CN108599863A - Overhead transmission line wind based on shallow-layer CNN disaggregated models waves monitoring and pre-warning system and method - Google Patents

Abstract

The embodiment of the present invention relates to the technical field of transmission line monitoring, in particular, to an overhead transmission line wind dance monitoring and early warning system and method based on a shallow CNN classification model, the overhead transmission line wind dance monitoring based on a shallow CNN classification model The early warning system includes a detection optical cable, a demodulator and a data processing device; there is an optical signal interaction between the detection optical cable and the demodulator, the demodulator is connected to the data processing device, and the demodulator is used to receive feedback from the detection cable in real time. Backward Rayleigh scattered light, photoelectric conversion and analog-to-digital conversion are performed on the backward Rayleigh scattered light, corresponding digital signals are obtained, and the digital signals are sent to a data processing device, and the data processing device is used to process each of the spatial sample points The digital signals are differentiated, framed, feature extracted, feature fused, and classified to identify abnormal signals and give early warning. The system and method can realize real-time wind dance monitoring and early warning of long-distance transmission lines.

Description

Overhead transmission line wind dance monitoring and early warning system based on shallow CNN classification model and its method

technical field

Embodiments of the present invention relate to the technical field of transmission line monitoring, and in particular, relate to a wind dance monitoring and early warning system and method for overhead transmission lines based on a shallow CNN classification model.

Background technique

The power system is the nerve center of modern society and economic operation, the source of power, and the master switch of social and economic operation. The power system is of great significance to the normal production of the society and the daily life of the people.

Transmission lines play an important role in the transmission of electric energy and are an important part of the power system. Windy weather will cause the dance of the transmission line, which will cause unnecessary accidents and disasters. For long-distance transmission lines, the amount of wind dance data is large. Most of the existing technologies take a long time to process these large amounts of data, and the real-time concurrent processing capability is weak. , so it is difficult to realize real-time wind dance warning.

Contents of the invention

In view of this, the present invention provides an overhead transmission line wind dance monitoring and early warning system and method based on a shallow CNN classification model, which can quickly process a large amount of wind dance data, has high real-time concurrent processing capabilities, and can realize real-time wind dance early warning.

In order to achieve the above object, the embodiment of the present invention provides a wind dance monitoring and early warning system for overhead transmission lines based on shallow CNN classification model, including: a detection optical cable, a demodulator and a data processing device; the detection optical cable and the solution There is optical signal interaction between the demodulators; the demodulator is communicatively connected with the data processing device;

The demodulator is used to generate a pulsed light signal, the detection cable is used to receive the pulsed light signal, and the demodulator is also used to receive all spatial sample points of the transmission line to be monitored fed back by the detection cable in real time performing photoelectric conversion and analog-to-digital conversion on the Rayleigh backward scattered light to obtain a corresponding digital signal, and sending the digital signal to the data processing device;

The data processing device is used to receive the digital signal, display the digital signal in real time through the display module, and differentiate the digital signal of each of the spatial sample points, and convert the digital signal after the difference in a preset time frame Carry out frame processing at intervals, extract time domain features and chromatogram features in the digital signal after frame processing; normalize the time domain features; take each frame signal as the minimum signal processing unit, normalize The time-domain features after optimization are fused with the chromatogram features to obtain fusion features, and the preset shallow convolutional neural network (CNN) classification model is used to classify and identify the fusion features, so as to identify abnormal signals and give early warning.

Optionally, the demodulator includes a light emitting element, an acousto-optic modulator, an optical amplifier, an isolator, a circulator, a polarizer and a detector;

The light-emitting element is fixedly connected to the acousto-optic modulator, the acousto-optic modulator is fixedly connected to the optical amplifier, the optical amplifier is fixedly connected to the isolator, and the isolator is fixed to the circulator connected, the circulator is fixedly connected to the detection optical cable, the circulator is also fixedly connected to the polarizer, the polarizer is fixedly connected to the detector, and the detector communicates with the data processing device connect;

The light-emitting element is used to generate a pulsed light signal, and the pulsed light signal is transmitted to the detection optical cable through the acousto-optic modulator, the optical amplifier and the circulator;

The circulator is also used to receive the Rayleigh backscattered light fed back by the detection cable;

The polarizer is used to receive the Rayleigh backscattered light, and perform polarization processing on the Rayleigh backscattered light;

The detector is used to receive the Rayleigh scattered light after polarization processing, convert the Rayleigh scattered light after polarization processing into a digital signal through demodulation and analog-to-digital conversion, and send the digital signal to the the data processing device.

Optionally, the circulator is surrounded by a first interface, a second interface and a third interface;

The first interface is fixedly connected to the isolator, and the pulsed optical signal enters the circulator through the first interface;

The second interface is fixedly connected to the detection optical cable, and the pulse light signal enters the detection optical cable through the second interface; the backward Rayleigh scattered light enters the detection optical cable through the second interface circulator;

The third interface is fixedly connected to the polarizer, and the Rayleigh backscattered light enters the polarizer through the third interface.

Optionally, the data processing device realizes the extraction of time-domain features and chromatogram features in the digital signal after frame processing in the following manner:

Obtain the short-term galloping maximum value, short-term galloping peak-to-peak value, short-term galloping average value, short-term galloping standard deviation and short-term galloping root mean square in the digital signal after the frame division processing, and the short-term galloping maximum value, the short-term galloping The peak-to-peak value of the short-term galloping, the mean value of the short-term galloping, the standard deviation of the short-term galloping, and the root mean square of the short-term galloping are used as time-domain features;

Short-time Fourier transform is performed on the digital signal after the frame processing, and the transformation result is mapped to twelve frequency points corresponding to the seven scales in the chromatogram to obtain the chromatogram features; wherein, the chromatogram features are:

Optionally, the data processing device performs normalization processing on the time-domain features in the following manner:

A 24-hour time sliding window is used to normalize the time-domain features of each frame signal in the time interval from the current moment to the past 24 hours.

Optionally, the data processing device fuses the normalized time-domain features and the chromatogram features in the following manner to obtain fusion features:

For each frame signal, the short-term galloping maximum value, the short-term galloping peak-to-peak value, the short-term galloping mean value, the short-term galloping standard deviation, the short-term galloping root mean square, and the chromatogram The graph features are compounded into a two-dimensional array of five rows and twelve columns and the two-dimensional array is used as a fusion feature; wherein, the maximum value of the short-term galloping is timeMax, the peak value of the short-term galloping is timeMaxMin, and the average value of the short-term galloping is timeMean, the standard deviation of the short-term dancing is timeStd, and the root mean square of the short-term dancing is timeRMS; the fusion feature is:

Optionally, the data processing device implements classification and identification of the fusion features using a preset shallow convolutional neural network (CNN) classification model in the following manner, so as to identify abnormal signals and give early warning:

The fusion feature is input to a classifier, and the classifier is integrated with a trained shallow convolutional neural network CNN classification model; wherein, the preset shallow convolutional neural network CNN classification model includes sequential connections The first convolutional layer, the first pooling layer, the second convolutional layer, the second pooling layer, the first fully connected layer, the second fully connected layer, the third fully connected layer and the fourth fully connected layer;

The classifier is used to classify and identify the fusion feature, and if the fusion feature is identified as an abnormal signal, a wind dance warning is issued.

The embodiment of the present invention also provides an overhead transmission line wind dance monitoring and early warning method based on a shallow CNN classification model. The method is applied to the above-mentioned overhead transmission line wind dance monitoring and early warning system based on a shallow CNN classification model. The wind dance monitoring and early warning system for overhead transmission lines of the CNN classification model includes a detection optical cable, a demodulator and a data processing device; there is an optical signal interaction between the detection optical cable and the demodulator; the demodulator and the data The processing device is communicatively connected, the method comprising:

The detection optical cable receives the pulsed light signal generated by the demodulator;

The demodulator receives in real time the Rayleigh backscattered light of all spatial sample points of the transmission line to be monitored fed back by the detection optical cable, performs photoelectric conversion and analog-to-digital conversion on the Rayleigh backscattered light, and obtains the corresponding a digital signal, sending the digital signal to the data processing device;

The data processing device receives the digital signal, displays the digital signal in real time through the display module, and differentiates the digital signal of each of the spatial sample points, and converts the digital signal after the difference according to a preset time frame Carry out frame processing at intervals, extract time domain features and chromatogram features in the digital signal after frame processing; normalize the time domain features; take each frame signal as the minimum signal processing unit, normalize The time-domain features after optimization are fused with the chromatogram features to obtain fusion features, and the preset shallow convolutional neural network (CNN) classification model is used to classify and identify the fusion features, so as to identify abnormal signals and give early warning.

Optionally, the step of extracting the time-domain features and chromatogram features in the digital signal after the frame processing includes:

Obtain the short-term galloping maximum value, short-term galloping peak-to-peak value, short-term galloping average value, short-term galloping standard deviation and short-term galloping root mean square in the digital signal after the frame division processing, and the short-term galloping maximum value, the short-term galloping The peak-to-peak value of the short-term galloping, the mean value of the short-term galloping, the standard deviation of the short-term galloping, and the root mean square of the short-term galloping are used as time-domain features;

Short-time Fourier transform is performed on the digital signal after the frame processing, and the transformation result is mapped to twelve frequency points corresponding to the seven scales in the chromatogram to obtain the chromatogram features; wherein, the chromatogram features are:

Optionally, performing normalization processing on the time-domain features; taking each frame signal as the minimum signal processing unit, merging the normalized time-domain features and the chromatogram features to obtain fusion features, using The preset shallow convolutional neural network CNN classification model classifies and recognizes the fusion features to identify abnormal signals and perform early warning steps, including:

Using a 24-hour time sliding window to normalize the time-domain features of each frame signal in the time interval from the current moment to the past 24 hours;

After the normalization process, for each frame signal, the short-term galloping maximum value, the short-term galloping peak value, the short-term galloping average value, the short-term galloping standard deviation, the short-term galloping mean value The square root and the chromatogram feature are compounded into a two-dimensional array of five rows and twelve columns and the two-dimensional array is used as a fusion feature; wherein, the short-term galloping maximum value is timeMax, and the short-term galloping peak value is timeMaxMin, The mean value of the short-term dance is timeMean, the standard deviation of the short-term dance is timeStd, and the root mean square of the short-term dance is timeRMS; the fusion feature is:

The fusion feature is input to a classifier, and the classifier is integrated with a trained shallow convolutional neural network CNN classification model; wherein, the preset shallow convolutional neural network CNN classification model includes sequential connections The first convolutional layer, the first pooling layer, the second convolutional layer, the second pooling layer, the first fully connected layer, the second fully connected layer, the third fully connected layer and the fourth fully connected layer;

The classifier is used to classify and identify the fusion feature, and if the fusion feature is identified as an abnormal signal, a wind dance warning is issued.

In the overhead transmission line wind dance monitoring and early warning system and method based on the shallow CNN classification model provided by the embodiments of the present invention, the demodulator can receive the Rayleigh backscattered light detected by the detection cable, and scatter the back Rayleigh light Photoelectric conversion and analog-to-digital conversion are performed on light to obtain corresponding digital signals. A large amount of long-distance transmission line wind dance data can be obtained through the distributed optical fiber sensing system composed of detection optical cables and demodulators, which is convenient for making training sets for training. The shallow CNN classifier for Fengwu signal classification, the data processing device performs difference, frame division, feature extraction, normalization and feature fusion processing on the digital signal, and uses the preset shallow convolutional neural network CNN classification model to integrate Features are classified and recognized. By integrating the pre-trained CNN classification model in the system, it can reduce the time required to identify wind dance anomalies compared with existing technologies. Wind dance warning.

Further, the acquisition of a large amount of data enables the classifier to solve problems more conveniently, in real time, and accurately.

Description of drawings

In order to illustrate the technical solutions of the embodiments of the present invention more clearly, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention, and thus It should be regarded as a limitation on the scope, and those skilled in the art can also obtain other related drawings based on these drawings without creative work.

Fig. 1 is a schematic structural diagram of an overhead transmission line wind dance monitoring and early warning system based on a shallow CNN classification model provided by an embodiment of the present invention.

Fig. 2 is a schematic structural diagram of a detection optical cable provided by an embodiment of the present invention.

Fig. 3 is a structural block diagram of a data processing device provided by an embodiment of the present invention.

Fig. 4 is a flow chart of the wind dance monitoring and early warning method provided by the embodiment of the present invention.

Fig. 5 is a working principle diagram of the 24-hour time sliding window provided by the embodiment of the present invention.

Icons: 1-detection optical cable; 2-demodulator; 21-luminous element; 22-acousto-optic modulator; 23-optical amplifier; 24-isolator; 25-circulator; 251-first interface; 252-second Interface; 253-third interface; 26-polarizer; 27-detector; 3-data processing device; 31-data receiving module; 32-display module; 33-data storage module; 34-wind dance abnormality monitoring and early warning module.

Detailed ways

The power system is the nerve center of modern society and economic operation, the source of power, and the master switch of social and economic operation. The power system is of great significance to the normal production of the society and the daily life of the people. Transmission lines play an important role in the transmission of electric energy and are an important part of the power system. Windy weather will cause the dancing of power lines, causing unnecessary accidents and disasters. Therefore, it is necessary to monitor and warn the wind dance of the transmission line.

The inventor's investigation and research found that for long-distance power transmission lines, the amount of wind dance data is relatively large. Most of the existing technologies take a long time to process these large amounts of data, and the real-time concurrent processing capabilities are weak, making it difficult to realize real-time wind dance early warning.

Aiming at the existing wind dance monitoring technology, the inventor found through research and analysis that the existing transmission line wind dance monitoring system includes a monitoring system based on network cameras, a monitoring system based on electrical sensors and a monitoring system based on fiber Bragg grating sensor arrays . The monitoring system based on the camera can realize the qualitative observation and simple quantitative monitoring of the wind dance signal, but it cannot obtain more accurate quantitative monitoring results, nor can it do quantitative analysis such as spectrum analysis; the monitoring system based on electrical sensors requires A large amount of power supply is often unable to meet this power supply demand in actual situations, and the electrical sensor has weak anti-interference ability and is not easy to maintain; the fiber Bragg grating sensor array has the advantages of quantitative monitoring, anti-interference, easy maintenance, and distributed monitoring. However, for distributed monitoring scenarios with many sensor nodes, the load capacity of transmitting sensing data is weak, so it is not suitable for long-distance monitoring scenarios such as wind dance monitoring of transmission lines.

For the analysis, processing and abnormal identification of wind dance signals, most of the existing technologies adopt mechanical modeling analysis such as finite element analysis, or directly artificially set a threshold value of wind dance signal gallop amplitude, and then analyze the wind signal whose gallop amplitude exceeds the threshold. Dance signal for early warning. For the method of mechanical analysis, more accurate mechanical measurement data and a suitable experimental environment are required in the early stage to produce the same abnormal wind dance signal as the real environment, and building such an experimental environment requires a lot of manpower, material resources and even time costs. For the method of artificially setting the early warning threshold, the set threshold is fixed, and it is only judged whether the signal is abnormal by whether the magnitude of the galloping amplitude exceeds the set threshold, regardless of whether the signal meets the low-frequency resonance characteristics, and it may be mistaken for the noise signal. It is recognized as an abnormal signal, so this method can only have a rough judgment on the abnormality of Fengwu, but cannot accurately identify the abnormality of Fengwu and then give an early warning.

The defects in the solutions in the above prior art are all the results obtained by the inventor after practice and careful research. Therefore, the discovery process of the above problems and the solutions to the above problems proposed by the embodiments of the present invention below , should be the inventor's contribution to the invention during the process of the invention.

Based on the above research, the embodiment of the present invention provides a wind dance monitoring and early warning system and method for overhead transmission lines based on a shallow CNN classification model, which can quickly process a large amount of wind dance data, has high real-time concurrent processing capabilities, and can realize real-time Wind dance warning.

In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments It is only a part of embodiments of the present invention, but not all embodiments. The components of the embodiments of the invention generally described and illustrated in the figures herein may be arranged and designed in a variety of different configurations.

Accordingly, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the invention. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

It should be noted that like numerals and letters denote similar items in the following figures, therefore, once an item is defined in one figure, it does not require further definition and explanation in subsequent figures.

Fig. 1 shows a schematic structural diagram of an overhead transmission line wind dance monitoring and early warning system based on a shallow CNN classification model provided by an embodiment of the present invention. As can be seen from the figure, the overhead transmission line wind dance monitoring and early warning system based on the shallow CNN classification model includes a detection optical cable 1 , a demodulator 2 and a data processing device 3 . Wherein, there is an optical signal interaction between the detection optical cable 1 and the demodulator 2 , and the demodulator 2 is connected to the data processing device 3 in communication.

Among them, the demodulator 2 is used to generate pulsed optical signals, the detection optical cable 1 is used to receive pulsed optical signals, and the demodulator 2 is also used to receive the back Rayleigh scattering of all spatial sample points of the transmission line to be monitored fed back by the detection optical cable in real time light signal. Please refer to FIG. 2 , which is a schematic structural diagram of a detection optical cable 1 provided by an embodiment of the present invention. In this embodiment, the detection optical cable 1 adopts an optical fiber composite overhead ground wire (Optical Power Grounded Waveguide). It can be seen from Fig. 2 that when the pulsed light signal L1 passes through the detection cable, it will produce backward Rayleigh scattered light L2, and when the wind load Fv (vibration signal) acts on the detection cable 1, the detection cable 1 will produce a strain effect and bounce effect and Poisson effect, which in turn lead to changes in the light intensity of the Rayleigh backscattered light, and the corresponding Wind Dance situation.

Please continue to refer to FIG. 1 , the demodulator 2 is also used to perform photoelectric conversion and analog-to-digital conversion on the backward Rayleigh scattered light to obtain a corresponding digital signal, and send the digital signal to the data processing device 3 .

The data processing device 3 is used to obtain these digital signals, display these digital signals in real time, and perform a series of analyzes on these digital signals, such as difference, frame processing, feature extraction, normalization processing, and feature fusion, and then use The preset shallow convolutional neural network CNN classification model classifies and recognizes the fusion features to identify abnormal signals and give early warning.

The system can convert analog signals into digital signals and conduct quantitative analysis, which solves the problem that most of the existing technologies can only conduct qualitative analysis. Through the quantitative analysis of digital signals, abnormal signals can be accurately identified in real time and early warning.

Please continue to refer to Fig. 1, as can be seen from the figure, the demodulator 2 includes a light emitting element 21, an acousto-optic modulator 22, an optical amplifier 23, an isolator 24, a circulator 25, a polarizer 26 and a detector 27, wherein the circulator surrounds A first interface 251 , a second interface 252 and a third interface 253 are provided. Further, the light emitting element 21 is fixedly connected to the acousto-optic modulator 22, the acousto-optic modulator 22 is fixedly connected to the optical amplifier 23, the optical amplifier 23 is fixedly connected to the isolator 24, the isolator 24 is fixedly connected to the first interface 251, and the second The interface 252 is fixedly connected to the detection optical cable 1 , the third interface 253 is fixedly connected to the polarizer 26 , the polarizer 26 is fixedly connected to the detector 27 , and the detector 27 is connected to the data processing device 3 in communication.

The light emitting element 21 is used to generate a pulsed light signal, and the pulsed light signal is transmitted to the detection optical cable 1 through the acousto-optic modulator 22 , the optical amplifier 23 , the isolator 24 , the first interface 251 and the second interface 252 . The second interface 252 receives the Rayleigh backscattered light fed back by the detection cable 1, the polarizer 26 receives the Rayleigh backscattered light through the third interface 253, performs polarization processing on the Rayleigh backscattered light, and the detector 27 receives The backward Rayleigh scattered light after polarization processing is converted into digital signals through demodulation and analog-to-digital conversion, and these digital signals are sent to the data processing device 3 .

It can be understood that there is real-time optical signal interaction between the demodulator 2 and the detection optical cable 1, please continue to refer to Figure 2, the detection optical cable 1 receives the pulsed optical signal generated by the demodulator 2 in real time, and the demodulator 2 receives the detection optical cable 1 in real time Feedback Rayleigh scattered light. Through this real-time optical signal interaction, the overhead transmission line wind dance monitoring and early warning system and method based on the shallow CNN classification model can obtain original and accurate optical signal data, laying the foundation for subsequent data analysis.

Further, please continue to refer to FIG. 1 , the data processing device 3 receives the digital signal sent by the detector 27 in real time. It can be understood that these digital signals correspond to all spatial sample points of the transmission line to be monitored. For example, the transmission line to be monitored has ten spatial sample points, and these digital signals correspond to the real-time digital signals of the ten spatial sample points. The functional modules of the data processing device 3 will be described below.

Fig. 3 shows a structural block diagram of a data processing device provided by an embodiment of the present invention. It can be seen from the figure that the data processing device 3 includes a data receiving module 31 , a display module 32 , a data storage module 33 and a wind dance anomaly monitoring and early warning module 34 . Among them, the data receiving module 31 is used to receive the digital signals sent by the detector in real time, and transmit these digital signals to the display module 32 and the data storage module 33 in two ways. Further, the display module 32 displays the wind dance monitoring in real time according to the digital signals. Data waveform, the data storage module 33 is used to store the received digital signal. Fengwu abnormal monitoring and early warning module 34 regularly reads the digital signal stored in the data storage module 33 (realized one-step reading of the digital signal), and performs differential, framing, feature extraction, fusion feature and Classify, and finally realize the online monitoring and early warning of Fengwu abnormalities.

Please refer to Fig. 4, which is a flow chart of the method for wind dance monitoring and early warning provided by the embodiment of the present invention. This method is applied to the above-mentioned overhead power transmission wind dance monitoring system. The specific process shown in Fig. 4 will be described in detail below:

In step S41, for each spatial sample point, the digital signal of the spatial sample point is differentiated to obtain a differential digital signal.

It should be noted that whether it is the real-time synchronous waveform display of the wind dance monitoring data waveform by the display module 32, or the asynchronous reading and processing of digital information by the dance abnormality monitoring and early warning module 34, it is necessary to obtain the original digital signal in time. The differential is performed to obtain a differential digital signal. This is because the acquired digital signal corresponds to the light intensity of the backward Rayleigh scattered light in the detection cable. The dance of the detection cable itself or the external disturbance causes the light intensity of the backward Rayleigh scattered light to change, if and only if The difference between the wind dance monitoring signals at adjacent times can reflect the galloping situation of the transmission line at the current moment compared with the previous moment, and then realize the galloping monitoring of the transmission line.

Step S42 , performing frame division processing on the differential digital signal at preset time frame intervals, to obtain a frame-divided differential digital signal.

In this embodiment, for each spatial point, the differential digital signal within 24 hours of a single point is divided into 1728 time frames. It can be understood that each time frame corresponds to a duration of 50s.

Step S43, extracting time-domain features and chromatogram features in the differential digital signal after framing.

For the extraction of time-domain features, calculate each short-term feature value in the differential digital signal after framing. In this embodiment, the short-term feature values include short-term galloping maximum value timeMax, short-term galloping peak value timeMaxMin, short-term galloping peak value timeMaxMin, short-term Dance mean timeMean, short-term dance standard deviation timeStd and short-term dance root mean square timeRMS. Among them, these five short-term eigenvalues can be used as time-domain features.

For the extraction of chromatogram features, short-time Fourier transform is performed on the differential digital signal after framing, and the transformation results are mapped to twelve frequency points corresponding to seven scales in frequency, and then the chromatogram features are obtained. Among them, The chromatogram features are:

In step S44, normalization processing is performed on the time-domain features by using each frame signal as the minimum signal processing unit.

Since the time-domain features of each frame signal are discrete, it is difficult to obtain the connection between discrete features, so it is necessary to normalize the time-domain features of each frame signal, and then make the discrete features continuous, which is convenient for subsequent classification recognition.

As shown in FIG. 5 , in this embodiment, a 24-hour time sliding window is used to normalize the time-domain features of each frame signal within the time interval from the current moment to the past 24 hours.

For example, the normalization process can be understood as, the signals in the original time sliding window are A1, A2 and A3, when the new signal A4 enters, the time sliding window becomes A2, A3 and A4, and at the same time, the new time window A4 is normalized. It should be noted that the eigenvalues of the framed signal put into the time sliding window are the statistical eigenvalues of the framed signal before the normalization process.

Taking the short-term dance maximum value timeMax as an example, the normalization formula is:

Among them, timeMax on the right side of the equation is the time domain feature before the normalization of the current framed signal, and the time domain feature on the left side of the equation is the time domain feature after normalization; The time-domain characteristics of the frame signal before normalization, N is the number of signal frames included in the 24-hour time range where the framed signal is located.

It can be understood that other short-term eigenvalues are also normalized by the above-mentioned similar formula, so no further description is given here.

Step S45, fusing the normalized time-domain features and chromatogram features to obtain fusion features.

For example, the short-term galloping maximum value timeMax, short-term galloping peak-to-peak value timeMaxMin, short-term galloping average timeMean, short-term galloping standard deviation timeStd, short-term galloping root mean square timeRMS and chromatogram features are combined into five rows and twelve columns of two-dimensional Array and use this two-dimensional array as the fusion feature Q, where the fusion feature Q is:

Step S46, using the preset shallow convolutional neural network (CNN) classification model to classify and identify the fusion features, so as to identify abnormal signals and give early warning.

Input the fused features to the classifier, where the classifier integrates the trained shallow convolutional neural network CNN classification model, and the classifier classifies and recognizes the fused features. If the fused feature is identified as an abnormal signal, a wind dance is issued early warning.

Among them, the preset shallow convolutional neural network CNN classification model includes an 8-layer neural network structure model composed of convolutional layers, pooling layers, and fully connected layers. Further, convolutional layer 1, pooling layer 1, convolutional layer 2, pooling layer 2, fully connected layer 1, fully connected layer 2, fully connected layer 3, and fully connected layer 4 are connected in sequence. It can be understood that the above The output of one layer is the input of the next layer, and the neural network structure model is used to classify the fusion features of the input Fengwu framed signal and to identify and warn of abnormalities.

Further, the parameter design of the convolutional layer and pooling layer is shown in Table 1.

Table 1

Further, the parameter design of the fully connected layer is shown in Table 2.

Table 2

Fully connected layer 1 Fully connected layer 2 Fully connected layer 3 Fully connected layer 4 Shallow CNN model 512units 512units 256units 2units

In a system that integrates a classifier, since the classifier is pre-trained, it only needs to input the required features to the classifier for recognition. For the case of multiple spatial points, multiple trained classifiers are used to work in multiple threads, which can realize real-time long-distance wind dance warning of transmission lines and improve concurrent processing capabilities. Furthermore, since a large amount of data can be obtained, a training set can be produced at any time to train a classifier, so the use of a classifier can solve problems more conveniently, in real time, and accurately.

In summary, the overhead transmission line wind dance monitoring and early warning system and method based on the shallow CNN classification model provided by the embodiments of the present invention can quickly process a large amount of wind dance data, have high real-time concurrent processing capabilities, and can realize real-time wind dance data. early warning.

In the several embodiments provided by the embodiments of the present invention, it should be understood that the disclosed devices and methods may also be implemented in other ways. The device and method embodiments described above are only illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show possible implementation architectures of devices, methods and computer program products according to multiple embodiments of the present invention, function and operation. In this regard, each block in a flowchart or block diagram may represent a module, program segment, or part of code that includes one or more Executable instructions. It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks in succession may, in fact, be executed substantially concurrently, or they may sometimes be executed in the reverse order, depending upon the functionality involved. It should also be noted that each block of the block diagrams and/or flowchart illustrations, and combinations of blocks in the block diagrams and/or flowchart illustrations, can be implemented by a dedicated hardware-based system that performs the specified function or action , or may be implemented by a combination of dedicated hardware and computer instructions.

In addition, each functional module in each embodiment of the present invention can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

If the functions are realized in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the essence of the technical solution of the present invention or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including Several instructions are used to make a computer device (which may be a personal computer, an electronic device 10, or a network device, etc.) execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk and other media that can store program codes. . It should be noted that, in this document, the terms "comprising", "comprising" or any other variation thereof are intended to cover a non-exclusive inclusion such that a process, method, article or apparatus comprising a set of elements includes not only those elements, It also includes other elements not expressly listed, or elements inherent in the process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article or apparatus comprising said element.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims (10)

1. a kind of overhead transmission line wind based on shallow-layer CNN disaggregated models waves monitoring and warning system, which is characterized in that including：It visits Survey optical cable, (FBG) demodulator and data processing equipment；There are optical signal interactions between the detecting optical cable and the (FBG) demodulator；The solution Instrument is adjusted to be communicated to connect with the data processing equipment；

The (FBG) demodulator is for generating pulsed optical signals, and the detecting optical cable is for receiving the pulsed optical signals, the demodulation Instrument is additionally operable to the backward Rayleigh scattering of all space sample points for the power transmission line to be monitored that detecting optical cable described in real-time reception is fed back Light carries out opto-electronic conversion and analog-to-digital conversion to the backward Rayleigh scattering light, obtains corresponding digital signal, the number is believed Number it is sent to the data processing equipment；

