensing: A review | |
US11562224B2 (en) | 1D-CNN-based distributed optical fiber sensing signal feature learning and classification method | |
Liu et al. | Traffic flow detection using distributed fiber optic acoustic sensing | |
Wang et al. | Pattern recognition using relevant vector machine in optical fiber vibration sensing system | |
Wu et al. | A novel DAS signal recognition method based on spatiotemporal information extraction with 1DCNNs-BiLSTM network | |
Hernández et al. | Deep-learning-based earthquake detection for fiber-optic distributed acoustic sensing | |
Tian et al. | Intrusion signal recognition in OFPS under multi-level wavelet decomposition based on RVFL neural network | |
WO2016198683A1 (en) | Method and apparatus for monitoring pipeline using an optical fiber sensor system | |
CN103226028B (en) | Method for identifying and detecting disturbance signals of phase-sensitive optical time domain reflectometer | |
CN112749637B (en) | SNN-based distributed optical fiber sensing signal identification method | |
Fouda et al. | Pattern recognition of optical fiber vibration signal of the submarine cable for its safety | |
CN111537056A (en) | Pipeline along-line third-party construction dynamic early warning method based on SVM and time-frequency domain characteristics | |
Zhao et al. | Markov transition fields and deep learning-based event-classification and vibration-frequency measurement for φ-OTDR | |
Wu et al. | Improved generalization in signal identification with unsupervised spiking neuron networks for fiber-optic distributed acoustic sensor | |
Timofeev et al. | Multimodal heterogeneous monitoring of super-extended objects: modern view | |
CN116838955A (en) | Two-stage oil and gas pipeline line interference identification method | |
Yan et al. | Mixed intrusion events recognition based on group convolutional neural networks in DAS system | |
Wang et al. | RVFL-based optical fiber intrusion signal recognition with multi-level wavelet decomposition as feature | |
Marie et al. | A hybrid model integrating MPSE and IGNN for events recognition along submarine cables | |
CN117496650B (en) | Distributed optical fiber intrusion early warning method and system based on environment embedding | |
Chen et al. | An AI-based monitoring system for external disturbance detection and classification near a buried pipeline | |
Qiu et al. | A new detection method based on CFAR and DE for OFPS | |
CN111539393B (en) | Oil-gas pipeline third-party construction early warning method based on EMD decomposition and LSTM |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |