CN115797772A - Classified early warning method and system for icing thickness of overhead transmission line - Google Patents
Classified early warning method and system for icing thickness of overhead transmission line Download PDFInfo
- Publication number
- CN115797772A CN115797772A CN202211566952.0A CN202211566952A CN115797772A CN 115797772 A CN115797772 A CN 115797772A CN 202211566952 A CN202211566952 A CN 202211566952A CN 115797772 A CN115797772 A CN 115797772A
- Authority
- CN
- China
- Prior art keywords
- overhead transmission
- transmission line
- icing
- early warning
- phase
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a classification early warning method and a classification early warning system for icing thickness of an overhead transmission line, wherein the classification early warning method comprises the following steps of: s1, collecting overhead transmission lines by using a phase-sensitive optical time domain reflectometer, and classifying the overhead transmission lines; s2, framing the ice coating data with different thicknesses in the overhead transmission line, demodulating the phase to obtain a phase difference frequency spectrum, and generating a two-dimensional image by using a Gelam angular field; s3, creating a CNN model, and training the CNN model by using the two-dimensional image; and S4, classifying whether the overhead transmission line is iced or not through the CNN model, and performing risk assessment and early warning on the iced overhead transmission line. The early warning system is not easily influenced by the environment in the icing monitoring of the overhead transmission line, has long service life and reliable data transmission and can accurately and comprehensively monitor.
Description
Technical Field
The invention relates to the field of optical fiber sensing technology and computer vision, in particular to a classification early warning method and system for icing thickness of an overhead transmission line.
Background
The ice coating of the power transmission line easily causes the mechanical and electrical properties of the power transmission line to be rapidly reduced, so that not only is great economic loss caused, but also the safe and stable operation of the power system is seriously influenced. Monitoring the running state of the transmission line to evaluate the threat caused by ice coating, and taking corresponding measures in time to inhibit the damage caused by ice coating is very important. The traditional power transmission line icing monitoring methods such as a weighing method, an image method, a lead inclination angle method and the like are easily influenced by the environment, the service life is short, the data transmission reliability is poor, the state of the power transmission line cannot be accurately and comprehensively monitored, and the like. The distributed optical fiber sensing technology has good electrical insulation, strong anti-electromagnetic interference capability and high sensitivity, can monitor the whole length of a line theoretically, is very suitable for working in severe environments such as high voltage, strong electromagnetic interference and strong corrosion of a power transmission line, and can guarantee the accuracy of measured data and the working stability of a monitoring system compared with other methods.
The early pattern recognition research also mainly adopts a threshold method or simple feature extraction to identify different vibration classes, and a classification algorithm is also simpler. Due to the influence of environment and region difference, the mode identification of the distributed optical fiber vibration sensor has higher false alarm rate and poorer timeliness, and cannot meet the practical requirements of the current industrial production. Since 2011, researchers have begun to propose various pattern recognition methods for distributed fiber optic vibration sensors based on machine learning. Various feature extraction methods emerge, such as wavelet packet analysis, wavelet entropy, hilbert transform, short-time energy methods, and the like. Pattern recognition classifiers are also constantly being updated, such as adaptive thresholds, support vector machines, bayesian classifiers, BP neural networks, radial basis function neural networks, and the like. In recent five years, the accuracy of mode identification of the distributed optical fiber vibration sensor is further improved, but from the characteristics to be classified, a large amount of work of the existing method adopts a characteristic vector (a one-dimensional characteristic vector of a time domain or a frequency domain) form. Although deep learning methods (1D CNN, RNN, LSTM, etc.) can directly process one-dimensional data, these networks tend to be more difficult to train, some studies are more difficult to apply, and there are no existing pre-trained networks.
The method cuts in from the angle of the influence of ice coating on the vibration characteristic of the transmission line, and is based on the principle that the thicker the ice coating is, the smaller the natural frequency of each order mode of the overhead conductor is, and the monotonous decreasing mapping relation exists between the natural frequency of each order mode of the overhead conductor and the ice coating thickness. And converting the time sequence into GAFs pictures, classifying the GAFs pictures by using the visual advantage of the CNN on the machine, and finally establishing a mapping relation between the GAFs pictures and the ice coating thickness.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the method and the system for classifying and early warning the icing thickness of the overhead transmission line are based on a convolutional neural network and distributed optical fiber sensing, and aim to solve the problems that the icing monitoring of the overhead transmission line is easily affected by the environment, the service life is short, the reliability of data transmission is poor, and the icing monitoring cannot be accurately and comprehensively carried out.
In order to solve the technical problem, the invention adopts a technical scheme that:
a classification early warning method for icing thickness of an overhead transmission line comprises the following steps:
s1, collecting an overhead transmission line by using a phase-sensitive optical time domain reflectometer, and classifying the overhead transmission line;
s2, framing the ice coating data with different thicknesses in the overhead transmission line, demodulating the phase to obtain a phase difference frequency spectrum, and generating a two-dimensional image by using a Gelam angular field;
s3, creating a CNN model, and using the two-dimensional image to train the CNN model;
and S4, classifying whether the overhead transmission line is iced or not through the CNN model, and performing risk assessment and early warning on the iced overhead transmission line.
In order to solve the technical problem, the invention adopts another technical scheme as follows:
the utility model provides an overhead transmission line icing thickness classification early warning system, includes:
the data acquisition module is used for acquiring the overhead transmission lines by using the phase-sensitive optical time domain reflectometer and classifying the overhead transmission lines;
the data analysis and preprocessing module is used for framing the icing data with different thicknesses in the overhead transmission line, demodulating the phase to obtain a phase difference frequency spectrum, and generating a two-dimensional image by using a gram angle field;
the model training module is used for creating a CNN model and training the CNN model by using the two-dimensional image;
and the processing and early warning module is used for classifying whether the overhead transmission line is iced or not through the CNN model and carrying out risk assessment and early warning on the iced overhead transmission line.
The invention has the beneficial effects that: the method uses the convolutional neural network to process the GAFs image generated after demodulation, fully utilizes the advantages of machine vision, and realizes the real-time detection of the icing condition of the overhead transmission line under the condition of ensuring the accuracy of the classification result. The convolutional neural network model provided by the invention does not need to carry out complex parameter adjustment and can carry out classification quickly and accurately. The early warning system is not easily influenced by the environment in the icing monitoring of the overhead transmission line, has long service life and reliable data transmission, and can accurately and comprehensively monitor.
Drawings
Fig. 1 is a flowchart of an icing thickness classifying and early warning method for an overhead power transmission line according to an embodiment of the present invention;
fig. 2 is a frame diagram of an icing thickness classifying and early warning system for an overhead transmission line according to an embodiment of the present invention;
fig. 3 is a structural diagram of a neural network model of an overhead transmission line icing thickness classification and early warning method according to an embodiment of the present invention and corresponding parameters thereof;
fig. 4 is an amplitude-frequency diagram with the same amplitude and different frequencies and a corresponding GAFs image for training of the method for classifying and early warning the icing thickness of the overhead transmission line according to the embodiment of the present invention;
fig. 5 is an amplitude-frequency diagram with different amplitudes and the same frequency and a corresponding GAFs image for training of the method for classifying and warning the icing thickness of the overhead transmission line according to the embodiment of the present invention;
fig. 6 is an amplitude-frequency diagram under the influence of interference frequency of the classification and early warning method for icing thickness of an overhead transmission line according to the embodiment of the present invention and a corresponding GAFs image for training.
Detailed Description
In order to explain technical contents, achieved objects, and effects of the present invention in detail, the following description is made with reference to the accompanying drawings in combination with the embodiments.
The noun explains:
RBS: a radio base station for providing a radio interface between a mobile station and a system is mainly composed of a radio transceiver.
GAFs: the gram angular field can convert time sequence data into space data, namely data similar to images, and a convolutional neural network can be used for feature extraction subsequently.
phi-OTDR: the phase-sensitive optical time domain reflectometer is a precise photoelectric integrated instrument made up by utilizing Rayleigh scattering produced when the light is transmitted in the optical fibre and back scattering produced by Fresnel reflection.
CNN: a deep learning method has a framework of a convolutional layer, a pooling layer and a full-link layer.
Batch Normal: and (3) batch regularization, namely normalizing the input batch data, and mapping the normalized batch data to normal distribution with a mean value of 0 and a variance of 1.
ReLU activation function: the modified linear unit, a piecewise linear function, changes all negative values to 0 and positive values, which is referred to as single-sided suppression.
The embodiment of the invention provides a classified early warning method for icing thickness of an overhead transmission line, which is characterized by comprising the following steps of:
s1, collecting an overhead transmission line by using a phase-sensitive optical time domain reflectometer, and classifying the overhead transmission line;
s2, framing the icing data with different thicknesses in the overhead transmission line, performing phase demodulation to obtain a phase difference frequency spectrum, and generating a two-dimensional image by using a Graham angular field;
s3, creating a CNN model, and training the CNN model by using the two-dimensional image;
and S4, classifying whether the overhead transmission line is iced or not through the CNN model, and performing risk assessment and early warning on the iced overhead transmission line.