The data processing equipment is shown the digital signal by display module for receiving the digital signal in real time Show, and difference is carried out to the digital signal of each space sample point, the digital signal of difference interframe at preset timed intervals will be completed The temporal signatures and chromatogram feature in digital signal after progress sub-frame processing, extraction sub-frame processing；To the time domain Feature is normalized；Using each frame signal as minimum signal processing unit, by after normalization temporal signatures and institute It states chromatogram feature to be merged to obtain fusion feature, using default shallow-layer convolutional neural networks CNN disaggregated models to described Fusion feature carries out Classification and Identification, to identify abnormal signal and carry out early warning.

2. the overhead transmission line wind according to claim 1 based on shallow-layer CNN disaggregated models waves monitoring and warning system, special Sign is that the (FBG) demodulator includes illuminating part, acousto-optic modulator, image intensifer, isolator, circulator, polarizer and detector；

The illuminating part is fixedly connected with the acousto-optic modulator, and the acousto-optic modulator is fixedly connected with the image intensifer, The image intensifer is fixedly connected with the isolator, and the isolator is fixedly connected with the circulator, the circulator with The detecting optical cable is fixedly connected, and the circulator is also fixedly connected with the polarizer, the polarizer and the detector It is fixedly connected, the detector is communicated to connect with the data processing equipment；

The illuminating part is for generating pulsed optical signals, and the pulsed optical signals are through the acousto-optic modulator and the image intensifer It is transmitted to the detecting optical cable with the circulator；

The circulator is additionally operable to receive the backward Rayleigh scattering light of the detecting optical cable feedback；

The polarizer carries out polarization manipulation for receiving the backward Rayleigh scattering light, to the backward Rayleigh scattering light；

The detector is used to receive the backward Rayleigh scattering light passed through after polarization manipulation, will be through by demodulation analog-to-digital conversion The backward Rayleigh scattering light crossed after polarization manipulation is converted into digital signal, and the digital signal is sent to the data processing Device.

3. the overhead transmission line wind according to claim 2 based on shallow-layer CNN disaggregated models waves monitoring and warning system, special Sign is that the circulator is surrounded with first interface, second interface and third interface；

The first interface is fixedly connected with the isolator, and the pulsed optical signals enter the annular by first interface Device；

The second interface is fixedly connected with the detecting optical cable, and the pulsed optical signals enter the detection by second interface Optical cable；The backward Rayleigh scattering light enters the circulator from the detecting optical cable by the second interface；

The third interface is fixedly connected with the polarizer, and the backward Rayleigh scattering light is entered described inclined by third interface Shake device.

4. the overhead transmission line wind according to claim 2 based on shallow-layer CNN disaggregated models waves monitoring and warning system, special Sign is, the data processing equipment be accomplished by the following way in the digital signal after sub-frame processing temporal signatures and The extraction of chromatogram feature：

Obtain sub-frame processing after digital signal in wave in short-term maximum value, wave in short-term peak-to-peak value, wave in short-term mean value, Standard deviation is waved in short-term and waves root mean square in short-term, and maximum value, the peak-to-peak value, described short waved in short-term are waved in short-term by described When wave mean value, standard deviation and the root mean square of waving in short-term waved in short-term as temporal signatures；

Short Time Fourier Transform is carried out to the digital signal after sub-frame processing, by transformation results seven in being mapped to chromatogram Corresponding 12 Frequency points of scale obtain chromatogram feature；Wherein, the chromatogram is characterized as：

5. the overhead transmission line wind according to claim 4 based on shallow-layer CNN disaggregated models waves monitoring and warning system, special Sign is that the data processing equipment is in the following manner normalized the temporal signatures：

It is small in current time to the past 24 to the temporal signatures of each frame signal using twenty four hours time slide window When time interval in make normalized.

6. the overhead transmission line wind according to claim 4 based on shallow-layer CNN disaggregated models waves monitoring and warning system, special Sign is, the data processing equipment in the following manner by after normalization temporal signatures and the chromatogram feature carry out Fusion is to obtain fusion feature：

For each frame signal, by it is described wave in short-term maximum value, it is described wave in short-term peak-to-peak value, it is described wave in short-term mean value, It is described to wave standard deviation in short-term, described wave root mean square in short-term and the chromatogram feature is combined into the two-dimemsional numbers of the row of the five-element 12 Group and using the two-dimensional array as fusion feature；Wherein, the maximum value of waving in short-term is timeMax, described to wave peak in short-term Peak value is timeMaxMin, and the mean value of waving in short-term is timeMean, and the standard deviation of waving in short-term is timeStd, described It is timeRMS to wave root mean square in short-term；The fusion feature is：

7. the overhead transmission line wind according to claim 6 based on shallow-layer CNN disaggregated models waves monitoring and warning system, special Sign is that the data processing device is accomplished by the following way using default shallow-layer convolutional neural networks CNN classification moulds Type carries out Classification and Identification to the fusion feature, to identify abnormal signal and carry out early warning：

The fusion feature is input to grader, the combining classifiers there are the default shallow-layer convolutional neural networks that training finishes CNN disaggregated models；Wherein, the default shallow-layer convolutional neural networks CNN disaggregated models include the first convolutional layer connected in sequence, First pond layer, the second convolutional layer, the second pond layer, the first full articulamentum, the second full articulamentum, the full articulamentum of third and the 4th Full articulamentum；

Classification and Identification is carried out to the fusion feature using the grader, if identifying, the fusion feature is abnormal signal, Send out wind dance early warning.

