Early warning method for transmission line icing based on optical fiber vibration data hierarchical context

Abstract

The present invention provides a method for early warning of transmission line icing based on optical fiber vibration data hierarchical context, including: installing optical fiber and phase-sensitive optical time domain reflectometer on the transmission line; establishing a time domain index expression module; establishing an optical fiber vibration data hierarchical context feature extraction module; establishing a hierarchical icing decision neural network model; using the optical fiber vibration data hierarchical context feature extraction module and the trained hierarchical icing decision neural network model, the collected 4-second optical fiber vibration data is analyzed to determine whether the transmission line has early icing, and if so, an early warning is issued. Through the hierarchical structure of the present invention, the change process between different periods of optical fiber vibration data can be described, and early warning of transmission line icing can be more effectively performed, thereby improving the efficiency of power grid inspection management.

Description

Technical Field

[0001] The present invention relates to the technical field of remote monitoring of power transmission lines, and in particular to a method for early warning of icing of power transmission lines based on the hierarchical context of optical fiber vibration data. Background Art

[0002] In the cold northern regions, ice on transmission lines often occurs, which can easily damage transmission line equipment, leading to huge economic losses and safety accidents. Therefore, it is necessary to deal with it as early as possible in the early stages when the ice on the transmission lines is thin. The optical fiber deployed on the transmission lines can quickly collect the vibration data of the corresponding lines, and the vibration mode of the transmission lines will change after ice is applied; therefore, it is very necessary to use optical fiber vibration data to remotely monitor the ice coverage of transmission lines, and to issue early warnings and deal with the early ice coverage of transmission lines, which can significantly improve the management efficiency of power grid maintenance companies.

[0003] At present, the main means of using optical fiber vibration data to collect the initial icing of transmission lines is to use phase-sensitive optical time-domain reflectometers to collect optical fiber vibration data on transmission lines and use artificial intelligence models to analyze the optical fiber vibration data. The main methods currently used include: 1) Directly use artificial intelligence models for processing. Since the frequency of optical fiber vibration data collected by phase-sensitive optical time-domain reflectometers is relatively high, the related time-frequency attributes generated are also more, and there are usually fewer samples reflecting the initial icing state of transmission lines, so this processing mode usually has limited accuracy; 2) Use Boost or Bagging integrated classifier methods. This method can relatively effectively deal with the problem of too high attribute dimensions of optical fiber vibration data; however, in real application scenarios, due to different transmission line lengths and assembly methods, the characteristics of optical fiber vibration data will change. Therefore, the optical fiber vibration data obtained in the laboratory is slightly different from the actual optical fiber vibration data. This model usually has good experimental results in the laboratory, but in actual applications, this method may fit a specific model and cause prediction failure. Especially in the case of initial icing of transmission lines, due to the low thickness of ice, its pattern is similar to that of the un-iced state, and the above two methods have poor performance; 3) Using a more complex deep neural network model to distinguish the core difference between initial icing and un-icing. The biggest problem with this model is that the number of "transmission line icing" samples is small, which is not enough to support the full training of larger neural networks, resulting in overfitting, making the warning accuracy fail to meet the requirements.

[0004] Therefore, it is necessary to propose a method that can adapt to the difference between the optical fiber vibration data of the transmission lines with thin initial ice coverage and those without ice coverage, adapt to the differences caused by the construction of different transmission lines, and accurately and early carry out early warning of transmission line ice coverage. Summary of the invention

[0005] In view of the problems existing in the prior art, the present invention provides a method for early warning of transmission line icing based on the hierarchical context of optical fiber vibration data. The method constructs a hierarchical context feature extraction module for optical fiber vibration data and a hierarchical icing decision neural network model, and utilizes the hierarchical characteristics of optical fiber vibration data to find the difference between the optical fiber vibration data of the initial thin ice coverage and the un-iced optical fiber vibration data of the transmission line, thereby realizing early warning of transmission line icing.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] The present invention provides a transmission line initial icing warning method based on optical fiber vibration data level context, comprising the following steps:

[0008] S1, installing optical fiber and phase-sensitive optical time domain reflectometer on the transmission line; the phase-sensitive optical time domain reflectometer is used to collect optical fiber vibration data, collect the optical fiber vibration data and historical data of the transmission line icing situation, and store them in the icing situation history data table VbList; obtain the number of elements VbNum of the icing situation history data table; initialize the optical fiber vibration data level context feature list CJList;

[0009] S2, establishing a time domain indicator expression module FEModel; the input of the time domain indicator expression module FEModel is the time domain indicator expression input array FEInput, and the output of the time domain indicator expression module FEModel is the time domain indicator expression output array FEOutput;

[0010] S3, establish a fiber vibration data level context feature extraction module TXModel; the input of the fiber vibration data level context feature extraction module TXModel is a floating point array TXInput of 4000 elements, the fiber vibration data level context feature extraction module TXModel uses the time domain index expression module FEModel to process the floating point array TXInput, and the output of the fiber vibration data level context feature extraction module TXModel is a fiber vibration data level context feature output result TXOutput;

[0011] S4, using the optical fiber vibration data level context feature extraction module TXModel to process all the contents of the ice cover condition form historical data table VbList, obtain the optical fiber vibration data level context feature, and store the obtained optical fiber vibration data level context feature in the optical fiber vibration data level context feature list CJList;

[0012] S5, establishing a hierarchical icing decision neural network model CCNN; using the optical fiber vibration data hierarchical context feature list CJList to train the hierarchical icing decision neural network model CCNN, to obtain a trained hierarchical icing decision neural network model CCNN;

[0013] S6, during actual monitoring, the phase-sensitive optical time domain reflectometer is used to collect 4-second optical fiber vibration data Test, and the optical fiber vibration data hierarchical context feature extraction module TXModel and the trained hierarchical icing decision neural network model CCNN are used to analyze the collected 4-second optical fiber vibration data Test to determine whether initial icing of the transmission line occurs. If so, an early warning is issued.

[0014] Preferably, S1 is specifically:

[0015] S101, installation of optical fiber and phase-sensitive optical time-domain reflectometers on transmission lines;

[0016] The phase-sensitive optical time domain reflectometer collects optical fiber vibration data at 1000 Hz, collects the optical fiber vibration data and historical data formed by the icing condition of the transmission line, and stores them in the icing condition form historical data table VbList; wherein each element of the icing condition form historical data table VbList has 2 fields:

[0017] Optical fiber vibration data array VbData: The optical fiber vibration data array VbData is a floating point array of 4000 elements. Since the acquisition frequency of the phase-sensitive optical time domain reflectometer is 1000 Hz, the optical fiber vibration data array VbData corresponds to optical fiber vibration data of 4 seconds in length;

[0018] Whether thin ice coverage appears VbObserved: It is an integer variable. 1 indicates that thin ice coverage appears. At this time, the ice thickness of the transmission line is greater than 0 and less than 5 mm. 0 indicates that there is no ice coverage.

[0019] S102, the number of elements in the ice cover form historical data table VbNum=the number of elements in the ice cover form historical data table VbList;

[0020] S103, establish a fiber vibration data level context feature list CJList, which is initially an empty list; each element structure of the fiber vibration data level context feature list CJList contains 6 fields:

[0021] The first level context feature Vb1 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector;

[0022] The second-level context feature Vb2 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector;

[0023] The third level context feature Vb3 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector;

[0024] The fourth level context feature Vb4 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector;

[0025] The fifth level context feature Vb5 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector;

[0026] Icing decision FBDecison of the fiber vibration data level context feature list CJList: an integer variable expressing whether icing occurs;

[0027] S104, constructing a fiber vibration data level context feature list counter CJCounter=1;

[0028] S105, creating an element CJListItem of the optical fiber vibration data level context property list CJList, the element CJListItem having 6 fields:

[0029] The first level context feature Vb1 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector, and the interior of the vector is all 0;

[0030] The second-level context feature Vb2 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector, and the interior of the vector is all 0;

[0031] The third level context feature Vb3 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector, and the interior of the vector is all 0;

[0032] The fourth level context feature Vb4 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector, and the interior of the vector is all 0;

[0033] The fifth level context feature Vb5 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector, and the interior of the vector is all 0;

[0034] Ice cover decision FBDecison=0;

[0035] S106, step S1 ends.

[0036] Preferably, S2 is specifically:

[0037] S201, establishing a time domain indicator expression module FEModel; the input of the time domain indicator expression module FEModel is a time domain indicator expression input array FEInput, and the time domain indicator expression input array FEInput is an array of 100 elements;

[0038] S202, the time domain index expression input array FEInput is calculated by the following formula:

[0039] (1) Time domain standard deviation index:

[0040] Standard deviation σ: measures the degree of dispersion of data, and the calculation formula is:

[0041]

[0042] Among them: The calculation formula of mean μ is:

[0043]

[0044] in: represents the i-th element in the time domain indicator expression input array FEInput; N=100, represents the number of elements in the time domain indicator expression input array FEInput;

[0045] (2) Time domain peak-to-peak index:

[0046] The peak-to-peak value PP represents the extreme value range in the input array FEInput expressed by the time domain indicator. The calculation formula is:

[0047]

[0048] in: Indicates that the time domain indicator expresses the maximum value of the elements in the input array FEInput; Indicates that the time domain indicator expresses the minimum value of the elements in the input array FEInput;

[0049] (3) Time domain skewness index:

[0050] Skewness Measures the symmetry of data distribution, and the calculation formula is:

[0051]

[0052] (4) Time domain margin index:

[0053] The margin R represents the relative size of the peak-to-peak value and the mean value, and the calculation formula is:

[0054]

[0055] (5) Time domain kurtosis index:

[0056] Kurtosis Describes the sharpness of the data distribution, and the calculation formula is:

[0057]

[0058] (6) Time domain maximum value index:

[0059] The maximum value Max represents the maximum element in the input array FEInput of the time domain indicator expression. The calculation formula is:

[0060]

[0061] (7) Time domain minimum index:

[0062] The minimum value Min represents the smallest element in the input array FEInput of the time domain indicator expression. The calculation formula is:

[0063]

[0064] (8) Time domain pulse indicators:

[0065] The pulse index P represents the pulse strength of the signal and is defined as:

[0066]

[0067] in: Indicates that the time domain indicator expresses the absolute value of the i-th element in the input array FEInput;

[0068] (9) Time domain waveform value index:

[0069] Waveform value A measure of the overall volatility of the signal, calculated as:

[0070]

[0071] (10) Time domain energy index:

[0072] Energy E represents the sum of the squares of all elements in the input array FEInput, expressed in the time domain. The calculation formula is:

[0073]

[0074] S203, establishing a time domain index expression output array FEOutput as an array of 10 elements;

[0075] S204, storing the 10 index results calculated in S202 in sequence into the time domain index expression output array FEOutput;

[0076] S205, outputting the time domain index expression output array FEOutput as the result of the time domain index expression module FEModel.

[0077] Preferably, S3 is specifically:

[0078] S301, establish a fiber vibration data level context feature extraction module TXModel; the input of the fiber vibration data level context feature extraction module TXModel is a floating point array TXInput of 4000 elements;

[0079] S302, set the first counter TXCounter1 of the optical fiber vibration data hierarchical context feature extraction module to 1; construct an optical fiber vibration data hierarchical context feature extraction module list variable TXList = empty list;

[0080] S303, the first temporary variable TXTemp1 of the optical fiber vibration data level context feature extraction module = takes out the (TXCounter1-1)×100th to (TXCounter1-1)×100+99th elements of the floating point array TXInput, which is 100 elements in total;

[0081] S304, the second temporary variable TXTemp2 of the optical fiber vibration data level context feature extraction module:

[0082] The first temporary variable TXTemp1 of the optical fiber vibration data level context feature extraction module is processed by using the time domain indicator expression module FEModel, the time domain indicator expression input array FEInput of the time domain indicator expression module FEModel = the first temporary variable TXTemp1 of the optical fiber vibration data level context feature extraction module, and the time domain indicator expression output array FEOutput is obtained; the second temporary variable TXTemp2 of the optical fiber vibration data level context feature extraction module = the time domain indicator expression output array FEOutput;

[0083] S305, adding the second temporary variable TXTemp2 of the optical fiber vibration data hierarchical context characteristic extraction module to the optical fiber vibration data hierarchical context characteristic extraction module list variable TXList;

[0084] S306, TXCounter1=TXCounter1+1;

[0085] S307, if TXCounter1 is less than or equal to 40, go to S303; otherwise go to S308;

[0086] S308, establish the optical fiber vibration data level context characteristic output result TXOutput, the structure of the optical fiber vibration data level context characteristic output result TXOutput is the same as an element structure of the optical fiber vibration data level context characteristic list CJList, including 6 fields:

[0087] The Vb1 field of the optical fiber vibration data hierarchical context characteristic output result TXOutput = the optical fiber vibration data hierarchical context characteristic extraction module list variable TXList; specifically, the Vb1 field of the optical fiber vibration data hierarchical context characteristic output result TXOutput is a list of 40 elements, and each element in the list is a 10-dimensional vector;

[0088] The Vb2 field of the fiber vibration data level context feature output result TXOutput = creates a new list whose value is to insert a 10-dimensional vector of all 0s before the first element of its vb1 field, and delete an element at the end of its vb1 field;

[0089] The Vb3 field of the fiber vibration data level context feature output result TXOutput = creates a new list whose value is to insert a 10-dimensional vector of all 0s before the first element of its vb2 field, and delete an element at the end of its vb2 field;

[0090] The Vb4 field of the fiber vibration data level context feature output result TXOutput = creates a new list whose value is to insert a 10-dimensional vector of all 0s before the first element of its vb3 field, and delete an element at the end of its vb3 field;

[0091] The Vb5 field of the fiber vibration data level context feature output result TXOutput = creates a new list, whose value is to insert a 10-dimensional vector of all 0s before the first element of its vb4 field, and delete an element at the end of its vb4 field;

[0092] Fiber vibration data level context feature output result TXOutput Icing decision FBDecison field: an integer variable expressing whether icing occurs;

[0093] S309, updating the fields of the optical fiber vibration data level context characteristic output result TXOutput established in S308:

[0094] The Vb2 field of the optical fiber vibration data level context characteristic output result TXOutput = the optical fiber vibration data level context characteristic extraction module list variable TXList - the Vb2 field of the optical fiber vibration data level context characteristic output result TXOutput;

[0095] The Vb3 field of the optical fiber vibration data level context characteristic output result TXOutput = the optical fiber vibration data level context characteristic extraction module list variable TXList - the Vb3 field of the optical fiber vibration data level context characteristic output result TXOutput;

[0096] The Vb4 field of the optical fiber vibration data level context characteristic output result TXOutput = the optical fiber vibration data level context characteristic extraction module list variable TXList - the Vb4 field of the optical fiber vibration data level context characteristic output result TXOutput;

[0097] The Vb5 field of the optical fiber vibration data level context characteristic output result TXOutput = the optical fiber vibration data level context characteristic extraction module list variable TXList - the Vb5 field of the optical fiber vibration data level context characteristic output result TXOutput;

[0098] The icing decision FBDecison field of the fiber vibration data level context characteristic output result TXOutput = 0;

[0099] S310, outputting the optical fiber vibration data hierarchical context characteristic output result TXOutput as the result output of the optical fiber vibration data hierarchical context characteristic extraction module TXModel.

[0100] Preferably, S4 is specifically:

[0101] S401, result processing counter CLCounter=1;

[0102] S402, result processing first temporary variable CLTemp1:

[0103] Take out the CLCounterth element of the ice cover status historical data table VbList, and assign it to the first temporary variable CLTemp1 of the result processing;

[0104] S403, result processing second temporary variable CLTemp2:

[0105] The optical fiber vibration data hierarchical context feature extraction module TXModel is used for processing, the floating point array TXInput of the optical fiber vibration data hierarchical context feature extraction module TXModel is input as the result processing first temporary variable CLTemp1, the optical fiber vibration data hierarchical context feature output result TXOutput is obtained by processing, and the result is assigned to the result processing second temporary variable CLTemp2;

[0106] S404, assigning the result processing second temporary variable CLTemp2 to the CLCounterth element of the optical fiber vibration data level context characteristic list CJList;

[0107] S405, taking out the VbObserved field of whether thin ice appears in the CLCounterth element of the ice condition history data table VbList, and assigning it to the ice decision FBDecison field of the CLCounterth element of the optical fiber vibration data level context feature list CJList;

[0108] S406, CLCounter=CLCounter+1;

[0109] S407, if CLCounter is less than or equal to VbNum, go to S402; otherwise go to S408;

[0110] S408, the processing of step S4 ends.

[0111] Preferably, S5 is specifically:

[0112] S501, establish a hierarchical ice cover decision neural network model CCNN, which includes the following structure:

[0113] It has 5 input branches, each of which outputs an output array of 100 elements after passing through an LSTM layer. After the 5 input branches are processed, the outputs are merged into a 100×5 feature map. For this feature map, a decision is formed using 2 convolutional layers, 1 maximum pooling layer, and a cubic MLP network.

[0114] S502, the five input branches of the hierarchical icing decision neural network model CCNN are respectively connected to the vb1, vb2, vb3, vb4 and vb5 fields of the optical fiber vibration data hierarchical context feature list CJList, and the decision output of the hierarchical icing decision neural network model CCNN is connected to the icing decision FBDecison field of the optical fiber vibration data hierarchical context feature list CJList;

[0115] S503, using the data of the optical fiber vibration data hierarchical context feature list CJList to train the hierarchical ice coverage decision neural network model CCNN, so that the hierarchical ice coverage decision neural network model CCNN has decision-making ability;

[0116] The processing of steps S504 and S5 ends.

[0117] Preferably, S6 is specifically:

[0118] S601, using a phase-sensitive optical time domain reflectometer to collect 4 seconds of optical fiber vibration data Test;

[0119] S602, a first temporary variable PDTemp1 for judging:

[0120] The optical fiber vibration data level context feature extraction module TXModel is used for processing, the floating point array TXInput of the optical fiber vibration data level context feature extraction module TXModel is input as the optical fiber vibration data Test, the optical fiber vibration data level context feature output result TXOutput of the optical fiber vibration data level context feature extraction module TXModel is obtained, and the result is assigned to the first temporary variable PDTemp1 used for judgment;

[0121] S603, inputting the first temporary variable PDTemp1 used for judgment into the hierarchical icing decision neural network model CCNN, and the five input branches of the hierarchical icing decision neural network model CCNN are respectively connected to the vb1, vb2, vb3, vb4 and vb5 fields of the first temporary variable PDTemp1 used for judgment;

[0122] S604, a second temporary variable PDTemp2 for judging:

[0123] The decision output of the hierarchical ice coverage decision neural network model CCNN is assigned to the second temporary variable PDTemp2 used for judgment;

[0124] S605, if the second temporary variable PDTemp2 used for judgment is equal to 0, go to S606; otherwise go to S607;

[0125] S606, output that there is no initial icing of the transmission line, and go to S608;

[0126] S607, indicating that the transmission line has initial icing, an early warning is issued, and the process goes to S608;

[0127] S608, end.

[0128] The transmission line initial icing warning method based on optical fiber vibration data hierarchical context provided by the present invention has the following advantages:

[0129] The hierarchical structure of the present invention can describe the change process between different cycles of optical fiber vibration data. On the one hand, this makes it possible for the neural network to find a specific level corresponding to the ice warning, and adapt to the changes in the characteristics of optical fiber vibration data caused by different construction methods and assembly methods; on the other hand, the difference calculated by the formula between the levels can be independent of the sample, and the difference between different cycles of optical fiber vibration data can be more clearly expressed, which is helpful for dealing with the situation where there are few samples of initial ice coverage of transmission lines. Through the present invention, early warning of ice coverage of transmission lines can be carried out more effectively, and the efficiency of power grid inspection management can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0130] Figure 1 A schematic flow chart of a method for early warning of icing of transmission lines based on the hierarchical context of optical fiber vibration data provided by the present invention. DETAILED DESCRIPTION

[0131] In order to make the technical problems, technical solutions and beneficial effects solved by the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0132] The present invention provides a method for early warning of transmission line icing based on the hierarchical context of optical fiber vibration data. The method constructs a hierarchical context feature extraction module for optical fiber vibration data and a hierarchical icing decision neural network model. The hierarchical features of optical fiber vibration data are established and the difference between the optical fiber vibration data of the initial thin ice and the un-iced optical fiber vibration data of the transmission line is found, thereby realizing early warning of transmission line icing.

[0133] See also Figure 1 The present invention provides a method for early warning of transmission line icing based on optical fiber vibration data level context, comprising the following steps:

[0134] S1, installing optical fiber and phase-sensitive optical time domain reflectometer on the transmission line; the phase-sensitive optical time domain reflectometer is used to collect optical fiber vibration data, collect the optical fiber vibration data and historical data of the transmission line icing situation, and store them in the icing situation history data table VbList; obtain the number of elements VbNum of the icing situation history data table; initialize the optical fiber vibration data level context feature list CJList;

[0135] S1 is specifically:

[0136] S101, installation of optical fiber and phase-sensitive optical time-domain reflectometers on transmission lines;

[0137] The phase-sensitive optical time domain reflectometer collects optical fiber vibration data at 1000 Hz, collects the optical fiber vibration data and historical data formed by the icing condition of the transmission line, and stores them in the icing condition form historical data table VbList; wherein each element of the icing condition form historical data table VbList has 2 fields:

[0138] Optical fiber vibration data array VbData: The optical fiber vibration data array VbData is a floating point array of 4000 elements. Since the acquisition frequency of the phase-sensitive optical time domain reflectometer is 1000 Hz, the optical fiber vibration data array VbData corresponds to optical fiber vibration data of 4 seconds in length;

[0139] Whether thin ice coverage appears VbObserved: It is an integer variable. 1 indicates that thin ice coverage appears. At this time, the ice thickness of the transmission line is greater than 0 and less than 5 mm. 0 indicates that there is no ice coverage.

[0140] S102, the number of elements in the ice cover form historical data table VbNum=the number of elements in the ice cover form historical data table VbList;

[0141] S103, establish a fiber vibration data level context feature list CJList, which is initially an empty list; each element structure of the fiber vibration data level context feature list CJList contains 6 fields:

[0142] The first level context feature Vb1 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector;

[0143] The second-level context feature Vb2 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector;

[0144] The third level context feature Vb3 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector;

[0145] The fourth level context feature Vb4 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector;

[0146] The fifth level context feature Vb5 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector;

[0147] Icing decision FBDecison of the fiber vibration data level context feature list CJList: an integer variable expressing whether icing occurs;

[0148] S104, constructing a fiber vibration data level context feature list counter CJCounter=1;

[0149] S105, creating an element CJListItem of the optical fiber vibration data level context property list CJList, the element CJListItem having 6 fields:

[0150] The first level context feature Vb1 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector, and the interior of the vector is all 0;

[0151] The second-level context feature Vb2 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector, and the interior of the vector is all 0;

[0152] The third level context feature Vb3 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector, and the interior of the vector is all 0;

[0153] The fourth level context feature Vb4 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector, and the interior of the vector is all 0;

[0154] The fifth level context feature Vb5 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector, and the interior of the vector is all 0;

[0155] Ice cover decision FBDecison=0;

[0156] S106, step S1 ends.

[0157] S2, establishing a time domain indicator expression module FEModel; the input of the time domain indicator expression module FEModel is the time domain indicator expression input array FEInput, and the output of the time domain indicator expression module FEModel is the time domain indicator expression output array FEOutput;

[0158] S2 is specifically:

[0159] S201, establishing a time domain indicator expression module FEModel; the input of the time domain indicator expression module FEModel is a time domain indicator expression input array FEInput, and the time domain indicator expression input array FEInput is an array of 100 elements;

[0160] S202, the time domain index expression input array FEInput is calculated by the following formula:

[0161] (1) Time domain standard deviation index:

[0162] Standard deviation σ: measures the degree of dispersion of data, and the calculation formula is:

[0163]

[0164] Among them: The calculation formula of mean μ is:

[0165]

[0166] in: represents the i-th element in the time domain indicator expression input array FEInput; N=100, represents the number of elements in the time domain indicator expression input array FEInput;

[0167] (2) Time domain peak-to-peak index:

[0168] The peak-to-peak value PP represents the extreme value range in the input array FEInput expressed by the time domain indicator. The calculation formula is:

[0169]

[0170] in: Indicates that the time domain indicator expresses the maximum value of the elements in the input array FEInput; Indicates that the time domain indicator expresses the minimum value of the elements in the input array FEInput;

[0171] (3) Time domain skewness index:

[0172] Skewness Measures the symmetry of data distribution, and the calculation formula is:

[0173]

[0174] (4) Time domain margin index:

[0175] The margin R represents the relative size of the peak-to-peak value and the mean value, and the calculation formula is:

[0176]

[0177] (5) Time domain kurtosis index:

[0178] Kurtosis Describes the sharpness of the data distribution, and the calculation formula is:

[0179]

[0180] (6) Time domain maximum value index:

[0181] The maximum value Max represents the maximum element in the input array FEInput of the time domain indicator expression. The calculation formula is:

[0182]

[0183] (7) Time domain minimum index:

[0184] The minimum value Min represents the smallest element in the input array FEInput of the time domain indicator expression. The calculation formula is:

[0185]

[0186] (8) Time domain pulse indicators:

[0187] The pulse index P represents the pulse strength of the signal and is defined as:

[0188]

[0189] in: Indicates that the time domain indicator expresses the absolute value of the i-th element in the input array FEInput;

[0190] (9) Time domain waveform value index:

[0191] Waveform value A measure of the overall volatility of the signal, calculated as:

[0192]

[0193] (10) Time domain energy index:

[0194] Energy E represents the sum of the squares of all elements in the input array FEInput, expressed in the time domain. The calculation formula is:

[0195]

[0196] S203, establishing a time domain index expression output array FEOutput as an array of 10 elements;

[0197] S204, storing the 10 index results calculated in S202 in sequence into the time domain index expression output array FEOutput;

[0198] S205, outputting the time domain index expression output array FEOutput as the result of the time domain index expression module FEModel.