As can be seen from the above description, the beneficial effects of the present invention are: the method uses the convolutional neural network to process the GAFs image generated after demodulation, fully utilizes the advantages of machine vision, and realizes the real-time detection of the icing condition of the overhead transmission line under the condition of ensuring the accuracy of the classification result. The convolutional neural network model provided by the invention does not need to carry out complex parameter adjustment and can carry out classification quickly and accurately. The early warning system is not easily influenced by the environment in the ice coating monitoring of the overhead transmission line, has long service life and reliable data transmission, and can accurately and comprehensively monitor.
Further, in step S1, the classifying the overhead transmission line data specifically includes:
dividing the data of the overhead transmission line into three types, namely non-icing, icing below a risk threshold and icing above the risk threshold, wherein the risk threshold is set as an axial bearing stress value which is smaller than the yield strength of steel;
icing load F on node i of power transmission pole tower in power transmission line i Comprises the following steps:
wherein n represents the number of members, ρ represents the ice coating density, h 2 Denotes the coefficient of variation of ice coating diameter with height,/ j Representing the length of a single component, and g represents the icing gravity of the ground wire;
the ground wire icing gravity g =9.8 × 0.9 pi δ (d + δ) × 10 -3 The unit is electric load, wherein d represents the outer diameter of the conducting wire and the grounding wire, and delta represents the thickness of the ice coating.
According to the description, the acquired ice coating data of the overhead transmission line with different thicknesses are framed, the phase difference spectrum is obtained through demodulation, and the GAFs are used for generating the two-dimensional image to provide the data sample for the deep learning model.
Further, the phase demodulation in step S2 specifically includes:
s201, performing intermediate frequency filtering on the original signals of the wireless base station acquired by the data acquisition card to suppress broadband noise and obtain intermediate frequency signals;
s202, IQ demodulation is carried out on the intermediate frequency signal, a complex intermediate frequency is constructed, and the amplitude and the phase of the complex intermediate frequency are extracted;
s203, performing sliding average on the amplitude along a time line by using a window with a proper width;
s204, dividing the n sampling points of each time row in the step S203 into phase discrimination intervals with equal width and connected end to end, and solving and storing the column index of the position with the maximum amplitude in all the intervals, which corresponds to the time row, in the result corresponding to the S203;
s205, extracting phases in a left interval and a right interval adjacent to the phases from the phases, using the phases as estimation of RBS phases at optical fiber positions corresponding to sampling points of the left interval and the right interval at the current moment, calculating a first phase difference, backtracking phases at a pair of column indexes of the left interval and the right interval at the previous row, which are adjacent to the left interval and the right interval at the current moment, using the phases as estimation of RBS phases at the optical fiber positions corresponding to the sampling points of the two intervals at the previous moment, calculating a second phase difference, and subtracting the phase difference at the previous moment from the phase difference at the current moment to obtain the variation of the phase difference;
s206, repeating the step S205, traversing from the 2 nd line and the 2 nd interval, and calculating the variation of the phase difference of the left interval and the right interval of each interval in a detection pulse period;
the matrix is summed cumulatively along the time axis and unwrapped to recover the change in fiber stretch over time between adjacent segments caused by external vibrations.
As can be seen from the above description, the demodulated GAFs images are convenient to be further processed in a convolutional neural network.
Further, the step S2 of generating the two-dimensional image by using the gram angular field specifically includes:
s207, scaling the data range of the phase difference spectrum to [0,1] through the following formula:
s208, converting the zoomed sequence data into polar coordinates, taking the numerical value as a cosine value of an included angle, and taking a time axis as a radius:
s209, acquiring the angle difference between corresponding angles of different time points according to the following formula:
in the formula (I), the compound is shown in the specification,
representing a one-dimensional time series.
As can be seen from the above description, the GAFs is used to convert the scaled one-dimensional sequence data from rectangular coordinates to polar coordinates, and then the angular differences between different points are calculated to identify the time correlations of different time points, so as to convert the time sequences into GAFs pictures for subsequent identification and processing on the CNN model.
Further, the CNN model in step S3 is specifically composed of an input layer, a 1-layer first convolution module, a 3-layer second convolution module, and an output module.
As can be seen from the above description, the CNN model has a shared convolution kernel, no pressure is applied to high-dimensional data processing, and the feature classification effect is good.
Further, the convolution module is composed of a convolution layer-BatchNormal layer-ReLU activation function layer.
As can be seen from the above description, the convolution module functions to automatically extract features. The BatchNormal layer can make the data distribution of each batch consistent, and simultaneously can avoid the disappearance of the gradient. By utilizing the unilateral inhibition property of the ReLU activation function, neurons in the neural network have sparse activation, particularly in a deep neural network model (such as CNN), when the model increases N layers, the activation rate of the ReLU neurons can be reduced by N times of 2 theoretically. Therefore, the convolution module consisting of the convolution layer-BatchNormal layer-ReLU activation function layer can improve the gradient flowing through the network, allow a larger learning rate and greatly improve the training speed.
Further, the output module is composed of a linear layer and an activation function, and when the output value of the activation function is 1, the corresponding icing thickness category is no icing; when the output value of the activation function is 2, the corresponding icing thickness category is that the icing thickness is lower than a threshold value; when the output value of the activation function is 3, the corresponding ice coating thickness category is that the ice coating thickness is higher than the threshold value.
From the above description, the activation function output dimension size is consistent with the category of ice coating thickness, with the value for each dimension representing the likelihood of a different category.
Further, the activation function is a softmax activation function.
From the above description, it can be known that the exponential Softmax function can draw a larger numerical distance with a large difference, in deep learning, the inverse propagation is generally used to solve the gradient, and then the gradient descent is used to perform a parameter updating process, and the exponential function is more convenient in derivation.
Further, in step S4, the user is prompted through a human-computer interaction interface.
According to the description, the human-computer interaction interface is adopted, so that a user can conveniently, conveniently and intuitively acquire information and operate.
The utility model provides an overhead transmission line icing thickness classification early warning system, includes:
the data acquisition module is used for acquiring the overhead transmission lines by using the phase-sensitive optical time domain reflectometer and classifying the overhead transmission lines;
the data analysis and preprocessing module is used for framing the icing data with different thicknesses in the overhead transmission line, demodulating the phase to obtain a phase difference frequency spectrum, and generating a two-dimensional image by using a gram angle field;
the model training module is used for creating a CNN model and training the CNN model by using the two-dimensional image;
and the processing and early warning module is used for classifying whether the overhead transmission line is iced or not through the CNN model and carrying out risk assessment and early warning on the iced overhead transmission line.
The classified early warning method and system for the icing thickness of the overhead transmission line can solve the problems that the icing monitoring of the overhead transmission line is easily influenced by the environment, the service life of equipment is short, the reliability of data transmission is poor, and the icing monitoring cannot be accurately and comprehensively monitored, and are explained by a specific implementation mode as follows:
example one
Referring to fig. 1, the invention discloses a classification early warning method for icing thickness of an overhead transmission line, which specifically comprises the following steps:
step one, preparing data
And (3) acquiring data of the overhead transmission line by using the phi-OTDR, and classifying the acquired data.
In this embodiment, in the phi-OTDR based on heterodyne coherent detection, the carrier signal frequency emitted from the internal radio frequency source is 200MHz, the balanced photodetector detects a backward rayleigh scattered light (RBS) signal generated by incident pulsed light continuously during the propagation along the optical fiber, and after completing photoelectric conversion, a band-limited intermediate frequency signal with a very narrow bandwidth and a center frequency of 200MHz is output, and the sampling frequency of the data acquisition card is 250MHz.
The deep learning model usually needs a large number of data samples, and the neural network is only continuously trained by a large number of samples
After optimization, stable and reliable parameters and models can be found. The collected icing data of the overhead transmission line with different thicknesses are subjected to framing, the phase difference spectrum is obtained through demodulation, a two-dimensional image is generated through GAFs, and the collected image data are subjected to screening and labeling, including manual screening and labeling.
The obtained data are classified specifically as follows: dividing the acquired data into three types, namely non-icing, icing below a risk threshold and icing above the risk threshold, wherein the threshold is set to be that the axial bearing stress value is smaller than the yield strength of steel; wherein, the icing load F on the node i of the power transmission pole tower i Comprises the following steps:
wherein n is the number of members; rho is the ice density; h is 2 The coefficient of variation of ice coating diameter with height; l. the j Is a single member length; g is the unit electric load of ice-coating gravity of the ground wire, g =9.8 × 0.9 pi delta (d + delta) × 10 -3 Wherein d is the outer diameter of the conducting wire and the ground wire, and delta is the ice coating thickness.
Since different wind speeds can affect the axial bearing stress, models of wind speed and maximum ice coating thickness are established, and the risk threshold is dynamically adjusted through the weather forecast on the local day. In the present embodiment, when the wind speed is 20m/s, the specification that the icing thickness is between 0 and 12mm is not to exceed the risk threshold, and the specification that the icing thickness exceeds 12mm is to exceed the risk threshold.
Step two, data preprocessing
And framing the acquired ice coating data of the overhead transmission line with different thicknesses, demodulating the phase of the ice coating data to obtain a phase difference spectrum, and generating a two-dimensional image by using GAFs.
Wherein, the phase demodulation specifically comprises:
s201, designing a band-pass filter to perform intermediate frequency filtering on RBS original signals collected by a data acquisition card, and inhibiting broadband noise to obtain intermediate frequency signals;
s202, IQ demodulation is carried out on the intermediate frequency signal, complex intermediate frequency is constructed, and the amplitude A and the phase phi of the complex intermediate frequency are extracted;
s203, performing sliding average on the amplitude A along a time line by using a window with a proper width;
and S204, dividing the n sampling points in each row in the step S203 into phase detection intervals with equal width and connected end to end. Calculating and storing the column indexes of the positions with the maximum amplitude in all the intervals in the result corresponding to the S203;
s205, extracting phases in two adjacent left and right intervals from phi, and calculating the phase difference as the estimation of the RBS phases at the optical fiber positions corresponding to the sampling points of the two intervals at the current moment. And backtracking the phases of a pair of column indexes of the two adjacent left and right intervals in the previous row, which are the same as the current time, and taking the phases as the estimation of the RBS phases at the optical fiber positions corresponding to the sampling points of the two intervals at the previous time to calculate the phase difference. Subtracting the phase difference of the previous moment from the phase difference of the current moment to obtain the variation of the phase difference;
s206, repeating the step S205, traversing from the 2 nd line and the 2 nd interval, and calculating the variation of the phase difference of the left interval and the right interval of each interval in a detection pulse period; the matrix is summed cumulatively along the time axis and unwrapped to recover the change in fiber stretch over time between adjacent segments caused by external vibrations.
Two-dimensional images are generated using GAFs, i.e., the scaled one-dimensional sequence data is converted from a rectangular coordinate system to a polar coordinate system by GAFs, and then the temporal correlation of different points in time is identified by taking into account the angular difference between the different points. The method specifically comprises the following steps:
s207, scaling the phase difference spectrum data range to [0,1] through the following formula:
s208, converting the scaled sequence data into polar coordinates, taking the numerical values as cosine values of an included angle and taking the time axis as a radius, wherein the specific formula is as follows:
s209, acquiring the angle difference between corresponding angles of different time points according to the following formula:
in practical application, the accuracy of network feature extraction is low and the robustness of the model is poor probably due to the influence of wind speed or other uncontrollable factors. Therefore to address this problem, the acquired data set is augmented by enhancing sample diversity. The expanded data set is mainly obtained by adding random signal noise with different intensities to the converted GAFs.
1. And performing random data enhancement on the converted GAFs image, enriching a sample set, manually screening the sample set, labeling category labels, storing the category-labeled pictures in corresponding folders, and making a data set. Fig. 3-5 show phase difference amplitude frequency plots for different ice coating thicknesses and corresponding GAFs images for the simulation case.
2. The data set was shuffled and 80% were randomly selected as the training data set and the remaining 20% as the test data set.
3. The input image is scaled to 224 x 224 size and normalized.
Step three, model building
And creating a CNN model, wherein the CNN model consists of an input layer, a 1-layer 7 × 7 convolution module, a 3-layer 3 × 3 convolution module and an output module, and the step size of each layer of convolution is 2. The convolution module is composed of a convolution layer-BatchNormal layer-ReLU activation function layer. The output module consists of a linear layer and a softmax activation function, and when the output value of the activation function is 1, the corresponding icing thickness category is no icing; when the output value of the activation function is 2, the corresponding icing thickness category is that the icing thickness is lower than a threshold value; when the output value of the activation function is 3, the corresponding ice coating thickness category is that the ice coating thickness is higher than the threshold value.
The Pythrch deep learning framework is applied to the classification of the icing thickness of the overhead transmission line, and the version is 1.10.0. Fig. 2 is a diagram showing a structure of a convolutional neural network model, and the convolutional neural network model suitable for classifying the icing thickness of the overhead transmission line is constructed through continuous debugging.
The CNN model consists of an input layer, a 1-layer 7 × 7 convolution module, a 3-layer 3 × 3 convolution module, and an output module.