8. a kind of overhead transmission line wind based on shallow-layer CNN disaggregated models waves monitoring and pre-alarming method, which is characterized in that the method Monitoring and warning system, base are waved applied to any overhead transmission line wind based on shallow-layer CNN disaggregated models of claim 1-7 It includes detecting optical cable, (FBG) demodulator and data processing dress to wave monitoring and warning system in the overhead transmission line wind of shallow-layer CNN disaggregated models It sets；There are optical signal interactions between the detecting optical cable and the (FBG) demodulator；The (FBG) demodulator is logical with the data processing equipment Letter connection, the method includes：

The detecting optical cable receives the pulsed optical signals that the (FBG) demodulator generates；

All space sample points of the power transmission line to be monitored of detecting optical cable feedback is backward auspicious described in the (FBG) demodulator real-time reception Profit scattering light carries out opto-electronic conversion and analog-to-digital conversion to the backward Rayleigh scattering light, obtains corresponding digital signal, will be described Digital signal is sent to the data processing equipment；

The data processing equipment receives the digital signal, and the digital signal is carried out real-time display by display module, And difference is carried out to the digital signal of each space sample point, the digital signal of difference interframe at preset timed intervals will be completed The temporal signatures and chromatogram feature in digital signal after progress sub-frame processing, extraction sub-frame processing；To the time domain Feature is normalized；Using each frame signal as minimum signal processing unit, by after normalization temporal signatures and institute It states chromatogram feature to be merged to obtain fusion feature, using default shallow-layer convolutional neural networks CNN disaggregated models to described Fusion feature carries out Classification and Identification, to identify abnormal signal and carry out early warning.

9. the overhead transmission line wind according to claim 8 based on shallow-layer CNN disaggregated models waves monitoring and pre-alarming method, special Sign is, the step of extracting the temporal signatures and chromatogram feature in the digital signal after sub-frame processing, including：

Obtain sub-frame processing after digital signal in wave in short-term maximum value, wave in short-term peak-to-peak value, wave in short-term mean value, Standard deviation is waved in short-term and waves root mean square in short-term, and maximum value, the peak-to-peak value, described short waved in short-term are waved in short-term by described When wave mean value, standard deviation and the root mean square of waving in short-term waved in short-term as temporal signatures；

Short Time Fourier Transform is carried out to the digital signal after sub-frame processing, by transformation results seven in being mapped to chromatogram Corresponding 12 Frequency points of scale obtain chromatogram feature；Wherein, the chromatogram is characterized as：

10. the overhead transmission line wind according to claim 9 based on shallow-layer CNN disaggregated models waves monitoring and pre-alarming method, It is characterized in that, the temporal signatures is normalized；Using each frame signal as minimum signal processing unit, will normalize Temporal signatures and the chromatogram feature later are merged to obtain fusion feature, using default shallow-layer convolutional neural networks CNN disaggregated models carry out Classification and Identification to the fusion feature, to identify abnormal signal and the step of carry out early warning, including：

It is small in current time to the past 24 to the temporal signatures of each frame signal using twenty four hours time slide window When time interval in make normalized；

After normalized, for each frame signal, maximum value is waved in short-term, described wave described peak-to-peak value, institute in short-term It states and waves mean value in short-term, described wave standard deviation in short-term, described wave root mean square in short-term and the chromatogram feature is combined into five Row 12 arrange two-dimensional array and using the two-dimensional array as fusion feature；Wherein, the maximum value of waving in short-term is TimeMax, the peak-to-peak value of waving in short-term is timeMaxMin, and the mean value of waving in short-term is timeMean, described to wave in short-term Dynamic standard deviation is timeStd, and the root mean square of waving in short-term is timeRMS；The fusion feature is：

The fusion feature is input to grader, the combining classifiers there are the default shallow-layer convolutional neural networks that training finishes CNN disaggregated models；Wherein, the default shallow-layer convolutional neural networks CNN disaggregated models include the first convolutional layer connected in sequence, First pond layer, the second convolutional layer, the second pond layer, the first full articulamentum, the second full articulamentum, the full articulamentum of third and the 4th Full articulamentum；

Classification and Identification is carried out to the fusion feature using the grader, if identifying, the fusion feature is abnormal signal, Send out wind dance early warning.