[0199] S3, establish a fiber vibration data level context feature extraction module TXModel; the input of the fiber vibration data level context feature extraction module TXModel is a floating point array TXInput of 4000 elements, the fiber vibration data level context feature extraction module TXModel uses the time domain index expression module FEModel to process the floating point array TXInput, and the output of the fiber vibration data level context feature extraction module TXModel is a fiber vibration data level context feature output result TXOutput;

[0200] S3 is specifically:

[0201] S301, establish a fiber vibration data level context feature extraction module TXModel; the input of the fiber vibration data level context feature extraction module TXModel is a floating point array TXInput of 4000 elements;

[0202] S302, set the first counter TXCounter1 of the optical fiber vibration data hierarchical context feature extraction module to 1; construct an optical fiber vibration data hierarchical context feature extraction module list variable TXList = empty list;

[0203] S303, the first temporary variable TXTemp1 of the optical fiber vibration data level context feature extraction module = takes out the (TXCounter1-1)×100th to (TXCounter1-1)×100+99th elements of the floating point array TXInput, which is 100 elements in total;

[0204] S304, the second temporary variable TXTemp2 of the optical fiber vibration data level context feature extraction module:

[0205] The first temporary variable TXTemp1 of the optical fiber vibration data level context feature extraction module is processed by using the time domain indicator expression module FEModel, the time domain indicator expression input array FEInput of the time domain indicator expression module FEModel = the first temporary variable TXTemp1 of the optical fiber vibration data level context feature extraction module, and the time domain indicator expression output array FEOutput is obtained; the second temporary variable TXTemp2 of the optical fiber vibration data level context feature extraction module = the time domain indicator expression output array FEOutput;

[0206] S305, adding the second temporary variable TXTemp2 of the optical fiber vibration data hierarchical context characteristic extraction module to the optical fiber vibration data hierarchical context characteristic extraction module list variable TXList;

[0207] S306, TXCounter1=TXCounter1+1;

[0208] S307, if TXCounter1 is less than or equal to 40, go to S303; otherwise go to S308;

[0209] S308, establish the optical fiber vibration data level context characteristic output result TXOutput, the structure of the optical fiber vibration data level context characteristic output result TXOutput is the same as an element structure of the optical fiber vibration data level context characteristic list CJList, including 6 fields:

[0210] The Vb1 field of the optical fiber vibration data hierarchical context characteristic output result TXOutput = the optical fiber vibration data hierarchical context characteristic extraction module list variable TXList; specifically, the Vb1 field of the optical fiber vibration data hierarchical context characteristic output result TXOutput is a list of 40 elements, and each element in the list is a 10-dimensional vector;

[0211] The Vb2 field of the fiber vibration data level context feature output result TXOutput = creates a new list whose value is to insert a 10-dimensional vector of all 0s before the first element of its vb1 field, and delete an element at the end of its vb1 field;

[0212] The Vb3 field of the fiber vibration data level context feature output result TXOutput = creates a new list whose value is to insert a 10-dimensional vector of all 0s before the first element of its vb2 field, and delete an element at the end of its vb2 field;

[0213] The Vb4 field of the fiber vibration data level context feature output result TXOutput = creates a new list whose value is to insert a 10-dimensional vector of all 0s before the first element of its vb3 field, and delete an element at the end of its vb3 field;

[0214] The Vb5 field of the fiber vibration data level context feature output result TXOutput = creates a new list, whose value is to insert a 10-dimensional vector of all 0s before the first element of its vb4 field, and delete an element at the end of its vb4 field;

[0215] Fiber vibration data level context feature output result TXOutput Icing decision FBDecison field: an integer variable expressing whether icing occurs;

[0216] S309, updating the fields of the optical fiber vibration data level context characteristic output result TXOutput established in S308:

[0217] The Vb2 field of the optical fiber vibration data level context characteristic output result TXOutput = the optical fiber vibration data level context characteristic extraction module list variable TXList - the Vb2 field of the optical fiber vibration data level context characteristic output result TXOutput;

[0218] The Vb3 field of the optical fiber vibration data level context characteristic output result TXOutput = the optical fiber vibration data level context characteristic extraction module list variable TXList - the Vb3 field of the optical fiber vibration data level context characteristic output result TXOutput;

[0219] The Vb4 field of the optical fiber vibration data level context characteristic output result TXOutput = the optical fiber vibration data level context characteristic extraction module list variable TXList - the Vb4 field of the optical fiber vibration data level context characteristic output result TXOutput;

[0220] The Vb5 field of the optical fiber vibration data level context characteristic output result TXOutput = the optical fiber vibration data level context characteristic extraction module list variable TXList - the Vb5 field of the optical fiber vibration data level context characteristic output result TXOutput;

[0221] The icing decision FBDecison field of the fiber vibration data level context characteristic output result TXOutput = 0;

[0222] S310, outputting the optical fiber vibration data hierarchical context characteristic output result TXOutput as the result output of the optical fiber vibration data hierarchical context characteristic extraction module TXModel.

[0223] S4, using the optical fiber vibration data level context feature extraction module TXModel to process all the contents of the ice cover condition form historical data table VbList, obtain the optical fiber vibration data level context feature, and store the obtained optical fiber vibration data level context feature in the optical fiber vibration data level context feature list CJList;

[0224] S4 is specifically:

[0225] S401, result processing counter CLCounter=1;

[0226] S402, result processing first temporary variable CLTemp1:

[0227] Take out the CLCounterth element of the ice cover status historical data table VbList, and assign it to the first temporary variable CLTemp1 of the result processing;

[0228] S403, result processing second temporary variable CLTemp2:

[0229] The optical fiber vibration data hierarchical context feature extraction module TXModel is used for processing, the floating point array TXInput of the optical fiber vibration data hierarchical context feature extraction module TXModel is input as the result processing first temporary variable CLTemp1, the optical fiber vibration data hierarchical context feature output result TXOutput is obtained by processing, and the result is assigned to the result processing second temporary variable CLTemp2;

[0230] S404, assigning the result processing second temporary variable CLTemp2 to the CLCounterth element of the optical fiber vibration data level context characteristic list CJList;

[0231] S405, taking out the VbObserved field of whether thin ice appears in the CLCounterth element of the ice condition history data table VbList, and assigning it to the ice decision FBDecison field of the CLCounterth element of the optical fiber vibration data level context feature list CJList;

[0232] S406, CLCounter=CLCounter+1;

[0233] S407, if CLCounter is less than or equal to VbNum, go to S402; otherwise go to S408;

[0234] S408, the processing of step S4 ends.

[0235] S5, establishing a hierarchical icing decision neural network model CCNN; using the optical fiber vibration data hierarchical context feature list CJList to train the hierarchical icing decision neural network model CCNN, to obtain a trained hierarchical icing decision neural network model CCNN;

[0236] S5 is as follows:

[0237] S501, establish a hierarchical ice cover decision neural network model CCNN, which includes the following structure:

[0238] It has 5 input branches, each of which outputs an output array of 100 elements after passing through an LSTM layer. After the 5 input branches are processed, the outputs are merged into a 100×5 feature map. For this feature map, a decision is formed using 2 convolutional layers, 1 maximum pooling layer, and a cubic MLP network.

[0239] S502, the five input branches of the hierarchical icing decision neural network model CCNN are respectively connected to the vb1, vb2, vb3, vb4 and vb5 fields of the optical fiber vibration data hierarchical context feature list CJList, and the decision output of the hierarchical icing decision neural network model CCNN is connected to the icing decision FBDecison field of the optical fiber vibration data hierarchical context feature list CJList;

[0240] S503, using the data of the optical fiber vibration data hierarchical context feature list CJList to train the hierarchical ice coverage decision neural network model CCNN, so that the hierarchical ice coverage decision neural network model CCNN has decision-making ability;

[0241] The processing of steps S504 and S5 ends.

[0242] S6, during actual monitoring, the phase-sensitive optical time domain reflectometer is used to collect 4-second optical fiber vibration data Test, and the optical fiber vibration data hierarchical context feature extraction module TXModel and the trained hierarchical icing decision neural network model CCNN are used to analyze the collected 4-second optical fiber vibration data Test to determine whether initial icing of the transmission line occurs. If so, an early warning is issued.

[0243] S6 is specifically:

[0244] S601, using a phase-sensitive optical time domain reflectometer to collect 4 seconds of optical fiber vibration data Test;

[0245] S602, a first temporary variable PDTemp1 for judging:

[0246] The optical fiber vibration data level context feature extraction module TXModel is used for processing, the floating point array TXInput of the optical fiber vibration data level context feature extraction module TXModel is input as the optical fiber vibration data Test, the optical fiber vibration data level context feature output result TXOutput of the optical fiber vibration data level context feature extraction module TXModel is obtained, and the result is assigned to the first temporary variable PDTemp1 used for judgment;

[0247] S603, inputting the first temporary variable PDTemp1 used for judgment into the hierarchical icing decision neural network model CCNN, and the five input branches of the hierarchical icing decision neural network model CCNN are respectively connected to the vb1, vb2, vb3, vb4 and vb5 fields of the first temporary variable PDTemp1 used for judgment;

[0248] S604, a second temporary variable PDTemp2 for judging:

[0249] The decision output of the hierarchical ice coverage decision neural network model CCNN is assigned to the second temporary variable PDTemp2 used for judgment;

[0250] S605, if the second temporary variable PDTemp2 used for judgment is equal to 0, go to S606; otherwise go to S607;

[0251] S606, output that there is no initial icing of the transmission line, and go to S608;

[0252] S607, indicating that the transmission line has initial icing, an early warning is issued, and the process goes to S608;

[0253] S608, end.

[0254] In order to verify the effectiveness of the description method of the present invention, the data of a power transmission line in Jilin Province in winter were introduced, and the optical fiber vibration data of 1000 time periods were extracted, including 100 thin initial ice samples, 50 of which were used as training samples and 50 as test samples; 900 un-iced samples, 450 of which were used as training samples and 450 as test samples; the comparison results between the method of the present invention and the traditional method are as follows:

[0255] Method Name Number of correct warnings for 50 ice samples 450 false alarms for un-iced samples Method of the present invention 50 1 Traditional method 1: decision tree 25 12 Traditional method 2: Boosting + neural network 28 8 Traditional method 3: Deep neural network 37 6

[0256] It can be seen from the above table that the method of the present invention is superior to other traditional methods in terms of the number of correct warnings and the number of false alarms.

[0257] The present invention provides a method for early warning of transmission line icing based on the hierarchical context of optical fiber vibration data, which can adapt to the difference between the vibration data of the initial thin ice and no ice on the transmission line, and the difference caused by the construction of different transmission lines, and provide early warning of early icing of the transmission line.

[0258] Specifically, the present invention constructs a fiber optic vibration data hierarchical context feature extraction module and a hierarchical icing decision neural network model, which are used to hierarchically establish the characteristics of the fiber optic vibration data and find the difference between the vibration data of the initial thin ice coverage and the un-iced transmission line, thereby realizing the early icing warning of the transmission line.

[0259] The hierarchical structure of the present invention can describe the change process between different cycles of optical fiber vibration data. On the one hand, this makes it possible for the neural network to find a specific level corresponding to the ice warning, and adapt to the changes in the characteristics of optical fiber vibration data caused by different construction methods and assembly methods; on the other hand, the difference calculated by the formula between the levels can be independent of the sample, and the difference between different cycles of optical fiber vibration data can be more clearly expressed, which is helpful for dealing with the situation where there are few samples of initial ice coverage of transmission lines. Through the present invention, early warning of ice coverage of transmission lines can be carried out more effectively, and the efficiency of power grid inspection management can be improved.

[0260] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be considered as the scope of protection of the present invention.