The convolution layer adopts a linear rectification function ReLU function as an activation function, when the input value is negative, the output is 0, and when the input value is more than 0, the output is the original value. The ReLU expression is:
f(x)＝max(0,x)。
and the pooling layer uses a max-pooling maximum pooling mode to take the maximum value of the feature points in the neighborhood. Max-posing can reduce the deviation of the estimated mean value caused by parameter errors of the convolutional layer, the three types of GAFs images identified by the invention have obvious texture characteristics, and more texture information is reserved by using Max-posing. And after pooling, performing local response normalization operation, creating a competitive mechanism for the activity of local neurons, so that the corresponding larger value of the local neurons becomes relatively larger, inhibiting other neurons with smaller feedback, and enhancing the generalization capability of the model.
Step four, model training
Training the CNN model using the two-dimensional images generated by the GAFs;
1. randomly setting the weight parameters of each layer in the network to be any value close to 0, and initializing a hyper-parameter;
2. setting the batch _ size to sequentially acquire train _ batch (including train _ data and train _ label) of each batch of batch training set, sending the train _ batch to a neural network, and performing gradient descent and iterative training at a learning rate of 0.0001;
3. calculating an actual output vector of the network;
4. comparing the elements of the target vector with the elements in the output vector, and calculating the output loss;
5. sequentially calculating the adjustment quantity of each weight and the adjustment quantity of the threshold;
6. adjusting the weight and the threshold;
7. and after M iterations, finishing training, and storing the weight value and the threshold value in corresponding folders. At this point, it can be considered that the respective weights have reached a stable value and a classifier has been formed.
The principle that the thicker the ice coating is, the smaller the natural frequency of each order mode of the overhead conductor is, and the monotonous decreasing mapping relation exists between the natural frequency of each order mode of the overhead conductor and the ice coating thickness is adopted. And converting the time sequence into GAFs pictures, classifying the GAFs pictures by using the visual advantage of the CNN on the machine, and finally establishing a mapping relation between the GAFs pictures and the ice coating thickness.
Step five, model testing
And classifying whether the overhead transmission line is iced or not through the CNN model, carrying out risk assessment on the iced overhead transmission line, and prompting a user through a human-computer interaction interface.
The trained neural network model meeting the precision requirement is tested for the identification accuracy by using a test set. The method comprises the following specific steps:
1. sending the sample data and the label of the test set into a model;
2. obtaining an output vector by using the trained model parameters;
3. an accuracy is calculated for the output vector using an accuracy function.
Example two
Referring to fig. 2, a classified early warning system for icing thickness of an overhead power transmission line is characterized by comprising:
the data acquisition module is used for acquiring the overhead transmission lines by using the phase-sensitive optical time domain reflectometer and classifying the overhead transmission lines;
the data analysis and preprocessing module is used for framing the icing data with different thicknesses in the overhead transmission line, demodulating the phase to obtain a phase difference frequency spectrum, and generating a two-dimensional image by using a gram angle field;
the model training module is used for creating a CNN model and using the two-dimensional image to train the CNN model;
and the processing and early warning module is used for classifying whether the overhead transmission line is iced or not through the CNN model and carrying out risk assessment and early warning on the iced overhead transmission line.
In summary, according to the classifying and early warning method and system for the icing thickness of the overhead transmission line, provided by the invention, the CNN model is established to classify whether the overhead transmission line is iced or not by utilizing the convolutional neural network and the distributed optical fiber sensing, and risk assessment and early warning are carried out, so that the problems that the overhead transmission line is easily influenced by the environment, the service life is short, the data transmission reliability is poor, and the monitoring cannot be accurately and comprehensively carried out in the icing monitoring process are solved.
It should be noted that, for the sake of simplicity, the above-mentioned method embodiments are described as a series of acts or combinations, but those skilled in the art should understand that the present invention is not limited by the described order of acts, as some steps may be performed in other orders or simultaneously according to the present invention. Further, those skilled in the art will appreciate that the embodiments described in the specification are presently preferred and that no acts or modules are necessarily required of the invention.
In the above embodiments, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.
The above description is only an embodiment of the present invention, and is not intended to limit the scope of the present invention, and all equivalent structures or equivalent processes performed by the present invention or directly or indirectly applied to other related technical fields are included in the scope of the present invention.
Claims (10)
1. The classification early warning method for the icing thickness of the overhead transmission line is characterized by comprising the following steps of:
s1, collecting overhead transmission lines by using a phase-sensitive optical time domain reflectometer, and classifying the overhead transmission lines;
s2, framing the ice coating data with different thicknesses in the overhead transmission line, demodulating the phase to obtain a phase difference frequency spectrum, and generating a two-dimensional image by using a Gelam angular field;
s3, creating a CNN model, and training the CNN model by using the two-dimensional image;
and S4, classifying whether the overhead transmission line is iced or not through the CNN model, and performing risk assessment and early warning on the iced overhead transmission line.
2. The method for classifying and early warning the icing thickness of the overhead transmission line according to claim 1, wherein the classifying the overhead transmission line in the step S1 specifically comprises:
dividing the overhead transmission line into three types, namely non-icing, icing below a risk threshold and icing above the risk threshold, wherein the risk threshold is an axial bearing stress value which is smaller than the yield strength of steel;
icing load F on node i of power transmission pole tower in power transmission line i Comprises the following steps:
in the formula, n representsNumber of members, p represents icing density, h 2 Denotes the coefficient of variation of ice coating diameter with height,/ j Representing the length of a single component, and g represents the icing gravity of the ground wire;
the ground wire icing gravity g =9.8 × 0.9 pi δ (d + δ) × 10 -3 The unit is electric load, wherein d represents the outer diameter of the conducting wire and the grounding wire, and delta represents the thickness of the ice coating.
3. The overhead transmission line icing thickness classification early warning method according to claim 1, wherein the phase demodulation in the step S2 specifically comprises:
s201, performing intermediate frequency filtering on the original signals of the wireless base station acquired by the data acquisition card, and inhibiting broadband noise to obtain intermediate frequency signals;
s202, IQ demodulation is carried out on the intermediate frequency signal, a complex intermediate frequency is constructed, and the amplitude and the phase of the complex intermediate frequency are extracted;
s203, performing sliding average on the amplitude along a time line by using a window with a proper width;
s204, dividing the n sampling points of each time row in the step S203 into phase discrimination intervals with equal width and connected end to end, and solving and storing the column index of the position with the maximum amplitude in all the intervals, which corresponds to the time row, in the result corresponding to the S203;
s205, extracting phases in a left interval and a right interval adjacent to the phases from the phases, using the phases as estimation of RBS phases at optical fiber positions corresponding to sampling points of the left interval and the right interval at the current moment, calculating a first phase difference, backtracking phases at a pair of column indexes of the left interval and the right interval at the previous row, which are adjacent to the left interval and the right interval at the current moment, using the phases as estimation of RBS phases at the optical fiber positions corresponding to the sampling points of the two intervals at the previous moment, calculating a second phase difference, and subtracting the phase difference at the previous moment from the phase difference at the current moment to obtain the variation of the phase difference;
s206, repeating the step S205, traversing from the 2 nd line and the 2 nd interval, and calculating the variation of the phase difference of the left interval and the right interval of each interval in a detection pulse period;
the matrix is summed cumulatively along the time axis and unwrapped to recover the change in fiber stretch over time between adjacent segments caused by external vibrations.
4. The overhead transmission line icing thickness classification early warning method according to claim 1, wherein the step S2 of generating the two-dimensional image by using the gram angle field specifically comprises the steps of:
s207, scaling the data range of the phase difference spectrum to [0,1] through the following formula:
s208, converting the zoomed sequence data into polar coordinates, taking the numerical value as a cosine value of an included angle, and taking a time axis as a radius:
s209, acquiring the angle difference between corresponding angles of different time points according to the following formula:
in the formula (I), the compound is shown in the specification,
representing a one-dimensional time series.
5. The overhead transmission line icing thickness classification early warning method according to claim 1, wherein the CNN model in step S3 is specifically composed of an input layer, a 1-layer first convolution module, a 3-layer second convolution module, and an output module.
6. The overhead transmission line icing thickness classification early warning method according to claim 5, wherein the convolution module is composed of a convolution layer-batch normal layer-ReLU activation function layer.
7. The overhead transmission line icing thickness classification early warning method according to claim 5, wherein the output module is composed of a linear layer and an activation function, and when the output value of the activation function is 1, the corresponding icing thickness category is no icing; when the output value of the activation function is 2, the corresponding icing thickness category is that the icing thickness is lower than a threshold value; when the output value of the activation function is 3, the corresponding ice coating thickness category is that the ice coating thickness is higher than the threshold value.
8. The overhead transmission line icing thickness classification early warning method according to claim 7, wherein the activation function is a softmax activation function.
9. The overhead transmission line icing thickness classification early warning method according to claim 1, wherein in step S4, a user is prompted through a human-computer interaction interface.