Claims

1. A transmission line initial icing warning method based on optical fiber vibration data hierarchical context, characterized in that: The following steps are involved: S1, installing optical fiber and phase-sensitive optical time domain reflectometer on the transmission line; the phase-sensitive optical time domain reflectometer is used to collect optical fiber vibration data, collect the optical fiber vibration data and historical data of the transmission line icing situation, and store them in the icing situation history data table VbList; obtain the number of elements VbNum of the icing situation history data table; initialize the optical fiber vibration data level context feature list CJList; S2, establishing a time domain indicator expression module FEModel; the input of the time domain indicator expression module FEModel is the time domain indicator expression input array FEInput, and the output of the time domain indicator expression module FEModel is the time domain indicator expression output array FEOutput; S3, establish a fiber vibration data level context feature extraction module TXModel; the input of the fiber vibration data level context feature extraction module TXModel is a floating point array TXInput of 4000 elements, the fiber vibration data level context feature extraction module TXModel uses the time domain index expression module FEModel to process the floating point array TXInput, and the output of the fiber vibration data level context feature extraction module TXModel is a fiber vibration data level context feature output result TXOutput; S4, using the optical fiber vibration data level context feature extraction module TXModel to process all the contents of the ice cover condition form historical data table VbList, obtain the optical fiber vibration data level context feature, and store the obtained optical fiber vibration data level context feature in the optical fiber vibration data level context feature list CJList; S5, establishing a hierarchical icing decision neural network model CCNN; using the optical fiber vibration data hierarchical context feature list CJList to train the hierarchical icing decision neural network model CCNN, to obtain a trained hierarchical icing decision neural network model CCNN; S5 is as follows: S501, establish a hierarchical ice cover decision neural network model CCNN, which includes the following structure: It has 5 input branches, each of which outputs an output array of 100 elements after passing through an LSTM layer. After the 5 input branches are processed, the outputs are merged into a 100×5 feature map. For this feature map, a decision is formed using 2 convolutional layers, 1 maximum pooling layer, and a cubic MLP network. S502, the five input branches of the hierarchical icing decision neural network model CCNN are respectively connected to the vb1, vb2, vb3, vb4 and vb5 fields of the optical fiber vibration data hierarchical context feature list CJList, and the decision output of the hierarchical icing decision neural network model CCNN is connected to the icing decision FBDecison field of the optical fiber vibration data hierarchical context feature list CJList; S503, using the data of the optical fiber vibration data hierarchical context feature list CJList to train the hierarchical ice coverage decision neural network model CCNN, so that the hierarchical ice coverage decision neural network model CCNN has decision-making ability; S504, the processing of step S5 ends; S6, during actual monitoring, the phase-sensitive optical time domain reflectometer is used to collect 4-second optical fiber vibration data Test, and the optical fiber vibration data hierarchical context feature extraction module TXModel and the trained hierarchical icing decision neural network model CCNN are used to analyze the collected 4-second optical fiber vibration data Test to determine whether initial icing of the transmission line occurs. If so, an early warning is issued.

2. The method for early warning of transmission line icing based on optical fiber vibration data level context according to claim 1 is characterized in that: S1 is specifically: S101, installation of optical fiber and phase-sensitive optical time-domain reflectometers on transmission lines; The phase-sensitive optical time domain reflectometer collects optical fiber vibration data at 1000 Hz, collects the optical fiber vibration data and historical data formed by the icing condition of the transmission line, and stores them in the icing condition form historical data table VbList; wherein each element of the icing condition form historical data table VbList has 2 fields: Optical fiber vibration data array VbData: The optical fiber vibration data array VbData is a floating point array of 4000 elements. Since the acquisition frequency of the phase-sensitive optical time domain reflectometer is 1000 Hz, the optical fiber vibration data array VbData corresponds to optical fiber vibration data of 4 seconds in length; Whether thin ice coverage appears VbObserved: It is an integer variable. 1 indicates that thin ice coverage appears. At this time, the ice thickness of the transmission line is greater than 0 and less than 5 mm. 0 indicates that there is no ice coverage. S102, the number of elements in the ice cover form historical data table VbNum=the number of elements in the ice cover form historical data table VbList; S103, establish a fiber vibration data level context feature list CJList, which is initially an empty list; each element structure of the fiber vibration data level context feature list CJList contains 6 fields: The first level context feature Vb1 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector; The second-level context feature Vb2 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector; The third level context feature Vb3 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector; The fourth level context feature Vb4 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector; The fifth level context feature Vb5 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector; Icing decision FBDecison of the fiber vibration data level context feature list CJList: an integer variable expressing whether icing occurs; S104, constructing a fiber vibration data level context feature list counter CJCounter=1; S105, creating an element CJListItem of the optical fiber vibration data level context property list CJList, the element CJListItem having 6 fields: The first level context feature Vb1 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector, and the interior of the vector is all 0; The second-level context feature Vb2 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector, and the interior of the vector is all 0; The third level context feature Vb3 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector, and the interior of the vector is all 0; The fourth level context feature Vb4 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector, and the interior of the vector is all 0; The fifth level context feature Vb5 of the optical fiber vibration feature is a list of 40 elements, each element in the list is a 10-dimensional vector, and the interior of the vector is all 0; Ice cover decision FBDecison=0; S106, step S1 ends.

3. The method for early warning of transmission line icing based on optical fiber vibration data level context according to claim 1 is characterized in that: S2 is specifically: S201, establishing a time domain indicator expression module FEModel; the input of the time domain indicator expression module FEModel is a time domain indicator expression input array FEInput, and the time domain indicator expression input array FEInput is an array of 100 elements; S202, the time domain index expression input array FEInput is calculated by the following formula: (1) Time domain standard deviation index: Standard deviation σ: measures the degree of dispersion of data, and the calculation formula is: ； Among them: The calculation formula of mean μ is: ； in: represents the i-th element in the time domain indicator expression input array FEInput; N=100, represents the number of elements in the time domain indicator expression input array FEInput; (2) Time domain peak-to-peak index: The peak-to-peak value PP represents the extreme value range in the input array FEInput expressed by the time domain indicator. The calculation formula is: ； in: Indicates that the time domain indicator expresses the maximum value of the elements in the input array FEInput; Indicates that the time domain indicator expresses the minimum value of the elements in the input array FEInput; (3) Time domain skewness index: Skewness Measures the symmetry of data distribution, and the calculation formula is: ； (4) Time domain margin index: The margin R represents the relative size of the peak-to-peak value and the mean value, and the calculation formula is: ； (5) Time domain kurtosis index: Kurtosis Describes the sharpness of the data distribution, and the calculation formula is: ； (6) Time domain maximum value index: The maximum value Max represents the maximum element in the input array FEInput of the time domain indicator expression. The calculation formula is: ； (7) Time domain minimum index: The minimum value Min represents the smallest element in the input array FEInput of the time domain indicator expression. The calculation formula is: ； (8) Time domain pulse indicators: The pulse index P represents the pulse strength of the signal and is defined as: ； in: Indicates that the time domain indicator expresses the absolute value of the i-th element in the input array FEInput; (9) Time domain waveform value index: Waveform value A measure of the overall volatility of the signal, calculated as: ； (10) Time domain energy index: Energy E represents the sum of the squares of all elements in the input array FEInput, expressed in the time domain. The calculation formula is: ； S203, establishing a time domain index expression output array FEOutput as an array of 10 elements; S204, storing the 10 index results calculated in S202 in sequence into the time domain index expression output array FEOutput; S205, outputting the time domain index expression output array FEOutput as the result of the time domain index expression module FEModel.