10. The utility model provides an overhead transmission line icing thickness classification early warning system which characterized in that includes:
the data acquisition module is used for acquiring the overhead transmission lines by using the phase-sensitive optical time domain reflectometer and classifying the overhead transmission lines;
the data analysis and preprocessing module is used for framing the icing data with different thicknesses in the overhead transmission line, demodulating the phase to obtain a phase difference frequency spectrum, and generating a two-dimensional image by using a gram angle field;
the model training module is used for creating a CNN model and training the CNN model by using the two-dimensional image;
and the processing and early warning module is used for classifying whether the overhead transmission line is iced through the CNN model and carrying out risk assessment and early warning on the iced overhead transmission line.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211566952.0A CN115797772A (en) | 2022-12-07 | 2022-12-07 | Classified early warning method and system for icing thickness of overhead transmission line |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211566952.0A CN115797772A (en) | 2022-12-07 | 2022-12-07 | Classified early warning method and system for icing thickness of overhead transmission line |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115797772A true CN115797772A (en) | 2023-03-14 |
Family
ID=85417740
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211566952.0A Pending CN115797772A (en) | 2022-12-07 | 2022-12-07 | Classified early warning method and system for icing thickness of overhead transmission line |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115797772A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116576924A (en) * | 2023-07-13 | 2023-08-11 | 陕西协成测试技术有限公司 | Detection alarm device for long-distance power transmission line |
| CN116863251A (en) * | 2023-09-01 | 2023-10-10 | 湖北工业大学 | Distributed optical fiber sensing disturbance recognition method |
-
2022
- 2022-12-07 CN CN202211566952.0A patent/CN115797772A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116576924A (en) * | 2023-07-13 | 2023-08-11 | 陕西协成测试技术有限公司 | Detection alarm device for long-distance power transmission line |
| CN116863251A (en) * | 2023-09-01 | 2023-10-10 | 湖北工业大学 | Distributed optical fiber sensing disturbance recognition method |
| CN116863251B (en) * | 2023-09-01 | 2023-11-17 | 湖北工业大学 | Distributed optical fiber sensing disturbance recognition method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115797772A (en) | Classified early warning method and system for icing thickness of overhead transmission line | |
| CN103323749B (en) | The partial discharge diagnostic method of multi-classifier information fusion | |
| CN111442827B (en) | Optical fiber passive online monitoring system for transformer winding vibration | |
| CN111505632B (en) | Ultra-wideband radar action attitude identification method based on power spectrum and Doppler characteristics | |
| CN113049084B (en) | Attention mechanism-based Resnet distributed optical fiber sensing signal identification method | |
| CN108564005A (en) | A kind of human body tumble discrimination method based on convolutional neural networks | |
| CN111679166A (en) | Switch cabinet partial discharge fault multi-source information fusion detection early warning system and method based on wireless transmission technology | |
| CN103617684A (en) | Interference type optical fiber perimeter vibration intrusion recognition algorithm | |
| CN108680796A (en) | Electromagnetic information leakage detecting system and method for computer display | |
| Zhang et al. | Device-free wireless localization and activity recognition with deep learning | |
| CN109270346A (en) | Polarization state waveform signal processing method at a kind of OPGW lightning stroke | |
| Li | A new robust signal recognition approach based on holder cloud features under varying SNR environment | |
| CN112883802A (en) | Method for identifying destructive event of pipeline optical fiber vibration safety early warning system | |
| CN112949487A (en) | WiFi-based ship cab personnel number detection method and system | |
| CN111181634B (en) | Distributed optical fiber vibration signal rapid positioning method | |
| CN114137358B (en) | Power transmission line fault diagnosis method based on graph convolution neural network | |
| CN117353811B (en) | Multi-working-condition monitoring and analyzing method for electric power optical communication system | |
| CN114519372A (en) | One-dimensional range profile target identification method based on support vector machine | |
| CN114062839A (en) | Railway power line fault positioning device and method thereof | |
| CN113916302A (en) | Overhead traveling crane online detection and diagnosis system and method based on artificial intelligence technology | |
| CN112305378B (en) | Photoelectric composite submarine cable fault location system and fault location method | |
| CN107341519B (en) | Support vector machine identification optimization method based on multi-resolution analysis | |
| CN106226845B (en) | A method of identifying Hail distribution region from OPGW vibration signals | |
| Xiang et al. | Signal recognition method of x-ray pulsar based on cnn and attention module cbam | |
| CN112329535B (en) | CNN-based quick identification method for low-frequency oscillation modal characteristics of power system |
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 |