4. The early warning method for early icing of transmission lines based on optical fiber vibration data level context according to claim 1 is characterized in that: S3 is specifically: S301, establish a fiber vibration data level context feature extraction module TXModel; the input of the fiber vibration data level context feature extraction module TXModel is a floating point array TXInput of 4000 elements; S302, set the first counter TXCounter1 of the optical fiber vibration data hierarchical context feature extraction module to 1; construct an optical fiber vibration data hierarchical context feature extraction module list variable TXList = empty list; S303, the first temporary variable TXTemp1 of the optical fiber vibration data level context feature extraction module = takes out the (TXCounter1-1)×100th to (TXCounter1-1)×100+99th elements of the floating point array TXInput, which is 100 elements in total; S304, the second temporary variable TXTemp2 of the optical fiber vibration data level context feature extraction module: The first temporary variable TXTemp1 of the optical fiber vibration data level context feature extraction module is processed by using the time domain indicator expression module FEModel, the time domain indicator expression input array FEInput of the time domain indicator expression module FEModel = the first temporary variable TXTemp1 of the optical fiber vibration data level context feature extraction module, and the time domain indicator expression output array FEOutput is obtained; the second temporary variable TXTemp2 of the optical fiber vibration data level context feature extraction module = the time domain indicator expression output array FEOutput; S305, adding the second temporary variable TXTemp2 of the optical fiber vibration data hierarchical context characteristic extraction module to the optical fiber vibration data hierarchical context characteristic extraction module list variable TXList; S306, TXCounter1=TXCounter1+1; S307, if TXCounter1 is less than or equal to 40, go to S303; otherwise go to S308; S308, establish the optical fiber vibration data level context characteristic output result TXOutput, the structure of the optical fiber vibration data level context characteristic output result TXOutput is the same as an element structure of the optical fiber vibration data level context characteristic list CJList, including 6 fields: The Vb1 field of the optical fiber vibration data hierarchical context characteristic output result TXOutput = the optical fiber vibration data hierarchical context characteristic extraction module list variable TXList; specifically, the Vb1 field of the optical fiber vibration data hierarchical context characteristic output result TXOutput is a list of 40 elements, and each element in the list is a 10-dimensional vector; The Vb2 field of the fiber vibration data level context feature output result TXOutput = creates a new list whose value is to insert a 10-dimensional vector of all 0s before the first element of its vb1 field, and delete an element at the end of its vb1 field; The Vb3 field of the fiber vibration data level context feature output result TXOutput = creates a new list whose value is to insert a 10-dimensional vector of all 0s before the first element of its vb2 field, and delete an element at the end of its vb2 field; The Vb4 field of the fiber vibration data level context feature output result TXOutput = creates a new list whose value is to insert a 10-dimensional vector of all 0s before the first element of its vb3 field, and delete an element at the end of its vb3 field; The Vb5 field of the fiber vibration data level context feature output result TXOutput = creates a new list, whose value is to insert a 10-dimensional vector of all 0s before the first element of its vb4 field, and delete an element at the end of its vb4 field; Fiber vibration data level context feature output result TXOutput Icing decision FBDecison field: an integer variable expressing whether icing occurs; S309, updating the fields of the optical fiber vibration data level context characteristic output result TXOutput established in S308: The Vb2 field of the optical fiber vibration data level context characteristic output result TXOutput = the optical fiber vibration data level context characteristic extraction module list variable TXList - the Vb2 field of the optical fiber vibration data level context characteristic output result TXOutput; The Vb3 field of the optical fiber vibration data level context characteristic output result TXOutput = the optical fiber vibration data level context characteristic extraction module list variable TXList - the Vb3 field of the optical fiber vibration data level context characteristic output result TXOutput; The Vb4 field of the optical fiber vibration data level context characteristic output result TXOutput = the optical fiber vibration data level context characteristic extraction module list variable TXList - the Vb4 field of the optical fiber vibration data level context characteristic output result TXOutput; The Vb5 field of the optical fiber vibration data level context characteristic output result TXOutput = the optical fiber vibration data level context characteristic extraction module list variable TXList - the Vb5 field of the optical fiber vibration data level context characteristic output result TXOutput; The icing decision FBDecison field of the fiber vibration data level context characteristic output result TXOutput = 0; S310, outputting the optical fiber vibration data hierarchical context characteristic output result TXOutput as the result output of the optical fiber vibration data hierarchical context characteristic extraction module TXModel.

5. The method for early warning of transmission line icing based on optical fiber vibration data level context according to claim 1, characterized in that: S4 is specifically: S401, result processing counter CLCounter=1; S402, result processing first temporary variable CLTemp1: Take out the CLCounterth element of the ice cover status historical data table VbList, and assign it to the first temporary variable CLTemp1 of the result processing; S403, result processing second temporary variable CLTemp2: The optical fiber vibration data hierarchical context feature extraction module TXModel is used for processing, the floating point array TXInput of the optical fiber vibration data hierarchical context feature extraction module TXModel is input as the result processing first temporary variable CLTemp1, the optical fiber vibration data hierarchical context feature output result TXOutput is obtained by processing, and the result is assigned to the result processing second temporary variable CLTemp2; S404, assigning the result processing second temporary variable CLTemp2 to the CLCounterth element of the optical fiber vibration data level context characteristic list CJList; S405, taking out the VbObserved field of whether thin ice appears in the CLCounterth element of the ice condition history data table VbList, and assigning it to the ice decision FBDecison field of the CLCounterth element of the optical fiber vibration data level context feature list CJList; S406, CLCounter=CLCounter+1; S407, if CLCounter is less than or equal to VbNum, go to S402; otherwise go to S408; S408, the processing of step S4 ends.

6. The method for early warning of transmission line icing based on optical fiber vibration data level context according to claim 1, characterized in that: S6 is specifically: S601, using a phase-sensitive optical time domain reflectometer to collect 4 seconds of optical fiber vibration data Test; S602, a first temporary variable PDTemp1 for judging: The optical fiber vibration data level context feature extraction module TXModel is used for processing, the floating point array TXInput of the optical fiber vibration data level context feature extraction module TXModel is input as the optical fiber vibration data Test, the optical fiber vibration data level context feature output result TXOutput of the optical fiber vibration data level context feature extraction module TXModel is obtained, and the result is assigned to the first temporary variable PDTemp1 used for judgment; S603, inputting the first temporary variable PDTemp1 used for judgment into the hierarchical icing decision neural network model CCNN, and the five input branches of the hierarchical icing decision neural network model CCNN are respectively connected to the vb1, vb2, vb3, vb4 and vb5 fields of the first temporary variable PDTemp1 used for judgment; S604, a second temporary variable PDTemp2 for judging: The decision output of the hierarchical ice coverage decision neural network model CCNN is assigned to the second temporary variable PDTemp2 used for judgment; S605, if the second temporary variable PDTemp2 used for judgment is equal to 0, go to S606; otherwise go to S607; S606, output that there is no initial icing of the transmission line, and go to S608; S607, indicating that the transmission line has initial icing, an early warning is issued, and the process goes to S608; S608, end.

Images